Based on the Kelvin ship wake mode, the ocean wave mode and the two-scale mode for electromagnetic scattering from rough surfaces, we propose a new method of simulation for Kelvin ship wake imaging by multi-look direction SAR. We have simulated the VV normalized radar backscattering cross-section (NRCS) of the ship wakes from different look directions in 2-D space firstly. The results show that the transverse waves can be viewed clearly when the radar look direction is parallel to the ship's track direction, while the divergent waves can be viewed clearly when the radar look direction is perpendicular to the ship's track direction. When the radar look direction is neither perpendicular nor parallel to the ship's track direction, Kelvin arms are imaged as a bright line and a dark line according to the relationship between spread direction of crest waves and ship's track direction. The waves whose track direction is more parallel to the ship's track direction can be imaged as a brighter line. Moreover, the simulation results also draw this conclusion another, when radar look direction is more near perpendicular to the ship's track direction, the angle between two Kelvin arms is smaller as a result of these circumstances gunpoint at the stern wave scattering cross-section is as large as bow wave. It's show that the simulated results agree quite well with the ERS-1/2 images.

Scene simulation is a necessary work in near field passive microwave remote sensing. A 3-D scene simulation model of microwave radiometric imaging based on ray tracing method is present in this paper. The essential influencing factors and general requirements are considered in this model such as the rough surface radiation, the sky radiation witch act as the uppermost illuminator in out door circumstance, the polarization rotation of the temperature rays caused by multiple reflections, and the antenna point spread function witch determines the resolution of the model final outputs. Using this model we simulate a virtual scene and analyzed the appeared microwave radiometric phenomenology, at last two real scenes of building and airstrip were simulated for validating the model. The comparison between the simulation and field measurements indicates that this model is completely feasible in practice. Furthermore, we analyzed the signatures of model outputs, and achieved some underlying phenomenology of microwave radiation witch is deferent with that in optical and infrared bands.

How to cull shadows and extract needed information accurately is particularly significant. For major remote sensing applications, it may be preferable that shadows are minimized and the detailed information in high-resolution satellite imagery is clear. Firstly this paper reviews some of basic methods of detecting and removing shadows, and outlines their disadvantages. Then taking Nanjing city as study area, we propose a novel method combing spatial-distribution relation with classification to detect building shadows from IKONOS imagery. When detecting and extracting shadows, a majority index based on neighborhood analysis is provided, and a 5-meter buffer analysis is operated after supervised classification. When removing the shadows, a piecewise linear contrast stretch and histogram match are used. The results show that the accuracy of shadows detection and extraction is 92.3%, but texture analysis is 88.1%, and the detail information within shadows regions is enhanced, and there are no bright edges around shadows regions by applying the techniques developed in this paper.

The Chinese moderate imaging spectra radiometer (CMODIS) is one of main payload in the third space ship SZ-3 of China launched in March 2002. CMODIS worked about a half of year on orbits. In this paper, first, the properties and characteristics of CMODIS are briefly introduced. Second, the technique of data procession algorithms are discussed in detail and the marine environment products by CMODIS will be presented in the paper. Finally, its potentiality for marine application is mentioned. The results show that CMODIS has its latent capability for the application of marine environment detection, management and protection of marine resources, and the national rights and interests. Meanwhile some suggestions are proposed to modify the next generation moderate imaging spectra radiometer in the future plan.

Land cover data are essential to most global change research objectives, including the assessment of current environmental conditions and the simulation of future environmental scenarios that ultimately lead to public policy development. Chinese Academy of Sciences generated a nationwide land cover database in order to carry out the quantification and spatial characterization of land use/cover changes (LUCC) in 1990s. In order to improve the reliability of the database, we will update the database anytime. But it is difficult to obtain remote sensing data to extract land cover change information in large-scale. It is hard to acquire optical remote sensing data in Chengdu plain, so the objective of this research was to evaluate multitemporal ENVISAT advanced synthetic aperture radar (ASAR) data for extracting land cover change information. Based on the fieldwork and the nationwide 1:100000 land cover database, the paper assesses several land cover changes in Chengdu plain, for example: crop to buildings, forest to buildings, and forest to bare land. The results show that ENVISAT ASAR data have great potential for the applications of extracting land cover change information.

The study site is selected in a Malaysian tropical rainforest area, which consists of a mixture of plain, hilly and mountainous terrain. Digital Elevation Model (DEM) images were generated from nine RADARSAT-1 imageries (F, S and W beam modes) which make up six stereo pair combinations. The DEM accuracies for all the stereo combinations have been validated and compared to each other. The results show that numerous factors affect the final DEM accuracy. In flat areas, the final DEM accuracy is highly correlated to the stereo intersection geometry of the different image combinations. The higher the stereo intersection angle of the same beam mode, the better the accuracy of the final DEM.

Terrestrial net primary production (NPP), as an important component of carbon cycle on land, not only indicates directly the production level of vegetation community on land, but also shows the status of terrestrial ecosystem. What's more, NPP is also a determinant of carbon sinks on land and a key regulator of ecological processes, including interactions among tropic levels. In the study, three existing models are combined with each other to assess net primary production in Haihe Basin, China. The photosynthetically active radiation (PAR) model of Monteith is used for the calculation of absorbed photosynthetically active radiation (APAR), the light utilization efficiency model of Potter et al. is used for determining the light utilization efficiency, and the surface energy balance system (SEBS) of Su is used into Potter's model to describe water stress in land wetness conditions. To assess NPP, We use NOAA-AVHRR data from November 2003 to September 2004 and the corresponding daily data of temperature and hours of sunshine obtained from meteorological stations in Haihe Basin, China. After atmospheric, geometrical and radiant corrections, every ten days NOAA data are processed to become an image of NDVI by means of the maximal value composition method (MVC) in order to eliminate some noises. Using these data, we compute NPP in spring season and spring season of 2004 in Haihe Basin, China. The result shows, in Haihe Basin, NPP for spring season is averaged to 336.10gC•m^-2, and 709.16 gC•m^-2 for autumn season. In spatial distribution, NPP is greater in both ends than in middle for spring season, and decrease increasingly from north to south for autumn season. Future work should rely on the integration of high and low resolution images to assess net primary production, which will probably have more accurately estimation.

Thanks to the shift of mapping companies into a full digital workflow (digital camera and digital processing solution), new generation of products are now available to the end-users. Such new products take all benefit of new paradigm introduced by full digital workflow allowing automation of production processes. Dense Digital Surface Model and TrueOrtho products are such new generation products and were recently introduced into China.

Spectrum of healthy green vegetation shows idiographic features of "peak and valley", the spectral curve will vary when crop's biochemical status changes (e.g. disease harmed). Normalized Difference Vegetation Index (NDVI) is an important vegetation index and has been proved to be very useful to vegetation change detection, vegetation classification and some parameters calculation. Based on the differences of spectra information and characteristics between multi-temporal hyperspectral images, a new adjustable vegetation index, Multi-Temporal NDVI (MT-NDVI), is provided in this paper. Comparing to the classification of Spectral Angle Mapper (SAM), mapping and analysis using MT-NDVI data can be well utilized for monitoring and recognizing crop disease from multi-temporal airborne PHI (Pushbroom Hyperspectral Imager) image data acquired at the same field. The applicable result shows that MT-NDVI is suitable way to extract crop disease information and estimate disease degrees.

Satellite-based hyper-spectral imaging became a reality in November 2000 with the successful launch and operation of the Hyperion system on board the EO-1 platform. Hyperion is a pushbroom imager with 220 spectral bands in the 400-2500 nm wavelength range, a 30-meter pixel size and a 7.5 km swath. The objective of this research is to use the Hyperion image for deriving the Rocks/Minerals Information. This paper introduces a complete processing flow from raw Hyperion image to geological theme information map, including radiance calibration and correction, atmospheric correction and geometrical correction, feature extraction and Selection, and spectral mapping by multi-method on the base of referring to the USGS spectral library and ground radiometric measurement data. The study explored the utility of Hyperion data in alteration mineral mapping. Two Hyperion images of the BeiYa in the northwest of YunNan was acquired and evaluated for alteration zone mapping. The results show that the alteration zones in the study area can be identified from Hyperion data very efficiently. The mineralogical and lithologic information extracted from Hyperion data is largely consistent with the geological map and previous research results.

The evaluation index system of urban land intensive use potential is built with IKONOS remote sensing data as the main data source. Based on this, quantitative evaluation model of land intensive use potential is constructed through BP neural network method. This model is applied to evaluate the potential of land intensive use in Shijiazhuang. It has been proved that this kind of evaluation system, i.e. remote sensing image as the main data resource and BP model as the evaluation method, to evaluate urban land intensive use potential, is more effective, more practicable, more convenient and the evaluation result is more objective.

