川西高原多时相干涉雷达土地覆盖分类研究  

Study of Land Cover Classification Using Multi-Temporal InSAR Data in West Plateau of Sichuan

作　　者：彭立[1,2] 杨武年[1,2] 黄瑾[1,2] 

机构地区：[1]成都理工大学管理科学学院,成都610059 [2]成都理工大学地学空间信息技术国土资源部重点实验室,成都610059

出　　处：《西南大学学报（自然科学版）》2016年第5期125-132,共8页Journal of Southwest University(Natural Science Edition)

基　　金：国家自然科学基金项目(41071265);教育部博士点基金项目(20105122110006)

摘　　要：获取了川西高原区多时相Sentinel-1星载合成孔径雷达(SAR)影像,引入了地形辐射校正技术,发展了一种阴影双视向补偿算法,综合分析后向散射系数、纹理特征、相干系数、数字高程的统计特征,采用支持向量机进行了土地覆盖分类.结果表明,对地形起伏大的高原山区,该方法能够有效提取常绿针叶林、混交林、灌木林、天然草地、水体等地物信息,分类总精度为83.62%.This paper used multi-temporal SAR images of Sentinel-1 satellite for land cover classification. Radiometric terrain correction was introduced to process the SAR images. A dual-visual-direction shadow compensation algorithm was developed here. Based on the statistical analysis of backscattering coefficient, textural features, coherence coefficient and digital elevation for different kinds of land covers, the Support Vector Machine （SVM） was applied to the classification. The results show that this method can effectively extract the features of evergreen coniferous forest, mixed forest, shrub forest, grassland and water in greatly undulate terrain plateau, with an overall accuracy of 83.62%.

关 键 词：多时相 星载合成孔径雷达 分类 遥感 植被 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]