Research On Structural Analysis And Detection Of LIDAR Point Cloud For Overhead Transmission Lines

As a crucial component of the power system,overhead transmission lines(OTLs)play an important role in the transmission and distribution of electricity,requiring regular inspection and maintenance.Traditional manual inspection is inefficient,costly,and dangerous,and is no longer able to meet the needs of modern power inspection.UAVs equipped with Light Detection and Ranging(LIDAR)provide a new solution for modern power inspection.LIDAR technology can rapidly acquire high-precision point cloud of power corridors,but it also generates massive non-OTLs points.Therefore,this thesis is dedicated to achieving structured analysis and detection of OTLs from complex power corridors,so as to quickly and accurately detect the location and status of OTLs,thereby promoting the automation and modernization of power inspection.The main research work and achievements include the following three aspects:(1)A multi-scale grid filtering and fast connected component analysis algorithm based on KD-tree for extracting OTLs is proposed.The multi-scale grid filtering gradually removes ground and object points at different scales based on feature differences,cutting off the connectivity between OTLs and other categories.Then,an improved fast connected component analysis method is proposed to further remove the residual non-OTLs points.The proposed method is demonstrated to be universal and reliable through multiple power corridor,and can accurately and quickly extract complete OTLs in power corridors.(2)A fast and accurate method for structured analysis of power lines is designed.Firstly,a multi-feature combination entropy weighting method is proposed to extract power lines by extracting its key features.Then,the power lines are separate into multiple segments using the point cloud bounding box,and the rapid Euclidean clustering technique is used to quickly divide the power lines into multiple individual lines.Finally,two improved mathematical combination models are proposed to fit a single power line and reconstruct broken power line.The experiments show that the method still has good extraction accuracy for power lines with large span and high drop,and different power line combination models can be flexibly selected according to the requirements of efficiency and accuracy.Moreover,reconstructing the broken power line can improve the continuity and integrity of the power line point cloud.(3)A template matching-based method for detecting insulators is proposed.Firstly,precise extraction of poles is achieved through elevation normalization correction and fast Euclidean clustering.Then,intensity filtering and local entropy filtering are employed in point cloud pre-processing to remove most of the pole body and crossarm points.Finally,in order to address the low accuracy and slow speed of traditional sample consensus initial alignment algorithms,the method is improved by adding distance constraint relationships between sampling point pairs and adaptively adjusting parameters for insulator detection.Experimental results show that this method can accurately detect insulators in towers even when texture information on the insulator surface is incomplete and the density distribution is uneven.The overall accuracy of insulator identification can reach 97.31%,and the comprehensive accuracy from the point of view of points can reach 94.84%,providing refined data support for power grid inspection.