Algorithm for Effectively Segmenting High-speed Rail Wheels Defects

Defects will reduce the performance and the quality of the wheels in different degree. Therefore, accurate identification of defects can separate broken wheels and reduce cost. Due to the uneven illumination and the high reflectivity of the wheel, the images of the wheels taken by the camera tend to produce more reflective areas. These reflective areas increase the misclassification rate of the images in the segmentation process. In order to reduce background interference and make the segmentation more accurate, this paper uses the improved Otsu algorithm that limits the gray range in recognition process to segment the defects. The arithmetic in this paper can also speed up the Otsu by the dichotomy. Experimental results show that the improved arithmetic used in this paper can find a better segmentation threshold and has good real-time performance.