为解决强海杂波条件下虚警率高、杂波多普勒较宽、信杂比低或低速目标落入杂波多普勒通道时海面目标难以检测的难题,提出了基于深度卷积网络(Faster R-CNN)的海面目标检测算法。利用深度卷积神经网络自动提取特征的能力,对输入含有目标的海面回波样本进行一系列非线性操作,逐层提取样本中目标抽象的特征;然后利用提取的特征对未知目标样本进行检测和定位,检测是否含有目标以及目标的位置。最后在实测南非海杂波数据集上进行实验验证,所提方法在虚警率为10-3 时,海面目标的检测率高达57. 98??,比传统的恒虚警率检测提高约28??,比稀疏可调Q 小波变换检测方法提高了21??,验证了该方法的准确性和有效性,为海面目标检测提供了新的技术途径。

In order to solve the problem that it is difficult to detect the marine target when the target under the condition of strong sea clutter with high false alarm rate, wide clutter Doppler, low signal-to-clutter ratio or target with low speed falls into the clutter Doppler channel, a marine target detection algorithm based on the deep convolution network (Faster R-CNN) is proposed. Using the capability of deep convolutional neural network to extract features automatically, a series of nonlinear operations are performed on the input marine echo samples containing targets and the abstract features of the targets in the samples are extracted layer by layer. Then the extracted features are used to detect and locate the unknown target samples to detect whether they contain the target and the location of the target. Finally, the experimental verification is carried out on the actually measured the council for scientific and industrial research (CSIR) sea clutter data set. When the false alarm rate is 10-3, the detection rate of marine target is as high as 57. 98?? with the proposed method, which are about 28% and 21% higher than the traditional constant false alarm ratio (CFAR) detection and the sparse tunable Q-factor wavelet transform (TQWT) detection method. The improvement proves the accuracy and effectiveness of the method and provides a new technical approach for marine target detection.

TN959