| With the continuous development of automation technology,Automatic Guided Vehicle(AGV)as the main equipment for material handling in the industrial environment,are playing an increasingly important role in warehousing,logistics,industrial production and other fields.Due to the continuous development of computer vision technology and the improvement of related hardware performance in recent years,visual guidance has gradually become a research hotspot of AGV guidance.This paper takes AGV as the research object,and combines computer vision related technology,mainly studies AGV visual navigation and positioning,multi-branch path recognition and obstacle detection.In view of the requirements of precise navigation and real-time monitoring in AGV visual guidance process,the method of wired visual navigation assisted by two-dimensional label positioning is proposed in this paper.The AGV is driven by a differential velocity drive,and the camera is mounted obliquely to gain a larger view.The image of the navigation line is captured by the camera,and then the navigation center line is fitted by the image processing module to calculate the angle deviation and distance deviation.The AGV main controller controls the speed of two differential wheels through the deviation passed by the image processing module to correct the AGV's running state.In order to realize real-time monitoring of AGV by upper computer,two-dimensional tags are placed in the navigation belt.By identifying the position information in two-dimensional tags,calculating the actual position coordinates,and uploading them to upper computer,the real-time positioning requirements of AGV can be realized in the process of AGV visual navigation.Experimental results show that this method can achieve the path tracking error within 10 mm,and real-time monitoring of the approximate position of the AGV in the upper computer.Aiming at the real-time and robustness requirements of multi-branch path recognition in AGV visual guidance process,this paper proposes a multi-branch path recognition algorithm based on PCA-LDA and SVM.The recognition process is mainly divided into three parts: image preprocessing,path feature extraction and path recognition.Firstly,the image is denoised,morphologically processed and converted into binary image.Then the binary image is dimensionally reduced by PCA and extracted by LDA.Finally,the data processed by PCA-LDA are taken as the sample of SVM classification,and the parameters of SVM model are optimized based on Grey Wolf algorithm.Compared with other multi-branch path recognition algorithms,the algorithm in this paper not only ensures the recognition accuracy,but also has better real-time performance and robustness.In terms of AGV positioning and parking,this paper uses visual feedback method for real-time ranging,and obtains deceleration time and acceleration through distance calculation formula,so that AGV can accurately locate and park.In the aspect of obstacle detection,combining with the distance measuring principle of binocular vision,an obstacle detection algorithm based on parallax graph is be proposed by this paper,which mainly involves the calibration,correction and stereo matching of binocular camera.The obstacles in front of AGV can be extracted and their azimuth information can be determined by setting the disparity threshold and binarizing the disparity map formed by stereo matching.The experimental results show that the ranging error rate is about 1.5%,which meets the industrial requirements.In order to improve the ability of AGV obstacle avoidance,this paper uses frame difference method to determine the motion state of obstacles,and makes relevant obstacle avoidance strategies according to its motion state,which has good real-time performance.Finally,the experiment is carried out on the AGV experimental platform,based on the algorithm proposed in this paper and the upper computer is developed by Qt and OpenCV.Experimental results verify the effectiveness of the proposed visual guidance correlation algorithm. |
PDF Full Text Request