摘要
有效的目标跟踪需要积极的传感器节点对运动目标群实行跟踪。与单目标跟踪相比,聚类在效能上有显著提高。本文提出准确的相干和非相干运动模式下目标的聚类,采用隐式动态时间框架来评估在创建连接组件加权图的目标关系史。该算法采用目标跟踪中定位算法的关键特征,即估计当前和预测的位置来确定移动目标的方向和距离的关系。模拟结果显示,通过动态调整历史窗口大小和预测目标之间的关系,可以显著提高聚类的准确性并减少运算时间。
Effective target tracking needs sensor node positive for the moving object clusters follow. Compared with single target tracking, clustering has significant improvement in performance. This paper presents accurate coherent and non coherent motion mode target clustering, using the implicit dynamic time framework to evaluate in the weighted graph to create the connection component of the target relation history. The proposed algorithm employs key features of localization algorithms in target tracking, namely, estimated current and predicted locations to determine the relational directions and distances of moving targets. Our simulation results show that by dynamically adjusting the relationship between historical window size and predict the target can significantly improve the clustering accuracy and reduce the computation time.
出处
《舰船科学技术》
北大核心
2015年第5期95-99,共5页
Ship Science and Technology
基金
国家自然科学基金资助项目(61303192)
关键词
聚类算法
目标跟踪
无线传感器网络
加权图
clustering algorithms
target tracking
wireless sensor networks
weighted graph