基于混合注意力CNN的角反射器阵列部署

Deployment of corner reflector arrays based on a hybrid attention convolutional neural network

  • 摘要:
    目的 针对海面角反射器(简称角反)阵列排布问题,提出一种基于混合注意力卷积神经网络(CNN)的优化角反阵列的部署方法。以提升角反对复杂目标的干扰能力,增强海上作战平台的电磁防护效能。
    方法 首先,采用弹跳射线法对不同排布下的角反阵列几何模型进行电磁散射特性求解,得到一维距离像;其次,利用CLEAN算法从一维距离像中提取散射中心,构建包含散射中心与雷达入射角的数据集;然后,搭建融合混合注意力机制CNN的角反阵列排布预测网络模型,将数据集输入模型进行训练,以实现对角反排布的智能预测;最后,测试网络模型对输入不同的舰船一维距离像的预测能力,并将原始舰船与预测阵列排布的一维距离像进行对比分析。
    结果 结果显示,该网络的训练损失为0.00027,且准确率相比原始CNN网络提升5.52%,表明模型具有较高的预测精度。在舰船目标简单散射特征场景下,由该方法布设的角反阵列一维距离像与舰船呈现高相关性,皮尔逊相关系数达到0.9137;在少数复杂情况下,该方法仍能借助耦合散射中心实现干扰规划。
    结论 研究证明基于混合注意力CNN的角反阵列部署方法能够有效优化角反的排布,提升其对目标的干扰能力,可为海上作战平台的电磁防护提供新的技术途径。

     

    Abstract:
    Objective This study proposes an optimization method for the deployment of corner reflector arrays based on a hybrid attention CNN (convolutional neural network), aiming to address the deployment challenges of such arrays on the sea surface. The objective is to enhance the electromagnetic interference capability of corner reflectors against complex targets and to improve the electromagnetic protection effectiveness of maritime combat platforms.
    Method First, the electromagnetic scattering characteristics of corner reflector arrays under various geometric arrangements were analyzed using the shooting and bouncing ray (SBR) method, generating one-dimensional range profiles. Next, the CLEAN algorithm is employed to extract scattering centers from these profiles, producing a dataset containing scattering center positions and radar incidence angles. A predictive model for corner reflector array deployment is then constructed by integrating a CNN with a hybrid attention mechanism. The dataset is fed into the model for training, enabling intelligent prediction of corner reflector deployment. Finally, the model's predictive performance is tested using various ship range profiles, and the predicted corner reflector arrays are compared with actual range profiles to assess prediction accuracy.
    Results The results show that the network achieves a training loss of 0.00027 and improves accuracy by 5.52% compared to the original CNN model, demonstrating its superior prediction accuracy. In scenarios with simple ship scattering characteristics, the one-dimensional range profile of the corner reflector array generated by this method shows a high correlation with that of the ship, with a Pearson correlation coefficient of 0.9137. Even in certain complex scenarios, this method can still formulate effective interference strategies by exploiting the effects of coupled scattering centers.
    Conclusion The research demonstrates that the corner reflector array deployment method based on a hybrid attention CNN effectively optimizes the arrangement of corner reflectors, enhances the ability to interfere with radar detection of targets, and provides a novel technical approach for the electromagnetic protection of maritime combat platforms.

     

/

返回文章
返回