基于SNN-AP聚类的扩展目标量测集划分方法

摘要

To solve the multiple extended target measurement set partitioning problem in cluttered environment when the densities of measurements are different,this paper introduced a new clustering method based on affinity propagation (AP) and proposed a novel measurement partition algorithm.Firstly,it adopted local outlier factor detection to preprocess the measurement set,and removed the clutter by threshold method,at the same time to solve the large density difference of measurement set of extended target tracking problem,it introduced a similarity measure based on shared nearest neighbor(SNN) method,considering the influence of surrounding measurement,the algorithm found the clustering center gradually through iteration with the two information,avoiding the choice of initial clustering number.Simulation results show that,this algorithm reduces the operation time with tiny effect on tracking performance compared with the traditional measurement set partitioning methods.%针对杂波环境下且量测密度差别较大的多扩展目标量测集划分问题,引入近邻传播聚类技术,提出了一种新的量测集划分算法.该算法首先采用局部异常因子检测对量测为杂波的程度进行度量,通过设定阈值的方法进行杂波滤除;同时对于目标量测密度差别较大的问题,引入一种基于共享最近邻的相似度度量方法;考虑了周围量测的影响,通过迭代传递两个信息量逐步寻找聚类中心,避免了对初始聚类个数的选择.仿真实验表明,与传统量测集划分算法相比,所提算法在保证扩展目标跟踪性能的同时,有效减少了算法的运算时间.