Research On AGV Environment Perception System Based On Binocular Vision

With the steady development of my country’s manufacturing industry,China has become the world’s largest manufacturing country.At this stage,the manufacturing industry is in a critical period of digital transformation and intelligent upgrading.The upgrading of the manufacturing industry chain is bound to be inseparable from the support of intelligent equipment.AGV automatic guided car is a kind of intelligent equipment,which plays an important role in the industrial production process.This paper conducts an in-depth study on the environment perception module,one of the key technologies in the AGV automatic navigation system.By analyzing the current research status of AGV navigation technology,an AGV environment perception system based on binocular vision is proposed,which is closely combined with the deep learning target detection algorithm and The binocular stereo vision ranging algorithm realizes the fusion perception of the surrounding environment.The experimental results show that the binocular visual environment perception system built in this paper has good real-time performance and accuracy,and has a certain practical value in the scene.The main research contents of this paper are as follows:(1)By comparing the characteristics of different target detection algorithms,according to the requirements of the working scene,the YOLOv4 network architecture is used to build the target detection algorithm network model.Collect the image production data set in the working scene,train and model the network model,and realize the recognition of the target object and the acquisition of the coordinates of the target area.(2)Research the mathematical principle of camera imaging and binocular vision ranging,analyze the projection process of 3D target point to 2D imaging point,and derive the projection formula.Select the visual sensor reasonably,make a calibration data set,and use the calibration tool to calibrate the binocular camera.Use Open CV development tools to perform stereo correction on the left and right cameras to eliminate imaging errors caused by camera materials or manufacturing processes.(3)In the process of binocular matching,the image is preprocessed to improve the quality of the image.Analyzing the advantages and disadvantages of traditional stereo matching algorithms,based on the traditional AD-Cencus algorithm,a stereo matching algorithm is proposed to construct a matching window of adaptive size.In different regions of an image,different matching windows are used to calculate the cost,so as to ensure the matching.On the basis of the accuracy of the algorithm,the real-time performance of the algorithm operation is improved.(4)Based on the common SPI hardware connection scheme,a communication protocol for edge AI computing platform and microcontroller is designed.On the basis of 3 communication buses and 1 chip selection line,3 GPIO ports are reserved for data sending and receiving notification between nodes to ensure the integrity of data sending and receiving.The layered architecture design of the communication protocol software layer is carried out to improve the stability and reliability of the data transmission process.Classify and design different messages,and define the meaning of communication protocol commands to exchange information.(5)Reasonably select the edge AI computing platform,configure the development environment of the computing platform,integrate and transform the binocular visual environment perception system model,and deploy it on the computing platform.Collect a large amount of data to test and analyze the binocular visual environment perception system to verify the real-time performance and accuracy of the system.