Research On Multi-camera Video Stitching Algorithm For Monitoring Scene

With the continuous improvement of science technology and economic level,the rapid development of network communication,the requirement of obtaining visual information is more and more perfect.The information obtained by a single visual sensor device can not meet the demand.People hope to obtain panoramic video image information with high definition and high resolution.Image mosaic technology can expand the field of vision without reducing the resolution,but in image mosaic,high precision and high accuracy mosaic algorithm often has the problem of poor real-time performance,which can not meet the requirements of a variety of scenes.In order to solve this problem,it is necessary to reduce the operation time while ensuring the accuracy and robustness of the algorithm.This paper focuses on the accuracy,robustness and running time of stitching algorithm.The main work are as follows:1?Firstly,the research background and significance of video image mosaic algorithm are introduced.For image registration,the ORB(Oriented FAST and Rotated BRIEF)algorithm is improved because of its poor robustness and low matching accuracy.The specific method is to build scale pyramid layer,detect feature points on each layer of image,and add scale feature vectors to feature points.In image fusion,because there are many moving objects in the monitoring scene,an image fusion algorithm combining the best suture and Poisson fusion is proposed to reduce the image stitching trace.Using the best stitching line to segment the overlapping regions of mosaic images can avoid moving objects,and then using Poisson fusion to make the transition region smoother.2?According to the factors that affect the quality of stitching image,a method of mosaic image quality evaluation based on differential information entropy is proposed.Compared with information entropy and structural similarity index,this method can better reflect the stitching seam and double shadow phenomenon in stitching images,and the evaluation results are closer to subjective quality evaluation results3?Designing image mosaic simulation experiments to verify the effectiveness of the proposed algorithm.The experimental results show that:(1)The matching accuracy of the improved ORB algorithm has been improved,and the running time has been significantly reduced compared with SIFT algorithm and SURF algorithm;(2)The results of image fusion test show that the improved fusion algorithm is more effective and effective than other existing methods in stitching seams and double shadow phenomena,which can make stitching images transition naturally.