In remote sensing, angle information and polarization attach great importance to the precise discrimination of land objects. This paper deals with the performance of soil reflectance spectrum dependent on such factors as incidence zenith angle, wave band, polarization, water content and soil type. Our results show that soil reflectance spectrum in 2π space is strongly affected by incidence zenith angle and water content, as well as by the other factors. In the end, the cause of such spectrum performance is analyzed.

Resourcesat-1(IRS-P6) was built by Indian Space Research Organisation and launched in October, 2003. Resourcesat-1 carries three imaging sensors--a moderate resolution camera Advanced Wide Field Sensor (AWFS), a moderate resolution Linear Imaging Self Scanning-III device (LISS-III) and a high resolution Linear Imaging Self Scanning-IV device (LISS-IV) to get different spatial resolution image. Apparent reflectance of LISS-III image is calculated in Poyang Lake, south of China. The mean Digital Number (DN) in Red Band image of part of this district is 108.9, and Stdev is 62.72. The mean apparent reflectance of this image is 0.17, and Stdev is 0.093. Apparent reflectance can be used in the normalized difference vegetation index (NDVI) calculation, atmospheric correction, and others. NDVI value which is calculated by apparent reflectance can represent surface status accurately.

This study sought to develop a modified change vector analysis(CVA) using normalized multi-temporal data to detect urban vegetation change. Because of complex change in urban areas, modified CVA application based on NDVI and mask techniques can minify the effect of non-vegetation changes and improve upon efficiency to a great extent. Moreover, drawing from methods in Polar plots, the extended CVA technique measures absolute angular changes and total magnitude of perpendicular vegetation index (PVI) and two of Tasseled Cap indices (greenness and wetness). Polar plots summarized change vectors to quantify and visualize both magnitude and direction of change, and magnitude is applied to determine change pixels through threshold segmentation while direction is applied as pixel's feature to classifying change pixels through supervised classification. Then this application is performed with Landsat ETM+ imageries of Wuhan in 2002 and 2005, and assessed by error matrix, which finds that it could detect change pixels 95.10% correct, and could classify change pixels 91.96% correct in seven change classes through performing supervised classification with direction angles. The technique demonstrates the ability of change vectors in multiple biophysical dimensions to vegetation change detection, and the application can be trended as an efficient alternative to urban vegetation change detection and classification.

In order to extract objective information more effectively, cloud should be removed from these remote sensing images contaminated by clouds. High accurate and automatic detection of clouds in satellite multi-spectral data lays a good foundation for the cloud classification or cloud removing. This paper is established in studying a simple, quick, automatic and efficient cloud detection algorithm based on spectral characteristic for different earth's surface. The authors take account of the spectrums specificity of different objects (cloud, snow, desert, land, plateau, vegetation, water and so on) and the MODIS instrument channel characteristic. We experiment on a lot of MODIS data including different times (spring, summer, autumn and winter) and different earth's surface (snow, desert, land, plateau, vegetation, water and so on). The experimental results indicate that the multi-spectral synthesis algorithm of the composite normalization algebraic operation for cloud detection is very useful. It can detect most clouds of MODIS data, including very thin cloud, on the different earth's surfaces, especially snow and desert.

Traditionally, field-based investigation is the main method to investigate greenbelt in urban environment, which is costly and low updating frequency. In higher resolution image, the imagery structure and texture of tree canopy has great similarity in statistics despite the great difference in configurations of tree canopy, and their surface structures and textures of tree crown are very different from the other types. In this paper, we present an automatic method to detect tree crowns using high resolution image in urban environment without any apriori knowledge. Our method catches unique structure and texture of tree crown surface, use variance and mathematical expectation of defined image window to position the candidate canopy blocks coarsely, then analysis their inner structure and texture to refine these candidate blocks. The possible spans of all the feature parameters used in our method automatically generate from the small number of samples, and HOLE and its distribution as an important characteristics are introduced into refining processing. Also the isotropy of candidate image block and holes' distribution is integrated in our method. After introduction the theory of our method, aerial imageries were used ( with a resolution about 0.3m ) to test our method, and the results indicate that our method is an effective approach to automatically detect tree crown in urban environment.

The airborne hyperspectral remote sensing data, such as PHI, OMIS, has the virtues of high spatial and spectral resolution. Hence, from the view of target classification we can consider that it can provide the ability of discriminating targets more detailedly than other data. So it's important to extract thematic information and update database using this kind of data. Whereas, the hyperspectral data has abundant bands and high between-band correlation, the traditional classification methods such as maximum likelihood classifier (MLC) and spectral angle mapper (SAM) have performed poorly in thematic information extraction. For this reason, we present a new method for thematic information extraction with hyperspectral remote sensing data. We perform classification by means of combining the self-organizing map (SOM) neural network which is considered as full-pixel technique with linear spectral mixture analysis (LSMA) which is considered as mixed-pixel technique. The SOM neural network is improved from some aspects to classify the pure data and find the mixed data. And then the mixed data are unmixed and classified by LSMA. The result of experiment shows that we can have the better performance in thematic information extraction with PHI by this means.

In order to evaluate the potential of ENVISAT/ASAR (ENVIronment SATellite /Advanced Synthetic Aperture Radar) imagery with multi-incidence angle and multi-polarization for Poyang lake wetland study, a microwave canopy backscatter model for herbaceous vegetation was used to simulate the ASAR backscatter from wetland vegetation at two different growing status. Field measurements from areas covered with Miscanthus floridulus acquired in middle April and in May, 2005 were averaged respectively as model inputs to simulate ASAR backscatter from different vegetation components with seven incidence angles (15° to 45°). In April, 2005, Miscanthus floridulus still has not spread adequately, HV polarization is more sensitive to vegetation canopy parameters than HH and VV; HH is more suitable for retrieving surface moisture content and roughness; and VV cannot differentiate the influence of vegetation from surface effects. Inundation results in significant increment in HV backscatter; for HH, it depends on vegetation growing status, for instance, a reduction in areas of low vegetation and an increase in areas of high vegetation; for VV polarization, enhancement in total backscatter is insufficient to be used in flood detection. However, as the height of Miscanthus floridulus is more than 1 meter in late May, the simulated ASAR backscatter is not sensitive to the variations of vegetation parameters, surface parameters, or even inundation.

To meet the demand of large-scale agricultural monitoring system with remote sensing, extracting crop area planted must be rapid, precise and reliable. In this paper, winter wheat identification with MODIS data in 2004 is taken as example in North China. Applying spectral analysis and integrating genetic algorithm with neural network (GA-BP) is proposed, which gives attention to two optimization algorithm, genetic algorithm and back propagation algorithm. According to the spectral and biological characteristics of winter wheat, Red, Blue, NIR, ESWIR, LSWI, EVI are selected as characteristic parameters. Then GA-BP algorithm is used for winter wheat identification. Results show that compared with maximum likelihood and back propagation neural network classification algorithm, the GA-BP algorithm can not only run with better efficiency, but also achieve best accuracy of identification. Therefore, it is the operational method for agricultural condition monitoring with remote sensing and information service system at national level.

This paper introduces the outline of the railway route design in the three-dimensional stereo model by the railway route design principle and methods based on seamless stereo orthoimage database in MicroStation. The main work is the route design. Then the visualization of the railway route can be done according to the terrain analysis of Geopak and the senior modeling function of MicroStation. The main works are as follows: 1. The method of traditional railway route design is analyzed. 2. The main principle, technique and outline of the plane, longitudinal section and cross section design are analyzed based on seamless stereo orthoimage database. 3. The flow of the railway route design are investigated based on the seamless stereo orthoimage. The characteristics of the stereo geometry model of the seamless stereo orthoimage database can be presented asgeometry-rectified, y-parallax-free, geo-referenced accurate measurement and so on. It can be easily integrated into GIS, so the 3D measurement, design and analysis can be done by the model. The railway design can be carried on in a three-dimensional visualization and area survey seamless environment. MicroStation has better character of 3D-modeling, visualization and expansibility. The design process and result can be visualized. Geopak runs on MicroStation, and is integrated into MicroStation entirely. So it supports general interactive design and visualization. All-sided solving scheme is supported for the road design of the plane, longitudinal section and cross section in the Site module.

Satellite SAR data offers land managers an affordable, all-weather capability for detailed land cover mapping. Visual classification of these data may be more appropriate to the resource base in many developing countries and human interpreters can often overcome problems of speckle more effectively than automated classification procedures. We report work in progress on the visual interpretation of SAR data to classify land cover types within tropical savannas. Airborne L-band SAR data for a region in Belize, Central America is degraded to approximate the single polarisation hh and dual polarization hh/hv data that is expected from the ALOS PALSAR satellite sensor. Interpretations of these two types of data by multiple interpreters were compared to explore how the number of polarizations, the effective spatial resolution and the visual presentation of the SAR data affected the ability of interpreters to classify land cover. An average classification accuracy of 78% for hh and 85% for hh/hv data were achieved for all classes and interpreters. Denser high forest areas were accurately interpreted using both data sets, whilst a red-green colour composite of the hh/hv data allowed grass dominated areas to be separated from areas of savanna woodland. Conclusions are drawn about the benefits of certain presentations of backscatter data to assist visual interpretation.

