Pose-guided YOLO-CED: A two-stage framework for robust SMD-PCB defect detection in mobile manipulators.

Area of Science:

• Robotics and Automation
• Computer Vision
• Manufacturing Engineering

Background:

• Printed circuit board (PCB) and surface-mount device (SMD) stability are critical in industrial production.
• Mobile manipulators face challenges in automated PCB inspection due to pose uncertainty, SMD variations, and computational limits.
• Current methods struggle with stable defect identification across diverse categories.

Purpose of the Study:

• To propose a robust two-stage method for SMD-PCB defect detection using mobile manipulators.
• To enhance the accuracy and stability of defect identification in automated manufacturing inspection.
• To address limitations in current automated visual inspection systems for PCBs.

Main Methods:

• A two-stage approach combining target pose estimation with a novel YOLO-CED model.
• Integration of a color attention mechanism for adaptive feature weighting and robust pose estimation.
• Implementation of an adaptive gating mechanism, multi-sensory field feature acquisition, and edge sharpening for enhanced feature fusion.
• Utilizing a detail restoration branch for precise boundary localization and an L2-normalized self-supervised multi-head attention mechanism.

Main Results:

• The proposed YOLO-CED method achieved a mean average precision (mAP) of 81.7% on the HUSTPCB dataset, outperforming lightweight models.
• The algorithm demonstrated successful application on a mobile platform with a robotic arm for flexible PCB manufacturing inspection.
• Validation on the PKU-Market-PCB dataset confirmed the universality and transferability of the YOLO-CED approach.

Conclusions:

• The developed two-stage method effectively addresses challenges in mobile manipulator-based PCB defect detection.
• The YOLO-CED algorithm offers a promising solution for stable and accurate SMD-PCB defect identification in industrial settings.
• The system's performance highlights the potential of mobile manipulators in flexible and automated PCB manufacturing processes.