ID	原文	译文
4453	通过选煤厂实测数据分析表明，所提多源信息传输与分类算法在提高监测数据实时传输效率情况下，能够有效提高故障识别精度。	The analysis of the measured data of the coal preparation plant shows that the proposed multi-source information transmission and classification algorithm can effectively improve the fault recognitionaccuracy under the condition of improving the real-time transmission efficiency of the monitoring data.
4454	针对低轨物联网卫星系统的路由问题，提出了基于流量预测的物联网卫星节点动态缓存分配路由策略。	Aiming at the routing problem of low earth orbit (LEO) Internet of things (IoT) satellite systems, a dynamiccache allocation routing strategy based on traffic prediction for IoT satellite nodes was proposed.
4455	首先，分析低轨卫星覆盖区域内业务分布的时空特性，提出了端到端流量预测方法。	Firstly, the space-timecharacteristics of traffic distribution in the LEO coverage area were analyzed, and an end-to-end traffic prediction modelwas proposed.
4456	然后，根据流量预测结果，提出了动态缓存分配路由策略。	Then, according to the traffic prediction result, a dynamic cache allocation routing strategy was proposed.The satellite node periodically monitored the traffic load of the inter-satellite link, dynamically allocated the cache re-sources of each inter-satellite link between the neighboring nodes.
4457	卫星节点通过对星间链路的流量负载进行周期性监测，动态分配与邻居节点间各条星间链路的缓存资源，分为初始化和系统运行 2 个阶段。	The cache allocation process was divided into twophases, initialization and system operation.
4458	同时，提出了节点拥塞时的业务分流及数据分组转发策略，通过比较排队时延和转发时延的大小，决定数据分组是否需要进行重路由。	At the same time, the traffic offload and packet forwarding strategy when thenode was congested was proposed. By comparing the queuing delay and the forwarding delay, it was determined whetherthe data packet needs to be rerouted.
4459	仿真结果表明，所提路由策略有效地降低了分组丢失率及平均端到端时延，改善了业务在全网的分布情况。	The simulation results show that the proposed routing strategy effectively reducesthe packet loss rate and average end-to-end delay, and improves the traffic distribution in the whole network.
4460	针对视频自动描述任务中的复杂信息表征问题，提出一种多维度和多模态视觉特征的提取和融合方法。	In order to solve the problem of complex information representation in automatic video description tasks, amulti-dimensional and multi-modal visual feature extraction and fusion method was proposed.
4461	首先通过迁移学习提取视频序列的静态和动态等多维度特征，并采用图像描述算法提取视频关键帧的语义信息，完成视频信息的特征表征；	Firstly, multi-dimensionalfeatures such as static and dynamic attributes of the video sequence were extracted by transfer learning, and the imagedescription algorithm was also used to extract the semantic information of the key frames in the video. By doing this, thevideo features extraction was carried out.
4462	然后采用多层长短期记忆网络融合多维度和多模态信息，最终生成视频内容的语言描述。	Then, multi-layer long and short memory networks were used to fuse mul-ti-dimensional and multi-modal information, and finally generated a language description of the video content.