基于背景差分和Haar-like特征的客运站行人检测算法研究
详细信息 查看官网全文
- 英文论文题名：Research on Pedestrian Detection in Coach Station Based on Background Subtraction and Haar-like Feature
- 论文作者：江倩殷 ; 李熙莹 ; 黄秋筱 ; 卢林
- 英文论文作者：Jiang Qianyin ; Huang Qiuxiao ; Lu Lin ; Li Xiying ; School of Engineering ; SUN YAT-SEN University ; Guangdong Provincial Key Laboratory of Intelligent Transportation System ; Key Laboratory of Video and Image Intelligent Analysis and Application Technology Ministry of Pubil Security ; People's Republic of China
- 年：2016
- 作者机构：广东省智能交通系统重点实验室;中山大学工学院;视频图像智能分析与应用技术公安部重点实验室;
- 论文关键词：客运站 ; 背景差分 ; Haar-like特征 ; 行人检测
- 英文论文关键词：Coach Station ; Background Subtraction ; Haar-like Feature ; Pedestrian Detection
- 会议召开时间：2016-11-16
- 会议录名称：第十一届中国智能交通年会大会论文集
- 语种：中文
- 分类号：U495;TP391.41
- 学会代码：ZJTX
- 会议名称：第十一届中国智能交通年会
- 会议地点：中国重庆
- 主办单位：中国智能交通协会
- 学会名称：中国智能交通协会
- 页数：8
- 文件大小：1016k
- 原文格式：D
- 会议级别：全国
摘要
客运站客流量是一项重要的数据,快速、准确的行人检测是客流统计以及行为分析的基础。客运站人流密集,这对传统的行人检测算法提出了新的要求;而在提高检测准确度的同时,如何保证检测算法的实时性也是一大难题。因此,本文提出一种基于背景差分和Harr-like特征的行人检测算法,首先采用背景差分提取运动区域,将特征提取的区域大大缩小,然后在运动区域内提取Haar-like特征,通过事先训练好的Adaboost分类器实现行人的判断与检测。通过对客运站监控视频进行检测统计,本文的行人检测方法解决了背景差分准确率低和Haar-like特征检测的实时性差的问题,检测速度和准确性都达到了应用要求。
Objective:Passengervolume in coach station is important,while a fast and accurate algorithm for pedestrian detection is the foundation of passenger volume counting and personal behavior analysis.The coach station is always crowded,which proposes a new demand to traditional algorithm of pedestrian detection.It is also a big problem to ensure the real time of detection as well as the improvement of accuracy.Therefore,the algorithm of pedestrian detection based on background subtraction and Haar-like feature is proposed.Firstly,the algorithm of background subtraction is used to extract the moving regionsso that the areas of feature extraction isscaled-down a lot.Then the Haar-like feature is extracted in the moving regions.Finally,the Adaboostclassifier which is trained in advance is used to detect and judge a pedestrian.The algorithm proposed in this paper is useful in passenger volume counting in surveillance videos in coach station and solve the problems of low accuracy in background subtraction and weak real-timeperformance in Haar-like feature detection.It is proved that the speed and accuracy of the algorithm have reached the application requirements.
Objective:Passengervolume in coach station is important,while a fast and accurate algorithm for pedestrian detection is the foundation of passenger volume counting and personal behavior analysis.The coach station is always crowded,which proposes a new demand to traditional algorithm of pedestrian detection.It is also a big problem to ensure the real time of detection as well as the improvement of accuracy.Therefore,the algorithm of pedestrian detection based on background subtraction and Haar-like feature is proposed.Firstly,the algorithm of background subtraction is used to extract the moving regionsso that the areas of feature extraction isscaled-down a lot.Then the Haar-like feature is extracted in the moving regions.Finally,the Adaboostclassifier which is trained in advance is used to detect and judge a pedestrian.The algorithm proposed in this paper is useful in passenger volume counting in surveillance videos in coach station and solve the problems of low accuracy in background subtraction and weak real-timeperformance in Haar-like feature detection.It is proved that the speed and accuracy of the algorithm have reached the application requirements.
引文
[1]Olivier B.,Marc V.D.Vi Be:a universal Background subtraction algorithm for video sequences[C].IEEE Transactions on Image Processing,2011,20(6):1709-1724.
[2]陈虹颖.隧道环境下行人目标视频检测技术研究[D].重庆大学,2013.
[3]王明明.固定背景下单/多目标行人跟踪算法研究[D].山东大学,2013.
