Abstract: In order to track moving targets in real time and effectively, in this paper, a new Camshift target tracking algorithm is proposed, in which corner features and color features are fused adaptively. The invariance of the corner is combined in this algorithm, and the kernel probability density estimation of the Mean-Shift algorithm is used to calculate the probability density function for each feature. Bhattacharyya coefficient is adopted as similarity measure function, then using the ratio of similarity to achieve adaptive fusion of corner and color features. The new probability density distribution and the Camshift tracking algorithm will be combined to achieve the target tracking. Simulation results show that the algorithm is better than traditional Camshift algorithm, and tracks object more accurately.

Key words: Camshift, feature fusion, Harris conner, Bhattacharyya coefficient, adaptive

摘要： 为了能够实时有效地跟踪运动目标，提出了一种新的自适应融合角点特征和颜色特征的Camshift目标跟踪算法。该算法融合了角点的特征不变性，并采用Mean-Shift算法提供的非参数核密度估计的统计思想，计算各特征的概率密度函数，用Bhattacharyya系数作为相似性度量函数，利用相似性度量值之比自适应地融合角点特征和颜色特征，将得到的新的概率密度分布结合Camshift跟踪算法实现目标跟踪。测试结果表明，该算法比传统的Camshift算法跟踪效果更好，更准确。

关键词: Camshift, 特征融合, Harris角点, Bhattacharyya系数, 自适应