Traffic Sign Recognition for Autonomous Vehicle Using Optimized YOLOv7 and Convolutional Block Attention Module

The infrastructure and construction of roads are crucial for the economic and social development of a region,but traffic-related challenges like accidents and congestion persist.Artificial Intelligence(AI)and Machine Learning(ML)have been used in road infrastructure and construction,particularly with the Internet of Things(IoT)devices.Object detection in Computer Vision also plays a key role in improving road infrastructure and addressing trafficrelated problems.This study aims to use You Only Look Once version 7(YOLOv7),Convolutional Block Attention Module(CBAM),the most optimized object-detection algorithm,to detect and identify traffic signs,and analyze effective combinations of adaptive optimizers like Adaptive Moment estimation(Adam),Root Mean Squared Propagation(RMSprop)and Stochastic Gradient Descent(SGD)with the YOLOv7.Using a portion of German traffic signs for training,the study investigates the feasibility of adopting smaller datasets while maintaining high accuracy.The model proposed in this study not only improves traffic safety by detecting traffic signs but also has the potential to contribute to the rapid development of autonomous vehicle systems.The study results showed an impressive accuracy of 99.7%when using a batch size of 8 and the Adam optimizer.This high level of accuracy demonstrates the effectiveness of the proposed model for the image classification task of traffic sign recognition.

ANC: Attention Network for COVID-19 Explainable Diagnosis Based on Convolutional Block Attention Module

Aim: To diagnose COVID-19 more efficiently and more correctly, this study proposed a novel attention network forCOVID-19 (ANC). Methods: Two datasets were used in this study. An 18-way data augmentation was proposed toavoid overfitting. Then, convolutional block attention module (CBAM) was integrated to our model, the structureof which is fine-tuned. Finally, Grad-CAM was used to provide an explainable diagnosis. Results: The accuracyof our ANC methods on two datasets are 96.32% ± 1.06%, and 96.00% ± 1.03%, respectively. Conclusions: Thisproposed ANC method is superior to 9 state-of-the-art approaches.

MobileNet network optimization based on convolutional block attention module

Deep learning technology is widely used in computer vision.Generally,a large amount of data is used to train the model weights in deep learning,so as to obtain a model with higher accuracy.However,massive data and complex model structures require more calculating resources.Since people generally can only carry and use mobile and portable devices in application scenarios,neural networks have limitations in terms of calculating resources,size and power consumption.Therefore,the efficient lightweight model MobileNet is used as the basic network in this study for optimization.First,the accuracy of the MobileNet model is improved by adding methods such as the convolutional block attention module(CBAM)and expansion convolution.Then,the MobileNet model is compressed by using pruning and weight quantization algorithms based on weight size.Afterwards,methods such as Python crawlers and data augmentation are employed to create a garbage classification data set.Based on the above model optimization strategy,the garbage classification mobile terminal application is deployed on mobile phones and raspberry pies,realizing completing the garbage classification task more conveniently.

Numerical Analyses of Electromagnetic Forces on the ITER Blanket Module Shield Block During Major Disruptions

Electromagnetic（EM） load is one of the key design drivers for the blanket shield block（SB） and other in-vessel components. In this article, an EM analysis method was developed to address the EM force on the SB. The plasma currents, which vary spatially and temporally,are loaded as a filament at each time point. The standard blanket module No.04（BM04） under major disruption（MD） is selected to perform the analyses. The analyses results are validated by comparing currents on the passive structure. To better understand the effects of cooling channels and slits on the EM force, the case of SB without cooling channel and the case without slits are calculated to make comparisons. The results show that the slits play an important role in controlling the EM load on SB.

