摘要
为了改善雷达对分离后群目标的跟踪性能,提出了一种混合的群目标分离检测与外形估计算法。该算法首先利用单个随机矩阵计算群外形尺寸,然后根据其数值变化分析群的态势;当检测到群分离后,利用k-均值聚类算法对各分群进行聚类,采用最小二乘法将各分群的有效量测拟合成多个椭圆,最后进行航迹关联。该算法的特点是:基于聚类思想处理分群量测,并采用多个椭圆描述群目标的形状;相比于传统随机矩阵算法,估计结果收敛较快。仿真结果表明,相比于传统单个随机矩阵,该混合算法能够估计分离后的群目标形状;相比于半正定规划求解群形状参数,该混合算法的单次蒙特卡洛仿真时间减少了2个量级,证明了算法的有效性。
A hybrid algorithm for group split detection and extension estimation is proposed to improve the performance of tracking the separation of group targets.The traditional random matrix algorithm is employed to calculate the size of a group and to analyze the situation of the group.When a group splitting is detected,big subgroups are divided into some small subgroups by k-means clustering.The least square method is then employed to fit effective measurements of each subgroup into several ellipses,and trajectory association is finally conducted.The proposed algorithm deals with measurements of groups based on the clustering,and describes extensions of splitted subgroups with multiple ellipses according to clustering result.The algorithm has a convergence rate faster than that of the traditional algorithm.Simulation results and a comparison with the single random matrix algorithm show that the proposed algorithm accurately estimates the extensions of splitted subgroups.A comparison with the semi-positive programming algorithm shows that single Monte Carlo simulation time reduces by 2 orders of magnitude.These results prove the validity and practicability of the proposed algorithm.
作者
杜明洋
毕大平
王树亮
潘继飞
DU Mingyang;BI Daping;WANG Shuliang;PAN Jifei(College of Electronic Countermeasure,National University of Defense Technology,Hefei 230037,China)
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2018年第10期116-123,共8页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(61671453)
安徽省自然科学基金资助项目(1608085MF123)
