摘要： 扩展传统的MeanShift跟踪算法，使其能够实现特征和量阶自动选择。引入比率对数图及互信息方差实现特征的自动选取，同时提出一种新的量化方法，能够更显著地区分目标和背景。实验结果表明，该算法在多场景下具有较好的鲁棒性并能提高跟踪精度，可以适应光线变化、背景干扰、被部分遮挡或色彩质量较差的情况。

关键词: 目标跟踪, MeanShift算法, 量化, 特征选择

Abstract: This paper extends the typical MeanShift tracking method and makes the feature selection and quantification automatic. It presents an adaptive tracking algorithm that integrates color and shape features and improves the quantification method. Good features and quantification parameter are selected and applied to represent the target according to the descriptive ability of these features. The proposed method is been implemented and tested in different kinds of image sequences. Experimental results show that the tracking algorithm is robust and efficient in image sequences, it is fit for some environment, such as variable light, disturbed background, shaded object and worse colon quality.

Key words: object tracking, MeanShift algorithm, quantify, feature selection

中图分类号: 

• TP311.52