【摘要】 快速准确地从灾后影像中提取出受灾区域对于灾后救援具有重要意义。鉴于现有提取方法过度依赖滑坡在影像中的光学、纹理等特征的问题,研究了一种结合独立成分分析(ICA)与最小噪声比率变换(MNF)的变化检测方法,以单一时相的影像为基础,运用基于负熵最大化的Fast-ICA算法分离出两个时相影像相互正交的独立成分,并构建对应独立成分的差异影像,最后用最小噪声比率变换实现分布于各个差异影像上变化信息的集中,应用直方图阈值法得到了最终的滑坡灾害信息。选取了滑坡灾害前后两时相的高分辨率遥感影像数据进行实验,结果证实了方法的可行性。更多还原

【Abstract】 It is very important for disaster relief to extract disaster information from the disaster images quickly and accurately. Given the existing methods depends on landslide disasters’ texture and other information on images heavily,a new method is studied which combined the independent component analysis and the Minimum Noise Fraction Transformation. It based on the single image,separated the independent component image using the FastICA algorithm of maximum neg-entropy approximations and built the components’ difference images,focused the different singles which distributed all of difference images based on the MNF and got the change information by the threshold from histogram method finally. Experiment results demonstrated the method’s utility with images both before and after the disaster.更多还原