Abstract: Ramp winch locking system is an important means of avoiding the simultaneous operation of winches and people walking in the ramp. The traditional locking system has the problems of low accuracy, difficult installation and maintenance, and high hardware cost, and for these problems, a ramp winch locking system based on multi-channel image processing is proposed, and an intrinsically safe locking dection system is developed. The algorithms are categorized into object detection, tracking and occlusion detection. In the object detection, Slim-C3 was designed using GSConv to lighten the backbone and neck networks and to improve the accuracy of target detection by the decoupling head based on GSConv; in the target tracking stage, the tracking algorithm of second-level matching is studied to improve the speed without decreasing the accuracy. The experimental results show that the proposed object detection optimized the AP_0.5:0.95 for people and winch by 0.57% and 1.65% over YOLOv8s, respectively; the proposed tracking improved by 0.55% over YOLOv8s+DeepSort in IDF1; the precision and recall rates of the count were 96.43% and 93.10%, respectively.