光谱曲线概率的高光谱影像小目标探测算法

Spectral curve probability algorithm of hyperspectral image small target detection

针对高光谱影像小目标的探测,最常用的约束能量最小化算法探测率低、探测效果欠佳,其它的多数探测算法或模型也基于CEM。在研究小目标特性的基础上,提出高光谱影像小目标的光谱曲线概率探测算法。该算法是基于高斯分布理论,可以在目标光谱已知或未知条件下对小目标进行探测。经过定性实验和与CEM算法探测结果的定量比较分析得出,SCP算法对小目标探测率高、探测效果好;并能有效抑制背景,不再需要白化处理,降低算法的复杂性。SCP是一种简单、高效的高光谱影像小目标探测算法。

As for the spectral curve probability algorithm of small target detection of hyperspectral image,the constrained energy minimization（CEM）algorithm,which is the most frequently used,is low and the result of it is bad on detecting small targets of hyperspectral image.Other detection algorithms or models are also based on CEM and lack of substantive innovation.This paper proposes a spectral curve probability（SCP）algorithm on detecting small targets of hyperspectral image.The algorithm,based on the Gaussian distribution theory,can detect the small targets whether the target spectrum is known or not.After the qualitative experiments and quantitative analysis comparing the result of CEM algorithm,it proves that the correct detection ratio of SCP algorithm is higher,the result of it is better on detecting small targets and the algorithm can curb the background effectively so that no whitening reduces complexity of the algorithm.SCP is a simple and efficient algorithm on detecting small targets of hyperspectral image.