The objective of this paper is the description of the development and the validation, using airborne hyper-spectral imagery data, of a non-conventional technique for the vegetation information extraction. The proposed approach namely the universal pattern decomposition method (UPDM) is tailored for hyper-spectral imagery analysis, which can be explained using two analysis methods: spectral mixing analysis and multivariate analysis. For the former, the UPDM expresses the spectrum of each pixel as the linear sum of three fixed, standard spectral patterns (i.e., the patterns of water, vegetation, and soil); each coefficient represents the ratio of spectral patterns of three components. If we think of the UPDM as multivariate analysis, standard patterns are interpreted as an oblique coordinate system, and coefficients are thought of as the coordinates of a pixel's reflectance. The later explanation is much more comprehensible than the former for the reason of additional supplementary pattern presence when necessary. The vegetation index based on the UPDM (VIUPD) is expressed as a linear sum of the pattern decomposition coefficients. Here, the VIUPD was used to examine vegetation amounts and degree of terrestrial vegetation vigor; VIUPD results were compared with results by the normalized difference vegetation index (NDVI), and an enhanced vegetation index (EVI). This paper described the calculation of VIUPD, using AVIRIS airborne remotely sensed data. The results showed that the VIUPD reflects vegetation and vegetation activity more sensitively than the NDVI and EVI.

This paper analyzes some changed reasons and characteristics of close range photogrammetry when a non-metric digital camera replaced a traditional metric camera. Then it discusses some problems of digital close range photogrammetric network with variant interior elements, and gives some mended mathematical models of photogrammetric process, such as single photo resection and multi-photo intersection, multi-photo resection and intersection, self-calibration bundle adjustment. Calculating methods of these models are described. At last, two examples of application are given.

Edge-detection of LIDAR depth-image is an important work for further image analysis. Based on the theory of Canny algorithm, this paper discusses insufficiencies of Canny operator and proposes an improved method. Instead of using Gaussian smoothing filter, the improved algorithm carries on the smoothing operation by an adaptive median filter for the characteristics of LIDAR depth-image. As a result, it can not only eliminate noises effectively but also protect unclear edges. Gradient computation and determination of edge points are also improved, gradient magnitudes of pixels are calculated with first-order derivatives within eight neighborhoods instead of four, and the precision of edge location is enhanced consequently. Considering the deficiency of uniform threshold for the whole image in Canny operator and its non-objectivity in determining threshold values, the improved algorithm divides image into a number of sub-images and detects edges with adaptive threshold values respectively. Therefore, edge points with low height values are protected and adaptation of the algorithm is also improved. Datasets from urban areas were selected to test this algorithm. The results show that the improved algorithm can make up for the disadvantages of canny algorithm, and can detect edges of LIDAR depth-images effectively.

Raw LIDAR data is a irregular spacing 3D point cloud including reflections from bare ground, buildings, vegetation and vehicles etc., and the first task of the data analyses of point cloud is feature extraction. However, the interpretability of LIDAR point cloud is often limited due to the fact that no object information is provided, and the complex earth topography and object morphology make it impossible for a single operator to classify all the point cloud precisely 100%. In this paper, a hierarchy method for feature extraction with LIDAR data and aerial images is discussed. The aerial images provide us information of objects figuration and spatial distribution, and hierarchic classification of features makes it easy to apply automatic filters progressively. And the experiment results show that, using this method, it was possible to detect more object information and get a better result of feature extraction than using automatic filters alone.

Laser scanning is a fast and precise technique for resampling the earth into irregular pattern. The large number of laser points hitting planar facades in urban areas makes it possible to extract objects (buildings, vegetation, etc.) in the areas. This paper has presented a new approach to extract buildings through analyzing characteristic of contours generated from LIDAR point data, which has been proved to be a robust and reliable method.

The main objective of this study is to derive plant nitrogen (N) status and aboveground biomass via satellite remote sensing. To understand canopy spectral reflectance, the focus of the first part was set on the analysis of spectral signatures of winter wheat during its vegetation period under different N treatments. Spectral reflectance at different phenological stages, measured by a spectroradiometer (ASD HandHeld), is related to agronomy parameters like plant N, aboveground biomass and leaf area index (LAI). For this purpose, an extensive field survey was carried out in Huimin County in the North China Plain. For detection of plant N status of winter wheat and biomass on regional scale, hyperspectral (EO-1 Hyperion) and radar (Envisat ASAR) remote sensing data were obtained. First results of preprocessing of remote sensing data are presented in this contribution.

Synthetic Aperture Radar remote sensing has been effectively used in water compliance and enforcement, especially in ship detection, but it is still very difficult to classify or identify vessels in inland water only using existing SAR image. Nevertheless some experience knowledge can help, for example waterway channel is of great significance for water traffic management and illegal activity monitoring. It can be used for judging a vessel complying with traffic rules or not, and also can be used to indicate illicit fishing vessels which are usually far away from navigable waterway channel. For illicit vessel identification speed and efficiency are very important, so it will be significant if we can extract waterway channel directly from SAR images and use it to identify illicit vessels. The paper first introduces the modified two-parameter CFAR algorithm used to detect ship targets in inland waters, and then uses principal curves and neural networks to extract waterway channel. Through comparing the detection results and the extracted waterway channel those vessels not complying with water traffic rules or potential illicit fishing vessels can be easily identified.

This paper focuses on the problem of semiautomatic extraction of building information from high-resolution satellite images covering urban areas. This information includes buildings height, 2-D structure, and variation detection. An increasing number of applications require accurate and up-to-date cartographic and 3-D data. We introduce a set of accurate and automatic algorithms based on high-resolution remote sensing imagery such as Quickbird. Our method exploits the relationship between buildings height and their shadow in satellite images. Firstly we use our multiple-restriction method to extract the shadow information. Then we can adopt their relationship to compute building height information. In the process of building 2-D information extraction we introduce a new method about morphology used to do edge detection. After that we utilize the methods including image processing, image analyzing, and pattern recognition to detect building 2-D structure. Based on the statistical skewness of image we introduce the conception of variation coefficient. Using this algorithm we can make sure the geographic position of variation detection easily and quickly. Our method involves thresholds, most of them tuned with respect to practical situation and the physical characteristics of the image. Results are shown and discussed on different images.

Extraction of linear features is a classical problem in Remote Sensing image processing. In the last twenty years, it is still difficult to extract linear features embedded in extremely high noise or when the SNR (signal to noise) is low. In this paper, an adaptive algorithm based on beamlet transform is proposed to extract linear features from remote sensing image, which can detect lines with any orientation, location and length, the parameter can be adaptively determined by histogram of beamlet energy function distribution to avoid subjective setting. The experimental results show that the method proposed extract linear features accurately even from high noise remote sensing image and has a better performance. It can be suited to remote sensing images processing and in practice it has surprisingly powerful and apparently unprecedented capabilities.

Ship features in high-resolution spaceborne Synthetic Aperture Radar (SAR) imagery has crucial significance for ship classification from satellite. In this paper, we discuss the features of merchant Ships including oil tanker, container ship and bulk carrier in SAR imagery, which is comprised of geometrical feature, scattering feature, tonnage information with Radar Cross Section (RCS) and wake. The study show that the ship lengths measured from SAR imagery has a good correlation with the real lengths, but the correlation of ship beam is worse. Ship scattering feature has positive correlation with the ship structure, which maybe is a feature to distinguish container ship from other vessels. A new equation about ship length and its displacement in tons is presented in this paper. The relation suggested by Skolnik M.I between ship tonnage and RCS is tested but not validated. We also validate the means of extracting ship speed by ship turbulence wake in SAR imagery.

Automatic image registration is important for many multiframe-based image analysis applications. With an increasing number of images collected every day from different sensors, automated registration of multi-sensor/multi-spectral images has become an important issue. A wide range of registration techniques exists for different types of applications and data sources, however no algorithm is known that can accurately register multi-source images consistently. This research addresses this problem by investigating the development of a fully automatic registration system for remote sensing images. The development of this new automatic image registration method is based on the extraction and matching of common features that are visible in both images. The algorithm involves the following five steps: noise removal, edge extraction, edge linking pattern extraction and pattern matching.

