A SPATIAL FILTER BASED FRAMEWORK FOR TARGET DETECTION IN HYPERSPECTRAL IMAGERY

Target detection in hyperspectral imagery is of great interest in signal processing field. So far, a lot of algorithms have been developed for hyperspectral target detection. And the local methods are expected to outperform the global methods; however, the local methods have two fatal defects: the singular and rank deficient matrix, and the high time consumption. Furthermore, when the threshold is adjusted to get a higher detection probability in practical application, the false alarm rate will also increase, and even severely increase. A framework based on spatial filter is proposed and implemented to improve the detection performance, using the local detection statistics information. Experiments with hyperspectral dataset are conducted to show the performance improvement of the proposed framework.