This PDF file contains the front matter associated with SPIE Proceedings Volume 12588, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

For small and medium-sized electronic enterprises to produce plug-in circuit boards, due to the size of the enterprise, order requirements and other reasons, small and medium-sized electronic enterprises usually use manual visual inspection of plug-in circuit boards for inspection. Based on the Vision Developments Module of LabView software, we can detect the solder joint defects of plug-in circuit boards. According to the actual production of different types of circuit boards, the solder joints are classified, and then the representative qualified solder joints are extracted to make standard solder joint templates, which are matched with the images of the plug-in circuit boards to be tested through multiple template geometry matching to achieve automatic detection of solder joints with 93% detection accuracy.

With the development of computer-aided design, visual languages have been widely used as formal methods to represent various types of graphical models. Accordingly, many grammar systems have been proposed for the specification of visual languages. However, apart from shape grammar, most grammars focus on the abstract structures of the languages and ignore the semantic modeling of graph drawing. Furthermore, shape grammar supports generation rather than parsing, with its limited application scope. To address these problems, this paper proposes an enhanced grammar system based on Coordinate Graph Grammar (CGG). Different from traditional grammars, the enhanced system defines a new type of grammatical rule named shape rules to transform graphs into shapes by shape applications. In each shape application, the assertion set describes the range of validity, and shapes can be generated by translation, zoom, and rotation to a set of rule-based coordinates. With the combinations of shape applications and L-applications, the node-edge graph and drawn outline could both be specified, building a bridge between abstract structures and physical layouts of visual languages. An example is given to illustrate the application of the enhanced system in industrial design, where a Bauhaus-style baby cradle is generated by the combination of shape applications and L-applications.

Zhaling Lake is the first lake at the source of the Yellow River. The morphology of Zhaling Lake has changed frequently in recent decades. We conducted a visual analysis of Landsat remote sensing images and weather station observations in Maduo County from 1976-2021. We derived the space-time evolution of Zhaling Lake shoreline and area change trends. This paper obtained the basic characteristics such as the area and shoreline of Zhaling Lake by using water index and visual interpretation. Based on these basic characteristics, we calculated other shoreline characteristics. We analyze the morphological, meteorological, and other data characteristics of Zhaling Lake by visualization method. Research results show that the area of Zhaling Lake is highly variable. Its overall change shows a cyclic trend of recession-expansion-recession. A significant shrinkage of Zhaling Lake occurred from 1976 to 1986, from 544.177 km² to 537.32 km². Between 1986 and 1993, Zhaling Lake gradually expanded, adding 9.541 km². During 1993-2019, the area of Zhaling Lake showed a continuous trend of increase. In 2019, Zhaling Lake reached the largest area of 555.368 km² in nearly 45 years. During 2019-2021, Zhaling Lake begins to shrink slightly again. In 2021, Zhaling Lake's area shrinks to 549.41km².

The health management and maintenance of photovoltaic (PV) plants are inherent problems in the PV industry. The need to establish digital positioning for each PV string and PV module is urgent. This paper provides a complete end-to-end system for digital segmentation and localization of PV strings and modules on the aerial orthophotos. The system includes three main parts: (1) the dataset built from the images captured by Unmanned Aerial Vehicles (UAV) and corresponding image preprocessing techniques. (2) a modified Deeplab V3+ neural network is designed to extract the PV strings in the aerial orthophotos. (3) a PV module extraction algorithm is introduced to get the centroid of every PV module and the sliding window strategy is adopted to avoid the chopped PV strings problem. With the above process, the digital location information of PV panels can be correlated with the actual physical information. We conduct detailed experiments with actual scene data and different models. The extensive results confirm the accuracy and efficiency of the system proposed in this paper with comparative analysis.

Social media is a platform for people to share their lives and interact with others. Image sharing is an integral component. Privacy information will inevitably be compromised during the process of sharing whole photos. Sometimes, it is not even caused by the person who shares the image but the third party who forwards the image. In the majority of cases, we do not attempt to protect the entirety of the media; rather, we seek to protect the most crucial portion, whether it is the background or specific objects, which decreasing superfluous cryptography procedures. Unfortunately, current research on such field is extremely rare. In this research, we established a novel pixel-level image encryption technique relying on Panoptic FCN panoramic segmentation and chaos for client-intended images on social media sites. Our suggested technology is capable of automated picture encryption on either the whole images or user-selected areas, whether they are rectangular or irregular, which is suited for all region of interest (ROI) encryption. Relying on a novel coupled chaotic map, this universal new encryption method flattens the array of the image ROI into a series of pixels. The module of Panoptic FCN can be replaced by any other panoptic segmentation models which are stronger in performance. Statistical and cryptographic evaluations demonstrate that our technique preserves the high efficiency for practical applications.

