建筑物损毁情况是地震灾害评估的一项重要指标，利用遥感技术快速提取震后建筑物震害信息，对科学指导地震应急救援工作具有重要意义。本文利用2010年4月14日青海玉树7.1级地震前后玉树县结古镇团结村高分辨率遥感影像，结合像素光谱和空间特性的纹理、结构等多源信息，基于支持向量机（SVM）方法对地震前后建筑物信息进行分类提取，变化检测出建筑物损毁情况，并与面向对象多源信息复合的模糊分类法的分类精度、提取效率进行对比分析。研究结果表明，多源数据复合的SVM影像分类方法能够有效解决模糊分类影像破碎问题，地震前后两实相影像分类总精度达到77.53%和73.56%，提高了建筑物震害信息提取精度。

 The collapse of building is a critical measurement on the earthquake hazard assessment. Using remote sensing technology rapidly extracted earthquake building damages，and scientifically guiding earthquake emergency work has important significance. This paper illustrates the remote sensing images of unity village before and after 7.1 earthquake in Qinghai Yushu county on April 14，2010，using the SVM method to extract the buildings integrating the information of spectral，texture and structure during the earthquake，to detect the building damages，and comparing to the accuracy of image classification and the extracting efficiency based on fuzzy method with multi-source data. This shows that the image classification based on SVM method with multi-source data can solve the image classification fragmentation which is based on the fuzzy method. The total accuracy of classification reached 77.53% and 73.56% using the images of before and after earthquake；and improving the extracting accuracy of damage buildings.