Bayes-SAR Net: Robust SAR Image Classification with Uncertainty Estimation Using Bayesian Convolutional Neural Network
Bayes-SAR Net: Robust SAR Image Classification with Uncertainty Estimation Using Bayesian Convolutional Neural Network
复制标题
DOI：
10.1109/radar42522.2020.9114737
复制
发表时间：
2020-04
期刊：
影响因子：
--
通讯作者：
Dimah Dera;G. Rasool;N. Bouaynaya;Adam Eichen;Stephen Shanko;J. Cammerata;S. Arnold
中科院分区：
文献类型：
--
作者：
Dimah Dera;G. Rasool;N. Bouaynaya;Adam Eichen;Stephen Shanko;J. Cammerata;S. Arnold
合成孔径雷达(SAR)图像分类是一个具有挑战性的问题，其成像机理复杂，且存在随机相干斑噪声，影响雷达图像的解译。最近，卷积神经网络(CNN)被证明在计算机视觉任务中表现出比以前最先进的技术更好的性能，因为它们能够从数据中学习相关特征。然而，神经网络，特别是神经网络，通常缺乏不确定性量化，很容易被对手攻击欺骗。本文提出了贝叶斯-合成孔径雷达网络，这是一种贝叶斯网络，它可以进行稳健的SAR图像分类，同时量化网络在决策中的不确定性或置信度。贝叶斯-合成孔径雷达网络传播给定数据的网络参数的近似后验分布的前两个矩(均值和协方差)，并获得分类输出的预测均值和协方差。使用基准数据集Flevoland和Oberpfaffenhofen进行的实验表明，与SAR-Net同源数据相比，该方法具有更好的性能和对高斯噪声和对手攻击的鲁棒性。在对抗性扰动的情况下，贝叶斯-合成孔径雷达网络的测试准确率提高了约10%(级别≽0.05)。
Synthetic aperture radar (SAR) image classification is a challenging problem due to the complex imaging mechanism as well as the random speckle noise, which affects radar image interpretation. Recently, convolutional neural networks (CNNs) have been shown to outperform previous state-of-the-art techniques in computer vision tasks owing to their ability to learn relevant features from the data. However, CNNs in particular and neural networks, in general, lack uncertainty quantification and can be easily deceived by adversarial attacks. This paper proposes Bayes-SAR Net, a Bayesian CNN that can perform robust SAR image classification while quantifying the uncertainty or confidence of the network in its decision. Bayes-SAR Net propagates the first two moments (mean and covariance) of the approximate posterior distribution of the network parameters given the data and obtains a predictive mean and covariance of the classification output. Experiments, using the benchmark datasets Flevoland and Oberpfaffenhofen, show superior performance and robustness to Gaussian noise and adversarial attacks, as compared to the SAR-Net homologue. Bayes-SAR Net achieves a test accuracy that is around 10% higher in the case of adversarial perturbation (levels ≽ 0.05).