一种结合CEM的高光谱遥感影像目标检测算法

Target Detection Algorithm of Hyperspectral Remote Sensing Imagery Combined with CEM

提出了一种基于CEM(constrained energy minimum)的高光谱影像目标检测算法。利用虚拟维数实现高光谱影像端元数目的估计,进而获取目标端元。利用基于最大噪声分量算法对原始影像进行降维,在虚拟维数的基础上保留合理的主成分数量。根据提取的目标端元,利用一种改进的CEM算子进行目标检测。实验结果表明:该算法能够获得较高质量的目标检测效果。

A target detection algorithm based on CEM（ Constrained Energy Minimum） is proposed.The virtual dimensionality algorithm is employed to effectively estimate the number of endmembers,and the interested endmembers are extracted from hyperspectral images. MNF（ Maximum NoiseFraction） is introduced for the dimensionality reduction of original images. Based on the results of virtual dimensionality,the reasonable number of principle component is selected. According to the target endmembers,an improved CEM detector is used for target detection. Experimental results show that the proposed algorithm provides better performance of target detection for hyperspectral images.