A Monocular Vision Positioning Method For Overhead Transmission Line Inspection Robots Based On Optical Flow And Feature Matching

Positioning technology is one of the most important issues in the research field of overhead transmission line inspection robot.In order to provide an accurate feedback of the defect locations for maintenance,the robot needs to know its accurate position in the operation environment.The existing overhead transmission line inspection robots mainly adopt the methods based on GPS or wheel encoder,which have the disadvantages of poor positioning accuracy and insufficient environmental adaptability.They can only give a rough feedback of robot position,and cannot meet the requirements of operations which need high-precision,such as defect positioning.With the development of machine vision,vision-based positioning method has been widely used in the indoor and outdoor positioning of mobile robots.It makes full use of environmental information and has the advantages of low cost and high precision.Based on the robot platform existed,a visual-based positioning method for overhead transmission line inspection robot based on optical flow and feature matching is proposed according to the characteristics of operating environment of overhead ACCC wires.The position of the robot on the transmission line is determined by using the texture of the line.In order to extract more texture features and reduce the interference of the background,the research of wire detection and feature extraction were carried out.The local binary mode(LBP)and gray level co-occurrence matrix(GLCM)feature were fused for classifier training for wire detection and recognition,and the wire image was separated from the background.According to the characteristics of the surface texture distribution of ACCC wires,the parameters of different orientations were kept in the vector,and the optimized model was determined.In order to improve the stability of the system,the point and line feature extraction is extracted.In order to improve the stability and accuracy of the system in outdoor environment,the optical flow tracking and line feature matching were integrated for calculation.Because the optical flow tracking method was not robust enough and the accuracy was decreased obviously under the illumination change,the camera response model was integrated into the optical flow tracking method,and the camera response function was calibrated through sequences of images with different exposure.The data of optical flow tracking and feature matching are fused by Kalman filter to improve the accuracy.In order to test the performance of the system,after the robot system integration was completed,the experiments of the system in continuous motion mode and automatic detection mode were completed on the outdoor ACCC wire testbed,and the positioning accuracy was verified under transmission line with different angle.Compared with feature points matching method,the accuracy increased by 50.86%,compared with LK optical flow method,the accuracy increased by 65.53%.