Research On Lane Image Stitching Technologies Based On ORB Feature Match

Lane image stitching technology is a current research hotspot in the field of intelligent transportation.At this stage,the lane image captured by the lane camera has high resolution but a narrow field of view.The transportation department needs high-resolution,wide-view lane images.This thesis optimizes the existing stitching technology.This dissertation mainly studies the lane image stitching technology based on ORB feature match.Firstly,an improved GMS algorithm is proposed to reduce the impact of mismatch on image registration,so that false matching points can be effectively eliminated.Consequently,the weighted fusion algorithm combined with the optimal suture is used to solve the edge cracks of the stitched image,so as to obtain high-quality Lane stitched image.The main research work of this dissertation includes:1.This thesis uses the ORB feature extraction method to extract the feature points of the lane image,in the feature point matching stage,an improved GMS algorithm is applied to refining the feature points gotten by registration and then those mismatched pairs are removed.Experiments show that the algorithm effectively improves the accuracy of image registration.2.In order to solve the problem of cracks in the initial stitching image boundary,this thesis uses the weighted fusion algorithm combined with the optimal suture to fuse the stitched images.The experimental results show that the final mosaic image of this algorithm effectively solves the boundary crack problem,and the lane stitched image has a good visual appearance.3.A prototype system of lane image stitching is dissertation designed and implemented.In the image registration stage,the improved GMS algorithm is used to purify the coarse matching,the weighted fusion algorithm combined with the optimal suture line is used to fuse the stitched images,and high-quality Lane stitching images are obtained.