Target Tracking And Trajectory Clustering Based On Mean Shift

Moving target tracking is in a continuous video sequence, for each frames to find the picture of moving target. Trajectory clustering is to cluster trajectories through obtaining the goal of tracking target. Moving target tracking and moving object trajectory clustering are the basis on the scene of analysis and understanding the behavior of computer vision. It is also an important research field of computer vision. Although the analysis of target tracking and motion trajectory of many domestic and foreign scholars have done a lot of researchers try to solve these problems, but there are still many problems unsolved, such as occluded objects in complex scenes, light stability problems and change tracking trajectory clustering the time consuming problems and so on.This paper around the complex scene illumination changes or the presence of occlusion when the moving target tracking and vehicle trajectory clustering and analysis the problem carry on research, the concrete research contents and results are as follows:1, Discusses and analyses the principle and properties of mean shift target tracking algorithm of the classical mean shift algorithm and mean shift extended algorithm, and using the simple background of one target tracking based on mean shift algorithm to experimented and analyzed.2, In order to solve the problems of target tracking in RGB color feature large influence by light and mean shift algorithm in some interference or the existence of occlusion tracking effect is not accurate. This paper proposed to use HSV color space based on illumination do not changes color.Because of HSV(Hue) color feature component and an improved algorithm based on mean shift algorithm. This paper first uses the hue component color feature based on the illumination invariant feature,then the tracking window are divided into blocks, then the improved mean shift algorithm for each tracking, finally using the maximum to matching degree block weighted decided to target position.3, For the vehicle trajectory used video clustering and time consuming problem, put forward a kind of method with different features weights to solve these problems. In this paper, the average velocity, distance of vectors extracted,the center position of the moving object trajectories, and the trajectory of the initial acceleration, velocity of the four kinds of feature information, For the different weights, to cluster trajectories using Mean shift algorithm, and make a judge on the abnormal trajectory.