PCB Solder Joint Location And Defect Detection Based On Machine Vision

With the rapid development of electronic industry,printed circuit board(PCB)is becoming smaller and denser.As a result,the welding quality of SMT components on PCB has become one of the key factors affecting the quality of electronic products.Therefore,the solder joint defects detection in PCB has become an essential step in the electronic industry production.This paper develops a solder joint defect detection system for PCB components based on machine vision,including common defects such as bridging,welding leakage and so on.The work of this paper summarized as follows:1.According to the specific requirements of solder joint defect detection,this paper proposes the solder joint detection system scheme and the basic composition of this system,and also chooses the key hardware including light source,lens,camera,motion control card and so on.At last,this work builds an experimental verification platform.2.To deal with the poor stability,the rotation and scale sensitivity and low matching speed in many existing template matching methods,this paper develops an image matching method based on gradient vector to determine the position of electric components.In which,an image pyramid strategy is introduced to realize image matching strategy from coarse to fine for improving the matching speed.Meanwhile,this paper introduces a stopping calculation strategy in each matching process to further reduce calculation burden.Finally,a sub-pixel fitting is carried on the initial matching results to achieve higher matching accuracy.Experimental results show that the proposed localization algorithm meets the requirement of solder joint detection platform.3.This paper proposes a solder joint defect detection method based on deep learning to overcome the poor performance in some existing methods because of the hand-crafted feature extraction.First,this paper develop s a small convolutional neural network to classify images of solder leakage,bridging defects and normal solder joints.Then,the defects are located by combining classification results and solder location information.Experimental results show that the proposed method is effective for solder joint detection.4.Finally,the work develops a PCB solder joint detection system software via Microsoft Visual Studio 2010,and conducts many debugging and verification work.Experimental results show that the developed visual inspection system can achieve good results in solder joint defect detection and provide certain reference value for automatic visual inspection of PCB.