YOLOv10 기반 PCB 실시간 결함 탐지 및 통합 품질 관리 플랫폼 구현

초록

Traditional quality control in the printed circuit board manufacturing process has limitations of inconsistency, high initial costs, and low flexibility. To solve this, this paper proposes an integrated quality assurance platform combining the real-time object detection model YOLOv10 with a role-based web interface. The proposed system provides a workflow where Artificial Intelligence detects PCB defects in real-time, allowing administrators, operators, and others to review, correct, and analyze the results based on their roles. Experimental results achieved a detection performance of 0.90 mean average precision on a specific PCB dataset. This study proves that by organically integrating Artificial Intelligence automation and human verification, the efficiency and accuracy of existing processes can be improved.