Multi-view Video Stitching In The Surround View Parking Assist System

Multi-view video stitching techniques has very great practical importance and significant engineering application perspective which mainly used in the surround view parking assist system. The surround view parking assist system can provides real-time surrounding circumstance by which the problem of narrow view which is the main reason of accidents could be solved.On the basis of the research of the key technologies of the surround view parking assist system, one program to implement this system is designed in this thesis.Firstly, calibration images used for calibrating camera parameters are collected by four cameras which installed on the four direction of vehicle; then the corrected calibration images are transformed into bird’s-eye view images; Finally, the four bird’s-eye view images are registered, and the homography matrixes are determined by which the surround view is formed after image blending and stitching.(1)To meet the real-time requirement for video stitching in the surround view parking assist system, this paper presents a multi-view video stitching method based on the combination of the specific image stitching and the look-up table method.(2)A comparison of the advantages and disadvantages of various camera calibration algorithm indicated that the Zhang Zhengyou calibration method met the demand.(3)According to the analysis of three kinds of feature extraction and description methods which are SIFT, SURF and ORB,the ORB features was used for image registration.(4)Aiming at the problem of image fusion, the algorithm of image fusion was improved, and with improved algorithm, the fusion result is better.Experimental results show that the proposed method can be effectively used for multi-view video stitching with real-time requirement.