CASIA OpenIR  > 毕业生  > 硕士学位论文
Alternative TitleResearch & Application of Real-time ObjectTracking
Thesis Advisor卢汉清
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline模式识别与智能系统
Keyword目标跟踪 粒子滤波 主颜色 管道 贡献图 Object Tracking Particle Filters Dominant Color Pipeline Contributing Image
Abstract实时目标跟踪技术是计算机视觉领域中研究热点之一。在安全检测、机器人导航、人机交互、军事应用等领域中有着广泛的应用价值和前景。本文在总结了目前已有的主流跟踪方法之后,提出了一些有效的算法,并实现了一个实时的目标跟踪系统。论文的主要贡献归纳如下: (1) 在 HSV 颜色空间中采用带空间加权的直方图来描述颜色特征,并提出了一种基于主颜色的目标提取方法,可以有效的减少目标区域内背景噪声的干扰。提出结合该颜色特征与目标的表象特征进行跟踪,有效的提高了目标跟踪的准确性,能解决具有不规则形状和倾斜状态目标的跟踪问题。 (2) 在对以往众多目标跟踪算法进行分析的基础上,提出了一种改进的粒子滤波优化算法,并将其应用在实时的目标跟踪系统中。该算法利用多阶迭代选择性的使用粒子,减少了粒子的数量需求并提高了粒子的质量,有效的提高了目标跟踪的速度和精度。 (3) 在微小目标跟踪中,目标由于其面积太小而无法利用前景提取特征来描述,同时复杂的背景会给微小目标带来很大的干扰,因此需要应用特殊的跟踪和检测方法。本文在对可见光条件下微小运动目标的分析研究中做出了以下贡献:基于加权直方图的原理,提出了前背景贡献图的思想,并将此贡献图应用在管道流水跟踪框架中,从目标描述角度提高了对背景的鲁棒性;将粒子滤波的采样思想应用在微小目标跟踪中,解决特殊背景下的目标跟踪。实验证明,这些算法能够有效的改进对复杂背景下微小目标的跟踪性能。
Other AbstractObject tracking on real-time is one of the hottest research point in computer vision. It has broad application and future in security surveillance, robot navigation, human-computer interface, military application, and so on. Based on summary of main current tracking methods, this paper proposes some efficient algorithms and achieves the real-time object tracking system: (1) The paper proposes an object extraction method based on dominant color, which reduces influence of background noises in object area. It also uses histogram based on spatial weights in HSV color space. The paper also proposes to combine color and appearance features to track, and this improves tracking accuracy and solves problems with special shaped and inclined objects. (2) Based on analysis on many object tracking methods, the paper proposes an extended particle filters, and applies it in a real time tracking system. This approach uses particles selectively in a multi-level iterated frame. It reduces the quantity demand and improves the quality of particles, effectively improveing the speed and accuracy of object tracking. (3) In micro-object tracking, the targets are so small that their features are not easy to describe by foreground, and the background noises bring much influence to the targets, therfore spetial tracking methods are needed. This paper gives contribution during analysis of micro-object trakcing under visible lights: based on the theory of weightes histogram, foregrand and background contributing image is proposed and applied in pipeline tracking frame, improving the tracking robustness from object description; this paper applies particle filter method in micro-object tracking, using its un-sensitivity to complex background and solving object tracking under spetial background. Experiments prove that these algorithms effectively improve tracking capability to micro-objects under complex background.
Other Identifier200228014603557
Document Type学位论文
Recommended Citation
GB/T 7714
柳峰. 实时目标跟踪的分析研究与系统实现[D]. 中国科学院自动化研究所. 中国科学院研究生院,2005.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[柳峰]'s Articles
Baidu academic
Similar articles in Baidu academic
[柳峰]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[柳峰]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.