采用卷积神经网络的海面多目标检测研究
|
摘要
对海雷达多目标检测在军事领域有着重要的应用价值。为了提高海杂波下的目标检测性能,减少临近目标的影响,本文引入深度学习思想,提出了采用卷积神经网络的多目标检测方法。通过雷达实测数据的分析与训练,构造适用于处理一维回波数据的网络模型,引入定向惩罚技术加快自适应学习效率,优化网络超参数提升网络性能,实现了回波数据信杂比的较大改善,完成了海面多目标的有效检测。最后,基于实测数据对该方法进行了性能验证,实验结果显示了本方法的有效性。
Multi-target detection in the sea clutter of shipborne radar has significant value for the military. In order to improve target detection performance under sea clutter,and reduce the negative influence of adjacent targets,this paper compares technologies used for target detection,introduces deep learning idea into this field and proposes a multi-target detection model based on convolutional neural network. Through collection and analysis of the real data in radar,a network structure suitable for processing one-dimensional echo data was constructed by structure searching. Directional penalty method was developed to speed up the learning efficiency,and the network hyper-parameters were optimized to improve network performance,thus raised the signal to clutter ratio and achieved effective detection of multiple targets on the sea. Finally,the performance is verified based on the measured data by receiver operating characteristic curve and the signal to clutter ratio improvement factor,which shows the effectiveness of this model.
Multi-target detection in the sea clutter of shipborne radar has significant value for the military. In order to improve target detection performance under sea clutter,and reduce the negative influence of adjacent targets,this paper compares technologies used for target detection,introduces deep learning idea into this field and proposes a multi-target detection model based on convolutional neural network. Through collection and analysis of the real data in radar,a network structure suitable for processing one-dimensional echo data was constructed by structure searching. Directional penalty method was developed to speed up the learning efficiency,and the network hyper-parameters were optimized to improve network performance,thus raised the signal to clutter ratio and achieved effective detection of multiple targets on the sea. Finally,the performance is verified based on the measured data by receiver operating characteristic curve and the signal to clutter ratio improvement factor,which shows the effectiveness of this model.
引文
[1]曾浩,李洁,鉴福升.强海杂波环境下目标检测方法对比分析[J].雷达科学与技术,2015,13(1):33-36.Zeng Hao,Li Jie,Jian Fusheng.Comparative Analysis of Target Detection Methods in Strong Sea Clutter Background[J].Radar Science and Technology,2015,13(1):33-36.(in Chinese)
[2]李琪.海杂波背景下的目标检测算法[D].西安:西安电子科技大学,2013.Li Qi.Methods of Target Detection in the Sea Clutter Environment[D].Xi’an:Xidian University,2013.(in Chinese)
[3]Fernandez M,Martínez T,Vidal B.CA-CFAR Adjustment Factor Correction with a priori Knowledge of the Clutter Distribution Shape Parameter[J].International Journal of Interactive Multimedia and Artificial Intelligence,2017,4(4):7-13.
[4]刘瑜.基于自适应曲面拟合的红外图像背景估计[J].科技创新导报,2012,4:38-40.Liu Yu.Infrared Image Background Estimation Based on Adaptive Surface Fitting[J].Science and Technology Innovation Herald,2012,4:38-40.(in Chinese)
[5]Liu M,Shi J,Li Z,et al.Towards better analysis of deep convolutional neural networks[J].IEEE Transactions on Visualization and Computer Graphics,2017,23(1):91-100.
[6]洪新海,宋彦.结合CNN不同层信息的全变量建模人脸特征表达学习方法[J].信号处理,2017,33(8):1073-1081.Hong Xinhai,Song Yan.Total Variability Modeling for Face Representation Learning with CNN Multi-Layers[J].Journal of Signal Processing,2017,33(8):1073-1081.(in Chinese)
[7]杨予昊,孙晶明,虞盛康,等.基于卷积神经网络的高分辨距离像目标识别[J].现代雷达,2017,39(12):24-28.Yang Yuhao,Sun Jingming,Yu Shengkang,et al.High Resolution Range Profile Target Recognition Based on Convolutional Neural Network[J].Modern Radar,2017,39(12):24-28.(in Chinese)
[8]邹浩,林赟,洪文.采用深度学习的多方位角SAR图像目标识别研究[J].信号处理,2018,34(5):513-522.Zou Hao,Lin Yun,Hong Wen.Research on Multi-Aspect SAR Images Target Recognition Using Deep Learning[J].Journal of Signal Processing,2018,34(5):513-522.(in Chinese)
[9]Hornik K,Stinchcombe M,White H.Multilayer feedforward networks are universal approximators[J].Neural Networks,1989,2(5):359-366.
