Precise Positioning Approach To Pin-hole Structure Based On Monocular Vision And Error Analysis

The paper analyzes and launches research on precise visual positioning for live inspection robot for isolating circuit breaker in the intelligent substation,which includes algorithm,error analysis and calibration approach.The live inspection process of isolating circuit breaker linking fitting can be simplified to pin-hole assembly and disassembly.However,distinct from tradition pin-hole assembly and disassembly,the working environment of isolating circuit breaker linking fitting is at outdoor,the operation height is five to six meters and the operation process is under 100 kV live environment.So it is a great challenge on the operation process,algorithm,calibration and reliability of the visual system.The main content of the paper includes:(1)Focus on the pin-hole assembly and disassembly process for live inspection of isolating circuit breaker linking fitting in intelligent substation,the paper designs the plan in detail and makes lectotype selection for the pin-hole precise positioning system including camera,lens and light source.Shading and equipotential system is designed to satisfy the technological requirements because of changeable outdoor lighting condition and strong electromagnetic interference from high voltage.(2)The paper launches research on automatic vision calibration process for single robot and base relationship of dual robot system,establishing internal parameters of the camera,hand-eye relationship parameter and base relationship matrix of the two robots.The paper also analyzes and launch research on possible error source to satisfy the precise positioning requirements of pin-hole assembly at high altitude.(3)The paper analyzes different feature figures based on single ejection of the plane,and designs a feature with three isometric and isomeric circles based on precision requirements.It also verifies the effectiveness of the feature plan.(4)For the monocular vision algorithm for pin-hole cooperation,the paper designs image recognition algorithm including threshold segmentation based on Otsu method and edge extraction based on Canny operator.After elliptical contour is fitted,feature plane coordinates relative to camera coordinates is restructured based on the principle of three points forming a plane.Then the result is transformed to the base coordinates of up robot.To improve precision,an adjustment approach based on pixel coordinates is designed.It excludes influence from calculation error of distance along optic axis.After that the paper transform the positioning result to the pose for the robot used for hole task.The paper also designs two pose transformation approach,no matter the base relationship matrix of the two robots is available or not.The paper also discussed possible error source for visual positioning.(5)The paper does experiment on the possible error sources discussed before in the process of calibration and positioning and influence of shooting angle of the camera on calculation precision.Experiments are also done on convergence of the visual algorithm and precision under different light conditions.The result proves the precision and robustness of the algorithm.