ID 原文 译文
18805 仿真结果显示，在适中的BRD量测误差和观测站位置误差下，所提算法的目标定位性能能够达到CRLB。 Simulation results show that the proposed algorithm can achieve the CRLB in a moderate level of noises.
18806 针对地磁背景下磁偶极子目标跟踪过程中存在的地磁干扰与模型非线性的问题，该文提出一种基于差量磁异常的蒙特卡洛卡尔曼滤波(MCKF)跟踪方法。 In order to solve the problem of geomagnetic interference and model nonlinearity in the trackingprocess of magnetic dipole under geomagnetic background, Monte Carlo Kalman Filter (MCKF) trackingmethod based on differential magnetic anomaly is proposed in this paper.
18807 新的跟踪方法以传感器阵列测量磁场的差量作为观测信号，并利用蒙特卡洛卡尔曼滤波算法解决模型的非线性问题，实现磁偶极子目标的实时跟踪。 The new tracking method takes thedifference of magnetic field measured by sensor array as the observation signal, and uses Monte Carlo KalmanFiltering (MCKF) algorithm to solve the nonlinear problem of the model to realize the real-time tracking ofmagnetic dipole targets.
18808 通过仿真跟踪实验，结果表明该文算法较传统的扩展或无迹卡尔曼滤波算法在稳定跟踪过程中对目标特征参数的估计更精确； The simulation results show that the proposed method is more accurate than thetraditional Extended Kalman Filter (EKF) or Untracked Kalman Filter (UKF) in the stable tracking process.
18809 通过地磁背景跟踪实验，结果验证了该文算法较传统算法在低信噪比下的性能优势。 The results of real geomagnetic background tracking experiments show that the proposed algorithm has bettertracking performance under low SNR.
18810 针对目前基于度量学习的小样本方法存在特征提取尺度单一，类特征学习不准确，相似性计算依赖标准度量等问题，该文提出多级注意力特征网络。 Existing few-shot methods have problems that feature extraction scale is single, the learned classrepresentations are inaccurate, the similarity calculation still relies on standard metrics. In order to solve the above problems, multi-level attention feature network is proposed.
18811 首先对图像进行尺度处理获得多个尺度图像；其次通过图像级注意力机制融合所提取的多个尺度图像特征获取图像级注意力特征； Firstly, the multiple scale images areobtained by scale processing, the features of multiple scale images are extracted and the image-level attentionfeatures are obtained by the image-level attention mechanism to fusion them.
18812 在此基础上使用类级注意机制学习每个类的类级注意力特征。 Then, class-level attentionfeatures are learned by using the class-level attention mechanism.
18813 最后通过网络计算样本特征与每个类的类级注意力特征的相似性分数来预测分类。 Finally, the classification is performed byusing the network to compute the similarity scores between features.
18814 该文在Omni-glot和MiniImageNet两个数据集上验证多级注意力特征网络的有效性。 The proposed method is evaluated on the Omniglot dataset and the MiniImagenet dataset.