Water Body Extraction in Arid and Semiarid Areas Based on TM/ETM Remote Sensing

In this paper, using TM/ETM images of Jinghuiqu Irrigation District in semiarid areas of Shaanxi Province, China, evaluated the multi-band water body extraction methods, such as threshold method, water index method, and spectral relationship method and so on. The experiment found that these methods were ineffective when extracting water body from the images of semi-arid area with the background of the betrunked river, the residential areas and the soil, and the extracted water body were incomplete. This paper presented a comprehensive extraction method, using (ETM2 + ETM3) / (ETM4 + ETM5)> I and Tasseled Cap Wetness (KT_3 > threshold) model, accurately extracted out the water body. This method reduced the error rate of the extraction, and it can effectively extract the tributaries and small water body. The experiment showed that the method met the accuracy demand (96.86%), and results of classification are better than that of common methods.