Pedestrian Tracking Using Directed Scene Motion Pattern and Particle Filter

doi:10.7652/xjtuxb201412012

您当前的位置：

首页 >

文章列表页 >

Pedestrian Tracking Using Directed Scene Motion Pattern and Particle Filter

更新时间：2025-07-09

- Pedestrian Tracking Using Directed Scene Motion Pattern and Particle Filter
- Vol. 48, Issue 12, Pages: 74-79(2014)
- 作者机构：
  
  西安电子科技大学计算机学院,西安,710071
- 作者简介：
- 基金信息：
- DOI：10.7652/xjtuxb201412012
  CLC： TN911.73
- Online First：10 December 2014，
  
  Published：2014
- 稿件说明：
移动端阅览
Pedestrian Tracking Using Directed Scene Motion Pattern and Particle Filter[J]. 2014, 48(12): 74-79.
DOI：

Pedestrian Tracking Using Directed Scene Motion Pattern and Particle Filter[J]. 2014, 48(12): 74-79. DOI： 10.7652/xjtuxb201412012.

摘要

针对现有行人跟踪算法较少考虑场景运动模式信息的问题

建立一种面向有向场景运动模式的在线学习模型以描述区域行人的共有运动特性

并以此提出了一种新型的粒子滤波行人跟踪算法。通过对行人运动特性的选择性在线统计

探索在非高密度行人跟踪问题中场景模式信息和运动历史信息的运用方式。模型由一个表征行人运动状态相空间局域运动特性的二阶直方图矩阵来描述

并根据每个跟踪单元的加权投票实施更新。通过修正粒子转移后似然概率分布

该算法能够加速粒子向真实的后验分布收敛。通过对两个不同特点的公共数据集视频中的行人进行跟踪实验并与标准的粒子滤波算法结果比较

该算法的平均跟踪误差均低于标准粒子滤波平均跟踪误差的40%

且其运算速度可达6～15帧/s

满足近实时应用帧率。

Abstract

An online learning model oriented to directed scene motion patterns and based on particle filter is proposed to get over the problem that existing pedestrian tracking algorisms have little consideration on the scene motion patterns

to process the priori scene information and to improve the precision of pedestrian tracking. The model describes the local motion characteristic in phase space of the system state vector using a direction-speed motion histogram matrix

and is updated according to weighted voting of each track unit. The particle filter can accelerate the particles evolution to the true posterior distribution through modifying likelihood probability distribution of particle transfer. Experimental results on two pedestrian data sets with different characteristic forms and comparisons with the standard particle filter algorithm show that the average tracking error of the proposed approach is 40% lower than that of the standard particle filter algorithm

while its computation speed approaches the range of 6 to 15 frames a second

and that the proposed approach completely meets the conditions of real-time application.

关键词

Keywords

references

MUSSO C, OUDJANE N, GLAND F L. Improving regularised particle filters[M]∥ Sequential Monte Carlo Methods in Practice. New York, USA: Springer, 2001: 247-271.

GILKS W R, BERZUINI C. Following a moving target: Monte Carlo inference for dynamic Bayesian models[J]. Journal of the Royal Statistical Society: Series B Statistical Methodology, 2001, 63(1): 127-146.

ALI S, SHAH M. Floor fields for tracking in high density crowd scenes[M]∥ Computer Vision-ECCV. Berlin, Germany: Springer-Verlag, 2008: 1-14.

KRATZ L, NISHINO K. Tracking with local spatio-temporal motion patterns in extremely crowded scenes[C]∥Proceedings of the 2010 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, NJ, USA: IEEE, 2010: 693-700.

XU Feiming, LU Tong, WU Yirui. Robust object tracking using motion context in crowded scenes[M]∥ Advances in Multimedia Information Processing. Cham, Switzerland: Springer International Publishing, 2013: 550-560.

IDREES H, WARNER N, SHAH M. Tracking in dense crowds using prominence and neighborhood motion concurrence[J]. Image and Vision Computing, 2014, 32(1): 14-26.

ZHOU Bolei, WANG Xiaogang, TANG Xiaoou. Random field topic model for semantic region analysis in crowded scenes from tracklets[C]∥Proceedings of the 2011 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, NJ, USA: IEEE, 2011: 3441-3448.

ZHAO X, MEDIONI G. Robust unsupervised motion pattern inference from video and applications[C]∥Proceedings of the 2011 IEEE International Conference on Computer Vision. Piscataway, NJ, USA: IEEE, 2011: 715-722.

CAVIAR Project. CAVIAR database[DB/OL].(2005-09-30)[2014-06-01]. http:∥homepages.inf.ed.ac.uk/rbf/CAVIAR/.

UK EPSRC REASON Project. PETS2009 DataSet[DB/OL].(2009-10-07)[2014-06-01]. http:∥pets2009.net/.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Gaussian Diffracted Particle Filter

Improved Particle Filter Algorithm Based on Threshold De-Noising

Particle Filter Using Second-Order Central Difference

Quadrature Kalman Particle Filter

Review of the Application of EBSILON Software in Combined Heat and Power System Optimization

Related Author

宁小磊 1

王宏力 1

宁宇琪 2

张琪 1

王向华 1

覃征 1

杨新宇 2

栗江涛 2

Related Institution

Controlling and Testing Department, The Second Artillery Engineering College

Department of Computer Science and Technology, Tsinghua University

School of Electronics and Information Engineering, Xi'an Jiaotong University

⁰