Hyperspectral target detection based on sparse errors matrix

Abstract

Listen

Translate article

In this paper, a hyperspectral target detection algorithm based on sparse errors matrix is proposed for target detection. In HSI, the spectrum of target and background pixels lies in different subspaces. According to the compressed sensing theory, high-dimensional HSI data can be decomposed to a low rank matrix and a sparse errors matrix. The low rank matrix contains background spectra and the sparse errors matrix consists of target spectra. The sparse errors matrix can be recovered by solving an l1-norm minimization problem. Once the sparse error matrix is obtained, the target in HSI can be determined by the characteristics of the sparse errors matrix. Robust PCA (RPCA) is used to decompose the HSI data to a low-rank matrix and a sparse errors matrix. Comprehensive experiments on one data sets using both visual inspection and quantitative evaluation are carried out. The results from these data sets have indicated that the proposed approaches help to generate improved results in terms of efficacy and efficiency.