Research On 3D Environment Reconstruction Technology And Its Implementation On Vehicle Based On Monocular Camera

Vehicle 3D environment reconstruction technology is an important component in the vehicle's visual environment perception system based on monocular vision,which construct the environmental information in front of the intelligent vehicle through 3D reconstruction technology and provides clear visual navigation and outline to driver safety.In recent years,related technology research has been increasingly popular at home and abroad,which has become a research hotspot in the field of vehicle visual environment perception.Therefore,it is of great technical value and application prospects to carry out research on vehicle 3D environment reconstruction based on monocular vision.In this paper,vehicle 3D environment reconstruction technology is studied for sequence images.A new feature extraction algorithm and sequence image reconstruction algorithm are proposed for vehicle motion estimation,as well as implementing environmental perception in front of the vehicle and measuring the position of the vehicle in the environment.The experiments verify the accuracy and real-time performance of vehicle environment perception systems.The research work of this paper mainly includes:1.The research background and development status of the vision 3D reconstruction technology are reviewed.The problems of the vehicle 3D environment reconstruction are analyzed based on monocular vision,binocular vision and multi-view vision,and the research objectives of the subject is proposed.2.A distortion problem with images captured by the industrial camera.Therefore,the algorithm of camera calibration is studied.The traditional Zhang Zhengyou calibration method is used to obtain the relevant parameters of the industrial camera.The distortion of the image is corrected based on these parameters.3.A new feature extraction algorithm is proposed for solve the problems of algorithm step cumbersome,calculated amount large,real-time performance limited in the traditional motion sequence image reconstruction algorithm.In order to further speed up feature extraction,the feature extraction method is used based on overlapping regions.After the feature points extracted from the overlapping regions are roughly matched according to the method based on the nearest distance than the last close distance,and random sampling consistency algorithm is used to purify the matching pairs.Because of the random sampling consistency algorithm needs to be further improved,a bootstrap matching method is proposed based on a random sampling consensus algorithm.This method improves the accuracy of the matching.Finally,in the paper,the least squares method is used to optimize the parameters and obtain the transformation model for reconstruction.According to the model,the 3D environmental information in front of the vehicle is obtained in the scene.4.In this paper,the sequence image reconstruction algorithm and industrial camera parameter calibration algorithm are studied.According to the performance and typical scenes of industrial cameras,a 3D reconstruction experiment platform is designed based on monocular vision.Through testing of the vehicle on-site,the environment reconstruction in front of the vehicle is well achieved in general road condition.The experiments verify the accuracy and real-time performance of vehicle environment perception systems.Therefore,the 3D environment reconstruction in front of the vehicle lays a good foundation for design and application of visual auxiliary driving system based on the monocular vision.