基于Sentinel-1A后向散射特性的高原山区地物分类探讨——以安顺市为例  

Discussion on classification for Sentinel-1A SAR data in mountainous plateau based on backscatter features——A case study in Anshun city

作　　者：许璟[1] 安裕伦[2,3] 刘绥华[2,3] 韩可欣 

机构地区：[1]黄山学院建筑工程学院,安徽黄山245021 [2]贵州师范大学地理与环境科学学院,贵州贵阳550001 [3]贵州省山地资源与环境遥感应用重点实验室,贵州贵阳550001

出　　处：《贵州师范大学学报（自然科学版）》2016年第6期15-19,38,共6页Journal of Guizhou Normal University：Natural Sciences

基　　金：国家自然科学基金项目(41161002);贵州省省长基金项目(黔省专合字(2011)46号);贵州省科学技术基金(黔科合J字[2014]2127)

摘　　要：探讨星载Sentinel-1A雷达数据在高原山区进行地物分类的可行性。以贵州省安顺市为研究区,NDVI数据为辅助数据,利用单极化多时相的SAR数据提取的后向散射特征Sigma σ值,建立符合研究区实际情况的专家树决策系统,进行地物分类和目标识别。结果表明:雷达数据的分类精度基本能够达到实际应用要求,特别是水域和建设用地在图像上解译标志明显,分类精度高于TM数据,分别达到90%和80%;植被和耕地精度低于TM数据,分别为70%和65%;但总体上TM数据的细节表现不如SAR数据。在分类工作时结合传统影像可以有效弥补高原山区气候和地形因素带来的不足,同时也能够促进雷达数据在此类高原山区的应用和推广。The paper establishes expert decision system tree which in conformity with the practical situ-ation of Anshun city of Guizhou Province study area about object classification and recognition. By u-sing the Sigma backscatter extracted from multi-temporal SAR data and other data such as NDVI,we explore the feasibility of feature classification of the spaceborne Sentinel-1 A radar data in plateau mountainous. The result showed that the radar data classification accuracy can meet the requirements of practical application,especially in water bodies and buildings on the image which interpreting marks are obvious. Their classification accuracy is 90% and 80%,which is higher than TM data. The accu-racy of vegetation and cultivated land is 70% and 65%,which is lower than that of TM data. Overall, the details of the TM data are not as good as SAR data. The SAR data combined with traditional image can effectively make up for the inadequacy of climate and terrain factors in mountain plateau in the work of classification,it not only has great practical significance,but also can improve the application and promotion of radar data in such plateau mountainous areas.

关 键 词：高原山区 雷达数据 后向散射 地物分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]