A cloud model-based camshift target tracking

A cloud model-based Camshift algorithm is proposed to track the moving target with an amount of background information. It overcomes the unstable tracking problem caused by the reasonable rigid division of histogram in the traditional Camshift algorithm. The cloud model of the target block is established to extract the target and background concepts and to construct the certainty graph. The interference of the background information to the target tracking is weakened by averaging the weight of the target conceptual area with kernel function. The experimental results which are done by Matlab validate that, compared to the traditional Camshift and saliency Camshift algorithm, the accuracy of the proposed algorithm is increased by 51 and 9% for the target tracking without occlusion, and also increased by 54 and 15% with occlusion, respectively.