Detection Of Sparsity Hyperspectral Image Sub-pixel Based Targets

Hyperspectral imagery with high spectral resolution has the unique characteristic of acquiring spectral and spatial information simultaneously. Different materials are distinguishable by the spectral difference revealed in hyperspectral images. The unique characteristic of hyperspectral image brings the hyperspectral detection advantages when dealing with target detection problem under complex conditions. Because of the complex distribution of different ground objects and the limited spatial resolution of the hyperspectral images, a pixel in the hyperspectral images is usually composed of different land objects and the targets usually reside in the sub-pixel scale. Subpixel target detection as a difficulty for target detection has attracted the attention of researchers in recent decades. Our research mainly focuses on how to use the sparsity of hyperspectral data to enhance the detection. The main work and contributions of this dissertation are as follows:1. We study the spectral mixture model of hyperspectral remote sensing images, and then describe signal estimation and signal detection theory in detail. After that four kinds of classic subpixel target detection algorithms including CEM、CEM based on weighted correlation matrix、OSP、AMSD which are proved effective in the experiments of Chapter IV are deduced.2. Sparse decomposition model based on constrained linear mixed spectrum is provided in this thesis. We design and achieve four sparse unmixing algorithms including OMP、 ISMA、SUNSAL、SUNSAL based on iterative weighted LI. And then SUNSAL algorithm based on Ll/2regularization is put forward. The experimental result shows that performance of SUNSAL algorithm based on L1/2regularization is better and more stable than other algorithms when SNR of hyperspectral image is low.3. The paper further investigates the subpixel target detection based on the sparsity of hyperspectral data. By combining sparse unmixing algorithm and AMSD algorithm we get SU-AMSD algorithm, and then process and implementation of the SU-AMSD algorithm are described in detail. To improve the efficiency and accuracy of target detection, SU-AMSD algorithm based on Lib-IEA is proposed. The experimental result shows SU-AMSD algorithm based on spectral library is more accurate than four classic subpixel target detection algorithms mentioned in the dissertation. It is frustrating that performance of SU-AMSD algorithm based on Lib-IEA is not as good as expected.