基于光谱角累加的高光谱图像异常检测算法

针对传统方法中用作背景的像素中存在干扰像素的问题,提出一种基于光谱角累加的高光谱图像异常检测算法。通过计算测试区域待测像素光谱向量与其他光谱向量之间的夹角,并将其累加,得到图像中每个像素的异常程度;然后使用波段选择预处理方法进一步提高检测性能。HyMap高光谱数据验证表明,在虚警概率设为0.008时,检测概率达到0.73,即在提高异常检测可靠度的同时,降低了虚警概率。

Anomaly detection algorithm for hyperspectral imagery based on summation of spectral angles

As interference pixels influence the background features in traditional methods, a new anomaly de- tection algorithm for hyperspectral imagery is proposed based on the summation of spectral angles. The anoma- ly degree of each pixel in the imagery is acquired by calculating the angles between the spectral vector of test pixel and other spectral vectors in the test region, and accumulating the angles. Then, the pretreatment meth- od of band selection is used to further improve detection performance. HyMap hyperspectral data experiments show that the detection probability reaches 0. 73 when the probability of false alarm is set to be 0. 008. The re- liability of anomaly detection is improved, while the probability of false alarm is reduced.