Research On Intelligent Monitoring System Based On Panorama Mosaic Technology

With the acceleration of urbanization and population growth,the problem of security has been paid more and more attention,but traditional monitoring models have been unable to meet the safety requirements of large-scale and high-density crowds.This article starts from two a spects of expanding the monitoring perspective and intelligent monitoring and upgrades the traditional monitoring mode.Firstly,the improvement and implementation of panoramic stitching technology are completed through algorithm research,and the monitoring perspective is expanded.Then,intelligent analysis is added to the panoramic monitoring system to extract crowd density and crowd number information that are difficult for human eyes to directly obtain in the monitoring picture,which builds a complete panoramic intelligent monitoring system.First of all,the paper introduces the camera acquisition device,feature detection technology,RANSAC matching purification and geometric transformation of the image,so as to determine the camera layout of this experiment and provide a theoretical basis for subsequent research.In terms of multi-image stitching,this paper uses camera models and optical principles to estimate the coincident area,which effectively reduces the feature detection interval,speeds up the program running speed,and improves the accuracy of feature point matching.Because the traditional multi-image stitching algorithm uses the side image as the reference image,the panoramic image is seriously distorted and the effective area is lost.T his paper proposes an improved method of bidirectional registration,which reduces the number of perspective transformations of the image to improve the mosaic effect.In the stage of image fusion,by improving the gradual in-and-out algorithm,this paper reduces the weighted interval of pixels,effectively improves the running speed of the algorithm,and solves the problem of "ghosting" in the fusion region.In order to realize the panoramic monitoring system,this paper first adopts multi-process technology to solve the problem of simultaneous acquisition between the cameras,and then realizes the fast stitching of multiple images based on the key frame parameters.In order to further meet the requirements of real-time surveillance,CUDA programming is used to complete the GPU optimization of the stitching algorithm,which greatly improves the program running speed.The experimental results show that the method has good stability and strong real-time performance.Finally,the article studies the crowd analysis technology and combines it with panoramic stitching technology to build a panoramic intelligent monitoring system.After a comparative study,we decided to use the CSRNet opening source network model to complete the training of the crowd detection model in the server environment.Taking into account the limitations of the hardware environment,the experiment will be the spliced super-resolution image after downsampling processing,and then use the detection model to extract the crowd density and population number information.Experiments show that the system can achieve the effect of panoramic intelligent monitoring.