Intelligent Monitoring System For Detecting And Tracking Moving Targets

With the rapidly development of computer vision, pattern recognition, artificialintelligence technology, as an important part of intelligent video surveillance, movingtearget diction and tracking technology has value of theory and application.This paper is mainly to resolve the problem of how to improve the self-adaptive,real-time and robustness capability in moving target detection and tracking algorithm.The main content of this dissertation can be summarized as followings:Firstly，the ordinary techniques of moving target detection are briefly introduced.The morphological processing method as the basis for the simulation experiment isdescribed.A method of using the means of limited frames as the extract backgroundmodel.This paper solves the problem of detecting the moving target when there is nota ideal model in the video sequence..This paper solves the problem that the target’sdisplacement is too small to detect.This method improves the background model inreal time and reduces the problem of detecting the background as the target because ofthe slight changes in the background.Sum up the classification method of target tracking. Introduces RGB and HSVcolor space model and the conversion between each other, also introduces the colorquantization and histogram, using the HSV color space model as the observationmodel.Kalman Filter is applied to forecast status vectors in next frame according toobservation vectors in current frame. Setting the gain coefficient as a constant toimprove the system state equation and the observation equation during the calculation.Kalman Filter can reduce the search area and improve the matching efficiency. Basedon the depth study of Kalman filter tracking algorithm prediction, realizes a fast andeffective algorithm of the Kalman filter prediction.verify its feasibility through theformula and a series of experiments.This method reduces the running time, andimproves the algorithm real-time.The Mean shift algorithm based on color pattern matching is researched deeply.This method can solve the problem of detecting the moving target with partialocclusion in a simple environment, using this advantage combined with Kalman filterpredicting algorithm.It can reduce the search time in tracking moving objects.It alsocan solve the problem that Mean shift tracking algorithm can’t track the targetaccurately When the moving target moves too fast.An in-depth analysis of the basic theory of particle filter algorithm based onMonte Carlo method is given from Monte Carlo Sampling，Importance Sampling，Sequential Important Sampling，Particle Degeneracy and Particle Resampling.Process of particle filtering method is given. Using particle filter algorithm combinedwith Mean shift algorithm. It can solve the problem that Mean shift algorithm cannotaccurately locate the target accurately when there are objects around the target and its color features similar with the target. The experiments can prove that the combinationalgorithm is effective. Through the real-time monitoring of system variance,effectively improve the stability of the algorithm. Use the particle filter algorithmcombined with the Mean shift algorithm, to insure the robustness and real-timetracking.The research contents of this paper expands the application rangement of themoving target tracking method and improves the robustness and accuracy of targettracking method effectively.