3D Point-Cloud-Based Safety Distance Detection and Dynamic Warning for Machinery Operating Near Energized Equipment

Abstract

Manual supervision of cranes, aerial work platforms, and other machinery operating near energized equipment is vulnerable to visual occlusion, while fixed distance thresholds cannot reflect dynamic motion trends. This paper proposes a safetydistance detection and dynamic-warning method based on 3D pointcloud perception. First, LiDAR point clouds are fused with machine pose and joint-state information. Statistical outlier removal, coordinate transformation, and point-cloud segmentation are then used to construct a 3D point set of energized objects and a predicted hazard envelope of the operating machinery. Second, the minimum distance between the machinery and energized objects is predicted from the machine motion state, and the safety-distance threshold is dynamically corrected by jointly considering approach velocity, system response delay, ranging uncertainty, and occlusion duration. Finally, a three-level warning mechanism with hysteresis and a direction-selective interlock strategy are designed to provide graded risk alerts and active control.