On-line Inspection System for Finished Circuit Board Based on Machine Vision
Wang Junhong,Wang Xibin,Jiao Zhi,Du Peng,Wang Nan,Wang Yuran
DOI: https://doi.org/10.1109/cac53003.2021.9727343
2021-10-22
Abstract:Based on the rough positioning and precise positioning methods of visual detection, ROI multi-partition, BP feature analysis within the partition and multi-factor model detection methods are adopted to detect the surface defects of the components themselves and the welding quality of the components. The detection is based on hierarchical classification and block detection and machine learning function. The components that are prone to problems will be tested in the first order to improve the detection efficiency, eliminate the unqualified products as soon as possible, and send the qualified products to the next process as soon as possible to improve the production efficiency. The software design of the detection system is based on C++ programming language and Halcon visual library. After the image is collected, the image is preprocessed, and the image gray stretch is used to enhance the image contrast. By means of mean filtering, two-dimensional threshold segmentation and Sobel operator edge algorithm are adopted to quickly and accurately obtain the region of interest in the visual image background. According to the ORB algorithm and related algorithms, the template matching feature recognition is carried out, the area detection is carried out, the surface defects of components are detected, and whether there are flip-side welding defects or not. This testing system is suitable for on-line testing of PCB products in large quantities.