应用混合像元分解提取胡杨覆盖度信息1）

摘要

以内蒙古荒漠地区胡杨林为研究对象，应用混合像元分解方法从TM多光谱数据中提取了胡杨林植被覆盖度，并以高分辨率Quickbird影像分割结果作为真值进行精度评价，与传统的基于植被指数提取植被覆盖度的方法进行了对比。结果表明：基于几何顶点端元选取的混合像元分解方法可以有效提取胡杨植被覆盖度（ R2＝0．893，RMSE＝0．12），优于植被指数回归方法提取精度（R2＝0．574）。研究结果有助于开展荒漠地区较大范围的胡杨林动态监测和保护。%We developed a spectral unmixing method using TM multispectral data for Populus euphratica in Inner Mongolia desert area to extract vegetation coverage.The decomposition result was evaluated by comparing with tree crown segmentation re-sults from a Quickbird image.Then, the method was compared with the traditional vegetation index based method.Spectral unmixing based on geometric-vertex-based end-member selection method can effectively extract Populus euphratica vegeta-tion coverage (R2=0.893, RMSE=0.12), which is better than using traditional vegetation index based regression method ( R2=0.574) .Our results can improve dynamic monitoring and protection at large scalefor desert poplar.