A low contrast defect detection method for PCB surface based on manual labeling

摘要

In order to solve the detection problem of low contrast scratches on PCB board, a detection method based on manual marking was designed by using computer vision technology and OpenCV visual library. Mark the scratches on the PCB board with marker pen, take photos and sample the PCB board with ordinary camera, process the picture (grayscale processing, image filtering and edge detection, etc.), extract the edge with canny edge detection algorithm, and finally extract the ROI area on the defective PCB board. The experimental results show that this method can successfully detect low contrast scratches on PCB.