Image captioning, a cross-modal study, aims to generating a description for a given image, which plays an important role in many fields like image retrieval and computer-assisted instruction. Currently, the challenge in image captioning is the limited quality of generated descriptions including insufficient utilization of image feature information and the limited language learning ability of the decoder. In this paper, we address the above problems by constructing a semantic enhancement module and a multi-round decoding mechanism to enhance the decoding ability of the model, which uses the Transformer model as the primary structure. To validate the efficacy of the model, we conducted intensive experiments on the MSCOCO2014 benchmark and evaluated its performance using five evaluation metrics. The experimental results show that the proposed method in this paper has improved to varying degrees on all five-evaluation metrics.

Along with the rapid development of the air transportation industry, the impact of aircraft wake vortices on flight safety and airport capacity has become increasingly prominent. In this paper, we propose a transformer-based model to solve the problem of multiple LIDAR wake vortex detection and recognition in airports. By setting up multiple Doppler LIDARs in the near-Earth flight areas of different runways of Shenzhen Baoan Airport (SZX), a large amount of accurate wind field data is captured for wake vortex data collection. In the deep learning framework, the radial velocity sequence obtained from the LIDAR is used as the input of the transformer. Meanwhile, local meteorological information and LIDAR operating parameters are introduced into the model, providing prior knowledge at different observation points. The experimental results show that the model has unified modeling for different LIDAR wake vortex detection, and has obtained excellent recognition results.

Traditional scene text spotters aim to detect and recognize entire words or sentences in natural scene images, however, the detection and recognition of every single character is also as important as the spotting of unifying words or sentences in one image. There are few specialized methods to spot single character in scene text spotting, and some word-based methods can not recognize a series of characters in images if they can not be spelled as a correct word. In addition, some early models can only detect or recognize texts which are horizontal and distinctive. We realize that it is necessary to improve some existing models for achieving the goal of spotting characters, therefore, we propose a novel method based on an improved YOLOv5 model to accomplish the character-level spotting. It’s worth noting that this method can spots characters not only in regular texts but also in irregular texts (curved texts and oriented texts).

In view of the current image segmentation field, there are few studies on the segmentation of typical wire clamp components of transmission lines. Traditional image processing methods have low segmentation accuracy and require artificial design of feature extraction methods, which are usually only suitable for equipment of a certain structure with insufficient generalization. In this paper, an infrared image segmentation method based on Mask R-CNN (Mask region-based convolutional neural network) for typical guide-ground lines is proposed. Its structure takes Mask R-CNN model combined with FPN (Feature pyramid structure) as the basic framework, and uses RPN (Regional proposal network) to generate candidate regions. Features are extracted from each candidate region through RoI Align layer, and then connected to FC (Fully connected layer) to achieve target classification and bbox (bounding box) regression. A mask branch is also added to predict the segmentation mask. The design can integrate multi-scale and multi-level semantic information to improve the recognition rate when extracting image features. In addition, the network structure is optimized by single channel for infrared images to reduce the size of the model and make it more lightweight. Ablation experiments were performed on two GTX 2080Ti graphics cards to verify the effectiveness of the proposed structure, and the mAP (mean average accuracy) of 0.421 was achieved with an IoU (Intersection over Union) threshold of 0.5.

As the development of power system, more and more transmission line engineering are turning into cable design in city or some built-up areas. Three-dimensional design is a new design method which is more suitable for cable project than traditional CAD 2D design. This paper shows the several work we have made to develop the three-dimensional design and introduces the main design process for the cable three-dimensional design. Finally, a case of hybrid overhead line and cable project is been shown, which shown the advantages of the three-dimensional design for the cable project.

Sentiment analysis is the technique of automatically evaluating and classifying emotions (often positive, negative, or neutral) from textual data, such as written comments and social media posts. Sentiment analysis is a subfield of natural language processing (NLP) that employs machine learning to classify the emotional tone of textual input. The fundamental model concentrates on positive, negative, and neutral categories, but it can also include the speaker's underlying emotions (pleasure, anger, insult, etc.) and purchase intents. Complexity is added to sentiment analysis by context. For example, consider the exclamation "Nothing!" Depending on whether or not the speaker enjoys the product, the meaning can vary significantly. In order for a machine to comprehend "I like it," it must be able to decipher the context and determine what "it" refers to. In addition, sarcasm and sarcasm can be tricky because the speaker may express a favorable sentiment while intending the opposite.

