The interpretation and application of optic remote sensing images, including aerial photos and satellite images, had become the indispensable tool for railway engineering geological survey. But microwave remote sensing images, mainly SAR images, were rarely used in the application. In the paper, SIR-C SAR remote sensing data were used to identify different rocks and survey the lithological distribution around Qinghe River in the north part of Xinjiang Province. Firstly, the multi-frequency and multi- polarization data were used to produce the false color composite image for visual interpretation and lithological mapping by remote sensing digital image processing and enhancement. Secondly, the various backscattering data of containing abundant information about rock’s characteristics were input into the computer Neural Network Classifier. The overall accuracy of lithological classification is 62.6%.