MULTICATEGORY TARGETS DETECTION OF HYPERSPECTRAL IMAGERY BASED ON ADAPTIVE STRUCTURED BACKGROUND AND SHAPE-FEATURE SUBSPACE
自适应结构化背景和形状特征子空间高光谱图像多类目标检测

A new detection algorithm was presented to detect multicategory targets of known shape-feature and unknown spectral signature in unknown environment. Firstly, a point spread function was constructed via high-order moments of quadratic form of data samples to obtain adaptive structured background. Then, a priori shape-features of targets were utilized to construct a shape-feature subspace which is matched with high-dimension spectral signature space. Theoretic analysis and the results of experiment verify the effectiveness of the algorithm.