Design of Defect Detection Algorithm for Copper Foil Particle Substrate Based on Computer Vision

In the production process, the surface of CFPS (copper foil particle substrate) may be damaged due to equipment and man-made reasons, resulting in various defects. According to the type, size and quantity of defects, whether the board is qualified or not can be judged. Computer vision can adapt itself to the environment, and its highest research goal is to enable computers to observe and understand the world through vision like people. This paper studies the use of computer image processing technology to detect and identify CFPS defects. Wavelet analysis theory is applied to image denoising. In order to meet the application requirements under different production conditions, a lightweight CFPS defect detection model based on YOLOv5 is proposed to improve the detection efficiency. The CBAM (Convolutional Block Attention Module) mixed attention module is introduced to emphasize the important features in the two dimensions of space and channel and suppress the miscellaneous features, so as to promote the information flow in the network and ensure the detection accuracy. The research results show that the lightweight YOLOv5 model achieves the best detection result for four types of defects, and the AP values of scratches and stains are 20.02 % and 18.68 % higher than those of the second SSD, which further proves the excellent performance of the proposed method in CFPS defect detection tasks. The defect detection of CFPS board is realized, and the processing process is convenient and fast, which is suitable for industrial online production detection requirements.