Intelligent tool wear monitoring based on parallel residual and stacked bidirectional long short-term memory network

Effective tool wear monitoring (TWM) is essential for accurately assessing the degree of tool wear and for timely preventive maintenance. Existing data-driven monitoring methods mainly rely on complex feature engineering, which reduces the monitoring efficiency. This paper proposes a novel TWM model based on a parallel residual and stacked bidirectional long short-term memory (PRes–SBiLSTM) network. First, a parallel residual network (PResNet) is used to extract the multi-scale local features of sensor signals adaptively. Subsequently, a stacked bidirectional long short-term memory (SBiLSTM) network is used to obtain the time-series features related to the tool wear characteristics. Finally, the predicted tool wear value is outputted through a fully connected network. A smoothing correction method is applied to improve the prediction accuracy. The proposed model is experimentally verified to have a high prediction accuracy without sacrificing its generalization ability. A TWM system framework based on the PRes–SBiLSTM network is proposed, which has a certain reference value for TWM in actual industrial environments.     

A novel vision-based method for 3D profile extraction of wire harness in robotized assembly process

Automating stages for deformable objects in the production line, in which assembling a wire harness into a predefined position is a complex task owing to the specialized characteristics of the objects. Besides a few automatized systems proposed in the other studies to implement this task under simplified setup conditions, a significant portion of this process remains to be completed manually in industrial environments. To construct an automatic wire harness assembly system, the development of a method that can automatically detect the wire harness profile in a 3D environment and, consequently, guide robot arms to implement assembly tasks is indispensable. Therefore, this study presents an approach that satisfies this requirement, which not only proposes a deep learning-based system to detect the wire profile, but also improves the accuracy of the detected results through a correction method according to the depth values of contiguous areas. The verification of the approach in a robot system that highlights its usefulness and practicality demonstrates the potential of the proposed method to replace people and consequently, reduce labour costs in factory environments.     

基于深度学习的文本分类技术研究进展

随着深度学习技术的快速发展,许多研究者尝试利用深度学习来解决文本分类问题,特别是在卷积神经网络和循环神经网络方面,出现了许多新颖且有效的分类方法。对基于深度神经网络的文本分类问题进行分析,介绍卷积神经网络、循环神经网络、注意力机制等方法在文本分类中的应用和发展,分析多种典型分类方法的特点和性能,从准确率和运行时间方面对基础网络结构进行比较,表明深度神经网络较传统机器学习方法在用于文本分类时更具优势,其中卷积神经网络具有优秀的分类性能和泛化能力。在此基础上,指出当前深度文本分类模型存在的不足,并对未来的研究方向进行展望。     

深度学习在主动脉中膜变性病理图像分类中的应用

胸主动脉瘤和夹层（TAAD）是严重的心血管疾病之一,而中膜变性（MD）的组织学改变对疾病的诊断及早期干预具有重要的临床意义。针对病理图像的高度复杂性使得MD的诊断过程耗时费力且一致性差的问题,提出了一种基于深度学习的病理图像分类方法,并将其应用于四种MD病变类型以进行性能验证。该方法使用了一种改进的基于GoogLeNet的卷积神经网络模型,首先采用迁移学习来将先验知识应用于TAAD病理图像的表达,然后使用Focal loss和L2正则化来解决数据不平衡问题,从而进一步优化模型性能。实验结果表明,所提模型的平均四分类准确率达到98.78%,表现出较好的泛化性能。可见所提方法可以有效地提升病理学家的诊断效率。     

基于图卷积网络的微博新闻故事线抽取方法

微博作为人们获取和传播新闻事件的主要平台,隐藏着丰富的事件信息。从微博数据中抽取故事线能为用户提供一种直观的方式来准确理解事件演化,然而微博数据稀疏和上下文缺乏的特点为故事线抽取带来了挑战。因此,通过两个连续的任务从微博数据中自动抽取故事线：1）基于微博传播影响力对事件进行建模,并提取出首要事件;2）基于事件特征建立异构事件图,提出事件图卷积网络（E-GCN）模型来提升对事件间隐式关系的学习能力,从而实现事件的故事分支预测并链接事件。在真实数据集上从故事分支和故事线两个角度进行评测,结果表明所提方法在故事分支生成测评中,相较于贝叶斯模型、斯坦纳树和故事森林在F1值上,在Dataset1上分别高出28个百分点、20个百分点和27个百分点,在Dataset2上分别高出19个百分点、12个百分点和22个百分点;而在故事线抽取评测中,相较于故事时间线、斯坦纳树和故事森林在正确的边准确率上,在Dataset1上分别高出33个百分点、23个百分点和17个百分点,在Dataset2上分别高出12个百分点、3个百分点和9个百分点。     

