The Algorithms For Remote Sensing Data Based On The Satellite-earth Coordination Using Dictionaries

With the technique advances of obtaining remote sensing data, our country has established the observation mechanism based on a variety means of spatial information acquisition including satellite, unmanned aerial vehicle, stratosphere balloon, etc. However, it results in the massive remote sensing data. The time-delay of downloading and processing data might contribute to missing essential information. Based on the sparsity character of the remote sensing image, we have proposed the dynamic updating strategy in orbit which includes sparse representation on the satellite, downloading sparse coefficients, reconstructing image and dictionary learning in the ground. The design mainly utilize the coordination between satellite and ground station, i.e., that the sparse information from the satellite lays the foundation of dictionary learning in the ground and the dictionary from the earth station also devotes to the sparse representation on the satellite. These two procedures are not mutually exclusive. In fact, they are complementary to each other, with the information interaction.The thesis summarized the research with respect to the sparse theory, with studying the mathematic knowledge of some classical methodologies in the area. Meanwhile, we also concluded the pros and cons of these algorithms. As a sequence, three new algorithms have been proposed. Respectively, they are the dictionary learning algorithm based on particle swarm optimization, the dictionary learning algorithm based on dynamic incremental updating of atoms and the compressed sensing methodology based on the reference image as priors. With the qualitative and quantitative comparison, the experiments have demonstrated these algorithms are able to represent the remote sensing image sparsely and accurately. Generally speaking, these algorithms proposed by the thesis are qualified to meet the requirements of whole approach.