Research On The Positioning System Of The Wall-Climbing Robot Based On Monocular Vision

Spraying of exterior wall of building is a work with wide and many uses while basically staying at the stage of manual spraying.Scaffolding or high altitude basket is required in the process of construction with many disadvantages,such as high cost,long time,and potential danger.The wall climbing robot is adsorbed on the working surface by the adsorption mechanism and can move on the straight wall and replace manual operation.In order to realize automatic movement and operation of a wall climbing robot,the position feedback of the robot is very important.The vision positioning technology has the advantages of high precision,good real-time performance and convenience.So this paper uses visual positioning as a technical scheme and develops a system for the wall climbing robot to realize the positioning function within the range of the given wall area.First,the overall design of the positioning system is carried out,its working process is clarified,and the overall detection method of ROI tracking + luminescent marker is put forward.On the other hand,the requirements of the visual system are defined and the selection of camera and lens is made,and the software platform and hardware platform of the visual system are constructed.Then,the camera model and the principle and method of extracting world coordinates are described.The principle and algorithm of camera calibration are analyzed.The accuracy evaluation method and operation process of camera calibration are put forward and the calibration experiment is carried out,obtaining high precision camera parameters.And get a bird's eye view from the world coordinate system based on the calibration results.Then,this paper adopts Canny and LoG edge detection and Hough linear feature extraction algorithm and it is validated that the luminescent markers can be detected successfully under extreme light conditions.Based on this,the position and direction detection algorithm for single frame ROI images is proposed.Finally,the robot detection scheme is analyzed and a scheme to detect image coordinates and then extract world coordinates is adopted.The algorithm is designed and the experiment is carried out to verify the correctness and rapidity of the algorithm.Also,the accuracy evaluation experiment of the world coordinate detection is carried out,and the error is within the acceptable range.And the world coordinate continuous detection experiment is carried out with ROI image acquisition and realizes real-time image acquisition and the position and direction detection.