Research On Key Technologies Of Trackless AGV Based On Multi-ocular Global Visual Positioning

As an important equipment for automated transportation and intelligent manufacturing,Automatic Guided Vehicle(AGV)is widely used in industry,medical and military fields due to its advantages of low power consumption,low cost and high efficiency.AGV trackless guidance has the advantages of flexible route planning and strong environmental adaptability compared with tracked guidance,but AGV trackless guidance still has some problems,such as poor real-time positioning,low accuracy and poor operation robustness,which affect the promotion and application of trackless AGV in the market.This paper focuses on improving the positioning accuracy and path tracking reliability of the trackless AGV based on global vision guidance,and the main research contents are as follows:(1)Propose a self-adaptive perspective transformation method based on calibration plate,aiming at the problem that reference points need to be selected manually in the process of image perspective transformation to orthographic view.By extracting four corners of the calibration board,the image can be transformed from non-vertical top view perspective to orthographic view.This method improves the flexibility and adaptability of the image perspective transformation to orthographic view,and lays the foundation for the stitching of non-vertical top view image.(2)Propose an image stitching and constructing the global map method based on the calibration board,aiming at the problems that the images acquired by multiple cameras in non-vertical top view cannot be mosaicked by using the traditional image stitching algorithm and the lack of universality of constructing the global visual map.After transforming the perspective of the image into orthographic projection,the image mosaic and self-adaptive global visual map construction is realized by the information of the calibration board.The speed and effect of the proposed algorithm are better than those of traditional image Mosaic algorithms,with the 0.007% global visual map error.(3)Design a positioning method of AGV based on YOLOv5 s.On the basis of AGV recognition frame,two color blocks are used to extract AGV center point and determine the direction of AGV,and the scene model is established to eliminate the error caused by AGV height.Finally,multiple positions are fused to achieve high precision positioning of AGV.Experiments show that the positioning speed and accuracy of the global visual guidance AGV are improved,with less than 0.009% positioning error,and less than 2° Angle error.(4)Design relevant nodes and control methods of path tracking based on Robot Operating System(ROS)to realize automatic path tracking of AGV based on ROS platform.The reliability and feasibility of the proposed algorithm are verified by the path tracking experiment,with less than 0.016% path tracking error.