ID 原文 译文
17925 首先，构建了4尺度的特征金字塔网络分别独立预测目标，补充高分辨率细节特征。 Firstly, a four scales feature pyramid network is constructed to predict object independently and supplement detail features with higher resolution.
17926 其次，在特征金字塔结构的横向连接中融入注意力模块，产生显著性特征，抑制不相关区域的特征响应、突出图像局部目标特征。 Secondly, attention module is integrated into the horizontal connection offeature pyramid structure to generate salient features, suppress feature response of irrelevant areas and enhancethe object features.
17927 最后，在显著性系数的基础上构建了锚框掩膜生成子网络，约束锚框位置，排除平坦背景，提高处理效率。 Finally, the anchor mask generation subnetwork is constructed on the basis of salientcoefficient to the location of the anchors, to eliminate the flat background, and to improve the processingefficiency.
17928 实验结果表明，显著性生成子网络仅增加5.94%的处理时间，具备轻量特性； The experimental results show that the salient generation subnetwork only increases the processingtime by 5.94%, and has the lightweight characteristic.
17929 超大视场(U-FOV)红外行人数据集上的识别准确率达到了93.20%，比YOLOv3高了26.49%； The Average-Precision is 93.20% on the U-FOV infraredpedestrian dataset, 26.49% higher than that of YOLOv3.
17930 锚框约束策略能节约处理时间18.05%。 Anchor box constraint strategy can save 18.05% ofprocessing time.
17931 重构模型具有轻量性和高准确性，适合于检测超大视场中的多尺度红外目标。 The proposed method is lightweight and accurate, which is suitable for detecting multi-scaleinfrared objects in the U-FOV camera.
17932 为了解决Alpha稳定分布噪声下目标螺旋桨特征提取问题，该文提出一种基于分数低阶循环谱的特征提取方法。 In order to solve the problem of target propeller features extraction under Alpha stable distribution noise, a method based on fractional low-order cyclic spectrum is proposed.
17933 首先，从理论上推导出脉冲噪声条件下舰船辐射噪声分数低阶循环谱，指出分数低阶循环谱中出现峰值与螺旋桨特征的关系。 Firstly, the low-order cyclicspectrum of ship radiation noise in impulse noise is derived, and the relationship between the propeller featuresand the peak value in the fractional low-order cyclic spectrum is given.
17934 然后根据该关系，提出基于分数低阶循环谱的螺旋桨特征估计方法。 Based on this, a propeller feature estimation method based on fractional low-order cyclic spectrum is proposed.