基于DnCNN的海面目标一维距离像识别方法

针对低信噪比条件下海面目标分类识别精度差的问题,该文提出了一种基于去噪卷积神经网络(Denoising convolutional neural network,DnCNN)的海面目标高分辨一维距离像(High Resolution Range Profile,HRRP)识别方法。所提方法设计了一个海面目标分类识别模型,该模型通过其中的降噪模块提高信噪比。首先,分析了HRRP和二维图像的相似特性,将HRRP降噪转变为二维图像降噪。其次,利用深层次卷积层与批归一化层相结合的结构,提取图像深层次的噪声特征,最后采用残差学习技术,减轻深层次网络的学习负担的同时重构图像进行分类识别。实验结果表明,该模型可以有效提升低信噪比条件下的海面目标分类识别正确率,在不同信噪比条件下其识别性能均优于对比模型,具有良好的识别性能和鲁棒性。 

One-dimensional range profile recognition method of sea-surface targets based on DnCNN

Aiming at the problem of poor target classification and recognition accuracy under low signal-to-noise ratio(SNR) conditions, a high resolution one-dimensional range profile(HRRP) recognition method for sea-surface targets based on denoising convolutional neural networks(DnCNN) is proposed. The proposed method designed a sea surface target classification and recognition model, which improves the signal-to-noise ratio through the noise reduction module. First, the similar characteristics of HRRP and two-dimensional images are analyzed, and HRRP noise reduction is transformed into two-dimensional image noise reduction. Secondly, the deep-level convolutional layer and the batch normalization layer are combined to extract the deep-level noise features of the image, and finally the residual learning technology is used to reduce the learning burden of the deep network while reconstructing the image for classification and recognition. Experimental results show that the model can greatly improve the accuracy of sea-surface target classification and recognition under the condition of low signal-to-noise ratio. What’s more, with the features of good recognition performance and robustness, its recognition performance is also better than that of contrast model under different conditions of SNR.