The method for identifying sleeper digital code based on maximum stable extremum region (MSER) and normalized crosscorrelation (NCC) was proposed according to the imaging characteristic of sleeper digital code, which facilitates management and maintenance of sleeper. At first, the image was preprocessed by gaussian filtering. Then the MSER algorithm was used to generate an extreme value region based on the gray contrast of the digital code and the background,by which the preliminary localization of the digital code area could be realized. Besides, the geometric prior knowledge of digital code was used to remove the connected domains of characters that did not meet the conditions, so as to locate the digital code region accurately and reduce numerous search space for subsequent digital code recognition. Finally, NCC algorithm was used to calculate the digital code similarity metric values of the template image and the localization area to realize the identification of digital code. The experimental results show that the method can be used to accurately locate the digital code on the sleeper and recognize it efficiently with the recognition rate of 95.5% and the recognition time of 0.24 s.