Research on Brake Pad Surface Defect Detection Method based on Deep Learning

In view of the present artificial auto brake block defect detection efficiency is not high, the traditional image processing methods such as morphological feature for automobile brake pad background complex, coexistence of different defect types, unable to realize accurate positioning and accurate classification of several kinds of defects, this paper proposes a car brake pad surface defect detection method based on deep learning. The object detection YOLO algorithm is applied to the brake pad surface defect detection. Firstly, construct the brake pad defection dataset, and then enhance and label these images. In order to improve the detection effect when different types of defects coexist, we optimize the YOLO model by combining the SPP structure, and add the K-means clustering algorithm to optimize anchors positioning. The optimized YOLO network model is used to train and test the constructed brake pad surface defection dataset. Experiments show that the optimized YOLO model greatly improves the detection accuracy of brake pad defect types, and the defect location is more accurate. For the defective brake pads that have been detected, the ROI area can be selected according to the location, and the HALCON software can be used for further defect analysis to feedback the defect characteristics, so as to facilitate targeted improvement in the subsequent production process. This method can overcome the influence of complex background and subjective factors on brake pad defect detection, and has certain engineering reference value.