Target Tracking With Improved Camshift Based On Kalman Predictor

With the progress in computer vision research, the moving target tracking was used broadly, such as image processing, pattern recognition, medical diagnostics,security monitoring, intelligenttransportation and so on. In our lives, our living environment is more complex and easilychanging, so that the existing moving target tracking technology can easily be influenced outside. The outside influence may be the variousnoise, the same color of background disturbing, other objects covering seriously, changes of light intensity and other reasons. There are many problems and shortcomings in the practical application of the algorithm, so studying a target tracking method that can be used in a full range is of great prospects, and also is the researchers' challenging task.This paper describes the means of detecting moving targets, and compares the optical flow method, background subtraction method, a frame difference detection method. On this basis, the paper focuses on the use of an improved detection method which combines the optical flow method and three difference method.And then do morphological processingfor the image to remove noise that may be introduced during processing, as well as the phenomenon of empty, and make the detection of the image smoother. The experiments show that this method can improve the accuracy of detection, and effectively prevent the situation of target undetected, it can provide a source for targets tracking.In this paper, the study of moving target tracking, analyzes the meanshift algorithm and continuous adaptive mean shift algorithm, and made a detailed derivation of comparative advantages and disadvantages.When the above methodsare insufficient, the paper offers a new algorithm that was mixed several features such as color, texture,edgeand establishes mechanism for certain characteristics right, so that the template mixed several features can be adaptively updated. On the real-time tracking of moving targets, when the environment of the estimation changes, the algorithm based on contribution mechanismassigns different weights for the features adaptively.The next step, it uses the histogram to achieve the goal of real-time tracking. In this basis,because of the emergence of object occlusion, sudden acceleration and the same color interference, the target is easily lost, the paper proposed a method that can judgewhether the object was blocked and disturbed by the same color. It combined with the algorithm named Kalman to estimate where themoving object will appear in the next frame.Then using the improved algorithm Camshift to find the center of the target. Experiments show that the improved algorithm, make the tracking accuracy improved, the time shortened and the real-time met.