Image Sequence Moving Target Detection And Tracking

In recent years, developing intelligent monitoring which involves computer science,information technology, artificial intelligence and many other disciplines, has greateconomic value and development prospects in the political, economic, cultural, militaryand other fields. Moving target detection and tracking in image sequence has been a keytopic and a popular research direction in the field of intelligent monitoring.This thesis which is based on analyzing the characteristics of various backgroundmodeling algorithm, uses Gaussian mixture model (GMM) algorithm to construct thebackground image. Given the low accuracy and the poor real-time characteristics ofclassical GMM algorithm, this thesis proposes a new adaptive Gaussian mixturebackground modeling algorithm. Compared with the classical GMM algorithm, thealgorithm proposed in this thesis which not only are moving targets detected accuratelyin image sequences, but also accuracy and real time responsibility are improved byexperimental analysis.This thesis which is based on analyzing the characteristics of various moving targettracking algorithm, uses the Meanshift algorithm to accomplish moving target tracking.Given the Meanshift algorithm has the poor tracking results or even lose the movingtarget, which the background color is similar to the moving target or the moving targetis blocked, the thesis proposed a new method which based on Kalman-Meanshiftalgorithm. Compared with the classical Meanshift algorithm, the improved algorithmproposed in the thesis which not only are moving targets achieved tracking accurately,but also can effectively deal with the situation that moving target is blocked.