Poselets for Pedestrian Detection

HUANG Yan; YAN Jia; Zhang Hu; DENG De-xiang

doi:10.3788/OPE.20132111.2906

您当前的位置：

首页 >

文章列表页 >

Poselets for Pedestrian Detection

更新时间：2020-08-12

- Poselets for Pedestrian Detection
- Optics and Precision Engineering Vol. 21, Issue 11, Pages: 2906-2913(2013)
- 作者机构：
  
  武汉大学电子信息学院，湖北武汉 430079
- 作者简介：
- 基金信息：
- DOI：10.3788/OPE.20132111.2906
  CLC： TP391
- Received：09 May 2013，
  
  Revised：02 July 2013，
  
  Published Online：22 November 2013，
  
  Published：15 November 2013
- 稿件说明：
移动端阅览
黄炎, 颜佳, 张虎, 邓德祥. 多部位集合的人体检测[J]. 光学精密工程, 2013,21(11): 2906-2913

HUANG Tan, YA Jia, ZHANG Hu, DENG De-Xiang. Poselets for Pedestrian Detection[J]. Editorial Office of Optics and Precision Engineering, 2013,21(11): 2906-2913
黄炎, 颜佳, 张虎, 邓德祥. 多部位集合的人体检测[J]. 光学精密工程, 2013,21(11): 2906-2913 DOI： 10.3788/OPE.20132111.2906.

HUANG Tan, YA Jia, ZHANG Hu, DENG De-Xiang. Poselets for Pedestrian Detection[J]. Editorial Office of Optics and Precision Engineering, 2013,21(11): 2906-2913 DOI： 10.3788/OPE.20132111.2906.

摘要

采用启发式、有监督的部位筛选方法组成了一种多部位集合的检测模型，用于缓解遮挡和形变对人体检测造成的影响。该模型通过比较人体同部位上关节点间的Procrustes距离，在训练集中获取有着相似姿态的同部位样本；将梯度方向直方图（HOG）作为特征，由典型部位分别训练出判别模型；比较其在验证集上的检测效果，从中筛选出检测率高的部位和未检出的图片，再寻找对未检出图片检测率高的部位，由这些部位组成混合模型。用混合模型检测时，由Kullback-Leibler距离判断各部位在图片上的不同响应是否属于同一人，以此来确定人体的外接矩形框。在INRIA人体库上的测试表明，本文采用的模型在误检率（FPPI）为0.5时有81%的检测率，高于有77%检测率的Poselets算法。本文基于Poselets，结合HOG的特点采用了一套有监督的部位筛选机制，使得模型成员数大幅度减少，检测时间比原始方法下降了50%，同时取得了优于Poselets的检测效果。

Abstract

A part-based mixture model(Poselets) was ultilized to alleviate the problem of pedestrian detection under occlusion and articulation. Firstly

Procrustes distances between similar pedestrian parts were calculated to gather training exmaples with the same configurations. Then

common parts of frontal pedestrian described by Histogram of Oriented Gradient(HOG) were trained to get discriminative models. A test was carried out on a validation set to find out which parts were more accurate

and those pictures which were not detected formed the hard set.Other complementary parts were explored on the hard set

afterwards

all these parts formed the final Poselets detector. While detecting

Poselets were clusterd if the symmetrized KL-divergence between two Poselet activations were small. The bounding box of a pedestrian was inferred by the bounding boxes of the Poselets in a cluster. Based on the test of the INRIA pedestrian dataset

it is showed that the detection rate increases from 77% of Poselets to 81% in this paper while False Positive Per Image(FPPI) is 0.5. It is concluded that the part selection mechanism proposed in this paper promotes the detection rate with speedup rate about twice of the traditional method

meanwhile it reduces the number of models.

关键词

Keywords

references

DALAL N, TRIGGS B. Histograms of oriented gradients for human detection [C].International Conference on Computer Vision & Pattern Recognition, 2005,1:886-893.[2]MIKOLAJCZYK K, SCHMID C, ZISSERMAN A. Human detection based on a probabilistic assembly of robust part detectors [C].European Conference on Computer Vision(ECCV04), 2004, 3021:69-82.[3]LIN Z, DAVIS L S. A pose-invariant descriptor for human detection and segmentation[C]. European Conference on Computer Vision, 2008,5305,423-436.[4]ANDRILUKA M, ROTH S SCHIELE B.Discriminative appearance models for pictorial structures[J]. International Journal of Computer Vision,2012, 99(3):259-280.[5]DOLLAR P, BABENKO B, BELONGIE S, et al.. Multiple component learning for object detection[C].Computer Vision-ECCV 2008, 2008,5303:211-224.[6]FELZENSZWALB P F,GIRSHICK R B,MCALLESTER D.Cascade object detection with deformable part models [C]. IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2010:2241-2248.[7]BOURDEV L,MAJI S, BROX T, et al.. Detecting people using mutually consistent poselet activations[C].Computer Vision-EECV 2010, 2010,6316:168-181.[8]EVERINGHAM M, VAN G L, WILLIAMS, et al..The PASCAL Visual Object Classes (VOC) Challenge [J].International Journal of Computer Vision, 2010,88(2), 303-338.[9]BOURDEV L, MALIK J. Poselets: Body part detectors trained using 3D human pose annotations[C]. 2009 IEEE 12th International Conference on Computer Vision,2009: 1365-1372.[10]BOURDEV L.Poselets and their applications in high-Level computer vision[D]. California：Computer Science Graduate Division of the University of California Berkeley,2011.[11]DALAL N.Finding people in images and videos[D].France:the French National Institute for Research in Computer Science and Control,2006.[12]黄炎，范赐恩，朱秋平,等.联合梯度直方图和局部二值模式特征的人体检测[J].光学精密工程,2013,21 (4):1047-1053.HUANG Y, FAN C E, ZHU Q P, el at.. HOG-LBP pedestrian detection[J]. Opt. Precision Eng., 2013, 21(4): 1047-1053. (in Chinese)

Views

254

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Microscopic feature localization for mass precision assembly tasks

Crack detection and segmentation in CT images using Hessian matrix and support vector machine

Classification of real-time digital PCR amplification curves

Threat assessment of support vector machine optimized by Krill Herd algorithm

Related Author

CUI Shipeng

LU Shiqin

XU Zheng

YU Zhongyang

WANG Xiaodong

Zhongyang YU

Xiaodong WANG

Zheng XU

Related Institution

School of Mechanical Engineering， Dalian University of Technology

ICT Research Center， Key Laboratory of Optoelectronic Technology and System of the Education Ministry of China， Chongqing University

College of Optoelectronic Engineering， Chongqing University

Ji Hua Laboratory

Children's hospital， Fudan University

AI问答

Address：No.3888 Dong Nanhu Road, Changchun, Jilin, China Postal code：130033
Tel：0431-86176855 Email：gxjmgc@ciomp.ac.cn
Technical support is provided by Beijing Founder electronics co., LTD 吉ICP备11002662号-17 京公网安备11010802024621
It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.

⁰