With the rapid development of the automotive industry, vehicles are equipped with a variety of system functions. The realization of many of these functions depends on the stable, safe, and reliable positioning and timing information of the bottom layer. Vehicle-mounted satellite positioning systems are an important way for vehicles to obtain absolute positions and have been widely employed in other industries. However, the automotive industry has its special requirements, such as high positioning accuracy and confidence, extremely harsh vehicle regulations, reliability, and high safety, all of which need to be tested and evaluated on vehicle-mounted satellite positioning systems. By studying and putting forward the evaluation scheme for the vehicle-mounted satellite positioning system, this paper further ensures the accuracy, reliability, and stability of the time-space information provided by the system and supports the development of the automotive industry.

The convolutional neural network (CNN) in deep learning artificial intelligence (AI) has developed rapidly in recent years, delivering many achievements to other areas of economic life. Nevertheless, gaps in CNN-related research still exist in the field of object identification and detection in regard to active sonar images, as most research in this field is still dominated by classical algorithms. Therefore, this paper summarizes the YOLOV5 used, analyzes the existing network defects, and optimizes the identification and detection algorithms based on the YOLOV5 network framework. The practical detection sets a high requirement for the precision of the sonar pulse signals detected. Specifically, it requires the false alarm rate to be lower than the designed value and the errors in the detection parameters to be kept within the tolerable range. To increase the detection precision, this paper adds an attention enhancement module to the network based on the original YOLOV5, which significantly improves the detection parameter effects.

Grammatical error correction (GEC) aims to automatically identify and correct grammatical errors in a sentence. Neural machine translation (NMT) models are the mainstream approaches for the GEC task. However, the models require a large amount of data to be adequately trained, the variety of grammatical errors and the dependencies between errors in a sentence make it difficult for a single NMT model to correct multiple errors at once. In the work, we propose an ensemble approach for heterogeneous models, which integrates rule-based, NMT, and pre-trained language model-based GEC models through the recurrent generation approach, the approach can exploit the strengths of each model and cover a wider range of errors in a sentence. We also mitigate the scarcity of task-specific data for the GEC task through the data augmentation approach. We conduct extensive experiments on the NLPCC2018 shared task dataset to demonstrate the effectiveness of our proposed methods, and reaches the F0.5 value of 37.26, outperforming the best model in the shared task.

Vehicle target detection is a key research hotspot in the field of computer vision. At present, with the continuous development of deep learning and artificial intelligence, some excellent vehicle target detection algorithms such as YOLOv5, YOLOv4 and YOLOv3 have emerged. Therefore, in order to solve the problem of low accuracy of vehicle target detection, ensure the safety of vehicles on the road and achieve target detection more accurately. This paper provides a YoloV5-based method for detecting car objects and an improved algorithm that uses large-scale internal fusion techniques. Finally, the vehicle target detection accuracy of the improved YOLOv5 algorithm is effectively improved through experimental comparison and analysis. This is of great practical significance for promoting the development of target detection algorithms.

To solve the problem of identifying the presence of foreign objects in microfluidic chip images, an improved model is proposed for the feature of small foreign object targets. The attention mechanism is introduced to enhance the perceptiveness of the model in channel and space. The ResUnit module in the network is modified to enhance the feature information. Also choose diou as the loss function to improve the edge accuracy. The experimental results show that the improved YOLOx target detection algorithm has a significant improvement in foreign object detection in terms of accuracy, and the average precision (AP) reaches 99.12% on YOLOx, which is 0.7% higher than the original network. The results show that the improved algorithm based on YOLOx in this study can achieve foreign object detection in microfluidic chip images.

Aiming at the problem that it is difficult for the existing target detection algorithms to detect high-precision circuit experimental equipment in middle school, an improved YOLOX detection network model is proposed. Based on the YOLOX network model. Firstly, the ECA attention module is added to the feature extraction network to enhance the model's ability to perceive electrical experimental equipment; Secondly, the feature enhancement structure is added to enhance the semantic information of the obtained feature map and improve the detection ability of the target; Finally, EIoU is selected as the loss function to achieve high-precision positioning. The experimental results show that the improved network model mAP reaches 91.9%, which is 1.5% higher than the original network model, which proves that the improvement is effective and feasible.

