Research Of Video Surveillance System Based On Dynamic Background

With the development of video surveillance technology,video surveillance has gradually entered the direction of intelligent video surveillance from the traditional ‘people watch'.Intelligent video surveillance technology is to change the traditional video into automatic analysis video content,to judge the abnormal behavior of video,automatic alarm.Intelligent video technology has become an important research direction in video surveillance area.In intelligent video surveillance research,moving object detection algorithm is the based algorithm for moving object detection and recognition of object behavior.The Prerequisite for moving object detection algorithm is that background is fixed,which leading to the practical application in monitoring scene fixed defects.Doom's function is restrict.To solve the problem,thesis propose a series of algorithms and optimization method for the practical application,to meet practical application in full use of ball machine characteristics,and expand the range monitoring scenarios.The main research contents and innovations of this paper include:(1)The proposed system architecture based moving target detection motion scenarios,describes the hardware and software system structure,design and implement a ball machine configuration and motion detection software for configuring the software too much human involvement,the success rate is not high configuration the problem,automatic image registration algorithm configured as an auxiliary means.(2)For image registration and stitching algorithm,in-depth study of image registration algorithm of basic principles and methods to compare a variety of image registration,for panoramic scenes sports under the registration requirements of practical application,based on SURF algorithm as select core theory of image registration.Study the image stitching algorithm,in-depth analysis of the affine transformation in image mosaic application,and the application of the corresponding algorithm in the system to achieve the mosaic image.(3)For road monitoring moving objects more,there are a lot of false corners scene image corner extraction,resulting in poor image registration effect problem,a corner of the selection algorithm.Use moving object in the scene,stay a shorter time characteristics,the introduction of the attenuation factor,select the scene as a stable corner corner image registration,in order to achieve greater effect registration purposes.(4)proposed in the background generation algorithm based on the dynamic scene moving target detection,based on research on the moving object detection algorithm for which the typical background generation algorithm in depth analysis of the advantages and disadvantages of its algorithm,corresponding to its shortcomings algorithm.(5)The moving object detection algorithm for defect analysis of the time-consuming and space consumption,proposed the use of the ball machine characteristics,using a panoramic approach to reduce the occupation of space;analyzes the dome movement,image registration time-consuming problem of excessive proposed algorithm projection matrix calculation using the configuration process to achieve real-time scene to get quick registration.Finally,the corresponding implementation of the proposed algorithm and the performance analysis are discussed in this paper.