Research on the Classification Based on the Reconstruction of Solder Joint
|School||Harbin Institute of Technology|
|Course||Computer Science and Technology|
|Keywords||solder joints classification 3D feature support vector machines AdaBoost Boostrap|
With surface mounting technology (SMT) to a higher density, smaller size, more complex mixture of PCB-depth development, the solder joints acting as connecting bridges between the board and components play a decisive role. SMT assembly processes are stepped by solder paste printing, component placement, post reflow, and defects arising from solder paste printing process and post reflow processes take a critical part. It is important to discover solder joint defects in a timely manner for the assurance of board’s quality.Because the destructive testing damages the electronic device in some respects, so it is not adapt to carry out in pipelining. It is an exigent requirement for nowadays electronic industry to design non destructive ,effective and low-cost testing methods and technology. In view that most of methods are so dependent that recognition rates strongly tie to their illumination systems and limited number of defected solder joints could be obtained ,this thesis proposes a method of solder joints classification based on both 3 dimension feature and C support vector machines (C-SVM).Main work this paper completed are: the feature extraction based on the reconstruction of solder joint is imported to solve the drawback of two dimensional features which used in two dimension solder joint classification method can not perfectly describe the shape of solder joint, in other word, the two dimensional features don’t fully use the information of the solder joint. The result of the experiment shows the feature extraction based on the reconstruction of solder joint is a better method. Used the combination of SVM and Adaboost, Boostrap to classify the solder joint, has a better result than only use the SVM.