结合方向纹理熵的Haar-like特征在线boosting跟踪算法  

On-line boosting tracking based on direction texture entropy combined Haar-like feature

作　　者：胡松[1] 孙水发[1] 陈鹏[1] 但志平[1] 董方敏[1] 

机构地区：[1]三峡大学智能视觉与图像信息研究所,湖北宜昌443002

出　　处：《计算机应用研究》2014年第4期1260-1263,1269,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61102155;61272236;61272237);湖北省高等学校优秀中青年科技创新团队资助项目(T201002);湖北省教育厅青年科学基金资助项目(Q20111205)

摘　　要：针对传统基于Haar-like特征的on-line boosting跟踪算法(HBT)需要产生大规模随机特征、占用大量计算资源和存储空间的缺点,提出结合方向纹理熵的Haar-like特征在线boosting跟踪算法(HBTT)。HBTT算法利用灰度共生矩阵的熵获得目标纹理的方向信息,在此基础上有针对性地产生具有方向纹理信息的Haar-like特征,从而可有效避免无效随机特征的产生,减小特征池容量;更进一步,可根据目标纹理的复杂程度自动调整特征数量,使得算法更灵活。在跟踪过程中,在线学习模块可以使错误率较高的特征被结合了目标纹理方向信息的Haar-like特征所替换。与HBT算法比较,HBTT算法的跟踪误差降低了10%以上;在相同特征池容量下,置信度提高了2%以上。实验结果表明,该算法不仅具有较高的鲁棒性,而且在跟踪效率和性能上都有所提高。Large numbers of features are needed for tracking based on the conventional Haar-like feature based on-line boos- ting （HBT）. The vast amount guarantees the existence-of useful features. But it leads to expensive computing and memory re- quirements, To address this problem, this paper proposed a new method named HBTI＂ （Haar-like feature based on-line boos- ting tracking with texture information）. It extracted the texture direction and richness information by the entropy of gray level co-occurrence matrix （GLCM）. Then calculated Haar-like features along directions with richest texture. And also governed the employed feature number by the texture richness. In this way, not only the number of useless features was largely reduced, but also the total size of the feature pool. Experiments reveal the higher robustness and performance of this new HBTr method, In general, the tracking error is reduced by more than 10%, and the confidence of the tracking is increased by more than 2%.

关 键 词：目标跟踪 在线boosting算法 类HAAR特征 灰度共生矩阵 方向纹理熵 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]