计算机工程 ›› 2011, Vol. 37 ›› Issue (19): 22-25.doi: 10.3969/j.issn.1000-3428.2011.19.006

• 专栏 • 上一篇    下一篇

视频动态场景下运动物体的自适应跟踪算法

程乐凯a,b,翟素兰a,c,涂铮铮a,b,罗 斌 a,b   

  1. (安徽大学 a. 计算智能与信号处理教育部重点实验室;b. 计算机科学与技术学院;c. 数学科学学院,合肥 230039)
  • 收稿日期:2011-03-18 出版日期:2011-10-05 发布日期:2011-10-05
  • 作者简介:程乐凯(1972-),男,硕士研究生,主研方向:模式识别,视频图像处理;翟素兰,副教授、博士;涂铮铮,讲师、硕士; 罗 斌,教授、博士生导师
  • 基金项目:

    国家自然科学基金资助项目(60772122);安徽省教育厅自然科学研究基金资助重点项目(KJ2008A033, KJ2007A072);安徽省高等学校省级优秀青年人才基金资助项目(2009SQRZ221)

Adaptive Tracking Algorithm for Motive Objects Under Video Dynamic Scenes

CHENG Le-kai a,b, ZHAI Su-lan a,c, TU Zheng-zheng a,b, LUO Bin a,b   

  1. (a. Key Laboratory of Intelligent Computing & Signal Processing, Ministry of Education; b. School of Computer Science & Technology; c. School of Mathematical Sciences, Anhui University, Hefei 230039, China)
  • Received:2011-03-18 Online:2011-10-05 Published:2011-10-05

摘要:

根据积分思想和粒子滤波理论,提出一种运动物体自适应跟踪算法。建立多特征的跟踪观测模型,采用积分策略对多个特征模型进行自适合融合和更新,通过观测似然度更新粒子滤波所需粒子数量及其分布,并在动态场景下对运动物体进行跟踪。实验结果表明,该算法相对于传统的跟踪算法在跟踪精度和实时性方面有所提高,具有较好的鲁棒性。

关键词: 跟踪, 特征模型, 自适应模型融合, 观测似然度, 粒子滤波

Abstract:

An adaptive tracking algorithm based on integral fusion thoughts and particle filter theory is proposed for video objects in motion. More features measurement models are used and the introduction of integral strategy fuses and updates them adaptively. Based on observation likelihood, the algorithm updates the number of particles and proposal density, and applies particle filtering for tracking under dynamic scenes. Experiments results demonstrate that the proposed algorithm improves the tracking precision and real-time with more robustness, contrasting to traditional tracking algorithm.

Key words: tracking, characteristic model, adaptive model fusion, observation likelihood, particle filtering

中图分类号: