ID	原文	译文
54027	弹道目标在中段飞行的过程中，存在平动和微动两种运动方式。	During the mid-flight of the ballistic target, there are two movement modes: translational and micro-movement.
54028	由于平动会破坏微多普勒结构，为了获得目标的相关参数以便对目标进行识别，需要对平动分量进行补偿。	Since translation will destroy the micro-Doppler structure, in order to obtain the relevant parameters of the target in order to identify the target, the translation component needs to be compensated.
54029	针对该问题，本文提出了一种基于Shi-Tomasi角点检测算法的弹道中段目标平动补偿方法。	Aiming at this problem, this paper proposes a method of compensation for the mid-trajectory target translation based on the Shi-Tomasi corner detection algorithm.
54030	该算法从图像处理的角度出发，可以有效的提取出图像中的角点信息。	From the perspective of image processing, this algorithm can effectively extract the corner information in the image.
54031	之后，根据得到的角点信息进行回归分析，从而实现平动参数的估计。	Afterwards, regression analysis is performed according to the obtained corner point information to realize the estimation of translational parameters.
54032	根据仿真实验结果，该算法可以实现平动分量的补偿并具有较好的鲁棒性。	According to the results of simulation experiments, the algorithm can realize the compensation of translational component and has better robustness.
54033	近年来出现并迅猛发展的深度伪造(DeepFake)技术深刻改变了多媒体内容伪造的方式和水平，给网络空间内容安全带来了新的严峻挑战。	DeepFake technology, which has emerged and been developed rapidly in recent years, has profoundly changed the way and level of multimedia content forgery, posing new and severe challenges to content security in cyberspace.
54034	本文主要关注深度伪造中危害最大的视频换脸伪造，提出基于I3D(Inception3D)网络的眼部与口部双流检测方法。	This paper mainly focuses on the most harmful video face-swapping forgery among the deep forgeries, and proposes a two-stream detection method jointly exploiting eye and mouth artifacts based on the I3D(Inception3D) net.
54035	首先，针对现有大多数伪造检测方法忽略了视频中重要的时间信息的问题，将目前常用的仅具备空域感受能力的2D卷积拓展为I3D卷积，赋予网络同时感受空域和时域信息的能力。	Firstly, considering most of the existing forgery detection methods do not take into account the important time information in video and the current common 2D convolution merely has the spatial domain perception ability, we propose to use the I3D convolution, enabling the network with the ability to simultaneously learn spatial and temporal domains artifacts.
54036	同时，通过调整I3D网络结构使其从原有的多分类任务设计改进为更适合换脸取证二分类任务的高效网络。	Meanwhile, through adjusting the I3D network structure, it could be improved from the original multi-classification task design to an efficient network that is more suitable for the binary classification task in DeepFake forensics.