多摄像头系统的目标检测与跟踪方法研究  

Research of Target Detection and Tracking Method for Multi-camera System

作　　者：高翔[1] 朱婷婷[1] 刘洋[1] 

机构地区：[1]南京邮电大学自动化学院,江苏南京210023

出　　处：《计算机技术与发展》2015年第7期221-225,共5页Computer Technology and Development

基　　金：国家自然科学基金资助项目(61104216);江苏省自然科学基金(BK2012832)

摘　　要：针对运动目标检测的问题,提出了一种中线检测算法。利用人体两侧关于中线几乎完全对称的特点,首先通过背景差分技术得到目标人的二进制图像,然后利用最小中位数均方差的方法对于单个目标人进行中线检测,并利用多个摄像头之间单应矩阵对应关系,得到检测目标在各摄像头中的相应位置,实现目标在多个摄像头中的目标检测。在目标跟踪的问题上,采用了一种基于MCMC(Markov Chain Monte Carlo)粒子滤波的目标跟踪算法。算法中首先建立自回归运动模型及基于目标垂直投影直方图的观测模型,然后利用粒子滤波在各个摄像头中进行跟踪。中心处理机通过匹配算法融合各摄像头中的目标跟踪数据,实现多个摄像头环境下目标的有效跟踪。Aiming at the problem of motion target detection, propose a median detection algorithm, which makes use of both sides of the body on the median almost perfectly symmetrical characteristic. First, obtain the binary image of target person by background subtraction. Then, detect the independent target by applying the least median of squares. Next, using the corresponding relationship of homographic matrix between multiple cameras to get the position of the detected target in each camera, achieving person detection in multi-camera sys- tem. On the problem of target tracking, adopt a target tracking algorithm based on MCMC particle filter. First, this algorithm establishes an autoregressive motion model and an observation model based on the vertical projection histogram of the detected person. Next,using the particle filter track person in each camera. The central processor has merged the tracking data of each camera by matching algorithm,reali- zing target tracking of multi-camera system effectively.

关 键 词：目标检测 目标跟踪 背景差分 中线 MCMC粒子滤波 多视角协同 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]