基于密集Inception的单图像超分辨率重建方法

近几年,基于卷积神经网络（CNN）的单图像超分辨率（SR）重建方法成为了主流。通常情况下,重建模型的网络层数越深,提取的特征越多,重建效果越好;然而随着网络层数的加深,不仅会出现梯度消失的问题,还会显著增加参数量,增加训练的难度。针对以上问题,提出了一种基于密集Inception的单图像SR重建方法。该方法引入Inception-残差网络（Inception-ResNet）结构提取图像特征,全局采用简化后的密集网络,且仅构建每一个模块输出到重建层的路径,从而避免产生冗余数据来增加计算量。在放大倍数为4时,采用数据集Set5测试模型性能,结果显示与超深卷积神经网络的图像超分辨率（VDSR）相比,所提方法的结构相似性（SSIM）高了0.013 6;与基于多尺度残差网络的图像SR（MSRN）相比,SSIM高了0.002 9,模型参数量少了78%。实验结果表明,所提方法在保证模型的深度和宽度的情况下,显著减少了参数量,从而降低了训练的难度,而且取得了比对比方法更好的峰值信噪比（PSNR）和SSIM。     

基于双重权重偏差建模的无监督域适应

无监督域适应(Unsupervised Domain Adaptation,UDA)是一类新兴的机器学习范式,其通过对源域知识在无标记目标域上的迁移利用,来促进目标域模型的训练。为建模源域与目标域之间的域分布差异,最大均值差异(Maximum Mean Discrepancy,MMD)建模被广泛应用,其对UDA的性能提升起到了有效的促进作用。然而,这些方法通常忽视了领域之间对应类规模与类分布等结构信息,因为目标域与源域的数据类规模与数据分布通常并非一致。为此,文中提出了一种基于跨域类和数据样本双重加权的无监督域适应模型(Sample weighted and Class weighted based Unsupervised Domain Adaptation Network,SCUDAN)。具体而言,一方面,通过源域类层面的适应性加权来调整源域类权重,以实现源域与目标域之间的类分布对齐;另一方面,通过目标域样本层面的适应性加权来调整目标域样本权重,以实现目标域与源域类中心的对齐。此外,文中还提出了一种CEM(Classification Expectation Maximization)优化算法,以实现对SCUDAN的优化求解。最后,通过对比实验和分析,验证了所提模型和算法的有效性。     

基于深度卷积生成对抗网络的半生成式视频隐写方案

生成式隐写通过生成足够自然或真实的含密样本来隐藏秘密消息,是信息隐藏方向的研究热点,但目前在视频隐写领域的研究还比较少。结合数字化卡登格的思想,提出一种基于深度卷积生成对抗网络（DCGAN）的半生成式视频隐写方案。该方案中,设计了基于DCGAN的双流视频生成网络,用来生成视频的动态前景、静态后景与时空掩模三个部分,并以随机噪声驱动生成不同的视频。方案中的发送方可设定隐写阈值,在掩模中自适应地生成数字化卡登格,并将其作为隐写与提取的密钥;同时以前景作为载体,实现信息的最优嵌入。实验结果表明,该方案生成的含密视频具有良好的视觉质量,Frechet Inception距离（FID）值为90,且嵌入容量优于现有的生成式隐写方案,最高可达0.11 bpp,能够更高效地传输秘密消息。     

Automated construction of Wi-Fi-based indoor logical location predictor using crowd-sourced photos with Wi-Fi signals

Owing to the recent proliferation of smartphones and the SNS, a large number of images taken by smartphones at various places have been uploaded to SNSs. In addition, smartphones are equipped with various sensors such as Wi-Fi modules that enable us to generate an image associated with the sensory information that represents the context in which the image was captured. This study demonstrates the benefits of images associated with Wi-Fi signals in the automated construction of a Wi-Fi-based indoor logical location classifier that predicts a semantic location label of a user’s position for shopping complexes. In this study, a logical location class refers to the store class label in a shopping complex, such as Starbucks and H&M. Given a collection of images associated with Wi-Fi signals taken at a shopping complex and the complex’s floor plan, the proposed method first estimates the store label at which an image was taken by analyzing the image and crawled online images of branch stores. Then, the 2D coordinates of the images taken at branch stores on the floor coordinate system can be estimated using the floor plan. Subsequently, by using the Wi-Fi signals of the branch store images and their estimated 2D coordinates, we construct a transformation function that maps Wi-Fi signals onto the 2D coordinates, and we adopt this function to predict an indoor location class of an observed Wi-Fi scan from a smartphone possessed by an end user. The proposed transformation function comprises an ensemble of sub-functions designed based on CVAEs. Finally, we demonstrate the effectiveness of the proposed method for three actual shopping complexes.     

基于多尺度圆周频率滤波与卷积神经网络的遥感图像飞机目标检测方法研究

针对遥感图像飞机目标检测因目标尺度不一存在漏警、虚警等问题,该文基于遥感图像中飞机目标形状特征和灰度变化特点提出了一种多尺度圆周频率滤波(MSCFF)与卷积神经网络(CNN)相结合的MSCFF+CNN飞机目标自动检测算法.该算法首先采用多尺度圆周频率滤波器滤除遥感图像复杂背景,实现不同尺度飞机目标候选区域提取;然后,通过构建卷积神经网络(CNN)模型实现候选区域有效分类,最终精确确定飞机目标位置.最后,基于获取的真实遥感图像进行目标检测算法实验验证,经统计该算法的飞机目标检测率为94.38%,虚警率为3.76%,实验结果充分验证了该文算法的有效性,该算法可为机场监管、军事侦察等应用提供重要的技术支持.