ID 原文 译文
21555 首先，采用由粗到细Lucas-Kanade估计法逐帧提取视频中人体运动的光流特征。 First, the optical flow featuring body motion are extracted frame-by-frame from video with coarse-to-fine Lucas-Kanade flow estimation.
21556 然后，利用预训练模型微调的GoogLeNet神经网络分别逐层卷积并聚合给定时间窗口视频中外观图像和相应光流特征。 Then, the GoogLeNet neural network with fine-tuned pre-trained model is applied to convoluting layer-by-layer and aggregate respectively appearance images and the related optical flow features in the selected time window.
21557 接着，利用长短时记忆多层递归网络交叉感知即得含高层显著结构的时空流语义特征序列； Next, the multi-layered Long Short-Term Memory (LSTM) neural networks are exploited to cross-recursively perceive the spatio-temporal semantic feature sequences with high level and significant structure.
21558 解码时间窗口内互相依赖的隐状态；输出空间流视觉特征描述和视频窗口中每帧标签概率分布。 Meanwhile, the inter-dependent implicit states are decoded in the given timewindow, and the attention salient feature sequence is obtained from temporal stream with the visual feature descriptor in spatial stream and the label probability of each frame.
21559 其次，利用相对熵计算时间维每帧注意力置信度，并融合空间网络流感知序列标签概率分布。 Then, the temporal attention confidence foreach frame with respect to human actions is calculated with the relative entropy measure and fused with the probability distributions with respect to the action categories from the given spatial perception network streamin the video sequence.
21560 最后，利用softmax分类视频中行为类别。 Finally, the softmax classifier is exploited to identify the category of human action in the given video sequence.
21561 实验结果表明，与其他现有方法相比，该文行为识别方法在分类准确度上具有显著优势。 Experimental results show that this presented approach has significant advantages in classification accuracy compared with other methods.
21562 星载SAR大斜视聚束可以实现高分辨率宽覆盖成像和目标多方位信息获取，但斜视情况下大距离徙动会造成回波数据获取效率下降、波位选择困难等问题。 High squint spotlight mode of spaceborne SAR can be used to achieve high resolution and wide swath, and also can be used to acquire information of target from multi-azimuth. However, the considerable range migration can result in efficiency decreasing in data acquisition, and dilemma in system design.
21563 变PRI技术可以跟踪聚束成像过程中目标斜距变化，提高数据获取效率、降低波位选择难度。 This problem can be solved by the technology named PRI (Pulse Repetition Interval) variation which can track the slant range variation of the target during data acquisition.
21564 该文对星载大斜视聚束SAR的变PRI工作机理进行了研究，提出了一种PRI变化序列迭代设计方法和波位选择策略，研究了样条插值和NUFFT两种非周期非均匀采样重建方法，并首次通过机载飞行试验对所提出的变PRI的SAR系统工作体制进行了验证。 In this paper, the principle of PRI variation is studied, and methods of PRI sequence iterative design and system parameter selection are proposed. Two approaches to reconstructing the nonequal spaced and nonperiod data in azimuth sampling are compared. Finally, the first results of PRI variation mode of airborne SAR experiment with high slant spotlight mode are presented.