A visual monitoring equipment for coal spontaneous combustion temperature field in goaf is designed，The automatic operation system and self-cleaning device are designed according to the coal mine goaf site，An optimal solution algorithm for temperature anomaly area in goaf is developed and embedded software is formed，The collected data are analyzed through the model to obtain the optimal location of high temperature ignition source in the goaf, and the data are uploaded to the monitoring center through the transmission network.The system formed by the combination of the device and determination method can make visual monitoring and intelligent judgment of the danger area in goaf.

The use of virtual reality (VR) in urban area planning is examined in this dissertation. There are numerous allusions to Openstreetmap's urban building data. This VR technology allows users to enter a virtual environment using an input device (HTC Vive). Without actually carrying out the plan, the effectiveness of urban planning can be tested beforehand. VR will become more crucial in urban area planning, boosting urban construction, and assisting critical individuals in making decisions. There were three primary tasks for this project:
First task: Create a 3D urban model using the OpenStreetMap map file.
Unreal Engine 4 has imported a file with an OSM map (UE4). A data stream has been created to convert OSM files to UE4 uasset files, according to StreepMap and the RuntimeMeshComponent plugin.
Second task: Create an overview mode.
Users can highlight various items with various features on the overview map in this mode, including different building levels, the top speed limit, hotel star ratings, amenity categories, and retail categories.
Third task: Create a VR mode.
A virtual environment has been developed to enhance immersion and user experience. The HTC VIVE controller allows users to explore city maps independently. A new interactive interface has been created for this project. With the aid of this method, urban planning initiatives can be pretested before being put into action.
Contributions to science:
This project's main contribution is developing an application to aid pertinent staff members in managing and planning urban data. Based on this research, urban spatial planning could cut costs by using virtual reality 3D modeling, data integration, and VR interaction.

The current mobile office data transmission management involves the problems of image traffic and delay, which leads to the low efficiency of mobile office applications. Therefore, this paper proposes a new energy micro-training and mobile office application based on 5G. After standard processing of mobile office images, a communication model is built to transmit information by using the characteristics of the continuous and rapid decline of 5G peak bandwidth. The experimental data shows that the time consumed by the new method is controlled below 1s, which proves that the new energy micro-training and mobile office under 5G can handle higher computing and processing needs. Even if more files are superimposed, they can be transmitted quickly in a shorter time. The method in this paper can ensure that the data transmission has a good quality of network experience, which meets the needs of mobile offices.

The evaluation of grassland degradation is an important part of ecological conservation research, and rodent infestation is a significant factor in grassland degradation. The presence of a large number of mouseholes means that the environmental balance of grassland has been destroyed, so the coverage of mouseholes can be used as an evaluation method for grassland degradation levels. In this paper, the image segmentation method is used to segment the mousehole images, Upernet is used as the segmentation network, and Swin Transformer as the Backbone. FAM and FSM modules are added to the Upernet network to solve the target misalignment problem when upsampling the network. The mIoU is improved by 5.3% according to the experimental results.

This paper proposes an indoor house structure mapping system, which mainly studies the SLAM positioning, laser depth binocular camera locally built figure, robot path planning, and other issues. And the three-dimensional indoor autonomous robot is built based on called VINS system. With wheel odometry and sensor fusion, this paper realizes the robot chassis indoor high-precision positioning, indoor autonomous robots scanning walls. Finally, through the integration of several modules, the chassis robot can achieve the function of indoor independent three-dimensional drawing. This system is successfully applied on a 4-wheel mecanum small car. I take a 20 square meters underground parking lot in as the experiment area. The test shows that the system of displacement drift error within 2% and the ratio of the overall path, room-built figure area error within 3%, loopback positioning error is less than 0.1 m, the robot can work independently to complete the building figure. In the end, the system is evaluated and a commercial indoor 3D drawing scheme is proposed.

The data generated in the Internet age is increasing exponentially. Sometimes such a huge amount of data cannot be processed in time, and people cannot dig out useful information from it. In order to realize the efficient processing of massive data, this paper develops a big data real-time processing architecture based on the Hadoop platform, uses HBase as the database, and combines the C# programming language and the MapReduce programming mode to design a big data processing system, so that users can view and upload data through mobile devices. The data processing results of the cloud computing center. The performance test of MapReduce and various functional modules of the big data processing architecture is carried out. The test results show that MapReduce has certain advantages in processing big data, and the data processing time of each functional module increases with the increase of data volume.

