Abstract: ETM~+ thermal infrared image is used to detect cloud, snow and ice automatically around the Zhongshan Station in Antarctica. Better classification result is achieved by adding thermal radiation where brightness temperature information of ETM~+ is used and texture feature of Entropy derived from gray level co-occurrence matrix. Then we improved the method of the minimum distance classification-the weighted minimum distance classification. This algorithm improves the accuracy to 96.0% in the experimental satellite image of ETM~+ around Zhonshan Station of China in Antarctica.