[10]Goodfellow I,Bengio Y,Courville A,et al.Deep learning[M].Cambridge:MIT Press,2016.
[11]Prusa J D,Khoshgoftaar T M.Improving deep neural network design with new text data representations[J].Journal of Big Data,2017,4(1):7-22.
[12]Sutskever I,Martens J,Dahl G,et al.On the importance of initialization and momentum in deep learning[C]∥International Conference on Machine Learning,2013:1139-1147.
[13]Scardapane S,Comminiello D,Hussain A,et al.Group sparse regularization for deep neural networks[J].Neurocomputing,2017,241(12):81-89.
[14]Glorot X,Bengio Y.Understanding the difficulty of training deep feedforward neural networks[C]∥Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics,2010:249-256.
[2]李琪.海杂波背景下的目标检测算法[D].西安:西安电子科技大学,2013.Li Qi.Methods of Target Detection in the Sea Clutter Environment[D].Xi’an:Xidian University,2013.(in Chinese)
[3]Fernandez M,Martínez T,Vidal B.CA-CFAR Adjustment Factor Correction with a priori Knowledge of the Clutter Distribution Shape Parameter[J].International Journal of Interactive Multimedia and Artificial Intelligence,2017,4(4):7-13.
[4]刘瑜.基于自适应曲面拟合的红外图像背景估计[J].科技创新导报,2012,4:38-40.Liu Yu.Infrared Image Background Estimation Based on Adaptive Surface Fitting[J].Science and Technology Innovation Herald,2012,4:38-40.(in Chinese)
[5]Liu M,Shi J,Li Z,et al.Towards better analysis of deep convolutional neural networks[J].IEEE Transactions on Visualization and Computer Graphics,2017,23(1):91-100.
[6]洪新海,宋彦.结合CNN不同层信息的全变量建模人脸特征表达学习方法[J].信号处理,2017,33(8):1073-1081.Hong Xinhai,Song Yan.Total Variability Modeling for Face Representation Learning with CNN Multi-Layers[J].Journal of Signal Processing,2017,33(8):1073-1081.(in Chinese)
[7]杨予昊,孙晶明,虞盛康,等.基于卷积神经网络的高分辨距离像目标识别[J].现代雷达,2017,39(12):24-28.Yang Yuhao,Sun Jingming,Yu Shengkang,et al.High Resolution Range Profile Target Recognition Based on Convolutional Neural Network[J].Modern Radar,2017,39(12):24-28.(in Chinese)
[8]邹浩,林赟,洪文.采用深度学习的多方位角SAR图像目标识别研究[J].信号处理,2018,34(5):513-522.Zou Hao,Lin Yun,Hong Wen.Research on Multi-Aspect SAR Images Target Recognition Using Deep Learning[J].Journal of Signal Processing,2018,34(5):513-522.(in Chinese)
[9]Hornik K,Stinchcombe M,White H.Multilayer feedforward networks are universal approximators[J].Neural Networks,1989,2(5):359-366.
[10]Goodfellow I,Bengio Y,Courville A,et al.Deep learning[M].Cambridge:MIT Press,2016.
[11]Prusa J D,Khoshgoftaar T M.Improving deep neural network design with new text data representations[J].Journal of Big Data,2017,4(1):7-22.
[12]Sutskever I,Martens J,Dahl G,et al.On the importance of initialization and momentum in deep learning[C]∥International Conference on Machine Learning,2013:1139-1147.
[13]Scardapane S,Comminiello D,Hussain A,et al.Group sparse regularization for deep neural networks[J].Neurocomputing,2017,241(12):81-89.
[14]Glorot X,Bengio Y.Understanding the difficulty of training deep feedforward neural networks[C]∥Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics,2010:249-256.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |