Reference-Free PCB Defect Detection Based on Two-Stage Segmentation and Rule Verification

This paper proposes a novel reference-free detection method for several defects of Printed Circuit Boards (PCB), including short, open circuit, spur, and conductor hole integrity. Firstly, an improved Fuzzy-c means (FCM) algorithm is developed for the PCB image segmentation to obtain the initial deblurred and denoised binary image. After that, we utilize the fast active contour segmentation method to obtain the final binary image. On this basis, the rule verification method is designed considering the PCB process characteristics to detect the line through the connectivity component and area feature analysis. For the conductor hole defects, the proposed method of region combination features to achieve PCB image conductor hole detection according to the geometric characteristics of the circle. Experimental results show that the proposed method achieves a promising detection performance with low computational costs.