In order to address the redundancy of image in the detection process of rail surface defects, an algorithm of matching for the redundancies of rail surface images was proposed. At the beginning, the rail surface area was extracted by using the vertical projection method. And then, the location information of defects was obtained through the image preprocess and binarization on the rail surface. Next, the morphological information of the rail surface defects was achieved in the horizontal projection method. At last, the defect location information and morphological information were matched on the basis of the improved fuzzy matching algorithm. The experiment results verify that this algorithm can effectively identify the redundancy information of image, and the accuracy rate is as high as 97.5%.