Research On 3D Reconstruction Method Of Vehicle Scene Based On Binocular Vision

The vehicle's 3D environment reconstruction based on binocular vision is an important part of the vehicle's visual environment perception system.This method can provide an accurate three-dimensional local area map for intelligent vehicles by 3D reconstruction technology,so that the decision-making control system of intelligent vehicle can control the intelligent vehicle according to the current driving environment.This method is conducive to improving the level of traffic safety on the roads and improving the traffic jam of the roads.Therefore,in recent years,research on 3D reconstruction technology for intelligent vehicles has received widespread attention from domestic and foreign researchers,and has become a research hotspot in the field of vehicle's visual environment perception.Therefore,it is of great significance to carry out research on the method of vehicle 3D environment reconstruction based on binocular vision.In this paper,the 3D environment reconstruction method of vehicles based on binocular vision is studied,adopting a self-defined feature extraction and description algorithm and a feature circle-matching algorithm,designing a 3D vehicle environment reconstruction system based on binocular vision,realizing the perception of the surrounding environment of the vehicle and the location of the vehicle's position in the environment,verifying the real-time and accuracy of the reconstruction results through experiments,proposing a new vehicle environment 3D reconstruction and motion estimation method based on multi-cameras,which can reconstruct the surrounding environment of the vehicle.The research work of this article mainly includes:1.In this paper,the research background and significance of 3D reconstruction of vehicle scene based on binocular vision,the research status of intelligent vehicle visual perception at home and abroad,the research progress of binocular vision 3D reconstruction algorithm,and the development and research status of binocular SLAM algorithm is reviewed.2.In this paper,Zhang Zhengyou calibration method was used to calibrate the binocular camera through experiments,and the basic theory of binocular stereo vision applied in the experiment was analyzed and implemented.3.For the traditional binocular visual 3D reconstruction methods,which are cumbersome in procedure,large in computation,and limited in real-time,this paper adopt a self-defined feature extraction algorithm based on the actual needs of the vehicle environment perception in general road environment.Compared with other algorithms,the self-defined feature extraction algorithm has simple steps,less computational complexity,and more abundant image feature information.It not only ensures the real-time requirements of the vehicle environment perception in the general road environment,but also ensures sufficient image feature information.4.For the traditional feature point matching algorithm has large error and is time-consuming,this paper adopt a feature point circle-matching algorithm which combining before-after matching and stereo matching,and then the random sample consensus(RANSAC)algorithm is used to eliminate false matching pairs and optimizing other parameters,which ensures the accuracy of camera motion estimation and matching.5.Based on the self-defined feature point extraction algorithm and feature point circle-matching algorithm studied in this paper,this paper adopts KITTI's vehicle driving environment data set,designing a simulation experiment platform for the research on the 3D reconstruction method of vehicle environment based on binocular vision,verifying various parts' experiment and finally realizing a point cloud generated map of the 3D reconstruction of the vehicle environment based on binocular vision.And the final experimental data shows the reliability and real-time performance of the relevant algorithms,which lays a good foundation for the design and application of the intelligent vehicle's visual assistance driving system based on binocular vision.6.In addition,a vehicle environment 3D reconstruction and motion estimation method based on multi-cameras is proposed in this paper,which can reconstruct the surrounding environment of the intelligent vehicle,since binocular cameras can only monitor the scene in front of the vehicle,that is,only the environment in front of the vehicle can be reconstructed.