基于卷积神经网络的动目标检测

Moving Target Detection Based on Convolutional Neural Network

合成孔径雷达(SAR)动目标检测技术是雷达信号处理领域中的重要技术。文中利用深度学习高维特征提取能力,提出了一种基于卷积神经网络(CNN)的多通道SAR地面动目标检测算法,并针对雷达实测数据较少、动目标样本难以获得的问题,提出了基于仿真-实测混合样本集的网络训练方法完成网络的高精度训练。实测数据检测结果表明,此类方法能够有效地完成地面动目标检测,与传统动目标检测方法相比,具有显著的优势。

Synthetic aperture radar(SAR)moving target detection is a key technology in radar signal processing.In this paper,a multi-channel SAR ground moving target detection algorithm based on convolutional neural network(CNN)is proposed by using deep learning high-dimensional feature extraction capability.Aiming at the problem that radar measured data is less and moving target samples are difficult to obtain,a network training method based on simulation-measured mixed sample set is proposed to complete the high-precision training of the network.The measured data detection results show that this method can effectively complete the ground moving target detection,has significant advantages compared with the traditional moving target detection method.