基于多维特征的高超声速目标分类识别算法  

Classification algorithm of hypersonic target based on multi-dimensional characteristics

作　　者：高羽婷 李享 孙鹤枝 尹可 李超[2] 李翰夫 杨冬[1] 李劲东[1] GaoYuting;Li Xiang;Sun Hezhi;Yin Ke;Li Chao;Li Hanfu;Yang Dong;Li Jindong(China Academy of Space Technology,Beijing 100094,China;Harbin Institute of Technology,Harbin 150006,China)

机构地区：[1]中国空间技术研究院,北京100094 [2]哈尔滨工业大学,哈尔滨150006

出　　处：《战术导弹技术》2023年第2期81-87,共7页Tactical Missile Technology

摘　　要：针对高超声速目标天基探测中分类识别困难的问题,提出了一种基于多维红外特征的高超声速目标分类识别算法。在对主动巡航式和被动滑翔式两类不同机动类型的典型高超声速目标多维红外特征进行对比研究的基础上,结合高斯混合模型进行机器学习聚类分析,开发了基于多维特征的高超声速目标分类识别算法,构建了高超声速目标多维特征混合识别模型。基于天基探测全链路模型仿真的数据进行测试验证,其结果证明了所提出的高超声速目标分类识别方法的有效性。在检测到的高超声速目标运动特性未知的情况下,该算法仅通过多维红外特征可快速对两型高超声速目标(主动巡航式/被动滑翔式)进行分类识别,为后续跟踪和拦截高超声速目标争取了时间。Focusing on the difficulty in the space-based detection of hypersonic targets,a classification algorithm of hypersonic targets is proposed based on multi-dimensional infrared characteristics under spacebased observation conditions.Comparative research on multi-dimensional characteristics of typical hypersonic targets with two different maneuver types(active cruise type and passive glide type)is carried out.On this basis,a machine learning clustering analysis is conducted with Gaussian mixture model to develop a classification algorithm of hypersonic targets based on multi-dimensional characteristics.A classification model has been introduced,which is capable of rapid classification of typical hypersonic targets.By using the simulation data from full-link space-based detection model,the validation results indicate the effectiveness of the proposed classification algorithm of hypersonic targets based on multi-dimensional characteristics.This algorithm is able to classify typical hypersonic targets with two different maneuver types(active cruise type and passive glide type)by only depending on the multi-dimensional infrared characteristics and without considering the kinetic characteristics of the targets,which makes time for the later tracking and intercepting of hypersonic targets.

关 键 词：高超声速目标 天基探测 多维特征 目标分类识别 聚类算法 高斯混合模型 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]