A Novel Multi-Stream Fusion Network for Underwater Image Enhancement

Due to the selective absorption of light and the existence of a large number of floating media in sea water, underwater images often suffer from color casts and detail blurs. It is therefore necessary to perform color correction and detail restoration. However,the existing enhancement algorithms cannot achieve the desired results. In order to solve the above problems, this paper proposes a multi-stream feature fusion network. First, an underwater image is preprocessed to obtain potential information from the illumination stream, color stream and structure stream by histogram equalization with contrast limitation, gamma correction and white balance, respectively. Next, these three streams and the original raw stream are sent to the residual blocks to extract the features. The features will be subsequently fused. It can enhance feature representation in underwater images. In the meantime, a composite loss function including three terms is used to ensure the quality of the enhanced image from the three aspects of color balance, structure preservation and image smoothness. Therefore, the enhanced image is more in line with human visual perception.Finally, the effectiveness of the proposed method is verified by comparison experiments with many stateof-the-art underwater image enhancement algorithms. Experimental results show that the proposed method provides superior results over them in terms of MSE,PSNR, SSIM, UIQM and UCIQE, and the enhanced images are more similar to their ground truth images.

Machine-Learning Based Packet Switching Method for Providing Stable High-Quality Video Streaming in Multi-Stream Transmission

Broadcasting gateway equipment generally uses a method of simply switching to a spare input stream when a failure occurs in a main input stream.However,when the transmission environment is unstable,problems such as reduction in the lifespan of equipment due to frequent switching and interruption,delay,and stoppage of services may occur.Therefore,applying a machine learning(ML)method,which is possible to automatically judge and classify network-related service anomaly,and switch multi-input signals without dropping or changing signals by predicting or quickly determining the time of error occurrence for smooth stream switching when there are problems such as transmission errors,is required.In this paper,we propose an intelligent packet switching method based on the ML method of classification,which is one of the supervised learning methods,that presents the risk level of abnormal multi-stream occurring in broadcasting gateway equipment based on data.Furthermore,we subdivide the risk levels obtained from classification techniques into probabilities and then derive vectorized representative values for each attribute value of the collected input data and continuously update them.The obtained reference vector value is used for switching judgment through the cosine similarity value between input data obtained when a dangerous situation occurs.In the broadcasting gateway equipment to which the proposed method is applied,it is possible to perform more stable and smarter switching than before by solving problems of reliability and broadcasting accidents of the equipment and can maintain stable video streaming as well.

High-Performance Transmission Mechanism Design of Multi-Stream Carrier Aggregation for 5G Non-Standalone Network

Multi-stream carrier aggregation is a key technology to expand bandwidth and improve the throughput of the fifth-generation wireless communication systems.However,due to the diversified propagation properties of different frequency bands,the traffic migration task is much more challenging,especially in hybrid sub-6 GHz and millimeter wave bands scenario.Existing schemes either neglected to consider the transmission rate difference between multistream carrier,or only consider simple low mobility scenario.In this paper,we propose a low-complexity traffic splitting algorithm based on fuzzy proportional integral derivative control mechanism.The proposed algorithm only relies on the local radio link control buffer information of sub-6 GHz and mmWave bands,while frequent feedback from user equipment(UE)side is minimized.As shown in the numerical examples,the proposed traffic splitting mechanism can achieve more than 90%link resource utilization ratio for different UE transmission requirements with different mobilities,which corresponds to 10%improvement if compared with conventional baselines.

Study on Multi-stream Heat Exchanger Network Synthesis with Parallel Genetic/Simulated Annealing Algorithm

The multi-stream heat exchanger network synthesis (HENS) problem can be formulated as a mixed integer nonlinear programming model according to Yee et al. Its nonconvexity nature leads to existence of more than one optimum and computational difficulty for traditional algorithms to find the global optimum. Compared with deterministic algorithms, evolutionary computation provides a promising approach to tackle this problem. In this paper, a mathematical model of multi-stream heat exchangers network synthesis problem is setup. Different from the assumption of isothermal mixing of stream splits and thus linearity constraints of Yee et al., non-isothermal mixing is supported. As a consequence, nonlinear constraints are resulted and nonconvexity of the objective function is added. To solve the mathematical model, an algorithm named GA/SA (parallel genetic/simulated annealing algorithm) is detailed for application to the multi-stream heat exchanger network synthesis problem. The performance of the proposed approach is demonstrated with three examples and the obtained solutions indicate the presented approach is effective for multi-stream HENS.

