电子学报 ›› 2014, Vol. 42 ›› Issue (4): 672-678.DOI: 10.3969/j.issn.0372-2112.2014.04.008

• 学术论文 • 上一篇    下一篇

融合热释电红外传感器与视频监控器的多目标跟踪算法

李方敏, 姜娜, 熊迹, 张景源   

  1. 武汉理工大学信息工程学院光纤传感技术与信息处理教育部重点实验室, 湖北武汉 430070
  • 收稿日期:2013-07-25 修回日期:2013-12-05 出版日期:2014-04-25
    • 作者简介:
    • 李方敏 男.1968年6月出生,湖南涟源人.教授、博士生导师,中国计算机学会高级会员,传感器网络专委会委员.1990年、1997年和2001年分别在华中理工大学、国防科技大学和浙江大学获工学学士、工学硕士和工学博士学位.主要从事无线自组织网络、新型网络体系结构、嵌入式系统等方面的工作.E-mail:lifangmin@whut.edu.cn;姜 娜 女.1989年1月出生,河北衡水人.2012年毕业于武汉理工大学信息学院,获工学学士学位,现为该校信息与通信工程专业硕士研究生.研究方向为信号处理、模式识别及应用等相关研究.
    • 基金资助:
    • 国家自然科学基金 (No.61170090)

Multi-Object Tracking Scheme with Pyroelectric Infrared Sensor and Video Camera Coordination

LI Fang-min, JIANG Na, XIONG Ji, ZHANG Jing-yuan   

  1. Key Laboratory of Fiber Optical Sensing Technology and Information Processing, Ministry of Education, School of Information Engineering, Wuhan University of Technology, Wuhan, Hubei 430070, China
  • Received:2013-07-25 Revised:2013-12-05 Online:2014-04-25 Published:2014-04-25
    • Supported by:
    • National Natural Science Foundation of China (No.61170090)

摘要: 现有基于热释电红外传感器的多目标跟踪系统在目标之间距离较近或者轨迹相交的情况下存在着误差较大的缺点.针对此缺点,提出了一种新型的基于热释电红外传感器与视频监测器协同工作的多目标跟踪方案.该方案可以充分利用两种传感器的优势,弥补在目标跟踪中的不足.算法采用最小二乘法利用热释电信息进行定位,并通过从图像或热释电传感器信号的幅频特性中提取特征信息来校正联合概率数据关联算法的关联矩阵,有效避免了错误关联.实验表明,该方案在多目标交叉情况下跟踪误差仅为其它算法的八分之一到四分之一.

关键词: 热释电红外(PIR)传感器, 视频监控器, 目标跟踪, 联合概率数据关联算法

Abstract: The error tends to be significant in many existing pyroelectric infrared sensor based multi-object tracking systems when the measured objects get close to each other or their trajectories have intersections.To solve this problem,we proposed a multi-object tracking scheme by having pyroelectric infrared sensors and video cameras work cooperatively.This scheme takes the advantages of both kinds of sensors,which help to improve the performance compared to those using any kind of such sensors.In the proposed scheme,we first achieve coarse positioning using least square method with data collected by pyroelectric infrared sensors,and then we correct the incidence matrix in joint probabilistic data association with features extracted from the images or the frequency responses of pyroelectric sensors.The coarse positioning is further filtered by joint probabilistic data association algorithm to obtain the final fine result.Such a method prevents false association effectively.Experimental results show that the tracking error of the proposed scheme in multi-object crossover scenario reduces to a quarter,even to one eighth of the errors that exist in the compared schemes.

Key words: PIR (pyroelectric infrared radial) sensor, video camera, human body tracking, JPDA (joint probabilistic data association)

中图分类号: