Moving Object Tracking Method Based on Improved Camshift Algorithm

The traditional Camshift tracking algorithm is easy to fall into local maximum when the target is occluded. It is easy to fail to track target when similar color interference or occlusion happens, and can't recover from failure. Aiming at this problem, the Kalman predictor and the inter-frame difference are adopted to improve Camshift algorithm. Firstly Kalman predictor predicts the next location of the target frame image, and regard this position as the center of a target tracking to determine search area Camshift algorithm, then use the size of the search box of the moving target to determine whether the target is similar color interference or occluded. If not, update the Kalman predictor parameters by position Camshift algorithm obtained, If it is similar color interference, extracted the moving object position and size by inter-frame difference to update Kalman predictor parameters, If obscured, the predicted value of the Kalman predictor is to be the current position of moving target, and the size of search window is a fixed value at the same time, using the set of values to update Kalman predictor parameters. Experimental results show that the improved algorithm can accurately track the target when similar color interference or occlusion happens.