SIA OpenIR  > 光电信息技术研究室
基于多特征的自适应目标跟踪算法
Alternative TitleAutomatic TargetTracking Method BasedOn Multi-feature
赵微1,2
Department光电信息技术研究室
Thesis Advisor惠斌
ClassificationTP391.41
Keyword多特征 目标跟踪 Sift 自适应
Call NumberTP391.41/Z47/2013
Pages68页
Degree Discipline控制工程
Degree Name硕士
2013-05-28
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract目标跟踪研究目的是使用计算机根据输入的图像对实际的目标进行精确跟踪,当今运动目标跟踪算法大致分为基于模型的跟踪、基于区域的跟踪、基于轮廓的跟踪和基于特征的跟踪等四类。目前目标跟踪技术已广泛应用于智能机器人、人机交互、视频监控、智能交通以及军事应用领域中。因此对目标跟踪技术的研究具有非常重要的意义。由于目标的突然旋转和尺度变化、目标的遮挡、目标的背景复杂等情况使得跟踪准确性下降,针对这个问题,本文着重研究如何改善目标跟踪的效果,为目标跟踪在实际中的应用提供理论和技术支持。 本文采用基于特征的目标跟踪方法,从目标特征描述和跟踪算法两方面进行研究,提出一种基于多特征的自适应目标跟踪算法。 首先对目标特征描述方法进行研究,传统的目标特征描述方法比较单一而且不能够适应目标的光照变化、旋转变化和尺度变化等等,为了解决目标特征的准确描述,本文采取将SIFT特征和综合特征直方图相结合的方式对目标进行描述,使目标特征具有良好的鲁棒性和有效性。 其次,对自适应目标跟踪算法进行研究,经典的目标跟踪算法有Kalman滤波、粒子滤波和Mean Shift滤波等等,综合考虑不同跟踪算法的优缺点,确定以Kalman滤波跟踪为主,粒子滤波对其进行辅助跟踪的自适应目标跟踪方法,另外,考虑到粒子滤波计算量大的问题,本文用Mean Shift算法对粒子样本进行局部寻优,充分保证粒子的有效性。为了验证本文算法能否适应光照、旋转、尺度变化以及能够适应复杂背景、小目标、低对比度的情况进行了仿真实验,实验结果表明,本文提出的基于多特征的自适应跟踪算法在准确度和鲁棒性方面都取得了良好的实验效果,在低对比度和小目标的情况下仍然能够有效跟踪。 最后,进行高速实时视频跟踪系统的硬件平台的设计,开发了一种基于DSP+FPGA架构的视频跟踪硬件系统,详细介绍了系统的硬件电路设计,最终实现了图像的采集、存储、跟踪、显示等功能。 本文面向课题组在视频目标跟踪领域的重要课题。为目标的自适应跟踪提供了基础,有利于计算机视觉领域的发展。
Other AbstractTarget tracking is the using of computer to track the targets from the images that received. Video target tracking algorithms can be roughly classified into four types at present: model-based tracking, active comour-based tracking, region-based tracking, feature-based tracking. Target tracking has been applied very widely in the fields of intelligent robot, human-computer interaction, video surveillance, intelligent transportation, military applications and so on. Thus ,it is of great significance to study target tracking. The difficulties of target tracking lie in the sudden movements, the sudden change of the rotation and scale, the occlusion of targets , complex background. To solve some of the above tracking problems, this dissertation focuses on how to improve the performance of target tracking to provide a theoretical basis and technical support for the real application of object tracking. The main idea of this paper is target tracking based on feature matching, and researches on feature description and tracking algorithms are carried out separately to put forward a new adaptive target tracking algorithm based on many features. Firstly, we learned target feature description method. The traditional description methods of target features are usually unitary and not able to adapt to the illumination change, rotation, scale changes, etc.In order to achieve accurate description of target feature, this paper adopt the method combining the SIFT feature and the synthesize feature histogram to descript the target, which is very robust and efficient. Secondly, we leaned adaptive target tracking method. The classical algorithms of target tracking are Kalman filter, particle filter, Mean shift filter etc. Evaluating the Pros and cons of each algorithms and taking the robustness and real-time into account, this paper mainly uses the Kalman filter in target tracking, assisting with the particle filter. Considering the heavy computation of particle filter, this paper takes the Mean shift method to achieve local optimization of particle filter, thus guarantee the effectiveness. The adaptive target tracking algorithm, proposed in this paper, based on multi-feature is proved to acquire satisfying results by the simulation experiments. Finally, The hardware platform design of the high-speed realtime video tracking system is introduced simply in the last chapter of this paper. We design a hardware system of video tracking based on the framework of DSP+FPGA, and describe the hardware circuit design of the system in detail, and ensure the function of video collection, storage, tracking and display. The research of this topic is essential in the video target tracking field of the research group, provides a foundation of adaptive target tracking, and is beneficial to the development of the computer vision domain.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/10763
Collection光电信息技术研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
赵微. 基于多特征的自适应目标跟踪算法[D]. 沈阳. 中国科学院沈阳自动化研究所,2013.
Files in This Item:
File Name/Size DocType Version Access License
基于多特征的自适应目标跟踪算法.pdf(3227KB) 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[赵微]'s Articles
Baidu academic
Similar articles in Baidu academic
[赵微]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[赵微]'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.