Conventional human-visual and manual field pavement crack detection method and approaches are very costly, time-consuming, dangerous, labor-intensive and subjective. They possess various drawbacks such as having a high degree of variability of the measure results, being unable to provide meaningful quantitative information and almost always leading to inconsistencies in crack details over space and across evaluation, and with long-periodic measurement. With the development of the public transportation and the growth of the Material Flow System, the conventional method can far from meet the demands of it, thereby, the automatic pavement state data gathering and data analyzing system come to the focus of the vocation's attention, and developments in computer technology, digital image acquisition, image processing and multi-sensors technology made the system possible, but the complexity of the image processing always made the data processing and data analyzing come to the bottle-neck of the whole system. According to the above description, a robust and high-efficient parallel pavement crack detection algorithm based on Multi-Scale Space is proposed in this paper. The proposed method is based on the facts that: (1) the crack pixels in pavement images are darker than their surroundings and continuous; (2) the threshold values of gray-level pavement images are strongly related with the mean value and standard deviation of the pixel-grey intensities. The Multi-Scale Space method is used to improve the data processing speed and minimize the effectiveness caused by image noise. Experiment results demonstrate that the advantages are remarkable: (1) it can correctly discover tiny cracks, even from very noise pavement image; (2) the efficiency and accuracy of the proposed algorithm are superior; (3) its application-dependent nature can simplify the design of the entire system.

An approach for extraction and detection urban impervious surface was proposed in this paper, in which a decision tree classifier based on data learning algorithm was employed using Landsat TM/ETM data in 1988, 1994 and 2002 at same season. The feature subset was constructed with spectral, spatial and change information related to the characters of urban impervious surface. The samples from the higher spatial resolution image were dealt with CART algorithm. The extraction and change detection were performance with the decision tree classifier, and change information of 1994-2002 and 1988-1992 was verified by overlay analysis from GIS for the reasonability. The result of extraction impervious surface for six urban types was shown that the overall accuracy was 88.1% compared with 69.3% of MLC (maximum-likelihood Classifier) in 2002, and the detection accuracy for the five change types was 89.1% and 91.4% between 1994 and 2002, 1988 and 1994 respectively. The research has been demonstrated that the proposed approach is of capability for the change detection and can be achieved better accuracy using medium spatial resolution remotely sensed data.

Linear features are usually extracted from SAR imagery by a few edge detectors derived from the contrast ratio edge detector with a constant probability of false alarm. On the other hand, the Hough Transform is an elegant way of extracting global features like curve segments from binary edge images. Randomized Hough Transform can reduce the computation time and memory usage of the HT drastically. While Randomized Hough Transform will bring about a great deal of cells invalid during the randomized sample. In this paper, we propose a new approach to extract linear features on SAR imagery, which is an almost automatic algorithm based on edge detection and Randomized Hough Transform. The presented improved method makes full use of the directional information of each edge candidate points so as to solve invalid cumulate problems. Applied result is in good agreement with the theoretical study, and the main linear features on SAR imagery have been extracted automatically. The method saves storage space and computational time, which shows its effectiveness and applicability.

Based on the introduction to the characters and constructing flow of space semantic model, the feature space and context of house information in high resolution remote sensing image are analyzed, and the house semantic network model of Quick Bird image is also constructed. Furthermore, the accuracy and practicability of space semantic model are checked up through extracting house information automatically from Quick Bird image after extracting candidate semantic nodes to the image by taking advantage of grey division method, window threshold value method and Hough transformation. Sample result indicates that its type coherence, shape coherence and area coherence are 96.75%, 89.5 % and 88 % respectively. Thereinto the effect of the extraction of the houses with rectangular roof is the best and that with herringbone and the polygonal roofs is just ideal. However, the effect of the extraction of the houses with round roof is not satisfied and thus they need the further perfection to the semantic model to make them own higher applied value.

The water body information is accurately extracted from remotely sensed images with the method of normalized ratio index, and the water information is greatly enhanced through restricting the brightness of backgrounds. What's more, there is no noise formed by shadows in results. However, the spatial resolution of most images used for water extraction is usually not high enough to identify water body clearly. Fusion of remotely sensed images with different spatial resolution can solve this problem. Four data fusion methods such as Modified Brovey Transform (MBT), Multiplication Transform (MLT), Smoothing Filter-based Intensity Modulation Transform (SFIMT) and High Pass Filter Transform (HPTF) have been applied to merge ETM+ panchromatic band with multi-spectral band data. Normalized ration method is adopted to extract water body information from both original and merged images. The effect of data fusion and extracting result are validated and evaluated by qualitative analysis and quantitative statistical calculation. SFIMT model enjoys the best maintenance of spectral quality from the multi-spectral bands. On the other hand, MLT model has the highest spatial frequency information gain. In the data fusion algorithms, SFIMT is the optimization data fusion method appropriate to the normalized ration water extracting model.

Fuzzy edge is the essential character of image and the main obstacle for image comprehension. That is why the fuzzy mathematics applied to edge detection widely. But most of fuzzy algorithms could not express the fuzziness and random of edge of image precisely. Edge information with lower gray-level was lost, and the result lacks precision. Cloud model is a flexible and useful model for treating uncertain issues. It considers fuzziness and random at the same time, overcomes the limitation of classical fuzzy algorithms. The paper proposes Object-Cloud Algorithm (OCA). In OCA, we consider that objects in image are fuzzy objects with uncertain edge. The degree membership of each pixel belonging to an object is inverse proportion with the distance between pixel and center of the object. So each object can be represented with a cloud, which can be called "object-cloud". Based on "object-cloud", OCA replaces microcosmic pixels with macroscopical objects. Pixel is just element of "object-cloud", and edge information with lower gray-level can be preserved precisely. OCA has two steps. First, creating a cloud for each object in image. Uncertain point and uncertain polyline in image can be treated as uncertain area. Area-cloud includes "core" and "half-cloud loop" outside the core. It can be represented by model A(Area,En,He) based on cloud theory. En is entropy. He is hyperentropy. Generating core (Area) is the precondition of creating an uncertain area-cloud. Second, achieving the fuzzy edge of object in image. After creating "object-cloud" for each object in image, adjacent area can be represented by intersecting clouds. We can gain a new "boundary-cloud" by Boolean calculation between two or more intersecting clouds which have public boundary. The fuzzy edge between objects in image can be detected by digital characteristics of "boundary-cloud" and cloud calculation.

Level set evolution theory is introduced to bridge or dam detection above river in order to improve performance in case of very low contrast and faint targets feature in optical or radar imagery. Aiming at shortages like boundary leak, weak robust to noises existing in classical level set methods, and sub- or over- segmentation, irregular boundary with gap existing in traditional segmentation, an adaptive narrow band level set evolution model based on Chan-Vese model is presented to excellently extract river regions from radar imagery with faint edge and unwelcome effects, while greatly accelerate the curve evolution process. Furthermore, we propose a novel algorithm based on Narrow Band Level Set(NBLS) for detecting and simultaneously distinguishing bridge and dam. The algorithm is efficient, avoiding the disadvantages that medial-axis search methods are subjected to noises and are hard to process river branch with complex shape. Finally, feature-weighted decision rule is adopted to combine the detection results from the two binary classifiers form radar and optical imagery, in order to make use of complementary feature from different classifiers and to achieve higher accuracy of targets detection than single classifier. Experimental results demonstrate that our scheme proposed in the paper outperform some others, with the advantages of time-effectiveness and robust to noises.

With the development of the high-resolution remote sensing image, it is important to identify and extract the image information automatically. Texture analysis has been recognized as a useful method of improving the target identification and its accuracy. In this paper we mainly discussed texture patterns analysis of main geographical objects in QuickBird image. We analyze the texture's characteristic using Gray Level Co-occurrence Matrix (GLCM) and the Texture Measure of Gray Level Co-occurrence Matrix (TMGLCM). The texture pattern analysis takes TMGLCM as the main method. We establish a unify texture pattern use the TMGLCM parameter. The method is available after experiment.

Post-classification analysis is an important way for remotely sensed imagery change detection. In this paper, we propose a novel classification way for change detection using multispectral IKONOS imagery. The classification way is called after Enhanced Growing Self-Organization Map (EGSOM). The EGSOM is designed to solve two limitation of traditional Self Organization Feature Map (SOM). One is the training time of SOM is endless, the other is SOM's structure is fixed before train. EGSOM make use of Growing Self Organization Feature Map and the network's weights are initialized after hierachical clustering method. The method can save network-training time and make the network express input data correctly. Using EGSOM, we classify Multispectral IKONOS imagery and analyze the change detection result. The experiment shows the EGSOM can achieve better classification results than max likelihood method.

