Design And Implementation Of Automated Guided Vehicle Vision Navigation System Based On Deep Learning

Visual guidance is an important research direction in AGV automatic navigation system.Among several navigation modes commonly used in AGV vehicles,electromagnetic guidance is the most widely used guidance mode at present.The advantages of this guidance method are simple principle,simple communication and control in the system.The disadvantage of this guidance method is the extensibility of this guidance method is poor.In order to improve the adaptability of AGV vehicle vision navigation system to the environment,as well as the accuracy and efficiency of navigation system,a vision navigation system based on deep learning method is designed.After researching and analyzing the requirement of this subject,the following work has been carried out.Firstly,the AGV vehicle's operating environment and the software and hardware equipment platform have been studied and tested.Secondly,all the images are corrected to solve the distortion of the image data,and provide the correct image data for the follow-up work,thus ensuring the accuracy of navigation.The real-time detection of specific markers in the image is realized.According to the results of target detection,the task category of AGV vehicle is judged.Then,the edges of specific targets are segmented.The detection of feature points is completed.Finally,the AGV driving path is planned.This paper introduces the design and implementation process of the overall framework of visual navigation system,and tests the function and performance of the navigation system in the practical environment.The errors of angle adjustment are within3 degrees,the errors of forward direction translation adjustment are within 2 cm,and the errors of forward distance are within 3 cm.