Research On Key Technologies Of Panoramic Vision Assisted Driving System

With the vigorous development of China's economy and technology and the continuous improvement of people's living standard,cars are becoming more and more popular.Even in the rural streets and lanes,a row of cars can be seen.Although the number of domestic traffic accidents has been declining year by year in recent years,the continuous increase in the number of vehicles has also brought a lot of threats to traffic safety.Especially when there are many vehicles,it is particularly prone to collisions and scrapings when driving or reversing,while cause unnecessary trouble.In order to reduce the unnecessary trouble of driving,many relevant researchers at home and abroad have carried out extensive and in-depth research on the auxiliary driving system.Therefore,it is of practical significance to further study the automobile auxiliary driving system.This thesis is mainly about the key technology of panoramic vision assisted driving system,providing a feasible design plan for panoramic vision assisted driving system.The system has four 180-degree fish-eye lens cameras covering the entire field of view of the vehicle around the vehicle as the input for capturing video.The four images collected at the same time are processed into a 360-degree top view of the surroundings of the vehicle.The driver of the system directly checks the 360 degree panoramic bird's eye view around the car through the central control panel,so that the blind area can be eliminated,and the driver can easily drive and park the vehicle.In view of the functions of the above-mentioned panoramic vision assisted driving system,this article thoroughly studied the key technologies needed to realize this function,including the analysis of the imaging principle of the fisheye camera and the cause of the barrel distortion of the fisheye image based on the imaging principle of the camera.For the fisheye lens,this article uses the Zhang Zhengyou calibration method to obtain the camera internal reference matrix,and in-depth analysis of a variety of fisheye image distortion correction model,pointing out the limitations and deficiencies of each model in the experiment.At the same time,the integral correction distortion correction model is proposed based on the existing algorithms,and the video image is effectively distorted and the real-time performance of the algorithm is taken into account.Based on the distortion correction map,this thesis also studies the generation of panoramas.Firstly,the method of generating the vertical top view and the top view of the virtual view is introduced,and a method based on hemispherical top view generation is proposed.Then the related technical theory of image registration is introduced.Considering the requirement of performance,we use artificial registration to pick up feature points to conduct bird's eye view transformation and stitching to create a top view of the surrounding environment that accords with the interface design.After registration,the coordinate mapping relation table from fisheye image to panoramic image is saved,and the real-time processing of four images to panorama is completed through query relation table.Then,on the basis of the panoramic view,the overlapped region is fused and the exposure compensation is made for the light inhomogeneity according to the actual situation.The experimental results show that the panoramic vision parking system researched in this paper can process 4 image rectification and stitching in real time,and the effect is ideal,which can meet the needs of the system.