Spatial-Modulated Physical-Layer Network Coding Based on Block Markov Superposition Transmission for Maritime Relay Communications

As an alternative to satellite communications,multi-hop relay networks can be deployed for maritime long-distance communications.Distinct from terrestrial environment,marine radio signals are affected by many factors,e.g.,weather conditions,evaporation ducting,and ship rocking caused by waves.To ensure the data transmission reliability,the block Markov superposition transmission(BMST)codes,which are easily configurable and have predictable performance,are applied in this study.Meanwhile,the physical-layer network coding(PNC)scheme with spatial modulation(SM)is adopted to improve the spectrum utilization.For the BMST-SMPNC system,we propose an iterative algorithm,which utilizes the channel observations and the a priori information from BMST decoder,to compute the soft information corresponding to the XORed bits constructed by the relay node.The results indicate that the proposed scheme outperforms the convolutional coded SM-PNC over fast-fading Rician channels.Especially,the performance can be easily improved in high spatial correlation maritime channel by increasing the memory m.

Differential modulation based on space-time block codes

A differential modulation scheme using space-time block codes is put forward. Compared with other schemes, our scheme has lower computational complexity and has a simpler decoder. In the case of three or four transmitter antennas, our scheme has a higher rate a higher coding gain and a lower bit error rate for a given rate. Then we made simulations for space-time block codes as well as group codes in the case of two, three, four and five transmit antennas. The simulations prove that using two transmit antennas, one receive antenna and code rate of 4 bits/s/Hz, the differential STBC method outperform the differential group codes method by 4 dB. Useing three, four and five transmit antennas, one receive antenna, and code rate of 3 bits/s/Hz are adopted, the differential STBC method outperform the differential group codes method by 5 dB, 6. 5 dB and 7 dB, respectively. In other words, the differential modulation scheme based on space-time block code is better than the corresponding differential modulation scheme

MAKING FULL USE OF OUTDATED CHANNEL ESTIMATES FOR BLOCK BY BLOCK ADAPTIVE MODULATIONS

This paper explores the potential to use accurate but outdated channel estimates for adaptive modulation. The work is novel in that the research is conditioned on block by block adaptation. First,we define a new quantity,the Tolerable Average Use Delay (TAUD),which can indicate the ability of an adaptation scheme to tolerate the delay of channel estimation results. We find that for the variable-power schemes,TAUD is a constant and dependent on the target Bit Error Rate (BER),average power and Doppler frequency; while for the constant-power schemes,it depends on the ad-aptation block length as well. At last,we investigate the relation between the delay tolerating per-formance and the spectral efficiency and give the system design criterion. The delay tolerating per-formance is improved at the price of lower data rate.

基于改进YOLOv5s的轻量级绝缘子缺失检测

针对现有绝缘子缺失检测模型计算复杂度高和小目标难以检测等问题,提出一种基于改进的YOLOv5s轻量级检测模型。首先,移除主干网络中的C3模块来减少模型的参数量。其次,在多尺度特征融合网络中引入卷积块注意力机制来提高复杂背景下模型的特征提取能力。同时,采用加权双向特征金字塔网络结构对特征进行双向跨尺度加权融合,提升网络在遮挡物、相似目标干扰下目标的检测性能。最后,选用SIoU损失函数提升网络的收敛速度和检测精度。实验结果表明,所提模型的平均精准率为96.8%,浮点运算数为2.8 GFLOPS,而原始YOLOv5s在保证97.4%的平均精准率下的浮点运算数为16.3 GFLOPS。相较于原始模型,所提模型对小目标、遮挡目标以及模糊等场景有着较强的鲁棒性,且在保证近似检测精度的同时极大减少了计算量。