Performance Evaluation Methods for Multi-stream Plate-Fin Heat Exchanger

Mathematical model of cross type multi-stream plate-fin heat exchanger is established.Meanwhile,mean square error of accumulative heat load is normalized by dimensionless,and the equations of temperature-difference uniformity factor are improved.Evaluation factors above and performance of heat exchanger are compared and analyzed by taking aircraft three-stream condenser as an example.The results demonstrate that the mean square error of accumulative heat load is common result of total heat load and excess heat load between passages.So it can be influenced by passage arrangement,flow inlet parameters as well as flow patterns.Dimensionless parameter of mean square error of accumulative heat load can reflect the influence of passage arrangement to heat exchange performance and will not change dramatically with the variation of flow inlet parameters and flow patterns.Temperature-difference uniformity factor is influenced by passage arrangement and flow patterns.It remains basically unchanged under a certain range of flow inlet parameters.

Multi-Stream CNN-Based Personal Recognition Method Using Surface Electromyogram for 5G Security

As fifth generation technology standard(5G)technology develops,the possibility of being exposed to the risk of cyber-attacks that exploits vulnerabilities in the 5G environment is increasing.The existing personal recognitionmethod used for granting permission is a password-basedmethod,which causes security problems.Therefore,personal recognition studies using bio-signals are being conducted as a method to access control to devices.Among bio-signal,surface electromyogram(sEMG)can solve the existing personal recognition problem that was unable to the modification of registered information owing to the characteristic changes in its signal according to the performed operation.Furthermore,as an advantage,sEMG can be conveniently measured from arms and legs.This paper proposes a personal recognition method using sEMG,based on a multi-stream convolutional neural network(CNN).The proposed method decomposes sEMG signals into intrinsic mode functions(IMF)using empirical mode decomposition(EMD)and transforms each IMF into a spectrogram.Personal recognition is performed by analyzing time–frequency features from the spectrogram transformed intomulti-streamCNN.The database(DB)adopted in this paper is the Ninapro DB,which is a benchmark EMG DB.The experimental results indicate that the personal recognition performance of the multi-stream CNN using the IMF spectrogram improved by 1.91%,compared with the singlestream CNN using the spectrogram of raw sEMG.

多源流理论视角下美国职业生涯教育立法政策的变迁

多源流理论为解释政策过程和分析政策变迁提供了一种视角和框架。美国职业生涯教育立法政策经历了三个时期:一是以片段化的引导性立法政策支持职业指导和职业生涯发展的探索期;二是以特色化的规制性立法政策推进职业生涯教育实践的确立期;三是以系统化的策略性立法政策深化职业生涯教育改革的深化期。运用多源流理论分析这一变迁历程会发现:美国时代变动引发的失业危机与传统教育效能单一之间的冲突构成了“问题识别”;职业生涯教育思想在政策共同体中的演化促成了“政策阐明”;公众舆论与政党执政理念对职业生涯教育的影响产生了“政治活动”;马兰德的“过程软化”与打开“政策之窗”则最终促成了“政策之窗与溪流的结合”。回顾美国职业生涯教育立法政策的颁布与调整,有两点值得思考和借鉴:一是立法政策在发展趋势、人群面向、目标定位和功能发挥上有明显变化;二是立法政策规范和保障是影响美国职业生涯教育实践的重要推动因素,而职业生涯教育立法政策的颁布又是特定时期多种因素长期共同作用促成的结果。

