Research On Panoramic Video Stitching And Reconstruction Technology

With the continuous improvement of social informatization,video is an important way for modern people to obtain information,and its related technologies have become a relatively popular research field at present.Panoramic video stitching and reconstruction technology has also received more and more attention as one of the future development directions of video technology.With the development of science and technology,people are no longer satisfied with watching videos on flat-panel displays to obtain scene information.Panoramic video stitching and reconstruction technology came into being in this context.Panoramic videos are generally composed of videos shot by multiple cameras facing at different angles,and their field of view is very wide.In order to take advantage of this feature,the reconstruction of panoramic video is often done in combination with virtual reality technology.Users watch panoramic videos by wearing a head-mounted device with a built-in screen.Since the screen will cover the entire field of view and the display content will be updated synchronously with changes in the head posture,the user will have an immersive viewing experience.This technology has broad application prospects in the fields of security monitoring,remote surgery,driving simulation and robot teleoperation.This paper introduces the research status of panoramic video stitching and reconstruction technology at home and abroad,builds a software and hardware system,and realizes the stitching and reconstruction of panoramic video.The main work is as follows:(1)Firstly,the static images are detected by SIFT,SURF and ORB algorithms,and the detected feature points are matched by the FLANN algorithm.Then,the RANSAC algorithm is used to filter the matching pairs of feature points and calculate the homography matrix to realize the image registration.According to the registration results,the SURF algorithm is selected as the feature point detection algorithm for image stitching.(2)Use cylindrical projection to map two images to the inside of the same cylindrical surface.Through experimental comparison and analysis,the graph cut method is selected as the best seam search algorithm for image stitching.On the basis of the graph cut method,the gradual fusion method is selected as the image fusion algorithm for image stitching through experimental comparison and analysis.(3)The stitching of panoramic still images is completed.In the process of panoramic video splicing,the first frame of the video to be spliced is extracted for preprocessing,and the parameter data obtained in the preprocessing process is applied to the real-time splicing of subsequent video frames,which greatly improves the splicing speed and realizes the splicing of panoramic videos.(4)An experimental system was built,using FFmpeg and streaming media server to realize the encoding and streaming of panoramic video.A virtual reality environment is created in Unity,in which panoramic video playback and perspective follow-up functions are implemented.Finally,the panoramic video is watched through the 3Glasses D4 head-mounted device,the reconstruction of the panoramic video is completed,and the experimental verification is carried out.The experimental results show that the built software and hardware system can realize the splicing and reconstruction of panoramic video in the virtual reality environment,and provide the viewer with an immersive viewing experience.