Researches on sparse representation and on-board processing for massive spatial information
Mar 14, 2014Author：
On-board intelligent processing for spatial information stands at the core ofspatial information network sciences. It is of great strategic significance on Remote Sensing Information (RSI) transmission with high efficiency, real-time RSI content-based translation and understanding, RSI sharing within spatial networksas well as its various applications and services in society. Accordingly, the goal of this project is to develop new theories and new methodologies for massive RSI on-board intelligent processing, and find the solutions to the bottleneck issuesthat largely hinder the technical development of RSI transmission andunderstanding. Specifically, this project will research the issues related tomassive RSI sparse representation, efficient RSI compression, multi-source RSI on-board description, automatic cloud detection and object enhancement, real-time on-board detection for sea targets, and accurate target location. Along this research line, new systematic theories and methodologies will be developed for RSI on-board intelligent processing, including multi-source coupled dictionarylearning, batch-mode RSI sparse representation, adaptive dictionary learning for content-based RSI compression, deep learning for sea target detection, and so on.Furthermore, to tightly gear to the application demands for fast response of in-orbit satellites, the study of this project will provide a series of real-time,robust and advanced key techniques for RSI on-board intelligent processing. In summary, the practice of this project will provide the core theories and key techniques for the construction of new-generation intelligent spatial networks,and help drive the leap-forward development of spatial information network techniques with sound technical foundation.