Design and Implementation of an Eye-in-Hand Vision-Guided QR Sorting System on Dobot MG400

This research presents a self-developed programming framework using Python and OpenCV, without manufacturer-provided software, to control a Dobot MG400 for automated pick-and-place tasks. The system handles mock workpieces labeled with QR code stickers. Custom algorithms were developed for QR code detection and edge-preserving image enhancement, while manual planar homography calibration was applied to accurately convert image-plane coordinates into real-world robot coordinates. An eye-in-hand camera mounted on the robot captures images and sends them to a computer for processing. The calculated coordinates are transmitted back to the robot to perform pick-and-place operations at predefined locations. Experimental validation included 22 cycles under varying object positions and lighting conditions, achieving a 100% success rate. The average QR code detection confidence reached 99.34% ± 0.32%, demonstrating highly consistent identification. The average positioning error was 0.20 ± 0.12 mm, with deviations between 0.057 mm and 0.397 mm, indicating sub-millimeter accuracy and strong repeatability. The average cycle time was approximately 6.8 seconds per operation. Real-time trajectory visualization enabled quantitative motion analysis. Results confirm the system’s reliability, precision, and scalability for modern manufacturing applications.