Research On Fast Registration Method Of Vehicle Panoramic Image

With the rapid increase of the number of cars,due to the driver's visual blind area,the problems of safe driving become increasingly prominent,and the traffic accidents are also increasing.At present,the vehicle panoramic system is considered to be one of the effective means to mitigate the problem,has become the research hot spot of the intelligent transportation and image processing field in recent years,which attracted the attention of many scholars.Whether the vehicle panoramic system is workable,the image registration is a key link.Therefore,the rapid image registration algorithm has been carried on deeply in the paper,the main research content as follows:Firstly,in view of the traditional scale invariant feature transform(SIFT)algorithm in feature extraction and description problem of large amount of calculation and poor real-time performance,and puts forward a kind of SIFT based on region partition is a quick registration method.Firstly,the matching image and the image to be matched are divided into several uniform sub graphs.The feature type of the local sub graph is determined by calculating the information entropy value and the set threshold of each sub graph.Then,the selected sub graph is SIFT Extraction and feature vector dimension reduction processing to generate PCA-SIFT descriptor;Finally,two images are registered and purified.The registration time is reduced by 15%-25% when the registration accuracy is above 90%.Secondly,in view of the traditional Harris corner detection process,manually enter a single threshold may appear the phenomenon of corner clustering,pseudo corner and so on.In this paper,an improved Harris corner detection method for image registration is proposed.Firstly,the image is divided into 3?3 non overlapping sub graphs.According to the contrast of each sub graph,the threshold of each sub image is set.Secondly,the NCC algorithm is used to match the detected corners.Finally,the RANSAC algorithm is used to eliminate the false matching points in rough matching.The algorithm makes the detection of the angular distribution more uniform,and effectively increase the number of matching points in the image registration,can effectively improve the accuracy of image registration.Thirdly,in order to verify the application effect of the image registration algorithm in the vehicle panoramic system,using the directshow technology and two USB camera toset up the system experiment platform.We used the improved SIFT algorithm and Harris corner detection algorithm to extract feature points for the image frames of the acquisition of the system and compare with the effect of the image registration.Experimental results show that using the improved registration algorithm can make the image fusion effect better,and reduce the time of the image mosaic effectively.All in all,the study of image registration algorithm can improve the efficiency about20%,reduce effectively the time consumption,and lay a foundation for developing the practical vehicle image panoramic system in the paper.