In recent years, meta-learning has become a mainstream technique for few-shot learning, and it has been widely used and achieved good results in computer vision and image processing. Based on this powerful empirical performance, we are interested in using Meta-learning frameworks in NLP to deal with the task of few-shot learning (FSL). However, due to the sparse sample size, sample-level comparisons based on other expressions are highly susceptible to interference, leading to serious overfitting problems. To achieve classification tasks, we suggest a novel Adaptive Cross-Capsule Network (ACCN) for learning generalized representations. A dynamic routing technique is utilized with the concept of a prototype network to train the support set to generalize the generalized representations of each category. The support set and the query set can fully interact dynamically to capture the essential semantic aspects of the query set following a successful non-parametric cross-attention method. Experimental results show that ACCN proposed in this paper is well adaptive to the intention classification task under additional categories, which obtain SOTA results on FewRel Datasets, which also can perform significantly better than the original classification system on Huffpost Datasets. This provides a crucial foundation for this study.

Many transformer modules, have been applied to computer vision. However, the transformer can extract the distal connections of human skeleton points and apply the attention mechanism to the problem of predicting human motion pose. We introduce a transformer module in the joint dimension. In addition, the Encoder module of the transformer is improved. Finally, our method achieves impressive results on benchmark datasets, including short- and long-term predictions of FNTU, confirming its effectiveness and efficiency.

With the continuous development of computer technology and human-computer interaction system, virtual reality technology has been widely used in all aspects of social life.The application of virtual reality technology in psychotherapy provides a new idea for the intervention of eating disorders whose main symptoms are bulimia nervosa and binge eating disorder.In this paper, the application of virtual reality technology in the treatment of bulimia nervosa and binge eating disorder was discussed by reviewing relevant studies at home and abroad, and the future research was prospected.

The traditional monitoring means of airport asphalt runway construction has problems such as irregular construction operation, unreliable process control and unscientific quality assessment. To meet the higher requirements of runway construction quality monitoring process, this paper analyzes the mechanized asphalt pavement construction process and the root causes affecting the construction quality, researches the application of Internet of Things (IoT) technology in asphalt pavement mechanized construction monitoring information system, and also develops the overall structure design of asphalt pavement monitoring system based on IoT technology, completes the development of a remote monitoring system based on the Web through serial communication, network protocol and database design. Finally, the system was analyzed in the test results of the west runway overhaul project of Capital International Airport. The results showed that the system has good hardware seismic resistance, good data integrity and real-time performance, and high reliability. The active and effective use of the airport asphalt runway construction management system can reflect the runway construction process comprehensively, while it is important to promote the traditional construction monitoring to advanced automated real-time process monitoring and management.

This paper proposes a method combining binocular vision and deep learning to identify and locate ripe tomatoes in greenhouses. First, the CBAM attention mechanism module is added to the YOLO V3 model to improve the robustness of the YOLO V3 model to the greenhouse environment, and then the tomato results identified by the improved YOLOV3 CBAM are fused with the three-dimensional information obtained by the binocular stereo camera. to obtain the threedimensional position information of the tomato fruit. After testing, the model has an accuracy of 89.15% for tomato recognition, the AP is 86.17%, and the F1 value is 82%. The relative error of the tomato fruit positioning is less than 1.5%. Finally, the model was arranged in the greenhouse to test the tomato picking robot, which verifies the practicability of the method.

As the most intuitive and reliable monitoring quantity of concrete dams, deformation can comprehensively reflect the service performance of dams in real time. By constructing a real-time prediction model, it has important guiding significance for the identification and response of deformation anomalies in the operation of water conservancy projects. In this paper, a deep learning algorithm: long-term and short-term memory neural network (LSTM), combined with attention mechanism, is used to construct the deformation prediction model of concrete dam. Through engineering examples, the MSE of LSTM model with attention mechanism is 0.69, and the MAE is 0.67. Compared with the stepwise regression model, the recurrent neural network model (RNN) and the LSTM model without attention mechanism, the errors are reduced. LSTM can better mine the long-term and short-term dependencies in deformation sequences, and use the attention mechanism to influence the global and local relationships between factors, highlighting the contribution of main factors to deformation.

Face recognition has always been a popular research task in the field of computer vision, which aims to identify the people by analyzing the relationship between the local features of the face (nose, mouth, eyes, etc.), and has been widely used in public security, mobile smart devices, transportation and many other fields. Depending on whether there is external occlusion, face recognition task mainly includes unoccluded face recognition and more challenging occluded face recognition. Through a detailed literature survey and analysis, this paper firstly introduces the representative unoccluded face recognition methods from five perspectives: based on geometric features, based on global features, based on local features, based on FaceNet and based on elastic graph matching. The classical methods and principles of occluded face recognition are further introduced, and the above-mentioned representative face recognition algorithms are quantitatively compared and analyzed. Finally, we discuss the remaining problems and future development directions in the field of face recognition.