Remote sensing is a powerful tool for precision forestry, providing the forestry industry with spatial information on environment impacts, growth and yield, site variables and damage assessment. Nonetheless the extraction of information from remotely sensed imagery is presently labor intensive requiring highly qualified remote sensing experts, making this information source expensive and slow. With the improvement of spatial resolution, very high resolution remote sensing image are now a competitive alternative to aerial photography and field visits in forest resource survey. In recent years, numerous classification methods were described in the literature and they can be classified into two large classes: traditional pixel-based classification and object-oriented image analysis method. Traditional pixel-based classification techniques either supervised methods or unsupervised method all based on spectral analysis of individual pixels and significant progress has been achieved in recent years. However, these approaches have their limitations since the problem of mixed pixels is indeed reduced, but the internal variability and the noise within land cover classes are increased the improved spatial resolution. In order to improve the classification accuracy, object-oriented image analysis concept has been proposed. This paper explores the use of object oriented image analysis approaches in mapping forest resource and introduces a fast and robust segmentation algorithm--mean shift. The study is based on SPOT-5 image covering the national forest park of Tian'eshan, Zixing City, Hunan, China. Image processing included geometric and atmospheric correction and image segmentation and classification using spectral and spatial information to separate 5 classes. 86.5342% overall accuracy was achieved with this approach. In additional, object oriented image analysis method is compared with traditional pixel based method. The results show the importance, capabilities and challenges of object oriented approaches in providing detailed and accurate information about the physical structure of forest areas.

Urban growth induces urban spatial expansion in many cities in China. There is a great need for up-to-date information for effective urban decision-making and sustainable development. Many researches have demonstrated that satellite images, especial high resolution images, are very suitable for urban growth studies. However, change detection technique is the key to keep current with the rapid urban growth rate, taking advantage of tremendous amounts of satellite data. In this paper, a multi-scale object-oriented change detection approach integrating GIS and remote sensing is introduced. Firstly, a subset of image is cropped based on existing parcel boundaries stored in GIS database, then a multi-scale watershed transform is carried out to obtain the image objects. The image objects are classified into different land cover types by supervised classification based on their spectral, geometry and texture attributes. Finally a rule-based system is set up to judge every parcel one by one whether or not change happened comparing to existing GIS land use types. In order to verify the application validity of the presented methodology, the rural-urban fringe of Shanghai in China with the support of QuickBird date and GIS is tested, the result shown that it is effective to detect illegal land use parcel.

An efficient edge detection for remote sensing image based on Wold-like decomposition in random field is presented in this paper. In such assumption that the image field is a realization of a 2-D homogeneous random field, image can be decomposed into a sum of two mutually orthogonal, spatially homogeneous components, namely deterministic in the prediction theory sense and purely indeterministic in the prediction theory. The Wold decomposition can be described by "periodicity," "directionality," and "randomness," approximating what is indicated to be the three most important dimensions of human perception. So, the remote sensing images are firstly decomposed into two components: deterministic component and indeterministic component. On the basis of Wold-like decomposition a new approach of low image processing, SUSAN algorithm, is estimated and recommended in the edge detection on the periodicity component, which presents the structural information convenient for detecting edge. Then this paper made some improvement of the approach in edge detection. The experiments show that the results of edge detection through Wold decomposition are better than that of no Wold decomposition. Simultaneously, the Wold texture modal is applicable to a wide variety of texture types, from structural to stochastic texture. And this modal gives a unified, perfect description of texture in natural images.

The researches on automatic extracting features information of farm crop, forest and territory resources etc. from remote-sensing image are being done at home and abroad. The paper put forward a new approach based on existed many approach, that merge the category based on the foundation of the traditional cluster. The new approach includes two steps: Firstly, choose a small scale to cluster and not need a category corresponding with a integrate region in image space. Secondly, take the category generated in the first step as object, then merge the category, form the corresponding relationship of the image region and category. The Step is a core of the new approach, it systematize the image segmentation based on multidimensional histograms and form a new approach, which by creating the Separating, Closeness (borderline connect, borderline disconnect and integrate)Tighten Index in characteristic space, creating Gathering, Capturing, Equality, Average between clusters and Average in the cluster Index in image space, Division and classification do constantly between the characteristic space and image space, changing the shortage of single space approach, extracting the features information more effective.

This paper introduces a new semi-automatic method to extract regions of interest (ROI) from aerial and satellite images. It uses a geometric active contour, which is based on level sets method, to segment and extract the interesting area which has complicated topological structures. A simple Mumford-Shah functional is defined as an energy function and the Gaussian PDF is adopted to represent the distribution of intensity. A specified polygon is used as both initiate data and the range in which contour evolve. We give some examples of ROI extraction on various aerial and satellite images. Experimental results show that our proposed method has strong capability to deal with the extraction of areas which has a complex topological structure.

Automatic feature extraction for road information plays a central role in applications related to terrains. In this paper, we propose a new road extraction method using the one-class support vector machine (SVM). For a manually segmented seed road region, only a part of pixels are really road, some pixels locating on the sideway, shadows of the building, and the cars etc., are not really road pixels. The one-class SVM is used to estimate a decision function that takes the value +1 in a small feature region capturing most of the data points in the seed road area, and -1 elsewhere. Since the road pixels in the satellite image have the similar properties, such as the spectral feature in multi-spectral image, the novelty pixel is discriminated by the estimated decision function for road segmentation. Many computation experiments are undertaken on the IKONOS high resolution image. The results demonstrate that the proposed method is effective and has much higher computation efficiency than the standard pixel-based SVM classification method.

The maximum likelihood classification (MLC) is one of the most popular methods in remote sensing image classification. Because the maximum likelihood classification is based on spectrum of objects, it cannot correctly distinguish objects that have same spectrum and cannot reach the accuracy requirement. In this paper, we take an area of Langfang of Hebei province in China as an example and discuss the method of combining texture of panchromatic image with spectrum to improve the accuracy of CBERS02 CCD image information extraction. Firstly, analysis of the textures of the panchromatic image (CCD5) made by using texture analysis of Gray Level Coocurrence Matrices and statistic index. Then optimal texture window size of angular second moment, contrast, entropy and correlation is obtained according to variation coefficient of each texture measure for each thematic class. The chosen optimal window size is that from which the value of variation coefficient starts to stabilize while having the smallest value. The output images generated by texture analysis are used as additional bands together with other multi-spectral bands(CCD1-4) in classification. Objects that have same spectrums can be distinguished. Finally, the accuracy measurement is compared with the classification based on spectrum only .The result indicates that the objects with same spectrum are distinguished by using texture analysis in image classification, and the spectral /textural combination improves more than spectrum only in classification accuracy.

In this paper, the imaging principle of Light Detection And Ranging (LIDAR), and, in particular, the effect factors of intensity of return pluses of LIDAR are analyzed in detail. It is shown that, in one scan band, the intensity of LIDAR return pulses is determined by the ground undulation. The distance information of LIDAR can reflect the actual ground situation, and in the paper, the operator of slope is used to describe the ground undulation, which can be calculated based on LIDAR distance information. A new way is put forward to remove the pulse noise of LIDAR intensity image based on orientation gradient of distance image. The experiments prove that the algorithm can not only remove the pulse noise effectively but also preserves the edge information of Intensity Image better.

Networks play the role of a high-level language, as is seen in Artificial Intelligence and statistics, because networks are used to build complex model from simple components. Recently Bayesian Networks, one of probabilistic networks, are a powerful data mining technique for handling uncertainty in complex domains. However, in the classification domain it was not paid attention to by researchers until the simplest form of Bayesian Networks, Naive Bayesian Network, turned up. In this paper, Naive Bayesian Network is applied to texture classification of aerial image. In order to validate the utility of Naive Bayesian Classifier, six hundred and eighty-four aerial images are used in the experiment and results demonstrate Naive Bayesian Classifier needs less computational costs than maximum likelihood method during classification and outperforms maximum likelihood method in the classification accuracy. Therefore, it is an attractive and effective method, and it will lead to its wide application.

Compression techniques are required to transmit the large amounts of high-resolution synthetic aperture radar (SAR) image data over the available channels. Common Image compression methods may lose detail and weak information in original images, especially at smoothness areas and edges with low contrast. This is known as "smoothing effect". It becomes difficult to extract and recognize some useful image features such as points and lines. We propose a new SAR image compression algorithm that can reduce the "smoothing effect" based on adaptive wavelet packet transform and feature-preserving rate allocation. For the reason that images should be modeled as non-stationary information resources, a SAR image is partitioned to overlapped blocks. Each overlapped block is then transformed by adaptive wavelet packet according to statistical features of different blocks. In quantifying and entropy coding of wavelet coefficients, we integrate feature-preserving technique. Experiments show that quality of our algorithm up to 16:1 compression ratio is improved significantly, and more weak information is reserved.

In the investigation of China present road state and optimum designing of road, the linear data collection and computing of relevant parameters are very important. Though the methods of collecting data are feasible, but the methods of computing design parameters of road are lacking. This paper discusses that the theory of condition adjustment applies in road data collecting and optimum designing of road. The method goes as follows: after the road line types, mileages, curvatures of two endpoints of the road have been known, considering the constraint conditions on the distances between road and buildings on both sides of the road, we adjust all parameters of line type using the theory of Condition adjustment so that they satisfy a series of objective requirements while considering the condition that the curvatures of road should be continuous. This method has proved its availability by Visual C++ programming finally.

