飞行态势感知中目标分群方法研究

发布时间：2018-11-16 10:26

【摘要】：目标分群作为态势感知需要解决的重点和难点,是确定目标实体之间关系的重要依据,是后期数据融合系统实现的基础。本文主要研究目标分群的算法,并针对其缺陷给出其算法改进思路与具体实现。论文的主要工作如下::首先,在深入分析数据融合、态势评估与态势感知三者按层次结构具体化的关系基础上,指出目标分群在态势感知存在的重点与难点问题,研究现有的目标分群方法,着重分析解决目标分群算法之一的层次聚类算法。其次,分析了经典层次聚类算法的特点与功能,分析层次聚类算法中Rock算法、Cure算法和Chameleon算法在相似度计算中的优劣,对在相似度计算中有明显优势的Chameleon算法进行分析,详细研究了其基本概念、数学模型、实现流程,通过理论分析与仿真实验发现了Chameleon算法的局限性。最后,从Chameleon算法的局限性入手,按照算法两阶段的流程,将基于密度峰值的DPC算法引入Chameleon算法第一阶段,将社团结构学科引入Chameleon算法第二阶段,对于算法的两阶段进行改进,提出了可用于解决目标分群问题的Chameleon改进算法。针对Chameleon改进算法,详细介绍了其算法模型、算法流程,给出了具体的仿真实验。实验结果表明,Chameleon改进算法较传统Chameleon算法具有对输入参数敏感程度低且可以处理多形状数据集的特点。
[Abstract]:As the focus and difficulty of situational awareness, target clustering is an important basis to determine the relationship between target entities and the basis of the later data fusion system. In this paper, the algorithm of target clustering is studied, and the improvement and realization of the algorithm are given. The main work of this paper is as follows: firstly, on the basis of in-depth analysis of the relationship among data fusion, situation assessment and situational awareness according to hierarchical structure, this paper points out the key and difficult problems of target clustering in situational awareness. The existing target clustering methods are studied, and the hierarchical clustering algorithm, which is one of the target clustering algorithms, is analyzed. Secondly, the characteristics and functions of classical hierarchical clustering algorithm are analyzed. The advantages and disadvantages of Rock algorithm, Cure algorithm and Chameleon algorithm in similarity calculation are analyzed. The Chameleon algorithm, which has obvious advantages in similarity calculation, is analyzed. The basic concept, mathematical model and implementation flow are studied in detail. The limitations of Chameleon algorithm are found through theoretical analysis and simulation experiments. Finally, starting with the limitation of the Chameleon algorithm, according to the two-stage flow of the algorithm, the DPC algorithm based on the peak density is introduced into the first stage of the Chameleon algorithm, and the discipline of community structure is introduced into the second stage of the Chameleon algorithm, and the two stages of the algorithm are improved. An improved Chameleon algorithm is proposed to solve the problem of target clustering. For the improved Chameleon algorithm, the algorithm model and algorithm flow are introduced in detail, and the specific simulation experiments are given. Experimental results show that the improved Chameleon algorithm is less sensitive to input parameters than the traditional Chameleon algorithm and can deal with multi-shape data sets.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：E926.4;TP301.6

【参考文献】