In recent years, the aging process of China's society has accelerated, and the problem of old-age care has aroused widespread concern in society. The subway plays an increasingly important role in the urban public transportation system. Therefore, an intelligent sign system is designed. The three-dimensional environment of public space is displayed as a seamlessly connected plane environment by using virtual reality technology and quasi-physical elements consistent with the knowledge and experience of the elderly, in which the vision of public space is matched. The ASM model is used to quantitatively analyze the availability of the intelligent sign system from four dimensions: vision matching, utility-oriented analysis, interaction experience and visual effects, and proposes suggestions to optimize the subway sign system for the elderly, and puts forward design principles for the intelligent sign system.

In recent years, facial landmark detection has assumed an important role in various fields. However, the current facial landmark detection algorithms are still lacking in recognition accuracy. In order to solve the above problem, this paper uses Ghost bottleneck to replace the original bottleneck on the basis of the original model of PFLD model, and adds and improves the CBAM attention mechanism. The improved PFLD model increases the ability of the model to extract facial landmark and improves the accuracy of the algorithm. The improved model has high accuracy and low parametric number and improves the accuracy of facial landmark detection. It also provides a new idea for facial landmark detection task.

Nowadays, Alzheimer's Disease (AD) has become a massive problem for middle-aged and older adults. Although due to its long incubation period and early mild symptoms, patients have a more extended period and more possibilities to check out, it is still hard for patients and doctors to diagnose in early routine examinations. This article provides a new method to help the doctor to diagnose Alzheimer's Disease in the early phase. We use transfer learning in deep learning to help diagnose Alzheimer's Disease early in developing Computed Tomography (CT) brain images. Using three pre-trained models, ShuffleNet, DenseNet, and NASNet-mobile as the transfer learning training model and convolution neural networks. We made some improvements to make it more relevant to the actual situation. DenseNet has best performance (87.36%) among the three models. We set the output into four classes: the four stages of Alzheimer's are widely recognized (Mild Demented, Moderate Demented, Very Mild Demented).

Wildfire, also known as forest fire, is fire that usually occur in forests and are difficult to control. If it could be detected and suppressed at an early stage (mainly smoke and flames), it has important meaning for reducing the loss. With the attention of relevant researchers, wildfire detection technology has become more and more advanced, from traditional manual monitoring to traditional target detection to sensor detection and infrared detection, etc. The various detection methods involved still have problems such as slow detection speed, low accuracy, easy interference and high cost. In this paper, SSD, an advanced target detection method, was chosen from deep learning algorithms. Three independent SSD networks are built with VGG16, MobileNet v2, and EfficientNet b3 as the backbone. The experimental results show that the mAP (mean Average Precision) of VGG16-SSD is 95.34%, which is 4.76% higher than MobileNet v2-SSD and 4.53% higher than EfficientNet b3-SSD. Therefore, VGG16-SSD can effectively detect wildfires in the early stages.

Focusing on applications to the three-dimensional (3D) garment computer aided design (CAD)system, a human features recognition method based on 3D scale invariant feature transformation (SIFT) is proposed in this paper. First of all, pre-processing is performed on the 3D scanned human body, which are the noise reduction and the conversion into point cloud format. Then the 3D scale-invariant feature transformation constrained by directional gradient constraints is used to extract the feature points of the human point cloud model, and the measurement results are recorded. Finally, according to definitions of reference points for garment anthropometry and the actual measurement value corresponding to the human body, the comparison and analysis of diverse recognition algorithms is given. Simulation results show that the proposed method in this paper is valid and effective.

In view of the poor accuracy of the traditional method to evaluate the ability of network security talents in the power industry, this paper proposes a method to evaluate the ability of network security talents in the power industry based on artificial intelligence. Through the establishment of talent ability evaluation system, it uses artificial intelligence algorithm to calculate the index weight value. According to the index corresponding weight to build the evaluation model,it determines the membership range of the index, and calculates the comprehensive score of the assessor's ability. Then, the paper compares the comprehensive score with the ability classification table and determines the ability level of the assessor to realize the evaluation of talent ability. The effectiveness of the designed evaluation method is verified by the demonstration of comparative experiments. The experimental results show that the evaluation results of the design method are consistent with the actual appraisal results in the process of evaluating the attack and defense penetration index of the network security staff. Therefore, it can be proved that the evaluation method of network security talents in power industry based on the artificial intelligence has high accuracy and objectivity, which is more realistic and helpful for the company to form a personnel management pattern with clear talent positioning and full use of talents.

