Intelligent Video Surveillance System Based On Moving Object Detection And Tracking

Along with the progress and development of electronic technology, communication technology and streaming media technology, the scope of application in net-intelligent video surveillance system is more and more widely, and video monitoring system has also experienced a development process from analog to digital, networked, intelligent. Detection and tracking technology on moving object is an important part in computer vision and also the foundation to realize the intelligent monitoring system, then studying on it becomes popular in recent years because of its wide application prospects. Through the tireless efforts of relevant scholars for decades, detection and tracking technology on moving object has achieved great progress. But the detection and tracking on video object is still facing many difficulties in practical application, a real stability, practical monitoring system still need to design more robust core algorithms. This paper focuses on the method of detection and tracking of object and improves correlation algorithms according to the common problems in visual tracking. The main work of this paper includes as follow:First, this paper uses PBAS background difference algorithm for the reason that moving target is easy to be affected by dynamic background, illumination changes and other issues on object detection. This algorithm can overcome the light, sloshing trees, water ripples and other external perturbation very well through closed-loop feedback control and adaptive parameter and adapt to the dynamic changing background effectively. At the same time, a ghost removed algorithm based on Pixel-Based Adaptive Segmenter is proposed to obtain a satisfied effect in moving target detection under complex background because origin algorithm is easy to produce ghost when initialization modeling,.Secondly, the particle filtering method has some shortcomings although it performs excellent in tracking such as large amount of calculation and the particle degeneracy among time and the problem with the conventional resampling upon the face of the particle diversity loss. Aiming at the above problems, this paper proposes an improved particle filter algorithm on clustering and particle swarm optimization. It improves the distribution of the samples, speeds up the convergence of the particle set and makes the performance of particle filter greatly through the optimization of the particle set,.Finally, according to the moving object is easy to be affected by target deformation and complex background on tracking process, single visual information can not describe the object fully and target tracking is not stable in the motion under changing environment, this paper adopts a method of multi features fusion in target tracking, in this method, the color, texture, motion edge feature of the object is described by histogram model. The information is fused together through three kinds of feature weights adaptive fusion strategy, it makes the tracking algorithm can adjust the weight of three kinds of features adaptively according to the tracking situation and realizes the complementary information among information. In this way, it solves the problem of tracking failure when using single information in dynamic changing scene and ensures the tracking accuracy.