Design Of Mobile Robot Positioning System Based On Machine Vision

With the rapid development of science and technology,the positioning method based on vision sensors is gradually becoming popular.Due to the unique advantage of flexibility,high precision,rapidity and non-contact,the method is used to solve the problem of mobile robot autonomous positioning.However,the application level of vision position system of mobile robot is relatively low in China with the problem of accuracy,robustness and poor real-time performance.According to the above research status,a monocular vision position scheme based on Apriltag label is designed in this paper from the perspective of application.In the first place,the design and implementation of algorithm are completed on Windows platform by using Opencv.After that,the algorithms of image processing are transplanted and optimized by using CCS.Finally,the function of vision positioning is accomplished on DM8148 platform of embedded DSP.The paper is organized as follows:Firstly,considering the functional requirements and application environment of system,a single frame monocular scheme of vision position based on point feature is determined.Furthermore,the determination of label,the selection of camera and DSP are completed to realize hardware construction of the system.Secondly,the closed-loop coordinate system is constructed based on the analysis of theoretical basis of visual positioning and the positioning principle of P4 P algorithm.The calibration of camera inner parameters and hand-eye parameters is completed by using Zhang algorithm and Tasi algorithm respectively.Thirdly,comparing the characteristics of the multiple of filtering algorithm,image reprocessing is completed by Gauss filtering algorithm.Based on gradient clustering of edge points,the weighted least squares method is used to fit line,and then a sub-pixel detection method of the edge line is implemented rapidly.Line segment sequences which constitute quadrilateral are researched by depth-first algorithm,and corner point is extracted by calculating the intersection point of line segment.Position is realized by P4 P algorithm as the quadrilateral has been decoded and distinguished.And then the positioning accuracy of experimental analysis system called relative world coordinate system is designed.Finally,image processing algorithms is transplanted by some method,such as,C language,Emcv transplantation and Opencv rewriting and so on.The algorithm is optimized by component level,C level and assembler level.A positioning method with machine vision for mobile robot is implemented on platform on embedded DSP platform.