A multi-channel face liveness detection method based on multi-scale feature fusion is proposed to solve the problems of poor stability, poor generalization, and poor robustness against unknown attacks of existing face liveness detection models. Firstly, the method uses a multichannel residual network and introduces the center differential convolution and SimAM attention module in the residual block to improve the feature extraction ability and stability of the model. Secondly, the information contained in the feature map at different scales is further mined by multiscale feature fusion at the end of each channel. Finally, the network is supervised by using cross modal focal loss as an aid to binary cross entropy loss. Extensive evaluations in two publicly available datasets demonstrate the effectiveness, generalization, and robustness of the proposed method against unknown attacks.

The development of intelligent manufacturing promotes the intellectualization of traditional navigation technology. Because actor-critic (AC) algorithm is difficult to converge in the actual application process, this paper uses the optimization algorithm of this method, which is called deep deterministic policy gradient (DDPG). Through the use of experience playback and dual network design, the learning rate can be greatly improved compared with the original algorithm. Because curiosity strategy has more advantages in alleviating sparse reward problem, this paper also takes curiosity mechanism as an internal reward exploration strategy and proposes the DDPG method based on improved curiosity mechanism to solve the problem that robots lack external reward in some complex environments and tasks cannot be completed. The simulation and real experiment results show that the proposed method is more stable when completing the navigation task and performs well in the long-distance autonomous navigation task.

Aiming at the problems of poor convenience, poor scalability, and low authentication rate in traditional authentication technology using physical contact authentication methods such as magnetic cards and passwords, this paper explores the accuracy and convenience of the practical application of MR neural network in personal identity authentication. In the MR wearable device, the neural network person identity authentication method is studied flexibly and quickly to detect and identify the person. The 3D information of the face is collected and preprocessed by the depth camera, and the MR identity authentication data set is established. The neural network Resnet model is used for face detection and face feature vector extraction, and the Euclidean method is used to compare the feature vectors and label the characters. The neural network authentication algorithm is mapped to the MR wearable device, and the deep face information in the scene is identified, matched, and labeled by using the unique spatial mapping of MR technology and the camera of the MR wearable device. It solves the problems of low flexibility, poor reliability of face information, and weak recognition stability in traditional identity authentication methods, enabling MR technology to provide a more intelligent identification and labeling method for person identity authentication.

Aiming at the shortcomings of traditional disease recognition systems that require high shooting environment and large number of samples, this research designs a set of AR-assisted recognition schemes based on HOG-SVM. Under the premise of a small amount of material, due to the introduction of AR technology in the diagnostic system to assist shooting, this solution is better than other methods in terms of training time, recognition speed and average accuracy. Taking the Android terminal as an example, an AR-assisted HOG-SVM-based mobile vegetables disease identification system is implemented, which can quickly identify diseases and guide users to improve the quality of photographed pictures. Through the identification of disease spots in batches of images, the results of disease spot recognition are analyzed from the three aspects of disease accuracy, diseased leaf detection rate and disease spot location accuracy. Finally, AR technology and rapid identification scheme based on HOG-SVM are obtained. The combination can give faster training results and recognition results under the premise of small training samples. Its average accuracy is also higher than deep models such as YOLO v3, SSD 512, and Fast R-CNN. It is a more suitable method for disease identification on the current mobile terminal.

Due to the slow response and poor accuracy of traditional measurement error detection of the electric energy meter, the measurement error detection method of electric energy meter based on machine vision is studied. The minimum error method is used to segment the image threshold to form a binary image. The morphological refinement method is used to extract the image edge contour, combined with machine vision to refine the edge pixels, to achieve the measurement error detection of the instrument. The experimental results show that using the error detection method of machine vision, the detection results are consistent with the error detection results set by the system and the trend is the same. The accuracy also meets the requirements of relevant regulations, which improves the accuracy of electric energy meter measurement.

In this paper, the ideas of shape template matching, multi-sample template matching, and ROI area processing are combined into an edge contour template matching algorithm module. Based on the OpenCV open source vision library, the area to be removed in the ROI area is colored polygons, and the noise points of this shape feature need to be removed, and only the feature points with obvious edge contour features are extracted; as a result, the matching time is transferred to the mapping time, thereby reducing the total matching time.

