基于空谱联合特性的改进协同高光谱异常目标检测

Improved Collaborative Algorithm Based on Spatial-spectral Joint Characteristics for Hyperspectral Anomaly Detection

摘要:

传统的协同表示的高光谱异常目标检测算法忽略了图像的空间信息导致检测精度不高，因此，本文提出了一种基于空谱联合特性的改进协同高光谱异常目标检测算法。该算法以像元为中心构建一定大小的空间窗从而提取能够代表像元类别的空间信息，得到了像元的图像块灰度向量，将像元的光谱信息和空间信息相结合，提出了光谱协同和空间协同的概念，最后将二者进行加权和得到了算法最终的检测结果。该算法通过对三组真实的高光谱数据进行仿真实验，并与经典算法的检测结果进行对比，验证了该算法的有效性和先进性。

Abstract:

The traditional collaborative representation for hyperspectral anomaly target detection algorithm ignores the spatial information of the image and leads to low detection accuracy. Therefore, an Improved Collaborative Algorithm Based on Spatial-spectral Joint Characteristics for Hyperspectral Anomaly Detection algorithm is proposed. The algorithm constructs a certain size space window centered on the pixel to extract spatial information capable of representing the pixel class, and the image block vector of the pixel is obtained. Combining spectral information with spatial information, the concept of spectral collaboration and spatial collaboration is proposed, and weighting the two and obtaining the result of the algorithm. The algorithm is used to simulate the three sets of real hyperspectral data and compared with the existing algorithms, verifying the advancement of the algorithm.