摘要
大多数应用于视频监控领域的目标跟踪模式识别方法,都需要先对移动目标进行模式学习。但是这些方法不适合同时跟踪多个不同的目标,因为每一个移动目标的模式都应该是预先确定好的。因此,提出了一种新的基于粒子滤波和背景减除的无监督多运动目标检测与跟踪方法来解决这个问题。该方法能够自动探测和跟踪许多移动目标,没有任何学习阶段,也没有任何关于大小、性质或初始位置的先验知识。对多个视频测试集进行了实验验证,测试结果表明,该方法可以成功地处理复杂情况下的目标跟踪。与其他方法进行比较,结果显示提出的方法检测以及跟踪目标性能更好。
Most of the target tracking pattern recognition methods used in the field of video surveillance, both need to learn the model of moving target. However, these methods are not suitable for tracking multiple different targets at the same time, because the mode of each moving object should be determined in advance. Therefore, this paper proposed a new unsupervised multi moving target detection and tracking method based on particle filtering and background subtraction to solve this problem. The method is able to detect and track many moving objects automatically, without any learning phase, nor any prior know- ledge about the size, nature or initial position. The test results show that the proposed method can successfully deal with the target tracking in complex situations. Compared with other methods, the results show that the proposed method is better in detecting and tracking the target performance.
作者
李明杰
刘小飞
张福泉
翟萍
Li Mingjie;Liu Xiaofei;Zhang Fuquan;Zhai Ping(School of Information ＆ Intelligence Engineering,Sanya University,Sanya Hainan 572022,China;School of Software,Beijing Institute of Technology,Beijing100081,China;School of Information Engineering,Zhengzhou University,Zhengzhou 450001,China)
出处
《计算机应用研究》
CSCD
北大核心
2018年第8期2506-2509,共4页
Application Research of Computers
基金
海南省自然科学基金资助项目(20166233)
国家教育部博士点基金资助项目(20121101110037)
关键词
目标检测
目标跟踪
视频序列
粒子滤波
背景减除
颜色分布
无监督
鲁棒性
target detection
target tracking
video sequence
particle filter
background subtraction
color distribution
non-surveillance
robustness
