부산대학교 도서관

Automated optical inspection (AOI) is used to verify quality of printed circuit board (PCB) assembly and has been proposed for detecting counterfeit components and malicious "trojan" PCB modifications. Component pin localization and characterization is an important step in both of these processes. We present PinPoint: a computer vision algorithm which extracts pin information from surface-mount device (SMD) contours. PinPoint is robust against contour noise, component size, and package type. We evaluate PinPoint against a sample of SMD contours and show that it achieves remarkable performance. Our algorithm could serve as an efficient pin localization step in traditional assembly quality checks and can support future efforts to extract expensive-to-forge characteristics of SMD packages to improve optical assurance.