Map-building Research On Electric Transmission Line Based On Monocular Vision

The traditional power transmission line inspections,such as manual inspection and helicopter inspection,have the disadvantages of inefficiencies,high risk and low accuracy.To overcome these shortcomings,flying robots began to gradually assume the transmission line inspection tasks.The construction of the map of the transmission line scene is the most important task in the inspection task.The reasonable construction of the scene map is related to the positioning calibration and path planning of the flying robot inspection.In recent years,Simultaneous Localization and Mapping(SLAM)technology has been paid attention to by domestic and foreign research institutes and many exciting results have been obtained.It provides conditions for flight robot with vision sensor to do the power line inspection.The SLAM is a crucial research innovation direction.The monocular video eamera has the advantages of simple operation and low operation eost,and can well meet the need of flight robot for the power robot inspection.This paper mainly studies the method of constructing the transmission line scene map based on monocular vision.Image feature matching is the foundation of map construction.There are many algorithms for image matching,such as SIFT algorithm and SURF algorithm,but the real-time performance is poor.Therefore,Oriented FAST and Rotated BRIEF algorithm(ORB)is used in this paper.In view of ORB algorithm does not have the scale invariance in feature point matching,this paper proposes an improved ORB algorithm.Firstly,the Laplace method is used to delete the extremum.Then multi-scale space is generated and the stable extremum points are detected in the scale space.The ORB descriptors are used to describe the feature points,generating binary descriptor with rotation invariance.The feature points are matched by Hanming distance.Finally,the mismatch is removed by the improved RANSAC algorithm.The experimental results show that the algorithm proposed in this paper can effectively solve the problem that ORB does not have the scale invariance.When the image scale changes,the average accuracy of the feature point matching of the improved algorithm reaches about 83.9%,which is improved by about 58.5%compared with the ORB algorithm.At the same time,the matching speed between the improved algorithm and the ORB algorithm is roughly the same,and the improved algorithm retains the fast superiority of the original ORB algorithm.Therefore,the improved algorithm has application value.In this paper,the technology of camera position and pose estimation is studied deeply,and the incremental map is constructed according to the characteristics of rigid body motion.In view of the background of the transmission line scene,this paper proposes a camera position estimation method based on the object line frame model.This method is well applied to the position and attitude estimation process of single camera in the transmission line scene,and is optimized by graph optimization.Experiments show that this method not only completes the purpose of constructing monocular vision transmission line scene map,but also has higher accuracy and robustness than traditional methods.