PCB Defect Detection Based on Deep Learning Algorithm

Printed circuit boards (PCBs) are primarily used to connect electronic components each other. It is one of the most important stages in manufacturing products. A small defect PCB can make final product inoperable. Therefore, careful and meticulous detection steps necessary indispensable process. The methods generally be divided into manual inspection automatic optical (AOI). main disadvantage that speed too slow, resulting a waste human resources costs. Thus, order up production speed, AOI techniques have been adopted by many manufacturers. Most current mechanisms use traditional algorithms. These algorithms easily lead misjudgments due different light shadow changes caused slight differences placement or solder amount so qualified PCBs judged as defective products, which also reason for high misjudgment rate detection. In effectively solve problem misjudgment, re-judgment currently reinspection method manufacturers products AOI. Undoubtedly, need inspectors another kind labor cost. To reduce re-judgement, an accurate efficient mechanism based on deep learning algorithm proposed. This mainly establishes two models, classify defects product. When both models basic recognition capabilities, then combined model improve accuracy study, data provided Lite-On Technology Co., Ltd. were implemented. achieve practical application value industry, this research not only considers accuracy, but execution speed. fewer parameters construction model. results show about 95%, recall 94%. Compared with other modules, greatly improved. time image 0.027 s, fully meets purpose industrial application.