Abstract:
A new algorithm is proposed for small objects tracking, which uses both Gaussian Mixture Model(GMM) and Particle Filter(PF). Considering to the problem of missing in detecting on small objects, the algorithm uses the information of objects feed back by PF to instruct the GMM in background modeling. The results of experiment show that the algorithm can achieve auto-detecting and auto-tracking of multi-objects while overcoming the missing of small objects in former algorithms.
Key words:
background modeling,
Gaussian Mixture Model(GMM),
Particle Filter(PF)
摘要:
给出一种针对小目标跟踪问题的改进算法,该算法将背景建模与粒子滤波相结合对运动目标进行检测跟踪。其中,针对小目标在检测过程中容易出现的漏检问题,算法在进行背景建模时利用粒子滤波反馈的目标运动信息来指导混合高斯背景建模。实验结果表明,该算法能够自动地进行运动目标的跟踪,并且可以克服常规检测算法中小目标的丢失问题。
关键词:
背景建模,
混合高斯模型,
粒子滤波
CLC Number:
ZHOU Ku-Xin, ZHOU Jun, SONG Li, CHEN Li. Tracking Algorithm for Small Objects[J]. Computer Engineering, 2010, 36(16): 186-188.
周圣鑫, 周军, 宋利, 陈立. 一种针对小目标的跟踪算法[J]. 计算机工程, 2010, 36(16): 186-188.