Research On Techniques Of Visual Detection For PCB Surface Defects And Identification For Solder Paste Defects

With the rapid development of electronic technology,PCB(Printed Circuit Board)of smartphone tends to be high-density and high-precision.Apply visual detection and identification technology to judge whether there are missing hole,mouse bite,open circuit,short,spur and spurious copper defects and their locations before PCB soldering,identify the soldering deviation,excessive solder,insufficient solder and solider bridge defects and their defect levels after PCB soldering guarantee the quality of the smartphone;In this paper,aiming at the high-accuracy detection and identification tasks of main defects such as PCB defects,PCB solder paste welding deviation,excessive solder,insufficient solder,and solider bridge in the production process of mobile phones,researches are carried out from the aspects of PCB defect detection,solder paste soldering area detection,solder paste defect area segmentation,and solder paste defect identification;The main research contents include:(1)Research on the visual detection method of PCB bare board defects based on Faster RCNN＿FPN.Generative Adversarial Networks(GAN)is used to enhance the data set of the PCB bare board public data set,and the Faster RCNN＿FPN model formed by the combination of Faster RCNN and FPN is used to realize the defect location detection and type identification of the PCB bare board data set,and the accuracy of the detection is verified by experiments.(2)Research on detection of solder paste welding area based on template matching and morphology.Firstly,the PCB image is binarized by setting the grayscale threshold to obtain all the regions of interest,then the solder regions are roughly matched based on the fast template matching method,and finally each solder region is detected by the detection method based on mathematical morphology.The experimental results show that the method can realize the accurate detection of PCB solder paste soldering area.(3)Research on the segmentation of solder paste defect regions.Aiming at the defect of inaccurate initial cluster centers of SLIC(Simple Linear Iterative Clustering)segmentation algorithm,a mean-shift algorithm is introduced to select the initial cluster centers,and a mean-shift-based SLIC super-pixel segmentation algorithm is proposed to achieve solder paste welding deviation defect segmentation.Aiming at the defect that the segmentation accuracy of the level set segmentation algorithm is not high enough,a level set segmentation algorithm based on multi-feature fusion energy term in defect region is proposed to realize the fine segmentation of components with solder paste welding deviation defects.Aiming at the defect that the maximum entropy threshold segmentation method is difficult to accurately select the threshold,an improved salp swarm optimization algorithm is used to optimize the selection of the threshold,and an improved maximum entropy multithreshold segmentation algorithm is proposed to achieve segmentation of excessive solder,insufficient solder and solider bridge.Comparative experiments show that the proposed algorithm can accurately segment the solder paste welding defect area.(4)Research on solder paste defect identification technology.On the basis of realizing the precise segmentation of solder paste welding area,the calculation method based on area rules is used to realize the identification of solder paste soldering defects of excessive solder,insufficient solder and solider bridge,and the calculation method based on relative position judgment is used to realize the solder paste welding deviation defect and identification of defect levels.The experimental results show that the methods in this paper have a recognition accuracy rate of 99.40% for PCB defects,99.01% for solder paste welding defect types,and 97.03%for solder paste solder defect grades.