Abstract: In order to realize the automatic crack detection of train wheel,a crack recognition and measurement algorithm based on machine vision is proposed. It reduces the image noise,improves the contrast of the image features,proposes crack feature segmentation algorithm, extracts the aspect ratio of the crack area,and completes the crack classification and recognition according to the connected domain information.Finally,the method is used to measure the crack length.Experiments show that the proposed method can effectively and accurately detect the cracks,with recall of 98.56 %,accuracy of 91.12 % and identification speed of 48.5 ms.

Key words: fluorescent magnetic particle inspection, rain wheelset, connected domain, crack detection