The extraction of drainage system is necessary in many geoscience research fields. For instance, drainage system is an important indicator for structure and lithologic interpretation, sample sites in stream sediment geochemical exploration are designed according to drainage system, and drainage system needs to be recognized and masked in mineral alteration extraction. The drainage system in remote sensing image is generally extracted according to spectral features of water body. However, in the dry drainage systems, such as gullies and seasonal rivers in dry season and under prolonged dry condition, the method based on water body is not applicable. To tackle this problem, the authors propose a method based on independent component analysis (ICA). ICA is a signal decomposition technique that converts multispectral data to independent components which represent independent signal sources, thereby enhancing and separating the specific target in the image. The streambed system extracted by ICA may be still accompanied by noisy data. A series of methods are used to enhance image and remove noise, which include background suppression, morphological filtering and de-noising based on morphological features. The proposed method was tested with ASTER data from Huogeqi area of Urad Rear Banner in Inner Mongolia Autonomous Region, and the result was compared with that derived from supervised classification. The results indicate that the method proposed in this paper can be used to identify dry drainage system, and the recognition result is better than the traditional supervised classification method. The method put forward by the authors performs better in interference information reduction and de-noising, and training data are not needed in this method. In conclusion, the method proposed in this paper is ideal and practical in the extraction of dry drainage system.