传统的恒虚警(CFAR)检测器虽然在多数目标场景中具有不错的检测效果,但在强杂波背景下其检测性能会下降。基于深度学习的雷达目标检测方法可以提取深度特征用于目标检测,然而深度学习算法耗时长,很难应用于实际场景。针对上述问题,文中提出了一种基于深度神经网络的两步检测法实现海杂波背景下的目标检测。首先将雷达回波的距离-多普勒图以窗口滑动的方式截取小尺寸样本,通过全连接层网络(FCNN)进行初步检测,然后在检测区域截取更大尺寸的样本用以深度特征提取,通过多层卷积神经网络(CNN)区分目标与杂波,得到最终的检测结果。实验结果表明,与传统CFAR检测器和基于CNN的一步检测方法相比,文中所提方法保持较高检测概率的同时,能有效提升检测效率。

Although the traditional constant false alarm rate (CFAR) detector has a good detection effect in most target scenes, its detection performance will decrease in the strong clutter background. The radar target detection method based on deep learning can extract deep features for target detection. However, the deep learning algorithm is time-consuming and difficult to be applied to practical scenarios. To address this issue, in this paper a two-step detection method based on deep neural network is proposed to achieve target detection in the background of sea clutter. First, the range-Doppler map of the radar echo is used to intercept small-sized samples in a sliding window, and the fully connected layer network (FCNN) is used for preliminary detection, and then larger-sized samples are intercepted in the detection area for deep feature extraction. The target and clutter are distinguished through a multi-layer convolutional neural network (CNN), and the final detection result is obtained. The experimental results show that compared with the traditional CFAR detector and the one-step detection method based on CNN, the method proposed in this paper can effectively improve the detection efficiency while maintaining a higher detection probability.

TN957. 51