Research On Abstacle Recognition And Location Of Inspection Robot Based On Binocular Vision

High-voltage transmission lines are the main way to distribute power over long distances.The detection and maintenance of transimission lines is one of the key development plans for power transmission.With the development of machine vision,major universities and research institutes have tried to apply machine vision to line-tracking robots instead of sensors to detect line obstacles and environmental conditions,which has greatly improved the intelligence level of line-tracking robots.The control system of line-tracking robot is still immature,and it has the problems of single recognition type and low positioning accuracy.Therefore,an online obstacle recognition and positioning method based on binocular vision is proposed.Firstly,in view of the single problem of obstacle recognition for existing line-tracking robots,a visual inspection system for the recognition of obstacle types for high-voltage transmission lines is established.Based on Histogram of oriented dradient(HOG)feature extraction and support vector machine(SVM)classification recognition,designed a multi-cycle and multi-class recognition detection method.Each class of obstacles is trained separately.Based on SVM single-class recognition,It performs loop recognition detection,which solves the difficulty that a single classifier cannot identify multiple obstacles.It gets rid of the constraint that the multi-class recognition library can only be applied on MATLAB.Secondly,for the problem of low positioning accuracy of line-tracking robot obstacles,based on the analysis of monocular and binocular distance measurement,combined with the specific conditions of line hardware obstacles,a high-voltage transmission line hardware obstacle positioning Binocular vision system is established.The system is implemented by the cooperation of the left and right cameras.Based on the obstacle recognition of the left camera,the preset points are taken in conjunction with the identified obstacles,and then the distance is taken by the binocular camera,and as the robot advances,it constantly updates the distance between the obstacle and the lens.Finally,an online obstacle recognition and positioning system based on binocular vision was built on the existing line patrol robot platform.The binocular vision system consists of a multi-class recognition system and a real-time ranging system.Sensor information is combined to complete inspection tasks.After a series of experimental verifications,the experimental results show that the visual recognition positioning system built in this paper has fast positioning,high accuracy and real-time performance for the identification of line obstacles.It fully meets the online obstacle recogintion and positioning requirements of line patrol robots.