| With the rapid growth of transmission lines,the workload of transmission lines inspection increases rapidly.The way of traditional manned patrol is far from matching the requirement of the current smart grid construction.The usage of unmanned aerial vehicles(UAVs)based inspection way can reduce the workload and maintence costs,and also improve the inspection quality and scientific management level,which has significant research impact and wide application.This work studies on the intelligent UAVs based transmission lines inspection,and mainly focuses on the 3D mapping for transmission line scene and its application.The inspection task and robotic mapping techniques are explored,the line-based monocular SLAM(Simultaneous Localization and Mapping)problem and viewpoint planning problem are investigated.This work mainly disscusses some keypoints which include line representations,back-end optimization,loop closure detection and viewpoint selection problems.The main contributions include:(1)A graph-based line monocular SLAM algorithm is proposed which constructs an expressive map,and estimates more accurate poses than filter-based line SLAM.It solves the problems of mapping inconsistency,high computation cost and estimated worse results under the large-scale scene.(2)A loop closure detection method which uses image line features is proposed.Firstly,a binary LBD(Line Band Descriptor)is extracted,which improves the efficiency of subsequent procedure;secondly,a weight assigned method for the visual is presented to enhance the discrimination of each "word",it is named as TDI(term frequency&discrimination coefficient&inverse document frequency);thirdly,a whole loop closure detection algorithm is realized.The experiments show that the proposed algorithm achieves higher successful loop detection rate than traditional point-based method.(3)An improved incremental smoothing and mapping method is presented.Firstly,the high order unscented transformation is adopted to improve the original incremental smoothing and mapping method,which raised the status estimation precision;secondly,model the line based SLAM using factor graph,and the improved incremental smoothing and mapping method is employed to solve the line based SLAM problem;Finally,both simulated and real experiments verify the effectiveness and accuracy of proposed algorithm.(4)A viewpoint planning strategy for multi-rotor UAVs based transimission tower inspection is proposed.Firstly,construct safe inspection region according to the inspection rule to gurantee a safe inspection.Secondly,three factors are considered to make the evaluation criteria for candidate viewpoints which include fault type of equipment,covered information and observed quality.In correspondence,viewpoint entropy,related entropy and visibility indexes are quantified toconstruct the evaluation function,thus a lot best viewpoints can be selected according to the evaluation score;thirdly,path planning is necessary to perform on the safe inspection region to ensure each viewpoint planned can be visited.The experimental results verify that the proposed strategy is able to select the best viewpoint,from which the most information and the best viewing quality can be obtained to judge the status of equipment.(5)A 3D laboratory test-platform for overhead power line inspection and a transmission tower model are designed to simulate UAVs’ inspection of transmission lines.The proposed mapping algorithm and viewpoint planning strategy are testified on these platforms.The results indicate that in the images captured from transmission line scene,line features are abundant,the proposed line feature based loop closure detection method obtained a high successful loop detection rate,and the map reconstructed by the presented algorithm is able to express the environment structure significantly.Meanwhile,the viewpoint planning results verify that the viewpoint selection strategy is able to select the viewpoints which meet the requirements of inspection. |
PDF Full Text Request