| 引用本文: | 邴 政， 周良将， 董书航， 温智磊. 一种基于微多普勒和机器学习的无人机目标分类识别技术[J]. 雷达科学与技术, 2024, 22(5): 549-556.[点击复制] |
| BING Zheng, ZHOU Liangjiang, DONG Shuhang, WEN Zhilei. A Classification and Recognition Method for UAV Targets Based on Micro⁃Doppler and Machine Learning[J]. Radar Science and Technology, 2024, 22(5): 549-556.[点击复制] |
|
| 摘要: |
| 近年来，随着科技发展，无人机行业高速发展，空域的情况变得愈加复杂，消费级无人机的入侵事件日益增多，对无人机防控系统的需求越来越迫切。飞鸟和无人机都是典型的“低慢小”目标，二者的回波特征、运动模式都有极其相似之处，如何对二者进行有效的观测、区分和追踪已成为保障空中航路安全研究中的重要问题。本文针对传统方法难以将二者区分的问题，首先利用仿真建立无人机和飞鸟的数学及物理模型，并对二者的微动特征进行提取，利用卷积神经网络和集成学习方法对其进行区分，在此基础上对实测数据进行识别，证明了本文方法的有效性。 |
| 关键词: 微多普勒 机器学习 目标识别 卷积网络 集成学习 |
| DOI：DOI:10.3969/j.issn.1672-2337.2024.05.010 |
| 分类号:TN957.51 |
| 基金项目:国家民用航天项目资助（No.D040304） |
|
| A Classification and Recognition Method for UAV Targets Based on Micro⁃Doppler and Machine Learning |
|
BING Zheng, ZHOU Liangjiang, DONG Shuhang, WEN Zhilei
|
|
1. Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, China;2. National Key Laboratory of Microwave Imaging Technology, Beijing 100190, China;3. School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences, Beijing 100049, China;4. Qilu Aerospace Information Research Institute, Chinese Academy of Sciences, Jinan 250132, China
|
| Abstract: |
| The drone has been a key threat as its infiltration events are incredibly increasing, due to the rapid development of the drone industry, which makes it a critical demand to develop drone prevention systems. Birds and drones are both classic objects that have low attitude, slow speed and small scale, leading to similar echo characteristics and movement patterns. To ensure the safety of airlines, it has been an important issue to observe, separate and track them two. To solve the problem, firstly, the mathematical and physical model of the drone and the bird are established by simulation, and the micro?motion characteristics of both are extracted. Then, the two kinds of objects are distinguished using convolutional neural network and ensemble learning methods. Based on which, the real field data are processed, and the result has proved the validity of the method. |
| Key words: micro⁃Doppler machine learning object recognition convolutional neural network ensemble learning |
