Investigation On Active Infrared Defect Detection Method Of Exposed Solder Joints In Aerospace Electronics

The application environment of exposed solder joints in aerospace electronics is extremely harsh in the universe,so aerospace electronics has extremely high requirements for the reliability of solder joints.At present,the exposed solder joints are welded manually or by machine.Under the influence of environmental factors,virtual solder joints,false soldering,and tipping defects may occur.The solder joint defect detection is mainly by manual visual inspection,which has low efficiency and relatively low accuracy.Active infrared thermal imaging inspection is a non-destructive inspection technology,which has non-contact measurement and fast inspection speed,and can obtain internal and external structural information of solder joints.This thesis studies the infrared image defect detection of solder joints based on traditional image processing methods and based on the YOLO neural network model,and explores effective methods to achieve automatic inspection of solder joint defects.First,the principle of active infrared testing is introduced.Aiming at the detection of aerospace electronics exposed solder joints,A hardware platform for active infrared inspection of solder joints was built.The solder joint defects and corresponding infrared images are analyzed,and the three solder joint defect characteristics of holes,notches,and breaks are selected as the research goals.Then the difficulties in infrared inspection is analyzed.This thesis research on detecting solder joint defects basing on traditional image processing methods,and introduced image segmentation,Image denoising,edge detection,false edge removal and other image processing methods.then the specific process of extracting solder joint defects based on image processing is summarized.After experiments,it is found that the accuracy of the detection method needs to be improved and it has relatively limitations,witch leads to poor practicality.Aiming at the fact that there are few samples of infrared images of solder joints and the image structure is consistent,an active infrared detection method of solder joint defects basing on improved Tiny-YOLOv3 network is proposed.By introducing the Mobilenet model to enhance the feature extraction network of Tiny-YOLOv3,the automatic detection of various defects in the infrared image of the solder joint is realized.Experimental results show that the network improves the average accuracy value by 21.62% compared to Tiny-YOLOv3,reaching 82%,and the detection speed reaches basically the same,reflecting the application prospect of this method in solder joint detection.