The diversity of the spatial scale of landscape raises the requirement of multiscale analysis of remote sensing (RS) images. Usually the first step to analyze remote sensing images is image segmentation, in which the muitiscale effect should be taken into account to achieve satisfactory segmentation results. This paper describes an effective approach to segment remote sensing images in multiscale. Based on the fact that in a specific scale of a remote sensing image the same objects are similar, the image is first segmented in a small scale by uniting the most similar objects. After that, a set of multiscale objects with full topological relationship can be obtained. Based on the set of multiscale objects, the authors explore the application of this approach in object-oriented information extraction from remote sensing images.

The diversity of the spatial scale of landscape raises the requirement of multiscale analysis of remote sensing (RS) images. Usually the first step to analyze remote sensing images is image segmentation, in which the muitiscale effect should be taken into account to achieve satisfactory segmentation results. This paper describes an effective approach to segment remote sensing images in multiscale. Based on the fact that in a specific scale of a remote sensing image the same objects are similar, the image is first segmented in a small scale by uniting the most similar objects. After that, a set of multiscale objects with full topological relationship can be obtained. Based on the set of multiscale objects, the authors explore the application of this approach in object-oriented information extraction from remote sensing images.

This paper introduces a method regarding the remote sensing data for change detection by using GIS database. The concept of object-oriented has been used in this method to classify the remote sensing data. The objects of the classification not only can be single pixels of image but also can be pixel sets that represent GIS objects. The remote sensing data are classified with a supervised maximum likelihood classification. In order to reduce the workload and avoid the dependence on operator's experiences, the training areas are generated from the GIS database. Experiments show the method is effective on detecting the change of area objects.

Segmentation wooden member from range images is the basis of 3d reconstruction of wooden member and whole architecture because it provides reliable point clouds for modeling. This paper presents a segmentation strategy to extract point clouds of wooden member of ancient architecture from range image. Hybrid approach combining edge and region-based techniques is adopted in the paper to ensure a reliable and robust segmentation. First, range image is triangulated according to the implied topological relationships between point clouds. Second, filtering is processed by combination the two smoothing methods of λ|μ and Laplacian. Third, feature points are detected by local surface differential geometry properties, and feature edges are extracted according to a selective mechanism, so an initial, rough segmentation is provided based on edge information. And then, the initial edge-based segmentation is enhanced by region-based segmentation method. How to estimate the differential geometry properties robustly, and how to detect feature points and feature edges and so on are studied in the paper. Range images acquired from Forbidden City of China are used to test the segmentation strategy, and results prove its efficiency and robustness.

This paper presents a Pareto evolutionary artificial neural network (Pareto-EANN) approach based on the evolutionary algorithms for multiobjective optimization augmented with local search for the classification of remote sensing image. Its novelty lies in the use of a multiobjective genetic algorithm where single hidden layers Multilayer Perceptrons (MLP) are employed to indicate the accuracy/complexity trade-off. Some advantages of this approach include the ability to accommodate multiple criteria such as accuracy of the classifier and number of hidden units. We compared Pareto-EANN classifiers results of the classification of remote sensing image against standard backpropagation neural network classifiers and EANN classifiers; we show experimentally the efficiency of the proposed methodology.

In this paper, a novel algorithm based on Gabor-Wavelet and SVM is proposed. Compared to other algorithms of classification, our method fully exploited the external feature of remote sensing (RS) images extracted by Gabor-Wavelet. In detail, our algorithm can be divided into the following steps. Gabor wavelet decomposition is firstly applied to the fused RS images. The experiment shows that choosing 5 frequencies and 7 phases, that is, 35 Kernel functions, may reach good performance. Then we get a feature vector with several dimensions. Considering that Gabor filtering is not a orthogonal decomposition, principle component analysis (PCA) is used to reduce redundancy. Secondly, feature vectors are divided into two parts- training set and testing set. The training set is put into SVM classifier. The effect of the algorithm for different RS image type was compared. The result shows that our method performs better with fused images, only panchromatic images or multispectral images can not provide enough information for RS image classification.

Scalable Vector Graphics (SVG) is a newly developed format for efficient representation and transfer of vector graphics and raster images on the Internet. There have been a great number of researches conducted on the visualization of vector geospatial data in SVG. However, SVG format also has the advantages in the representation, transfer and processing of raster images. This paper begins with discussing the ways of representing remote sensing (RS) image in SVG and giving a visualization model of RS image according to the tree structure of SVG file and the multi-band characteristics of RS image data. Then experiments of RS image processing in SVG are conducted on a three-layer web platform, including spatial coordinate transformation, spatial enhancement and RS data fusion. The experiment results are given with code and the effects are demonstrated. And the possibility of further exploiting the potentials of SVG in RS image processing is discussed in the conclusion

This paper presents an automatic approach for building extraction and segmentation directly from Lidar point clouds without previous rasterization or triangulation. The algorithm works in the following sequential steps. First, a filtering algorithm, which is capable of preserving steep terrain features, is performed on raw Lidar point clouds. Points that belong to the bare earth and those that belong to buildings are separated. Second, the building points which may include some vegetation and other objects due to the disturbance of noise and the distribution of points are segmented further by using a Riemannian Graph. Then building segments are recognized by considering size and roughness. Finally, each segment can be treated as a building roof plane. Experiment results show that the algorithm is very promising.

This paper presents a 3D reconstruction method to model wooden member of ancient architecture from point clouds based on improved deformable model. Three steps are taken to recover the shape of wooden member. Firstly, Hessian matrix is adopted to compute the axe of wooden member. Secondly, an initial model of wooden member is made by contour orthogonal to its axis. Thirdly, an accurate model is got through the coupling effect between the initial model and the point clouds of the wooden member according to the theory of improved deformable model. Every step and algorithm is studied and described in the paper. Using the point clouds captured from Forbidden City of China, shaft member and beam member are taken as examples to test the method proposed in the paper. Results show the efficiency and robustness of the method addressed in the literature to model the wooden member of ancient architecture.

The aim of the filter processing to LIDAR dataset is to divide the dataset into ground points and non-ground points. So, the filtering of LIDAR dataset is a crucial step to obtain the DEM with high precision. Over the last few years, some algorithms have been developed to filter LIDAR data. This paper studies three filtering algorithms that are used in common at present, brings about some improvement to the MLS filtering algorithm which is one of the three filtering algorithms and gives two datasets as experimental data in order to compare and analyze the filtering results.

This study analyzed texture features in multi-spectral image data. Recent development in the mathematical theory of wavelet transform has received overwhelming attention by the image analysts. An evaluation of the ability of wavelet transform and other texture analysis algorithms in feature extraction and classification was performed in this study. The algorithms examined were the wavelet transform, spatial co-occurrence matrix, fractal analysis, and spatial autocorrelation. The performance of the above approaches with the use of different feature was investigated. Wavelet transform was found to be far more efficient than other advanced spatial methods.

Airborne laser scanning (ALS) is the most recently emerged technology for acquiring Digital Terrain Model (DTM). In order to generate DTMs from ALS data, the terrain points and off-terrain points should be classified. To meet the objective, a sweep line based filtering algorithm was proposed in this paper. This algorithm works on raw range data obtained by ALS system without interpolation. The algorithm includes 3 steps. First, the point cloud is divided into sweep lines. The second step is to classify terrain points and off terrain points within each sweep line by choosing suitable tolerance value of height difference and size of filtering window. Finally, the filtering results of each sweep line are collected. Experimental work is conducted on 5 data sets of various terrain environments. A quantitative assessment was conducted to compare this result with manual classification. Both qualitative and quantitative experiments show that sweep line based filter can remove most of the off-terrain points effectively. Compared with other algorithms, sweep line based filtering is featured by simplified the 2-dimentional problem into 1-dimentional problem where less computational resources are required.

Ant colony optimization algorithm is a novel simulated evolutionary algorithm, which provides a new method for complicated combinatorial optimization problems. In this paper the algorithm is used for solving the knapsack problem. It is improved in selection strategy and information modification, so that it can not easily run into the local optimum and can converge at the global optimum. The experiments show the robustness and the potential power of this kind of meta-heuristic algorithm.

