Research On Detection Of Solder Joint Based On Machine Vision

In the process of producing PCB boards,the welding of the electronic components is an important part of the quality of solder joints,also directly affect the quality of PCB boards.In order to ensure that high-quality PCB boards will be applied to the electronic products to improve product qualification rate,to avoid unnecessary losses,PCB boards to achieve the defect detection is also becoming increasingly important.The application of machine vision to the field of solder joint detection,instead of manual detection,to achieve the quality of solder joint testing,which will improve the detection accuracy,increase production,bringing great market competitiveness.This paper is based on the machine vision of the solder jo int defect detection work.The types of defects in the solder joint include connecting,pulling tip,leakage welding,tin and so on.Aiming at the overall design of the solder joint defect,the detailed detection method of defect solder joint is discussed in detail.The main contents are as follows:Firstly,the detection index of all kinds of defective solder joints was analyzed,and the overall scheme of the solder joint detection system was designed.By analyzing various hardware equipment composition and equipment parameters,the model selection of industrial camera,lens and light source was analyzed.The characteristics of the way and to adapt to the occasion were analyzed,selected the appropriate lighting to complete the hardware system to build.Secondly,the method of multi-exposure fusion image acquisition was studied.In view of the phenomenon of uneven image exposure in image acquisition,the image information was integrated by multi-exposure fusion,and various classical image fusion algorithms were introduced to analyze the fusion image quality evaluation index.Then,the classification method of solder joint defect was studied.The samples of qualified solder joints and defective solder joints were collected and classified into training samples and test samples.By using principal component analysis to reduce the dimension of high-dimensional data,and the training samples and test samples were respectively studied and tested by extreme learning machine,the classification results were obtained.The experimental results showed that combinating the principal component analysis with the limit learning machine had higher detection accuracy,better detection performance and less time than other machine learning classification methods.Finally,the three-dimensional reconstruction of the solder joint is studied,and the three-dimensional surface information of the solder joint is obtained,and the defect solder joint is further tested.The principle of geometric optics and the principle of reconstruction were introduced,and the steps of height extraction were analyzed.The method of focusing distance measurement was used to obtain the three-dimensional information of qualified solder joint and defective solder joint,and the data were compared.The results showed that the method can effectively obtain the height information of the solder joint and meet the testing requirements.