中文摘要：

针对复杂环境下,均值迁移算法只使用颜色特征跟踪目标鲁棒性差的问题,提出一种多特征自适应融合的MS目标跟踪算法。算法在跟踪场景的动态变化过程中,通过选择对目标和背景区分能力强的特征描述目标,建立多特征融合目标模型,并设置特征重要性权值。给出了多特征融合目标定位公式。通过动态评估不同特征在不同跟踪场景中的可靠性,对特征权值进行动态更新以及多特征自适应融合。依据不同特征的权值给出一种选择性模板更新机制,以减轻目标模型的漂移。实验结果表明,提出的算法在复杂场景下,具有更高的鲁棒性和跟踪效率。

英文摘要：

In order to solve the poor robustness problem due to using single feature in the target tracking process,a novel mean shift tracking algorithm based on adaptive multi-features fusion is presented in complex environment. In the process of the dynamic changes of the scene, several reliable features are selected according to their ability to distinguish target and background, by which the target model is represented,the multi-features model is established, and the feature importance weights are set. Mean shift target localization formula of multi-features fusion is proposed. By evaluating the reliability of the differ- ent features in different scenes dynamically, the feature weight is updated and the multiple features are fused adaptively. According to the weights of different features, a selective template updating mechanism is put forward to alleviate the model drifts. Experimental results indicate the robustness and effectiveness of the proposed method in complex scene.