The algorithm for retrieving the normalized water-leaving radiance (reflectance) from MODIS imagery is vital to quantitative determination of oceanic color parameters. The standard MODIS atmospheric correction algorithm, designed for open ocean case 1 waters, has been extended for use over turbid coastal waters. Failure of the standard algorithm over turbid waters can be attributed to invalid assumptions of zero water-leaving radiance at near-infrared wavelengths. In the present paper, these assumptions are replaced by assuming that the 748:869-nm ratios for aerosol reflectance and for water-leaving reflectance are spatial homogenous. The new algorithm separates the water-leaving radiance from the aerosol path radiance at 748 and 869nm for relative clear water pixels by solving a set of simple algebraic equations. The performance of the new algorithm is test for imagery of East China Sea. A comparison with in situ radiance spectra shows this atmospheric correction algorithm can achieve very high accuracy and is more rapid than previous algorithm. This new algorithm can be used to the image where no ideal clear waters pixel existed or clear case 1 waters just been covered by clouds in imagery.

Ocean color satellites measure radiance at the top of atmosphere, while water-leaving radiance at sea surface is needed to derive ocean color information. It is necessary to develop atmospheric correction algorithms to obtain the water-leaving radiance from satellite-measured radiance. A kind of ocean color hyperspectral sensor called Chinese Moderate Resolution Imaging Spectrometer (CMODIS) on "Shenzhou-3" spaceship was launched on March 25, 2002. The CMODIS with 30 channels is much different with Sea-viewing Wide Field-of-view Sensor (SeaWiFS). The software of SeaWiFS Data Analysis System (SeaDAS) cannot be used to process the CMODIS data. A practical program was developed to implement the atmospheric correction procedure for CMODIS. The program can compute the Rayleigh scattering radiance and aerosol scattering radiance to deduce the water-leaving radiance. It considers the multiple-scattering effects and atmosphere absorbing effects on radiative transfer model. The results were compared with those by SeaDAS program at SeaWiFS bands to check the performance of the program. The water-leaving radiance deduced from in-situ measurements were also used to evaluate the accuracy of atmospheric correction algorithm and the mean relative error is 9.83%. The results show that this program is effective in processing CMODIS data and easily to be modified to process other kinds of satellite hyperspectral ocean color data.

In the field of digital photogrammetry and computer vision, the determination of conjugate points in a stereo image pair, referred to as "image matching," is the critical step to realize automatic surveying and recognition. Traditional matching methods encounter some problems in the digital close-range stereo photogrammetry, because the change of gray-scale or texture is not obvious in the close-range stereo images. The main shortcoming of traditional matching methods is that geometric information of matching points is not fully used, which will lead to wrong matching results in regions with poor texture. To fully use the geometry and gray-scale information, a new stereo image matching algorithm is proposed in this paper considering the characteristics of digital close-range photogrammetry. Compared with the traditional matching method, the new algorithm has three improvements on image matching. Firstly, shape factor, fuzzy maths and gray-scale projection are introduced into the design of synthetical matching measure. Secondly, the topology connecting relations of matching points in Delaunay triangulated network and epipolar-line are used to decide matching order and narrow the searching scope of conjugate point of the matching point. Lastly, the theory of parameter adjustment with constraint is introduced into least square image matching to carry out subpixel level matching under epipolar-line constraint. The new algorithm is applied to actual stereo images of a building taken by digital close-range photogrammetric system. The experimental result shows that the algorithm has a higher matching speed and matching accuracy than pyramid image matching algorithm based on gray-scale correlation.

Clouds not only hide the ground but also cast their shadows on the ground, so a fast and automatic method to remove clouds and their shadows from the acquired satellite images is necessary. In this paper, a multispectral image fusion scheme to detect and remove clouds and their shadows is proposed. The entire algorithm is mainly composed of four steps: stationary wavelet transform, detection of clouds and their shadows, image fusion, inverse stationary wavelet transform. The simulation experiment shows that our method is valid and performs well.

A combination of good spatial and spectral resolution makes visible to shortwave infrared spectral imaging from aircraft a highly valuable technology for remote sensing of the earth's surface. Many applications require the elimination of atmospheric effects caused by molecular and particulate scattering; a process known as atmospheric correction, compensation, or removal. The Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) atmospheric correction code derives its physics-based algorithm from the MODTRAN4 radiative transfer code. A new spectral recalibration algorithm, which has been incorporated into FLAASH, is described. AISA (Airborne Imaging Spectrometer for Application) data covering the Eastern Sea in China was acquired and the results from processing AISA data with FLAASH are discussed.

Extremely precise texture modeling of complex 3D objects is a significant problem worth of much research. Comparing the laser scanning technology and the digital photogrammetry with their combination results, the optimal solution is the fusion of the data. Surveying points are used to model the main shapes while laser scanning captured the fine details. Photogrammetry is used to register the texture images with the geometry and produce ortho-texture. This paper explains a reliable method based on Direct Linear Transformation (DLT) and ortho-rectification method to model 3D objects combined with 3D laser scanning data and high-resolution image data. This novel technology can produce precise texture and improves the efficiency and quality of highly detailed texture modeling.

Texture is a very important feature in image analysis including content-based image retrieval (CBIR). A common way of retrieving images is to calculate the similarity of features between a sample images and the other images in a database. This paper applies a novel texture analysis approach, local binary patterns (LBP) operator, to 1m Ikonos images retrieval and presents an improved LBP histogram spatially enhanced LBP (SEL) histogram with spatial information by dividing the LBP labeled images into k*k regions. First different neighborhood P and scale factor R were chosen to scan over the whole images, so that their labeled LBP and local variance (VAR) images were calculated, from which we got the LBP, LBP/VAR, and VAR histograms and SEL histograms. The histograms were used as the features for CBIR and a non-parametric statistical test G-statistic was used for similarity measure. The result showed that LBP/VAR based features got a very high retrieval rate with certain values of P and R, and SEL features that are more robust to illumination changes than LBP/VAR also obtained higher retrieval rate than LBP histograms. The comparison to Gabor filter confirmed the effectiveness of the presented approach in CBIR.

Remote sensing image fusion has become one of hotspots in the researches and applications of Geoinformatics in recent years. It has been widely used to integrate low-resolution multispectral images with high-resolution panchromatic images. In order to obtain good fusion effects, high frequency components of panchromatic images and low frequency components of multispectral images should be identified and combined in a reasonable way. However, it is very difficult due to complex processes of remote sensing image formation. In order to solve this problem, a new remote sensing image fusion method based on frequency domain segmenting is proposed in this paper. Discrete wavelet packet transform is used as the mathematical tool to segment the frequency domain of remote sensing images after analyzing the frequency relationship between high-resolution panchromatic images and low-resolution multispectral images. And several wavelet packet coefficient features are extracted and combined as the fusion decision criteria. Besides visual perception and some statistical parameters, classification accuracy parameters are also used to evaluate the fusion effects in the experiment. And the results show that fused images by the proposed method are not only suitable for human perception but also suitable for some computer applications such as remote sensing image classification.

A new cloud-top height retrieval method is proposed by using polarizing remote sensing. In cloudy conditions, it shows that, in purple and blue bands, linear polarizing radiance at the top-of-atmosphere (TOA) is mainly contributed by Rayleigh scattering of the atmosphere's molecules above cloud, and the contribution by cloud reflection and aerosol scattering can be neglected. With such characteristics, the basis principle and method of cloud-top height retrieval using polarizing remote sensing are presented in detail, and tested by the polarizing remote sensing data of POLDER. The satellite-derived cloud-top height product can not only show the distribution of global cloud-top height, but also obtain the cloud-top height distribution of moderate-scale meteorological phenomena like hurricanes and typhoons. This new method is promising to become the operational algorithm for cloud-top height retrieval for POLDER and the future polarizing remote sensing satellites.

In this paper, a automatic approach of geometry precision checking for digital ortho-image (DOM), especially for a batch of DOM, is proposed. Automatic image matching between DOM with various resolution was performed in the DOM checking based on reference DOM with higher resolution. Automatic extraction of linear feature was used in the DOM checking based on reference vector map with higher precision. A lot of experiments testify that the method of this paper can improve the efficency and reliability of checking method of DOM image for geometry precision.

In this paper, a novel method for matching of synthetic aperture radar (SAR) image to optical image is proposed. The theory of support vector machine (SVM) is brief discussed and a new method of edge extraction via SVM for SAR image is presented. 2-D lifting wavelet transform is applied to construct the image pyramid. The multi-subarea strategy is adopted to enhance the reliability matching. Experimental results demonstrate that the method is feasible and could obtain high accuracy.

With the development of remote sensors, high-resolution satellite data such as QuickBird, IKONOS images have been available widely. Thus, remote sensing technologies can be successfully applied to more application areas such as extracting roads from the high-resolution images. This paper represents a total rectangle matching approach to extract straight roads of urban areas from high-resolution remote sensing images. The approach is based on the characteristics of the high-resolution remote sensing image, the knowledge about the road and the hit-miss transformation of mathematical morphology. It is implemented by matching rectangles from the inside to the outside to meet the optimized criterion through changing the threshold of image segmentation, the width and direction of rectangle. Experimental results on high-resolution satellite images demonstrate that the proposed approach can eliminate the influence of noise (trees, vehicles etc.), and extract the straight roads effectively.

