Hyperion高光谱影像坏线修复的局部空间-光谱相似性测度方法

发布时间：2019-02-16 03:16

【摘要】：Hyperion高光谱影像中的坏线将直接影响后续应用的准确性。针对Hyperion高光谱辐射率数据的特点,考虑影像中坏线像元与邻近像元在空间和光谱上的相似性,提出了一种局部空间-光谱相似性测度(local spectral-spatial similarity measure,LS3M),以实现对Hyperion高光谱数据的描述和坏线修复。LS3M由空间和光谱两部分的相似性测度构成,前者为欧氏距离度量,后者组合了Canberra距离和光谱相关角(spectral correlation angle,SCA)。考虑到Hyperion高光谱不同波段的辐射率特性,引入信息熵对SCA进行约束。针对相似像元的邻近搜索问题,引入相似度均值与方差对光谱相似性阈值进行动态调整。为验证该方法的有效性,选取了沙漠、草原、森林、城郊、沿海城市和内陆城市6种典型场景的Hyperion高光谱数据进行模拟坏线的定量误差分析和真实坏线的定性评价;通过与邻域均值法及常规光谱相似性测度的对比,证实LS3M法坏线修复精度更高,稳定性更好。
[Abstract]:The bad line in Hyperion hyperspectral image will directly affect the accuracy of subsequent applications. According to the characteristics of Hyperion hyperspectral emissivity data and considering the spatial and spectral similarity between bad line pixel and adjacent pixel, a local spatial spectral similarity measure (local spectral-spatial similarity measure,LS3M) is proposed. In order to describe Hyperion hyperspectral data and repair bad lines, LS3M is composed of spatial and spectral similarity measures, the former is Euclidean distance metric, and the latter combines Canberra distance and spectral correlation angle (spectral correlation angle,SCA). Considering the emissivity characteristics of Hyperion in different bands, information entropy is introduced to constrain SCA. To solve the proximity search problem of similar pixels, the similarity mean and variance are introduced to dynamically adjust the spectral similarity threshold. In order to verify the effectiveness of this method, the Hyperion hyperspectral data of desert, grassland, forest, suburbs, coastal cities and inland cities are selected to analyze the quantitative error of simulated bad lines and the qualitative evaluation of real bad lines. By comparing with the neighborhood mean method and the conventional spectral similarity measure, it is proved that the LS3M method has higher accuracy and better stability.
【作者单位】：中国地质大学(武汉)信息工程学院;中国矿业大学(徐州)环境与测绘学院;河海大学地球科学与工程学院;
【基金】：国家自然科学基金(41501459,41201341) 中国博士后科学基金(2013M542086) 中央高校新青年教师科研启动基金(CUGL140834)~~
【分类号】：P237

【相似文献】