Research On Target Detection And Ranging Technology Based On Binocular Vision

With the development of my country’s economy and technological progress,the number of families with cars has also increased.Although it provides more convenience for life,it is also accompanied by road congestion and traffic accidents.Car driving safety has become the current situation.One of the most important issues facing the development of automobile technology.The environment perception technology based on binocular vision is a research hotspot in the field of car safe driving.This article conducts in-depth research on target detection and ranging based on binocular vision,and proposes a deep learning network based on GA-Net and YOLOv4 Target detection and ranging technology.First,the four coordinate systems used in the geometric imaging of the camera and the conversion relationship between them are introduced.The principle of binocular stereo vision ranging is analyzed,and then the commonly used matching algorithms are summarized.Secondly,the principle,network structure and implementation process of GA-Net stereo matching are analyzed.Two data sets of KITTI-stereo2015 and Cityscapes are used for training and testing.The obtained disparity map is compared with the traditional SGM algorithm.The result is It shows that the disparity map processed by the GA-Net stereo matching algorithm is smoother,the structure is clearer,and the root mean square error is also significantly lower than the SGM algorithm.Thirdly,it studied and analyzed the algorithm principle,network structure,loss function of YOLOv4 model,using MS COCO data set training model,KITTI-Object training set to fine-tune model parameters,KITTI-Object test set and Cityscapes test set test model,using Average detection accuracy(AP),average average accuracy(m AP)and image frame rate per second(FPS)are three indicators to evaluate the recognition performance of the YOLOv4 model,and compare the results with the SSD target detection model.Finally,collect indoor and outdoor binocular images and videos.Use Zhang Zhengyou calibration method to calibrate the selected ZED binocular camera to obtain internal and external parameter information,and then use single-target and multi-target detection and ranging schemes to test the target detection and ranging algorithm proposed in this paper.Single target detection and The test results of ranging show that within the range of 15 m,the target ranging error is within 5.2%;in the multi-target detection and ranging test,the algorithm in this paper correctly recognizes the information of people,cars,chairs,etc.in the image,and calculates Out of their distance.The results of various experiments show that the target detection and ranging technology based on binocular stereo vision proposed in this paper can provide driving assistance technology for auto-driving.