How to recognize man-made objects from high-resolution remote sensing images has been considered an attractive and important research field in remote sensing applications undoubtedly. In this paper we try to present a feasible contour-based retrieval strategy of remote sensing images. The merit of our strategy is it can avoid the impact caused by the difficult of automatic manmade object discrimination so far and the deficiency of huge computational volume aroused by template matching. Besides, on the basis of analyzing the limitations of common descriptors such as Fourier descriptor and Hu invariant moments, invariant relative moments are adopted to describe shape feature of man-made objects in our retrieval strategy. After describing contour feature extraction method, feature matching method and retrieval process based on shape feature, a prototype system is also designed and implemented to prove the validity and accuracy of our strategy mentioned above. In our experiments three types of man-made objects with different shape feature, i.e., boat, oilcan and buildings with flat-roof, are selected as our research targets. Experimental results illustrate that our strategy is feasible and the corresponding retrieval performance is analyzed, followed by conclusions and future works.

By incorporating Mie theory derived physical characteristics of dust particles with radiative transfer model, a look up table (LUT) method was used to retrieve optical depth (at 0.55 μm) of dust storms. Self-defined spectral dependence (single scattering albedo, asymmetry factor and extinction efficiency) of dust aerosol was imbedded into atmospheric radiative model (SBDART, Santa Barbara DISORT Atmospheric Radiative Transfer) to model the effects of dust aerosol on TOA spectral radiance at band 31 and 32 of MODIS. Simulations showed that under dusty conditions (optical depth >0), brightness temperature difference (BTD, band 31-band 32) becomes negative, and monotonically decrease with the growth of optical depth, as can be used to identify and delineate dust storms in MODIS level 1B data. A MODIS scene of November 6, 2005 containing dust event was used to validate above method. By matching simulated optical depth and brightness temperature difference with satellite remotely-sensed brightness temperature difference, optical depth of dust was retrieved. The larger values appeared at the main dust regions, and small optical depth occurred at the edges of the dust. Finally, sensitivity analysis was conducted, application of this retrieval method were discussed.

Satellite's position is the major role for determining the baseline errors in interferograms. With high precision satellite position, the flat earth component can be removed. Envisat satellite bases on DORIS system for controlling and calculating its orbit within several centimeters precision. However, even several millimeters baseline error may cause fringes in the interferogram. This paper studied the baseline error profiles in interferograms and analyzed the Envisat satellite orbit characters. The linear part and the nonlinear part of the baseline errors in the interferograms have been studied. Fourier transformation for the two parts has been tested. With ASAR data sets, a case study has been made. The results showed that the orbits in ASAR products were not precise enough for interferometry and precise orbits from Delft and GFZ were more reliable. The linear part of the baseline error is easy to be discerned and mimicked but the nonlinear part is hard to derive and eliminate. The satellite orbit characters should be studied more to reveal the nonlinear part baseline errors.

Hyperspectral Remote Sensing (HRS) is one of the most significant recent achievements of Earth Observation Technology. Classification is the most commonly employed processing methodology. In this paper three new hyperspectral RS image classification methods are analyzed. These methods are: Object-oriented FIRS image classification, HRS image classification based on information fusion and HSRS image classification by Back Propagation Neural Network (BPNN). OMIS FIRS image is used as the example data. Object-oriented techniques have gained popularity for RS image classification in recent years. In such method, image segmentation is used to extract the regions from the pixel information based on homogeneity criteria at first, and spectral parameters like mean vector, texture, NDVI and spatial/shape parameters like aspect ratio, convexity, solidity, roundness and orientation for each region are calculated, finally classification of the image using the region feature vectors and also using suitable classifiers such as artificial neural network (ANN). It proves that object-oriented methods can improve classification accuracy since they utilize information and features both from the point and the neighborhood, and the processing unit is a polygon (in which all pixels are homogeneous and belong to the class). HRS image classification based on information fusion, divides all bands of the image into different groups initially, and extracts features from every group according to the properties of each group. Three levels of information fusion: data level fusion, feature level fusion and decision level fusion are used to HRS image classification. Artificial Neural Network (ANN) can perform well in RS image classification. In order to promote the advances of ANN used for HIRS image classification, Back Propagation Neural Network (BPNN), the most commonly used neural network, is used to HRS image classification.

A radar system of formation flying satellites, which consists of small, individual satellites with each having a standard Synthetic Aperture Radar (SAR) sensor, has the advantage that adds independent angle-of-arrival sample information in addition to range-Doppler data. The additional samples are useful for producing high resolution SAR image as well as extending the image swathwidth. To accomplish this aim, we propose an improved imaging algorithm based on traditional Kalman Filter procedure to combine the data from all sensors. The method strongly improves the resolution by incorporating prior knowledge and spatial information of multiple receivers, which is a scientific breakthrough in the case that the traditional matched filter constrains the improvement of SAR spatial resolution. It is a optimal method in the sense of mean square error and its computation cost is lower than the traditional Kalman Filter algorithm. Simulation results demonstrate the effectiveness of the proposed method.

The color composite Digital Mapping Camera (DMC) images are obtained by calibrating, mosaic and color composite fusion processing through the post-processing software, but sometimes the color transition in the transition area is not smooth enough and there are still residual radiometric differences. This paper, based on analyzing the characters and causes, proposes a self-adapting repair approach to remove such phenomenon. The approach first gives a strict proof about the feasibility to locate the exact position of seam-line and transition area, and adopts the principle of edge detection operators to detect the exact position automatically. Then it employs a multi-scale adjustment scheme which is mainly carried out by two different spatial scales and includes the "break" processing and "seam" removal. Experiments indicate that the approach can repair the color composite DMC images with residual radiometric differences effectively.

Quantitative measure of image information amount is of great importance in many image processing applications, e.g. image compression and image registration. Many commonly used metrics are defined mathematically. However, the ultimate consumers of images are human observers in most situations, thus such measures without consideration of internal mechanism of human visual system (HVS) may not be appropriate. This paper proposes an improved definition of mutual information between two images based on the visual information which is actually perceived by human beings in different subbands of image. This definition is both sensitive to the pixels' spatial location and correlates well with human perceptual feeling than mutual information purely calculated by pixels' grayscale value. Experimental results on images with different noises and JPEG&JPEG2000 compressed images are also given.

Three-line scanner CCD stereo surveying camera is now considered to be used not only in space borne vehicle but also airborne one. It can capture three images of a stereo pair on the same orbital pass. The imagery is characterized by invalidation of conventional collinear equation geometric model and impropriety of satellite borne sensor model. This paper addresses the method of simulating the change of exterior orientation parameters and making analog airborne three-line scanner (TLS) CCD image. Besides, we will present the accuracy test results processed with three kinds of camera geometric models. Results show that the precision of collinear-equation-based polynomial Model with root mean square error (RMSE) less than one pixel is better than other.

As wetland has been recognized as an important component of ecosystem, it is received ever-increasing attention worldwide. Poyang Lake wetlands, the international wetlands and the largest bird habitat in Asia, play an important role in biodiversity and ecologic protection. However, with the rapid economic growth and urbanization, landscape patterns in the wetlands have dramatically changed in the past three decades. To better understand the wetland landscape dynamics, remote sensing, geographic information system technologies, and the FRAGSTATS landscape analysis program were used to measure landscape patterns. Statistical approach was employed to illustrate the driving forces. In this study, Landsat images (TM and ETM+) from 1989 and 2000 were acquired for the wetland area. The landscapes in the wetland area were classified as agricultural land, urban, wetland, forest, grassland, unused land, and water body using a combination of supervised and unsupervised classification techniques integrated with Digital Elevation Model (DEM). Landscape indices, which are popular for the quantitative analysis of landscape pattern, were then employed to analyze the landscape pattern changes between the two dates in a GIS. From this analysis an understanding of the spatial-temporal patterns of landscape evolution was generated. The results show that wetland area was reduced while fragmentation was increased over the study period. Further investigation was made to examine the relationship between landscape metrics and some other parameters such as urbanization to address the driving forces for those changes. The urban was chosen as center to conduct buffer analysis in a GIS to study the impact of human-induced activities on landscape pattern dynamics. It was found that the selected parameters were significantly correlated with the landscape metrics, which may well indicate the impact of human-induced activities on the wetland landscape pattern dynamics and account for the driving forces.

This paper introduces the theories and methods for image fusion in remote sensing. Based on the IHS transform and wavelet packet transform, a scheme of remote-sensing images fusion aiming at protecting image's spectral characteristics is made in this paper. This scheme makes best use of the information in remote-sensing images to be fused and prevents the loss of image information. Through the experiment using SAR image and TM image, all the spectral characteristics, the textural feature and the spatial quality of image have been improved.