How to achieve effective exploration is a key issue in the training of Reinforcement learning. The known exploration policy addresses this issue by adding noise to the policy for guiding the agent exploring. However, it has two problems that 1) the exploration scale has low adaptability to the training stability due to the added noise from a fixed distribution and 2) the policy learned after the training may be locally optimal because the exploration is insufficient. Adaptive exploration policy addresses the first problem by adjusting the noise scale according to the training stability. But the learned policy may still be locally optimal. In this paper, we propose an adaptive exploration network policy to address this problem by considering exploration direction. The motivation is that the agent should explore in the direction of increasing the sample diversity to avoid the local optimum caused by insufficient exploration. Firstly, we construct a prediction network to predict the next state after the agent makes a decision at the current state. Secondly, we propose an exploration network to generate the exploration direction. To increase the sample diversity, this network is trained by maximizing the distance between the predicted next state from prediction network and the current state. Then we adjust the exploration scale to adapt to the training stability. Finally, we propose adaptive exploration network policy based on the new noise constructed by the generated exploration direction and the adaptive exploration scale. Experiments illustrate the effectiveness of our method.

This analytic study was performed with the focus on the target positioning of an unmanned aerial vehicle (UAV) and its error. It intended to enhance the positioning accuracy for the strike of naval guns at the targets at sea. A UAV’s target positioning model was first constructed after introducing five coordinate systems from camera coordinate system to geodetic coordinate system as well as the transformation of these coordinate systems. Subsequently, an error analysis model was established with the Monte Carlo method and based on the error sources affecting the accuracy of target positioning. In the end, a numerical simulation was conducted to quantitatively analyze the influence of these error sources on the target positioning accuracy. The simulation results proved that larger positioning errors must be attributed to the UAV’s attitude angle and the distance.

Traditional text classifiers that rely on supervised learning methods always require a large number of labeled documents. Labeling the documents often requires a certain amount of expertise to ensure the accuracy, which is time-consuming and costly. Therefore, a dataless text classifcation method around a small number of easily accessible label descriptions, ie, seed words,rather than surrounding the labeled documents to provide the supervision information for the classification task, shows a good development prospect. However, since the size of the seed word set is much smaller than the word set contained in the document , many documents do not contain any seed words or even contain some irrelevant seed words, which limits the effect of the seed word supervision. The manifold assumption suggests that highly similar texts tend to belong to the same category, so we maintain a local neighborhood structure for each document and construct a manifold regularizer to spread limited the supervised information between similar documents. We propose a Laplacian Nonnegative Matrix Factorization (LapNMF) method，adding the seed word prior information and document manifold into the framework of non-negative matrix factorization. And use the block corrdinate desent method to solve the problem. Experiments show that in most cases, our LapNMF performs better than the current weakly supervised classification methods, showing certain competitiveness.

Identification of traditional Chinese medicine is the core content in the practice teaching of traditional Chinese medicine. It requires students to master the identification method of traditional Chinese medicine and have the ability of clinical application. In daily teaching, due to the large loss of Chinese medicinal materials, the shortage of precious medicinal materials, the lack of living medicinal materials and the long observation period at each stage of living materials, the teaching effect is not good, which affects the improvement of students' ability to identify Chinese medicinal materials. Three-dimensional teaching resources can carry out three-dimensional simulation of the growth process of medicinal plants of Traditional Chinese medicine, Chinese medicine decoction pieces and their medicinal plants, help students build knowledge and improve their identification ability of medicinal materials. This paper summarizes the advantages of three-dimensional teaching resources. The development process of 3D teaching resources of Chinese medicinal materials was elaborated in detail. The development method and implementation process of 3D modeling, 3D animation, construction of virtual scene of medicinal herbs growing environment and interactive roaming of scene are emphasized. The optimization methods of model, animation and scene are discussed.

In the context of the digital age, digital sculpture technology has been widely used in my country compared to traditional sculpture design and production methods. ZBrush is a digital sculpting and painting software that has influenced and transformed the digital sculpting profession with powerful features and an intuitive workflow. This paper discusses the use of ZBrush software as a tool for new sculpture art creation, and further explores the performance characteristics of digital sculpture applications that are different from traditional creation methods. The experience and methods of using digital technology are extracted and summarized, which can provide some learning and reference for the development of digital sculpture art in the future.

The current situation of Museum data is generally faced with common problems such as scattered data resources, lack of high-quality cataloguing and labeling information, and difficult data management, resulting in the phenomenon of low data utilization. It is suggested that the establishment of a special institution to centrally manage Museum data assets, the unified design of Museum big data model, and the improvement of data analysis capability are the three measures to improve the big data capability. Two big data model design methods from technology to business and from business to technology are described. We study the three key points of building the museum big data model. The comprehensive application of the above measures, design methods and technical points can effectively ensure the continuous improvement of the museum's big data capability.