Research on Visual Inspection Method and Instrument of Solder Joint. Issue 3 (2022)

• Record Type: Journal Article
• Authors: Wenjie, Shi; Zhijiang, Zuo; Han, Li; Libo, Pan
• Is Part Of: IFAC-PapersOnLine. Volume 55:Issue 3(2022)
• Page Start: 131
• Page End: 136
• Publication Date: 2022
• Subjects: solder joint inspection -- visual detection platform -- deep learning -- YOLOv3 -- Machine vision; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2022.05.023
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: Due to the instability of stamping and welding process, there will be missing welding in the production process of auto parts, and multiple welding spots appear on the multi angle joint surface of the workpiece. However, the manual detection is inefficient and the accuracy is poor, and the leakage often occurs in the above situations. In view of the above problems, an intelligent detection algorithm of solder joints based on deep learning YOLOv3 framework is proposed. The machine vision equipment for inspecting solder joint is built, and the welding spot inspection experiment is carried out on the platform. A large number of welding spot images are collected and samples are enhanced to obtain different sample data. By comparing this method with the Hough circle transform for circle fitting of solder joints in the image, the experimental results show that the proposed deep learning-based YOLOv3 framework solder joint detection effect is better.