基于数据流的K-S变化检测的动态多目标规划算法

为了更加准确地判断环境是否发生变化并快速追踪动态多目标规划问题(dynamicmulti-objectiveoptimization problem,DMOP)当前时刻的Pareto前沿,提出了一种基于数据流的Kolmogorov-Smirnov(K-S)变化检测的动态多目标规划(DSK-SDMOP)算法。该算法以NSGA-Ⅱ为基础,通过数据流建立2个时刻的检验窗口,再利用K-S检验基于数据流的Pareto最优前沿是否发生变化,检测2个窗口的数据是否服从同一分布来判断环境是否发生变化,并就环境变化的剧烈程度实行相应的应答机制,以提高对环境的适应程度。利用基于数据流的K-S检测方法,对环境变化不会过于敏感,而且不用提前假设对应目标值的分布,易于操作。通过5个动态多目标规划标准测试函数对该算法进行测试,并和现有的2种算法进行对比分析,结果表明该算法处理动态多目标规划问题具有良好的性能。

非平衡数据流在线主动学习方法

数据流分类是数据流挖掘领域一项重要研究任务,目标是从不断变化的海量数据中捕获变化的类结构.目前,几乎没有框架可以同时处理数据流中常见的多类非平衡、概念漂移、异常点和标记样本成本高昂问题.基于此,提出一种非平衡数据流在线主动学习方法(Online active learning method for imbalanced data stream,OALM-IDS).AdaBoost是一种将多个弱分类器经过迭代生成强分类器的集成分类方法,AdaBoost.M2引入了弱分类器的置信度,此类方法常用于静态数据.定义了基于非平衡比率和自适应遗忘因子的训练样本重要性度量,从而使AdaBoost.M2方法适用于非平衡数据流,提升了非平衡数据流集成分类器的性能.提出了边际阈值矩阵的自适应调整方法,优化了标签请求策略.将概念漂移程度融入模型构建过程中,定义了基于概念漂移指数的自适应遗忘因子,实现了漂移后的模型重构.在6个人工数据流和4个真实数据流上的对比实验表明,提出的非平衡数据流在线主动学习方法的分类性能优于其他5种非平衡数据流学习方法.

5G-R高铁车站场景MIMO天线阵列结构与传输性能研究

面向5G-R高铁车站应用场景,在2155~2165 MHz频段采用射线跟踪仿真技术,对4种不同天线阵列排布方式的MIMO系统多层传输性能进行对比研究。通过研究采取不同收发端天线阵列排布方式时的信道相关性,计算每个接收点位所能达到的MIMO系统最大传输层数,获取各层数对应的信噪比,计算在不同调制方式、不同天线排布方式下每个接收点位的下行峰值传输速率,确定面向5G-R高铁车站场景的MIMO系统最佳传输层数。仿真验证结果表明,面向5G-R频段高铁车站应用场景,基站端采用线阵、车载端采用方阵排列,可取得最佳峰值传输速率。研究结果可为5G-R专网建设中高铁车站部署MIMO系统提供技术积累和理论依据。

基于双流YOLOv4的金属表面缺陷检测方法

目前有许多学者使用深度学习进行表面缺陷检测研究,由于这些研究大都沿用主流目标检测算法的思路,注重高级语义特征,而忽视了低级语义信息(色彩、形状)对表面缺陷检测的重要性,因此导致缺陷检测效果不够理想。为解决上述问题,提出了一种金属表面缺陷检测网络——双流YOLOv4网络,骨干网络分成两个分支,输入分为高分辨率图像和低分辨率图像,浅分支负责从高分辨率图像中提取低级特征,深分支负责从低分辨率图像中提取高级特征,通过削减两分支的层数和通道数来减少模型总参数量;为了强化低级语义特征,提出了一种树形多尺度融合方法(Tree-structured Multi-scale Feature Fusion Me-thod,TMFF),并设计了一个结合极化自注意力机制和空间金字塔池化的特征融合模块(Feature Fusion Module with Polarized Self-Attention Mechanism and Spatial Pyramid Pooling,FFM-PSASPP)应用到TMFF中。在东北大学热轧带表面缺陷数据集NEU-DET、金属表面缺陷数据集GC10-DET和伊莱特电饭煲内胆缺陷数据集Enaiter的测试集上对所提算法进行了测试,测得的map@50结果分别为0.80,0.66和0.57,相比大部分主流的用于缺陷检测的目标检测算法均有提升,且模型参数量仅为原YOLOv4的一半,速度与YOLOv4接近,可满足实际使用需求。

