采用混合像元分解方法提取“黑土滩”信息以实现玛多县“黑土滩”的遥感识别。以Roberts提出的多端元混合像元分解(MESMA)模型为基础，分析玛多县的地物组成并提取端元；针对MESMA在本研究区应用存在的不足，优化了分解策略和机制，并利用改进后的MESMA，结合"黑土滩"的形成机制提取研究区的“黑土滩”信息。最后，采用实测数据对本文“黑土滩”识别结果进行精度验证，得到的精度为82.81%，较MESMA方法的71.09%有明显提高，说明改进后的MESMA可以有效识别“黑土滩”。利用本文方法得到玛多县“黑土滩”面积为1.592×103 km2，主要分布在玛多县北部。该方法在保证"黑土滩"提取精度的同时，效率较传统的目视解译方法有了明显改善。该方法亦适用于其他复杂环境地表组分信息的提取。

The Multiple Endmember Spectral Mixture Analysis(ＭＥＳＭＡ)was used to extract the information to identify the “Black Soil Land” in a remote sensing image.Based on the MESMA proposed by Robert, the surface features in Maduo were analyzed while the endmembers were extracted. In consideration of the insufficient in applying MESMA to this research area, the decomposition strategy and its mechanism were optimized, then along with the formation mechanism of “black soil land”, modified MESMA was applied to extraction of the “black soil land” information. Identified results of “black soil land” was verified by the measured data, which shows the identifying accuracy is 82.81%, significantly higher than the accuracy of 71.09% from the MESMA and proves that the improved MESMA can identify the “Black Soil Land” effectively. On the basis of the improved MESMA , it indicates that the “black soil land” accounts for 1.592×103 km2 in Maduo County, and it mainly distributes in the northern Maduo County.In conclusion,the extracting efficiency for “Black Soil Land” has improved significantly as compared to that of visual interpretation method in guaranteeing the extraction accuracy. Preliminary analysis demonstrates that this method can also be applied to the extraction of surface composition information in other complex environments.