Target Detection and Position Measurement Based on Machine Vision for AGV

This paper proposes a strategy based on cooperative target to realize automatic guided vehicle (AGV) terminal detection and position measurement. Firstly, the enhanced YOLOv3 algorithm based on cross layer connection is employed to detect the interest area of the marker points, upon which more refined local features are restored to improve the adaptability to small targets. Then, a high-precision 3D reconstruction with the aid of the ellipse center coordinates is demonstrated to ensure the reliability and avoid the loss of targets due to the failure of landmarks caused by environmental factors. The experimental results revealed that the positioning accuracy error is less than 1.5mm and the misrecognition rate is superior to 1% whilst the AGV is within 5m from the target. Consequently, the proposed approach has the crucial application in rapid and stable AGV terminal vision positioning.