The Research Of Moving Object Detectionalgorithm In Intelligent Visual Surveillance System

With the development of intelligent visual surveillance, the research of moving object detection technique has entered a new dimension. Moving object detection is defined as the video image sequence to determine whether there are prospects of object motion, if there is, we need detection initial position to the target,which is the basis for identification and tracking to achieve its objectives. This paper completed the study and implementation of intelligent visual surveillance system moving object detection algorithm,based on the analysis and research of the moving object detection algorithm.The main thesis is as follows:(1) to consider the convenience and ease of use of intelligent visual surveillance, proposed that using the USB video capture card capture video,and get a clear video image. By OpenCV library for video stream processing, including pre-processing of video images and controlling the video stream when reading and writing video files.(2)In this paper,the extension of the background difference method to detect the moving object,and proposed to divide moving object into four stages,the initial background structure, scatter removal, connected components detection and background updating, respectively, studied the algorithm at different stages and propose the improvement of the background to build algorithms, scatter removal algorithm connected domain detection algorithm and background update algorithm.(3)For the demand of moving object detection in the intelligent visual surveillance,propose the improved background difference method.The algorithm includes video capture, activities, moving object detection, moving object identification and early warning activities in the whole process ,which can be transferred to any intelligent visual surveillance system, and has the versatility and the higher detection accuracy.(4)This paper focus on solving the three types of false alarm problem ,in the event of the moving object detection caused by the light slowly changes, sudden changes in light and scatter.The use of background updating algorithm to solve the light change,combined with the use of scatter removal algorithm and the connected domain detection algorithm to remove the bulk point, the algorithm is easy to understand and realize, effectiveness high.(5)code implementation and experimental results show are the important work of this article. This paper achieved a five algorithms proposed and the function-related activities of moving object detection,and verified the algorithm through collection of video. The moving object detection as the basis of an intelligent vision object classification, object tracking and behavior understanding, is widely used in traffic management, military scenes monitoring, sensitive areas such as public places, with a broad market prospect.