Lean Planning Of Three-dimensional Safety Viewpoints For Industrial Point Cloud Map

Industrial inspection imposes strict requirements on safety and efficiency.Effective planning of unmanned aerial vehicle(UAV)safe inspection viewpoints based on industrial point cloud maps is of great significance for improving maintenance quality and ensuring the stability and continuity of power supply.UAVs,with their advantages of small size,high flexibility,and low cost,have gained wide popularity in the field of power line inspection.Substation inspection is a typical application scenario of industrial inspection,where the main task of UAV inspection is to capture image information of the inspection objects using onboard cameras and perform subsequent processing.This paper focuses on the lean planning of three-dimensional safe viewpoints based on industrial point cloud maps,with a specific emphasis on viewpoint planning and path planning methods for substation point cloud maps.The research content of this paper is divided into four parts.Firstly,a deep neural network is used to segment the point cloud map and then voxelized to construct a safe airspace for UAVs by adjusting the voxel quantity based on the equipment voltage level and UAV size,enabling the UAVs to fly freely.To ensure a reasonable number of candidate viewpoints for inspection objects of different sizes,a voxel-based redundant iterative random sampling method is proposed.The viewpoint representation employs a fivedegree-of-freedom description,considering camera imaging geometry constraints and onboard camera pitch angle constraints in viewpoint generation.Secondly,a single gravitational potential field method is employed to determine the direction of candidate viewpoints,and the quality evaluation criteria for viewpoints are quantified based on voxel surface visibility.Thirdly,a single-point-single-capture viewpoint optimization strategy is proposed to generate one viewpoint for each inspection object.To reduce the number of waypoints,a single-point-multiplecapture viewpoint optimization method is further presented,allowing the capture of multiple inspection objects at a single spatial location.Lastly,to find the optimal flight path between waypoints within the UAV’s safe airspace,satisfying safety and feasibility requirements,a path planning method based on a priority queue and multiple random trees expansion is adopted,addressing the path planning problem between waypoints in complex environments.The proposed theoretical methods in this paper are experimentally validated,demonstrating the rationality and effectiveness of the algorithm for safe viewpoint planning in substation inspection using UAVs.The experiments verify that the proposed algorithms significantly improve the safety and efficiency of industrial UAV inspection operations.