Design And Implementation Of Real-time Panoramic Fusion Model Based On Image Fusion

Real-time panoramic fusion is a panoramic camera module composed of multiple cameras to capture video streams,using the relevant technology theory of image fusion,to achieve dynamic and efficient panoramic video effects.Unlike image fusion or live panoramic video broadcasting,real-time panoramic fusion emphasizes real-time,efficient,dynamic and mobile panoramic VR video.Video stream is essentially composed of image frames.Panoramic fusion can not be separated from image fusion related processing technology.However,the real-time panoramic fusion using the traditional image fusion model will have the problems of complicated process and low efficiency.Panoramic video live broadcasting is a hotspot in panoramic fusion.It has a lot of research and application in live broadcasting and videoconferencing.It solves the problem of camera parameters change in image fusion by customizing dedicated hardware and stereo shooting,avoids repetitive feature matching process and camera parameters calibration process,and has a wide range of applications in its related fields.But panoramic live video broadcasting needs customized special equipment to reduce the processing process of software through the harsh standards of hardware.Usually,the cost of hardware is very high,and it can not provide popular services.At the same time,its camera is fixed in the designated position in the scene,so it does not have dynamic mobility.With the rise of panoramic VR video technology,mobile real-time panoramic fusion technology with higher universality and lower hardware cost by software solutions has important research significance.Based on the theory of image fusion,this paper achieves a series of research results in real-time,efficient,dynamic and mobile panoramic VR.1.Analyzing the existing image fusion model and panoramic live video broadcasting model,combining with the related technology theory of object tracking in single camera video,aiming at the defects and shortcomings of the existing fusion model in real-time panoramic fusion,a new fusion model is proposed to meet theneeds of real-time panoramic fusion,and the efficiency of the fusion process is theoretically analyzed and validated by relevant data.2.In the non-ideal case of sudden change of camera parameters leading to invalidation of transformation matrix,the matching and screening of image and the calibration process of camera parameters in existing image processing are analyzed.Aiming at the particularity and requirement of real-time panoramic fusion,the original algorithm is improved and a new algorithm is proposed.3.Considering the concurrency of threads and the advantages of graphics card in matrix computing,the real-time panoramic fusion system is preliminarily designed and implemented,and the system is tested.