变化向量分析结合光谱解混的高光谱变化检测

针对多时相高光谱图像像素级的多类变化检测问题，提出变化向量分析和光谱解混相结合的多类变化检测方法.基于光谱变化向量分析，利用最大期望（EM）算法迭代求阈值，实现变化区域检测.对多时相高光谱图像分别提取端元，求解2个图像中变化区域像元的丰度.以相关系数为相似性判断准则，根据图像分类精细程度自适应确定阈值，实现多时相高光谱图像各端元对应类别的匹配和确定.对变化向量分析方法检测出的变化区域求丰度，根据丰度最大确定各像元类别.通过逐像元类别比较，判断类别变化信息.仿真数据和真实多时相高光谱图像的变化检测实验结果表明，与直接光谱解混分类后变化检测方法相比，采用提出的方法能够明显提高高光谱图像多类变化检测的精度，运行效率提高1倍以上.
Abstract: A multiple changes method was proposed based on change vector analysis and spectral unmixing technology in order to solve the problem of multiple changes detection in multi-temporal hyperspectral images. The changed and unchanged pixels were distinguished by change vector analysis, and the threshold was iteratively solved by using the expection maximization (EM) algorithm. Endmembers were respectively extracted from the two multi-temporal hyperspectral images, and the abundances were solved for each changed pixel in both two images. Taking the correlation coefficient as similarity criterion, the endmembers for the two images were matched with the threshold of similarity adaptively determined according to the fine degree of the ground classes in the images, and the category code of each endmember was assigned. The considered multiple-change detection problem was solved by comparing the classes of each pixel in the changed region after assigning the class of each pixel according to the maximum abundances. The proposed approach was tested on both simulated and real multitemporal images showing multiple-change detection problems. Experimental results show that the proposed method can greatly increase the accuracy for multiple-change detection compared with the change detection method based on the spectral unmixing directly, and the running efficiency is more than doubled.