非平衡概念漂移数据流主动学习方法

数据流分类研究在开放、动态环境中如何提供更可靠的数据驱动预测模型,关键在于从实时到达且不断变化的数据流中检测并适应概念漂移.目前,为检测概念漂移和更新分类模型,数据流分类方法通常假设所有样本的标签都是已知的,这一假设在真实场景下是不现实的.此外,真实数据流可能表现出较高且不断变化的类不平衡比率,会进一步增加数据流分类任务的复杂性.为此,提出一种非平衡概念漂移数据流主动学习方法 (Active learning method for imbalanced concept drift data stream, ALM-ICDDS).定义基于多预测概率的样本预测确定性度量,提出边缘阈值矩阵的自适应调整方法,使得标签查询策略适用于类别数较多的非平衡数据流;提出基于记忆强度的样本替换策略,将难区分、少数类样本和代表当前数据分布的样本保存在记忆窗口中,提升新基分类器的分类性能;定义基于分类精度的基分类器重要性评价及更新方法,实现漂移后的集成分类器更新.在7个合成数据流和3个真实数据流上的对比实验表明,提出的非平衡概念漂移数据流主动学习方法的分类性能优于6种概念漂移数据流学习方法.

融合三维人脸动态信息和光流信息的人脸表情识别

人脸表情识别在静态图像上取得了卓越的成效,但这些方法应用于视频或图像序列时,准确度和鲁棒性往往会受到影响。传统的方法通常无法基于空间信息和光流信息进行人脸表情的识别,然而这些辅助识别信息都是二维信息,没有考虑到人脸的表情变化是一种三维的变化过程。为充分挖掘人脸表情识别的深层语义信息,提出了一种基于三维人脸动态信息和光流信息相结合的融合表情识别方法。该方法构建基于人脸深度图像、光流图像和RGB图像的多流卷积神经网络,通过融合3种模态的信息进行人脸表情识别。所提方法在CAER,RAVDESS数据集上进行了充分验证,实验结果表明,其在表情识别性能上优于目前的主流方法,证明了其有效性。

新形态对称密码算法研究

安全多方计算、全同态加密和零知识证明是解决数据安全问题的重要密码技术,其快速发展和应用需求驱动,引发了新形态对称密码的设计与安全性分析.本文系统介绍适宜安全多方计算、全同态加密和零知识证明的新形态对称密码算法.首先梳理新形态对称密码算法的应用需求和设计目标;然后归类介绍新形态对称密码算法的设计理念和最新的安全性评估结果;最后总结探讨新形态对称密码算法的设计特点以及面临的问题.

基于FFmpeg多线程编码的智能交通监控系统设计

针对智能交通管理设备本身缺乏安全监管,传统视频监控延迟高、画质低、稳定性差的问题,提出一种基于FFmpeg的多线程编码视频流传输方案。通过FFmpeg调用h264_nvenc编码器,实现宏块行级的GPU多线程加速,降低编码延迟。使用Visual Studio 2019和QT15.5开发基于FFmpeg的音视频处理软件,对多路视频流进行封装、推流,并搭建Nginx流媒体服务器进行分发。通过实验表明,该系统整体的传输延迟低于1 s,且拥有良好的率失真特性,监控画面清晰、稳定性高,实现了对交通管理设备实时稳定的安全监控。

基于异构硬件的LSTM训练系统

在大数据时代,以LSTM为代表的深度神经网络模型具有处理海量数据的能力,在语言处理、语音识别、时序数据预测等领域表现优异。随着模型复杂度的提高,训练成本大幅提升。现有的LSTM训练系统使用了算子融合、多流等加速手段,但忽略了训练算子内部计算的可并行性,导致计算资源的利用率低,整体耗时长。为此,设计了基于细粒度模型划分和多流并行调度方法的LSTM训练系统TurboLSTM,在英伟达GPU和国产昇腾NPU这两种异构硬件上构建的全新底层训练算子实现了任务对计算资源的合理利用。与已有训练系统相比,在GPU上TurboLSTM的单算子训练时间缩短了23%,模型的整体训练时间缩短了17%,在NPU上TurboLSTM的单算子训练时间缩短了15%,且对计算资源的利用率显著提高。这表明提出的加速方案是高效的,具有良好的泛化能力。

