摘要: |
针对分布式多传感器航迹关联的特点,考虑采用K-means聚类的航迹关联算法。将来自各传感器的局部航迹与系统航迹进行关联,并将系统航迹作为初始聚类中心,避免了K-means算法本身依赖初始值的缺陷;提出将系统航迹与局部航迹的欧式距离以及其状态向量在1范数下的距离之和作为相似度测度;设定距离门限值,减少了极端数据对聚类结果的影响,并增加多义性处理。蒙特卡洛仿真实验表明,该算法在目标密集并且目标有交叉的情况下能以较小的代价得到较高的平均正确关联率。同时,该算法克服了最近邻域法的局部最优特性和关联正确率高度依赖特征阈值等局限性。 |
关键词: 航迹关联 聚类分析 K-means聚类 向量范数 正确关联率 |
DOI: |
|
基金项目:国家自然科学基金资助项目(91338107);四川省科技厅软科学研究项目(2016ZR0087) |
|
A distributed multi-sensor track association algorithm based on K-means clustering |
LI Su,WANG Yunfeng |
() |
Abstract: |
Aiming at the characteristics of distributed multi-sensor track association,a track association algorithm based on K-means clustering is adopted.The local tracks from the sensors are associated with the systems tracks and the system tracks are selected as the initial clustering centers to avoid the defect of the K-means algorithm itself relying on the initial values.The sum of the Euclidean distance and the 1-Norm of state vector between tracks are proposed as the similarity measure.The distance threshold is set to reduce the impact of extreme data on clustering results.And the polysemy processing is increased.Monte Carlo simulation results show that the algorithm can obtain a higher average correlation rate at a lower cost in the case of target-intensive.At the same time,the algorithm overcomes the local optimal characteristics of the nearest neighbor method and the limitation that the correlation accuracy is highly dependent on the feature threshold. |
Key words: track association clustering analysis K-means clustering vector norm correct association rate |