基于动态时间规整下密度聚类的轨迹识别研究  

Research on Improvement of DTW Maneuver Trajectory Recognition Based on DBSCAN

作　　者：吴达[1] 吕锐 杨宇 邓建军[1] 张保山 WU Da;LYU Rui;YANG Yu;DENG Jian-jun;ZHANG Bao-shan(Air and Missile Defense college,Air Force Engineering University,Xi’an 710051,China;Unit 92095 of the PLA,Taizhou 318050,China)

机构地区：[1]空军工程大学防空反导学院,西安710051 [2]解放军92095部队,福建台州318050

出　　处：《火力与指挥控制》2021年第10期73-78,共6页Fire Control & Command Control

摘　　要：为使地空导弹兵模拟训练系统获得的目标机动轨迹更加贴近实战、更有依据性,考虑对演习训练产生的轨迹数据进行分析,设计了DBSCAN(Density-Based Spatial Clustering of Applications with Noise,针对有噪声数据的基于密度的空间聚类)预分类的DBSCAN改进DTW(Dynamic Time Warping,动态时间归整)聚类算法来分析机动轨迹。并与未预分类的DBSCAN改进DTW聚类算法进行对比,发现其运行效率和分类准确度方面均较优。同时分析了数据结构和算法的参数对分类效果的影响,发现数据规模越大,DBSCAN预分类的DBSCAN改进DTW聚类算法的优势越明显。通过仿真得到了使DBSCAN预分类的DBSCAN改进DTW聚类算法发挥最高效率的参数。The DBSCAN(Density-Based Spatial Clustering of Applications with Noise)pre-classified DBSCAN improved DTW(Dynamic Time Warping)clustering algorithm is desinged to analyze maneuvering trajectory data generated in exercises and tranings.The purpose of that is to make the target maneuver trajectory obtained by the ground-to-air missile force simulation training system more close to actual combat and with more bases.The DBSCAN pre-classified DBSCAN improved DTW clustering algorithm is compared with the unpre-classified DBSCAN improved DTW clustering algorithm and it is found that the operating efficiency and classification accuracy of the former are better.Moreover,the influences of data structure and algorithm parameters on the classification effect are analyzed,and the proposed algorithm has more obvious advantages for the larger-scale data.As a result,the parameters with the highest efficiency in DBSCAN pre-classified DBSCAN improved DTW clustering algorithm are obtained through simulation.

关 键 词：机动识别 DBSCAN聚类 DTW距离度量 

分 类 号：TJ765.3[兵器科学与技术—武器系统与运用工程]