According to the characteristics of bearing characters, an automatic identification method based on machine vision is proposed. The detection software was developed on C++ and OpenCV computer vision library. Firstly, the bearing character image was segmented by threshold, then the edge contour was extracted by Sobel operator, then the circular character region was located by using the improved circle detection algorithm, and character segmentation was performed after the polar coordinate transformation. Finally, the improved character features were extracted, and the support vector machine (SVM) classifier was used for character recognition. Experimental results show that the method can effectively identify bearing characters, and the recognition rate reaches over 95%, which has broad application prospects.