| 摘要: |
| 车辆类型识别方法是智能交通系统的关键技术之一。利用深度学习的高维特征泛化学习能力，将改进的LeNet-5卷积神经网络用于基于交通微波雷达的大小车型分类识别。首先，以雷达触发前的N帧信号为基础，对雷达的回波信号进行分析并构建数据集；然后，分析LeNet-5卷积神经网络的特点；最后提出一种改进的LeNet-5卷积神经网络。实验结果表明，与传统的支持向量机方法相比，所提方法能够智能学习大小车的雷达时频信号特征，大小车型识别准确率达到97%以上，可为交通场景下的车型识别研究提供新的技术途径。 |
| 关键词: 智能交通系统 微波雷达 大小车型分类 深度学习 卷积神经网络 |
| DOI： |
|
| 基金项目:国家自然科学基金资助项目（61671069）；“勤信人才”培育计划（QXTCP A201902） |
|
| A Vehicle Size Classification Method Based on Echo Signal of Microwave Radar |
| CAO Lin,LI Jia,ZHANG Xinyi,WANG Dongfeng,FU Chong |
| (1a.Key Laboratory of Ministry of Education for Optoelectronic Measurement Technology and Instrument;1b.School of Information and Communication Engineering,Beijing Information Science&Technology University,Beijing 100101,China;1a.Key Laboratory of Ministry of Education for Optoelectronic Measurement Technology and Instrument,Beijing Information Science&Technology University,Beijing 100101,China;2.Beijing TransMicrowave Technology Co.,Ltd.,Beijing 100080,China;3.School of Computer Science and Engineering,Northeastern University,Shenyang 110169,China) |
| Abstract: |
| Vehicle type identification is one of the key technologies of intelligent transportation system.In this paper,the improved LeNet-5 convolutional neural network(CNN) is used to classify vehicles with different size based on traffic microwave radars by using the high-dimensional feature generalization learning ability of deep learning.Firstly,the radar echo signal is analyzed and the data set is constructed based on the previous N frame radar trigger signal.Then,the characteristics of LeNet-5 CNN are analyzed.Finally,an improved LeNet-5 CNN is proposed.Experimental results show that the proposed method can intelligently learn the radar time-frequency signal characteristics of the small vehicle compared with the traditional support vector machine method.The recognition accuracy rate of large and small models is more than 97%.It provides a new technical approach for vehicle identification research in traffic scenarios. |
| Key words: intelligent transportation system microwave radar vehicle size classification deep learning convolutional neural network |
