摘要
文中提出了一种基于动态聚类的地面移动目标低空跟踪方法。该方法根据目标跟踪的实际情况选择分类特征,确定聚类准则函数,采用动态聚类方法,依次对像平面上的象素进行子窗口聚类,然后对子窗口内的象素进行运动区域聚类,最后对运动区域的象素进行目标聚类,从而比较精确地获得被跟踪目标的重心位置,并将其作为下一帧的聚类中心。由此,目标跟踪的过程变成一个动态聚类的过程。该方法是一个非监督学习方法,简单实用,仿真实验也证明了该方法的有效性。
A low altitude tracking method for the ground moving target based on dynamic clustering is put forward in this paper.The classification characters are chosen and the clustering rule functions are constructed according to the actual situation of target tracking.By dynamic clustering,some pixels in the image plane are clustered into a sub-window and some pixels in the sub-window are clustered into a motion region.The exact gravity center of the tracked target is thus obtained.The gravity center of the tracked target is used as the clustering center in the next frame.Therefore the target tracking process becomes a dynamic clustering process.The simple and practical method is an unsupervised learning means.The simulation results show that the method is effective.
出处
《电子机械工程》
2012年第1期51-54,58,共5页
Electro-Mechanical Engineering
基金
上海市自然科学基金资助项目(10ZR1411500)
上海大学科技创新基金资助项目
关键词
动态聚类
聚类准则函数
目标跟踪
非监督学习方法
dynamic clustering
clustering rule function
target tracking
unsupervised learning means
