中图法分类号tp39中国图象图形学报

中图法分类号：TP391.4 文献标识码：A 文章编号：

论文引用格式:

利用背景加权和选择性子模型更新的视觉跟踪算法

黄安奇，侯志强，余旺盛，刘翔

空军工程大学学院 西安 710077Mean Shift算法对受背景干扰的目标无法进行有效跟踪并缺少有效的模型更新策略的问题，提出一种将背景加权和选择性子模型更新相结合的跟踪算法。方法：首先，在Mean Shift框架下，为了减少背景信息对目标定位的干扰，利用目标区域周围像素的颜色直方图定义背景加权系数，并将该系数只引入到目标模型的颜色直方图中，从而建立一个新的目标模型。然后，根据目标模型中每个分量匹配贡献度的大小选取需要更新的模型分量及其更新公式。结果：实验结果表明，文中算法能够抑制背景干扰，同时能对模型进行有效的选择性更新，克服了整体更新策略严重的模型漂移问题。结论：本文从模型描述和更新策略两个方面对传统Mean Shift算法进行了改进，实验结果表明本文算法具有较好的有效性和鲁棒性。

关键词：目标跟踪；均值漂移；背景加权；选择性子模型更新

A visual object tracking method based on weighted background and selective sub-model update strategy

Huang Anqi, Hou Zhiqiang, YU Wangsheng, LIU Xiang

The Information and Navigation Institute of Air Force Engineering University, Xi’an 710077, China

Abstract: Objective: The traditional Mean Shift tracking algorithm often loses target when the object is similar to the background and it lacks effective model update strategy. Aimed at those problems, a new visual object tracking method is proposed in this paper. Method: Firstly, in order to reduce the location error, coefficients based on the color histograms of the background pixels around the object are computed and introduced into the target model. Secondly, sub-models are selected and updated according to the match contributing degree in the current frame. Result: The experiment results show that the proposed method can well restrain background distraction, meanwhile it can effectively update the target model and eliminate drifting phenomenon. Conclusion: This paper proposes a visual object tracking method based on weighted background and selective sub-model update strategy, which improves the traditional Mean Shift tracking algorithm in object modeling and model update strategy. The experiment results show that the proposed method is effective and robust.

Keywords: object tracking; Mean Shift; weighted background; selective sub-model update

0 引 言

视觉跟踪(Visual Tracking)问题是计算机视觉领域中的一个热点问题。在各种各样的跟踪算法中，Mean Shift算法由于其计算复杂度低、调节参数少、稳健性较好和易于工程实现等优点已经广泛应用到各种实时