Design of Mobile Robot Vision Positioning System

Abstract

Against the backdrop of rapid advancements in intelligent manufacturing and robotics, enhancing the high-precision autonomous positioning capability of industrial robots in dynamic and complex environments has become a critical challenge. To address the pain points of large initial positioning errors and insufficient accuracy in traditional methods, this study designs and implements a high-precision, high-robustness mobile robot visual positioning system. The system adopts an Eye-in-Hand visual architecture, integrating HOG feature extraction with KCF kernel correlation filtering algorithms to achieve robust target tracking. By combining Blob region analysis with least squares circle fitting, it enables stable recognition and coarse positioning of circular targets, while utilizing binocular vision ranging technology for precise positioning. To improve system accuracy, an error modeling-based pose compensation method is proposed. Experimental results demonstrate that the proposed system effectively enhances the robot's positioning accuracy and attitude control performance, validating its effectiveness and robustness in engineering applications.