CASIA OpenIR  > 毕业生  > 硕士学位论文
Alternative TitleReal-time People-Counting System based on Feature Points Tracking
Thesis Advisor黄磊 ; 刘昌平
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计算机应用技术
Keyword行人检测 行人跟踪 特征点跟踪 人数计数 Pedestrian Detection Object Tracking Feature Point Tracking People Counting
Abstract近几年,随着社会的发展,公共安全已经越来越受到社会的关注和国家的重视。其中公共安全中一个重要的控制因素就是客流量, 利用统计的客流量信息,管理人员可以合理的调度人力、物力以及其它管理资源,并且能够在节省成本的同时,更好的控制公共安全。因此近几年,人数计数的相关研究已经逐渐增多。对于商场,大厦等,可以利用打卡来精准的实现人数的统计,但是对于一些室外开放式场合,如鸟巢奥林匹克公园,不可能用传统的打卡方法来实现人数计数;而计算机视觉的发展,使得基于视频的智能统计方法受到学者的关注。本文首先调研了大量的国内外参考文献资料,在此基础上,提出了一种基于目标检测和特征点跟踪的方法来解决这个问题。本文的主要工作和创新点包括: (1) 使用了基于LBP和EOH的融合特征进行行人检测。将行人区域进行多子区域划分,引入空间关联性,提高检测准确性; (2) 在跟踪阶段,使用不容易产生遮挡的全局特征点来进行跟踪,通过建立行人和特征点之间的模型,将对特征点的跟踪路径转化为行人跟踪路径; (3)计数阶段,使用了多拌线触发的计数规则,减少了对行人检测快速性的依赖,同时增加了计数的准确性;
Other AbstractWith the rapid development of the society, public security is becoming more and more important and being concerned by the government. Passenger flow is one of the most important factor in the public security area. With the information of the passenger of the flow, the manger can reasonable schedule the resources and provide a better management. In recent years, more and more research is concentrated on people counting. The method of people-counting based on radio machine can be used for the indoor scene like shopping malls、buildings. But for the outdoor scene like the Olympic Park, it is impossible to use the traditional method. On the other hand, Computer vision can be well used for solving this problem. In this article, a method based on object detection and point tracking are proposed to solve this problem. The main contribution of this thesis includes: (1) A new feature based on LBP and EOH are proposed for pedestrian detection. When extracting the feature of the image, the pedestrian area is divided into multi-blocks at first, and then the feature is extracted separately in each block, finally all the features are contacted together to describe the image; The accurate rate are greatly improved , while importing the space information. (2) Firstly the feature points of the image are tracked, and then a model is used for converting the tracking route of the feature points to the route of the related pedestrian. (3) A multi-counting line strategy is used for counting, this strategy can reduce the dependence on fast detection of people and improve the counting accurate rate.
Other Identifier200828014628035
Document Type学位论文
Recommended Citation
GB/T 7714
何鹏. 基于特征点跟踪的实时人数计数系统[D]. 中国科学院自动化研究所. 中国科学院研究生院,2011.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_20082801462803(3324KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[何鹏]'s Articles
Baidu academic
Similar articles in Baidu academic
[何鹏]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[何鹏]'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.