Abstract: For phenomenon of incomplete water extraction and false extraction in existing water extraction method based on remote sensing images, a water extraction method of mining area collapsed pond based on SPOT-5 multispectral images was proposed.On the basis of adding a available band by use of band synthetic, existing water extraction method was improved appropriately, and mining area water body was extracted by decision tree classifier and improved method with four-level extraction, so as to ensure the water body extraction integrity and reduce error rate. Finally, accuracy of the water body extraction was assessed by the extracted data. The test results showed that the water body extraction method based on decision tree classifier has higher accuracy and can meet with demand of practical application in mining area.