基于卷积神经网络的雷达航迹起始方法

Radar track initiation method based on the convolutional neural network

提出了一种基于卷积神经网络的雷达航迹起始方法。该方法将雷达的航迹起始问题转换为真假航迹的二分类问题,主要包括特征提取器、样本集生成器和分类器等3个主要模块以及训练和测试两个主要过程:特征提取器用于提取目标的运动参数作为分类特征,样本集生成器生成可用于分类的训练样本和测试样本,分类器实现训练和测试的样本分类过程。仿真实验和结果表明,这种基于卷积神经网络的雷达航迹起始方法相比传统航迹起始方法具有更优秀的航迹起始概率和虚警概率,并且具有更好的杂波背景环境适应能力。

A radar track initiation method based on convolutional neural networks.In this way,the radar track initiation problem is formulated to a binary classification of the true and false tracks,consisting of three key modules:feature extractor,sample set generator and classifier,and two main processes:training and testing.Kinetic characteristic parameters of the target are extracted as classification features by the feature extractor,the sample set generator generates training samples and testing samples that can be used for the classification,and the classifier implements the sample classification process.Simulation experiments and results show that the radar track initiation method based on the convolutional neural network has improvement in track initiation detection rate and false alarm rate,compared with traditional track initiation methods,and has better ability to adapt the clutter background environment.