Research And Implementation Of Vehicle-mounted Around View Monitor System Based On Embedded

The sales volume of new energy electric vehicles in China has reached 3.5 million in2021,a year-on-year increase of 1.6 times.Compared with conventional fuel vehicles,new energy electric vehicles not only use more environmentally friendly electrical energy in the power source,but also are equipped with advanced intelligent driving and assisted driving functions.Some assisted driving functions are realized by installing sensors such as millimeter-wave radar,lidar,and fisheye lens on the car.The car panoramic view system is one of the very important and useful functions.This function can provide the driver with a vision around the body,effectively reduce the blind spot of the driver’s field of vision and reduce the probability of accidents.Realize the vehicle panoramic surround view system,it is first necessary to obtain the images around the car body through the fisheye lens,and then input the image information into the brain-domain controller of the smart car for image processing,image correction,stitching and perspective transformation,etc.Finally,the panoramic surround view image is output to the central control screen of the car.In this thesis,the related image processing technologies and common algorithms required to realize panoramic surround view are firstly studied.The contents are as follows:（1）Aiming at the problem of fisheye image distortion,the imaging model and distortion model of fisheye camera are studied,and the camera calibration method and fisheye image distortion correction method are studied and analyzed.（2）Aiming at the perspective of panorama,the transformation model and common perspective transformation methods in image processing technology are studied.Perspective transformation of images using DLT direct linear transformation method（3）Research on image stitching technology for generating panoramic view,the feature extraction method of stitched images,the principles of SIFT and SURF algorithms are studied,and several methods of feature point matching and methods for eliminating gaps generated by image stitching are introduced.Common algorithms for image fusion.The feature matching method used in the traditional SIFT/SURF image registration algorithm is to match the feature points according to the Euclidean distance,and the matching effect is poor.In this thesis,an improved image registration algorithm based on SURF is proposed.The RANSAC algorithm is added to SURF to purify the paired points,which can filter out some feature points with inaccurate pairing,and the stitching effect is better.Finally,the implementation scheme of the vehicle panoramic surround view s ystem is given.Based on the Allwinner T507 hardware platform,an embedded Li nux operating system is built,and the Open CV visual function library is configured to the platform.According to the relevant image processing methods studied ab ove,the image processing algorithm flow for generating panoramic surround view images is designed.The experimental results show that this method can basically realize panoramic surround view.