|
In this presentation, two advanced feature extraction methods with fast and deep learning algorithms will be discussed for environmental monitoring in all-weather conditions with convergent and divergent thinking. One is the newly developed novel Spectral Information Adaptation and Synthesis Scheme (SIASS) and the other is the newly invented SMart Information Reconstruction (SMIR) method to support the Integrated Data Fusion and Mining (IDFM) research. Whereas the former is organized to generate cross-mission consistent ocean color reflectance by merging observations from several different satellites to recover the cloudy pixels, the latter is designed to reconstruct cloud contaminated pixel values from the time-space-spectrum continuum with the aid of a machine learning tool. For the purpose of demonstration, Lake Nicaragua located at Central America is selected as a study site which is a very cloudy area year round. In this case study, merging observations from MODIS-Terra, MODIS-Aqua, and VIIRS over Lake Nicaragua will be presented for the 2012-2015 time period. Then the performance of SMIR will be performed after the merging operation by reconstructing the missing remote sensing reflectance values derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) on board the Terra satellite over Lake Nicaragua. The SIASS algorithm is proven to have the capability not only in eliminating incompatibilities for those matchup bands but also in reconstructing spectral information for those mismatching bands among sensors. For the recovery of those missing pixel values after merging three satellite images, experimental results from SMIR show that the extreme learning machine may perform well with simulated memory effect due to linking the complex time-space-spectrum continuum between cloud-free and cloudy pixels. Final water quality assessment will be generated based on the integrative algorithm of the two with bio-optical models for eutrophication assessment in Lake Nicaragua.
|
