基于深度信念网络的列车故障音频诊断方法

铁路在交通运输行业有着举足轻重的地位，一旦列车发生故障将会导致严重的生命财产损失。由于列车发生故障的概率相对较低，因此难以捕获列车的故障样本。针对上述问题，提出了一种无监督学习的列车故障识别方法，通过检测列车音频信号来识别列车故障。该方法基于深度信念网络(DBN)，利用小波包分解提取检测信号的特征向量并将其作为DBN的输入，待网络充分训练后，由训练好的DBN识别当前列车的运行状况。现场监测实验结果表明，该方法能够在无监督的条件下有效识别列车故障，保障了列车的运行安全。     

基于深度学习的秀丽隐杆线虫显微图像分割方法

利用计算机实现自动、准确的秀丽隐杆线虫(C.elegans)的各项形态学参数分析,至关重要的是从显微图像上分割出线虫体态,但由于显微镜下的图像噪声较多,线虫边缘像素与周围环境相似,而且线虫的体态具有鞭毛和其他附着物需要分离,多方面因素导致设计一个鲁棒性的C.elegans分割算法仍然面临着挑战。针对这些问题,提出了一种基于深度学习的线虫分割方法,通过训练掩模区域卷积神经网络(Mask R-CNN)学习线虫形态特征实现自动分割。首先,通过改进多级特征池化将高级语义特征与低级边缘特征融合,结合大幅度软最大损失(LMSL)损失算法改进损失计算;然后,改进非极大值抑制;最后,引入全连接融合分支等方法对分割结果进行进一步优化。实验结果表明,相比原始的Mask R-CNN,该方法平均精确率(AP)提升了4.3个百分点,平均交并比(mIOU)提升了4个百分点。表明所提出的深度学习分割方法能够有效提高分割准确率,在显微图像中更加精确地分割出线虫体。     

The concept of smartness in cyber–physical systems and connection to urban environment

The next-generation systems are expected to be largely cyber–physical systems (CPSs) that autonomously control physical processes, through sensors and actuators typically in real-time feedback and cooperative control loops distributed among physical and cyber environments. The rapid technological advancements enhance the smartness of these CPSs, pushing their boundaries of performance and efficiency by embedding new information and communication technologies. However, to what extent CPSs should be smarter so that they do not compromise safety and security of safety critical systems? is an open research question. Towards this goal, the purpose of this study is to establish a grounded theory to analyse what makes these systems smart? and eventually, how to find a balance between smartness and safety risks? In this precinct, this article aims to develop a conceptual framework, define the dimensions and derive the characteristics that make CPSs smart. The proposed approach combines an automated informetric and systematic analysis of literature pertinent to the topic of smartness across anthropology, science, engineering and technology. The analysis of a case study building and the discussions presented herein support the connection between the existing understanding of CPSs and smartness offered by the building design approach in urban environment.     

Modeling the effect of deep traps on the capacitance–voltage characteristics of p-type Si-doped GaAs Schottky diodes grown on high index GaAs substrates

Numerical simulation, using SILVACO-TCAD, is carried out to explain experimentally observed effects of different types of deep levels on the capacitance–voltage characteristics of p-type Si-doped GaAs Schottky diodes grown on high index GaAs substrates. Two diodes were grown on (311)A and (211)A oriented GaAs substrates using Molecular Beam Epitaxy (MBE). Although, deep levels were observed in both structures, the measured capacitance–voltage characteristics show a negative differential capacitance (NDC) for the (311)A diodes, while the (211)A devices display a usual behaviour. The NDC is related to the nature and spatial distribution of the deep levels, which are characterized by the Deep Level Transient Spectroscopy (DLTS) technique. In the (311)A structure only majority deep levels (hole traps) were observed while both majority and minority deep levels were present in the (211)A diodes. The simulation, which calculates the capacitance–voltage characteristics in the absence and presence of different types of deep levels, agrees well with the experimentally observed behaviour.     