“村BA”何以进入全国和美乡村赛事政策议程?——基于多源流理论分析

全国和美乡村篮球大赛(村BA)政策的出台成为体育公共政策的一个经典案例。对“村BA”政策的研究既能为其他乡村体育赛事政策议程提供可行性的研究视角,也能为我国未来乡村体育赋能乡村振兴的系列政策制定提供重要参考。笔者基于多源流理论视角,以“村BA”赛事政策为研究案例,对其政策出台进行了全景式的剖析。研究表明:系列“村BA”焦点事件暴露出的乡村体育供需缺口,乡村振兴与乡村体育的天然合拍反映出的农民对美好生活向往构成了问题源流;中央政策持续跟进、人大代表与政协委员的提案议案与专家学者学理推论构成了政策源流;党的“体育为民”执政理念、纯粹而热烈的国民情绪与媒体力量构成了政治源流;最终“三源耦合”的动力作用快速打开了政策之窗。

改进型三级串联多股流LNG冷能发电系统性能研究

针对常见的三级串联朗肯循环液化天然气(LNG)冷能发电系统未能有效地利用工质冷量的问题,提出了一种改进型三级串联多股流联合循环(MS-3CC)系统。通过加入多股流换热器,该系统多级利用了LNG和系统内循环工质冷量,有效降低了系统复杂程度,提升了系统热力学性能和经济性。通过遗传算法优化计算,当天然气(NG)排气压力为5.5 MPa、系统内各级循环工质均选取丙烷时,MS-3CC系统的最大净输出功为6027 kW,相比三级串并联合循环(P-3CC)系统和单级联合循环(SCC)系统分别提高了5.4%和80.3%,而其年净收益则分别提高了16.9%和68.0%。比较了64种工质组合,发现采用CH_(2)F_(2)+CH_(2)F_(2)+C_(3)H_(8)时MS-3CC系统的净输出功最高。MS-3CC系统的提出为实际工业中更为有效地多级利用LNG冷量提供了参考。

多工序制造中的误差流模型及其应用研究进展

航空发动机结构件、汽车车身等复杂产品通常为多工序制造件,其制造过程中多种误差的产生和累积会对产品质量产生耦合影响,使产品质量控制处于暗箱状态。将误差流理论与精度控制理论、工程数据处理方法相结合,构建对误差的传递、累积、湮灭进行描述和分析的误差流模型对于多工序制造复杂产品质量控制具有重要意义。因此,总结和归纳了误差流模型的构建方法及其应用研究,提出了多工序制造的误差流模型质量控制总体框架,从小数据驱动的时间序列数学解析模型和大数据驱动的复杂网络统计分析模型两个方面对多工序制造的误差流模型进行了分类综述。最后,展望了误差流模型在高端装备和复杂产品等未来多工序制造领域中的发展趋势。

基于双分支多头注意力的场景图生成方法

针对已有场景图生成模型获取上下文信息有限的问题,提出一种有效的上下文融合模块,即双分支多头注意力(dual-stream multi-head attention, DMA)模块,并将DMA分别用于物体分类阶段和关系分类阶段,基于此提出基于双分支多头注意力的场景图生成网络(dual-stream multi-head attention-based scene graph generation network, DMA-Net)。该网络由目标检测、物体语义解析和关系语义解析3个模块组成。首先,通过目标检测模块定位图像中的物体并提取物体特征;其次,使用物体语义解析模块中的节点双分支多头注意力(object dual-stream multi-head attention, O-DMA)获取融合了节点上下文的特征,该特征经过物体语义解码器获得物体类别标签;最后,通过关系语义解析模块中的边双分支多头注意力(relationship dual-stream multi-head attention, R-DMA)输出融合了边上下文的特征,该特征经过关系语义解码器输出关系类别标签。在公开的视觉基因组(visual genome, VG)数据集上分别计算了DMA-Net针对场景图检测、场景图分类和谓词分类3个子任务的图约束召回率和无图约束召回率,并与主流的场景图生成方法进行比较。实验结果表明,所提出的方法能够充分挖掘场景中的上下文信息,基于上下文增强的特征表示有效提升了场景图生成任务的精度。