计算机与现代化

• 算法设计与分析 • 上一篇    下一篇

一种融合粒子滤波的多特征特定行人追踪方法

  

  1. 广东石油化工学院计算机与电子信息学院,广东茂名525000
  • 收稿日期:2014-07-18 出版日期:2014-11-03 发布日期:2014-11-05
  • 作者简介:余桂兰(1975-),女,广东饶平人,广东石油化工学院计算机与电子信息学院高级实验师,硕士,研究方向:人工智能与模式识别。
  • 基金资助:
    茂名市科技计划项目(2012B01040)

A Specific Pedestrian Tracking Method Based on Particle Filter and Multi-Feature Fusion

  1. College of Computer and Electronic Information, Guangdong University of Petrochemical Technology, Maoming 525000, China
  • Received:2014-07-18 Online:2014-11-03 Published:2014-11-05

摘要: 针对特定行人目标跟踪,提出了一种融合粒子滤波的多特征特定行人检测追踪方法。该方法借鉴空间金字塔匹配模型将行人在空间上进行划分,然后融合颜色特征和局部三值模式(LTP)纹理特征,提取目标行人各子区域信息,最后结合粒子滤波来综合判断目标行人的位置。实验结果显示,本文方法能够有效区分目标与背景,同时在目标行人被遮挡的情况下,算法能够有效跟踪。

关键词: 目标检测, 特定行人, 粒子滤波, 特征融合

Abstract: For the target tracking of specific pedestrians, a particular pedestrian tracking approach based on multi-feature fusion and particle filter is proposed. The approach uses the space pyramid matching model for reference to divide the target in space, then extracts color features and LTP features from each sub-region of the target, finally, integrates particle filter to determine the location of the target pedestrians. The experimental results show that the approach can effectively distinguish between target and background. Moreover, it can effectively track the target in the case of occluded.

Key words: target detection, specific pedestrian, particle filter, feature fusion

中图分类号: