S3DA: A 3D Point Cloud Based PCB Solder Defect Detection Algorithm

Yilongfei Xu, Zhewei Wang, Jinhao Liang, Yueling Zhang, Jincao Feng, Weikai Miao, Jiangtao Wang, Geguang Pu

Abstract: Surface Mount Technology (SMT) is prevalent in Printed Circuit Board (PCB) assembly, mainly using solder printing to connect the components and the board. During the process of solder printing, solder defects due to machine failure and environmental factors are widespread. Existing defect detection methods mainly use computer vision to detect solder defects. The main idea of this type of method is to obtain the image information and defect features of the PCB and use the machine learning model to identify the solder defects of the PCB. In actual industrial PCB assembly, the lack of illumination and the occlusion caused by other workpieces leads to incomplete input images for machine learning models, which makes existing methods unable to detect such occluded defects. In order to solve the above problems, this paper proposes a new algorithm for solder defect detection using 3D point cloud data. First, the point cloud data is obtained by scanning the 3D point cloud camera. Next, the point cloud data is denoised and filtered, and the area of interest is further screened to obtain the solder area to be calculated. Finally, using the idea based on integral summation, solder defects are identified by calculating the solder volume. This algorithm can automatically assist manual judgment and effectively identify possible defects in solder processing.

External IDs:dblp:conf/sfpvv/XuWLZFMWP24