SMD Defect Classification by Convolution Neural Network and PCB Image Transform

• Published in: 2018 IEEE 3rd International Conference on Computing, Communication and Security (ICCCS) pp. 180 - 183
• Main Authors: Young-Gyu Kim, Dae-Ui Lim, Jong-Hyun Ryu, Tae-Hyoung Park
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-10-2018
• Subjects: Brightness; Conferences; deep learning; Histograms; Image color analysis; Integrated circuits; Lighting; machine learning; SMT assembly defect; Testing
• DOI: 10.1109/CCCS.2018.8586818

Surface Mount Technology (SMT) is a manufacturing process in which chips are mounted on the surface of a printed circuit board (PCB). The automatic optical inspection system (AOI) has mainly used the learning-based method for the defect classification of the SMT process, and recently the CNN-based classification method has appeared. However, existing techniques do not consider the area margin of the part and uneven color distribution according to the position of the chip, so the classification accuracy decreases. In this paper, we propose a system that can extract the chip region and improve the color distribution by the input image transformation. We extract the correct chip area through vertical and horizontal projection, and the color improvement enhance the brightness value distribution of the chip image through local histogram stretching. By experimental result, we prove the performance of the proposed classification method.