[4]Toyama K,Krumm J,Brumitt B,et al.Wallflower:Principles and practice of background maintenance[J].International Conference on Computer Vision(Kerkyra,Greece),1999,9:255-261.
[5]Kaewtrakulpong P,Bowden R.A real time adaptive visual surveillancesystem for tracking low-resolution colour targets in dynamically changing scenes[J].Image&Vision Computing,2003,21(10):913-929.
[6]张书真,基于帧差和Otsu的红外行人分割算法[J].吉首大学学报(自然科学版),2013,34(4):59-76.
[7]姚倩,安世全,姚路.三帧差法和Mean-shift结合的行人检测与跟踪研究[J].计算机工程与设计,2014,35(1):223-227.
[8]魏玮,吴琪.三帧差结合改进高斯建模的运动目标检测算法[J].计算机工程与设计,2014,35(3):949-952.
[9]Dala N,Triggs B.Histogram s of oriented gradients for human detection[C].IEEE Computer Society Conference on Computer Vision and Pattern Recognition.San Diego,2005.
[10]姚源源,董小舒.基于改进HOG描述符的行人检测算法[J].指挥信息系统与技术,2013,4(4):55-69.
[11]Ninomiya H,Ohki H,Gyohten K.An evaluation on robustness and brittleness of HOG features of human detection[C].17th Korea-Japan Joint Workshop on Frontiers of Computer Vision.Ulsan,2011.
[12]Dang L,Bui B,Vo PD,et al.Improved HOG descriptors[C].Proceedings of the 3rd International Conference on Knowledge and Systems Engineering.Hanoi,2011:186-189.
[13]Viola P,Jones M.Fast and robust classification using asymmetric adaboost and a detector cascade[J].Advances in Neural Information Processing Systems,2002,2:1311-1318.
[14]Wu J,Yang S,Zhang L.Pedestrian detection based on improved HOG feature and robust adaptive boosting algorithm[C].Proceedings of the 4th International Congress on Image and Signal Processing(CISP),Shanghai,2011.
[15]Freund Y,Schapire R E.A decision-theoretic generalization of on-line learning and an application to boosting[J].Journal of computer and system sciences,1997,55(1):119-139.
[2]陈虹颖.隧道环境下行人目标视频检测技术研究[D].重庆大学,2013.
[3]王明明.固定背景下单/多目标行人跟踪算法研究[D].山东大学,2013.
[4]Toyama K,Krumm J,Brumitt B,et al.Wallflower:Principles and practice of background maintenance[J].International Conference on Computer Vision(Kerkyra,Greece),1999,9:255-261.
[5]Kaewtrakulpong P,Bowden R.A real time adaptive visual surveillancesystem for tracking low-resolution colour targets in dynamically changing scenes[J].Image&Vision Computing,2003,21(10):913-929.
[6]张书真,基于帧差和Otsu的红外行人分割算法[J].吉首大学学报(自然科学版),2013,34(4):59-76.
[7]姚倩,安世全,姚路.三帧差法和Mean-shift结合的行人检测与跟踪研究[J].计算机工程与设计,2014,35(1):223-227.
[8]魏玮,吴琪.三帧差结合改进高斯建模的运动目标检测算法[J].计算机工程与设计,2014,35(3):949-952.
[9]Dala N,Triggs B.Histogram s of oriented gradients for human detection[C].IEEE Computer Society Conference on Computer Vision and Pattern Recognition.San Diego,2005.
[10]姚源源,董小舒.基于改进HOG描述符的行人检测算法[J].指挥信息系统与技术,2013,4(4):55-69.
[11]Ninomiya H,Ohki H,Gyohten K.An evaluation on robustness and brittleness of HOG features of human detection[C].17th Korea-Japan Joint Workshop on Frontiers of Computer Vision.Ulsan,2011.
[12]Dang L,Bui B,Vo PD,et al.Improved HOG descriptors[C].Proceedings of the 3rd International Conference on Knowledge and Systems Engineering.Hanoi,2011:186-189.
[13]Viola P,Jones M.Fast and robust classification using asymmetric adaboost and a detector cascade[J].Advances in Neural Information Processing Systems,2002,2:1311-1318.
[14]Wu J,Yang S,Zhang L.Pedestrian detection based on improved HOG feature and robust adaptive boosting algorithm[C].Proceedings of the 4th International Congress on Image and Signal Processing(CISP),Shanghai,2011.
[15]Freund Y,Schapire R E.A decision-theoretic generalization of on-line learning and an application to boosting[J].Journal of computer and system sciences,1997,55(1):119-139.