Design and Implementation of a YOLO-Based Visual Positioning System for Train Static Weighing

Static weighting process is a critical component of industrial logistics, yet the prevailing approach to train positioning relies on manual visual guidance. This method is suboptimal, resulting in low efficiency and safety risks in harsh environments. To address these limitations, this study proposes a computer-vision-based positioning system for train static weighing. Industrial cameras are utilized to capture real-time images at both ends of the weighbridge, and a lightweight YOLO model is employed to detect couplers, wheels, and endpoints with an accuracy rate exceeding 98%. The process of camera calibration and the implementation of a pixel-world mapping model, founded on the principle of perspective transformation, is instrumental in the computation of the carriage’s actual positional deviation, in turn, serving as a critical guide for the execution of precise parking maneuvers. Empirical evidence from practical deployment indicates that the system enhances weighing efficiency by 75%, while concurrently facilitating fully unmanned, safer operations on site.