改进HSR-FCN的服装图像识别分类算法研究

目前网络上的服装图像数量增长迅猛，对于大量服装图像实现智能分类的需求日益增加。将基于区域的全卷积网络（Region-Based Fully Convolutional Networks，R-FCN）引入到服装图像识别中，针对服装图像分类中网络训练时间长、形变服装图像识别率低的问题，提出一种新颖的改进框架HSR-FCN。新框架将R-FCN中的区域建议网络和HyperNet网络相融合，改变图片特征学习方式，使得HSR-FCN可以在更短的训练时间内达到更高的准确率。在模型中引入了空间转换网络，对输入服装图像和特征图进行了空间变换及对齐，加强了对多角度服装和形变服装的特征学习。实验结果表明，改进后的HSR-FCN模型有效地加强了对形变服装图像的学习，且在训练时间更短的情况下，比原来的网络模型R-FCN平均准确率提高了大约3个百分点，达到96.69%。     

Deep Learning Based Process Analytics Model for Predicting Type 2 Diabetes Mellitus

Process analytics is one of the popular research domains that advanced in the recent years. Process analytics encompasses identification, monitoring, and improvement of the processes through knowledge extraction from historical data. The evolution of Artificial Intelligence (AI)-enabled Electronic Health Records (EHRs) revolutionized the medical practice. Type 2 Diabetes Mellitus (T2DM) is a syndrome characterized by the lack of insulin secretion. If not diagnosed and managed at early stages, it may produce severe outcomes and at times, death too. Chronic Kidney Disease (CKD) and Coronary Heart Disease (CHD) are the most common, long-term and life-threatening diseases caused by T2DM. Therefore, it becomes inevitable to predict the risks of CKD and CHD in T2DM patients. The current research article presents automated Deep Learning (DL)-based Deep Neural Network (DNN) with Adagrad Optimization Algorithm i.e., DNN-AGOA model to predict CKD and CHD risks in T2DM patients. The paper proposes a risk prediction model for T2DM patients who may develop CKD or CHD. This model helps in alarming both T2DM patients and clinicians in advance. At first, the proposed DNN-AGOA model performs data preprocessing to improve the quality of data and make it compatible for further processing. Besides, a Deep Neural Network (DNN) is employed for feature extraction, after which sigmoid function is used for classification. Further, Adagrad optimizer is applied to improve the performance of DNN model. For experimental validation, benchmark medical datasets were used and the results were validated under several dimensions. The proposed model achieved a maximum precision of 93.99%, recall of 94.63%, specificity of 73.34%, accuracy of 92.58%, and F-score of 94.22%. The results attained through experimentation established that the proposed DNN-AGOA model has good prediction capability over other methods.     

黄沙坪铅锌多金属矿的成矿规律及深部找矿远景

众所周知,矿物质的成分是多种多样的,社会的发展和科学的进步需要运用到多重金属矿物质,在专业人员的勘探与挖掘中,发现了黄沙坪铅锌多金属矿,这个矿区中有丰富的有色金属,这个矿的发掘为研究成矿规律提供了物质基础,同时也为深部找矿提供了可靠的依据。本文主要分析黄沙坪铅锌多金属矿的成矿规律及深部找矿远景。     

基于深度学习的人体姿态估计方法综述

基于深度学习的人体姿态估计方法旨在通过构建合适的神经网络，直接从二维的图像特征中回归出人体姿态信息。主要按照2D人体姿态估计到3D人体姿态估计的顺序，并从单人检测与多人检测、稀疏的关节点检测与密集的模型构建等方面，对近年来基于深度学习的人体姿态估计方法进行系统介绍，从而初步了解如何通过深度学习的方法得到人体姿态的各个要素，包括肢体部件的相对朝向和比例尺度、骨骼关节点的位置坐标和连接关系，甚至更为复杂的人体蒙皮模型信息。最后,对当前研究面临的挑战以及未来的热点动向进行概述，清晰地呈现出该领域的发展脉络。     

深度学习识别光网络单元故障的设计与应用

为解决依赖装维上门鉴别光网络单元故障带来的不便,可以从机器视觉入手实现自动化故障识别。近年,ImageNet挑战赛的成功推动了物体识别技术的跨越式发展,特别是基于卷积的深度学习技术在视觉识别方面已经达到人类水平,为光网络单元故障的自动识别提供了技术基础。文章对识别光网络单元的工作状态进行了研究,将设备工作状态分为7个场景,提出了利用手机APP采集图片识别故障的解决方案并投入了实际生产;重点阐述了深度学习模块的设计与实现,提出一种通过算法整合的方式综合运用物体检测和图像分类算法,分3阶段逐步求精,解决了图片过滤,光网络单元型号和状态识别等问题,实现了基于计算机视觉自动识别光网络单元故障。从数据上看产品的端到端准确率超过84%,识别速度达到10 FPS,月均提供服务超过1万人次,在减少用户等待的同时节约了人力资源。