Research On Multi-camera Video Image Real-time Stitching Algorithm Based On Specific Shot Angle

Our country has a vast territory with very long borders,so there are many land neighbors.At the same time,our country’s South China Sea is vast,spanning about 2,000 kilometers in the north-south direction and about 1,000 kilometers in the east-west direction.Our country has sovereignty disputes with many countries on land borders and in the South China Sea.Affected by geographical factors,these places are difficult to supervise comprehensively,and require a large amount of manpower and material resources.With the development of information technology today,it is possible to monitor remote areas in real time through digital equipment.In terms of social security,surveillance cameras have been widely installed in many public places,and when illegal incidents occur,criminal clues can be provided to the public security organs.It is difficult to comprehensively observe the characteristics of the object under a single camera,while under multiple cameras,under different angles,each camera can capture various feature information of the monitored object so that we can observe the object in all directions.However,due to the multi-camera,the amount of data that needs to be processed doubles.In order to achieve real-time performance,there are higher requirements for video splicing algorithms and computer performance.In this paper,the advantages and constraints of real-time video stitching of multi-camera are analyzed.Under the premise of cost control,the high quality and efficiency of video stitching are guaranteed and the performance requirements of hardware equipment are minimized.Aiming at the difficulty of video stitching in engineering application,this paper proposes a video stitching implementation scheme.This paper focuses on the following aspects of the research:1.In this paper,image stitching is realized based on SIFT feature extraction algorithm,and the high robustness of SIFT is used to improve the stability of the system in most application scenarios.SIFT algorithm needs to remove the edge points in the image to ensure the stability of the feature points.In this paper,Roberts edge detection operator is used to eliminate the unstable edge points in the image.The SIFT algorithm may produce a large number of feature points and many of them are useless feature points.The feature point matching algorithm proposed in this paper is combined with the feature points’ regularities of distribution to eliminate unnecessary feature points and a homography matrix solution method in the compromise between performance and accuracy is proposed.2.This paper analyzes typical image fusion algorithms and verifies their effects.Aiming at the problems of unnatural transition and ghosting after image fusion,an image fusion algorithm that can dilute ghosting is proposed based on the weighted average fusion algorithm.3.Due to the difference in image processing efficiency between hardware with different performance,there is a mismatch of frame rate between the output video of the monitoring camera and the video after video stitching.This paper analyzes and compares the typical key frame extraction algorithms and proposes a key frame extraction algorithm which combines equal interval key frame extraction algorithm with image feature key frame extraction algorithm.Under the premise of ensuring the maximum video frame rate,the frame rate of actual video displayed on the screen is improved.4.This paper designs and implements video stitching and tests the algorithm of video stitching in different application scenarios.The existing problems are analyzed at last.