融合Inception V1-CBAM-CNN的轴承剩余寿命预测模型

针对现有的滚动轴承剩余寿命(RUL)预测方法精度低、轴承健康指标(HI)构建困难等问题,提出了一种基于卷积神经网络(CNN)并融合Inception V1模块和卷积注意力机制模块(CBAM)的滚动轴承RUL预测模型。首先,在CNN中添加了CBAM机制,并进行了加权处理,在通道和空间维度对重要特征进行了强化,对次要特征进行了抑制,通过添加改进的InceptionV1模块,提高了CNN通道间信息交互水平,全面提取了退化特征;然后,进行了网络优化,采用全局最大池化(GMP)方法对模型进行了简化,采用Dropout和批量归一化(BN)方法,避免了过拟合,提高了精度,且克服了训练时出现的梯度消失问题;最后,对数据进行了处理,将降噪后的信号重组为三维张量,将其作为HI,构建了退化标签,引入了评价指标,采用PHM2012轴承数据集进行了实验验证,在3种工况下将其与深度神经网络(DNN)、CNN方法、结合注意力机制的残差网络方法(ResNet)进行了对比。研究结果表明:该方法在变负载条件下的平均RMSE为0.033,较其他方法的RMSE值分别降低了86%、78%和69%,在预测精度和泛化能力方面具有明显优势。

基于DenseNet和卷积注意力模块的高精度手势识别

非接触的手势识别是一种新型人机交互方式,在增强现实(AR)/虚拟现实(VR)、智能家居、智能医疗等方面有着广阔的应用前景,近年来成为一个研究热点。由于需要利用毫米波雷达进行更精确的微动手势识别,该文提出一种新型的基于MIMO毫米波雷达的微动手势识别方法。采用4片AWR1243雷达板级联而成的毫米波级联(MMWCAS)雷达采集手势回波,对手势回波进行时频分析,基于距离-多普勒(RD)图和3D点云检测出人手目标。通过数据预处理,提取手势目标的距离-时间谱图(RTM)、多普勒-时间谱图(DTM)、方位角-时间谱图(ATM)和俯仰角-时间谱图(ETM),更加全面地表征手势的运动特征,并形成混合特征谱图(FTM),对12种微动手势进行识别。设计了基于DenseNet和卷积注意力模块的手势识别网络,将混合特征谱图作为网络的输入,创新性地融合了卷积注意力模块(CBAM),实验表明,识别准确率达到99.03%,且该网络将注意力放在手势动作的前半段,实现了高精度的手势识别。

基于改进EfficientNet的红外图像光伏组件故障识别研究

光伏组件的故障会影响光伏阵列的输出性能,从而降低电站的发电效率,严重时甚至会危害电站的安全运行。传统的方法无法满足目前光伏组件故障检测快速性和正确率需求。因此,本文提出了一种基于改进EfficientNet的光伏组件故障识别方法。首先,利用采集到的光伏组件红外图像建立故障数据集,并利用图像分割和数据增强对数据集进行预处理;其次,基于EfficientNet网络构建故障识别模型,同时在模型中引入双通道注意力模块(CBAM),该模块能够抑制不必要特征的识别,增强模型对空间特征信息的提取能力,进而提高模型的识别准确率;最后,通过对比仿真实验证明模型的有效性和先进性。实验结果表明,该模型的故障识别准确率达到了90.83%,相较于原始的EfficientNet模型提高了2.83%,且模型大小仅为20.3 M,具有良好的实用性,能够满足光伏电站实际应用的需求。

基于机器视觉的垃圾分类算法研究与应用

垃圾分类识别算法是目前研究的热点问题,本文通过引入色块追踪模块Lab颜色模型对YOLOv3算法进行优化,利用优化后的算法搭建训练模型。并针对目前垃圾类别利用网络爬虫爬取日常生活中常见的垃圾图像并进行分类,形成数据集。其次通过优化的YOLOv3算法对处理好的数据集进行模型训练,将训练后的模型进行模型检测。最后通过实际测试,优化后的YOLOv3算法识别的平均准确率达到了94.33%,与原始算法相比,优化后的算法在稳定性和准确度上都有了明显的改善。

WGAN-GP结合CBAM-VGG16轻量化网络滚动轴承故障诊断

针对滚动轴承故障诊断中存在的故障数据不平衡且诊断效率低的问题,提出一种将改进的生成对抗网络(Wasserstein Generative Adversarial Network with Gradient Penalty,WGAN-GP)和轻量化卷积神经网络相结合的滚动轴承故障诊断方法。首先,利用连续小波变换(Continuous Wavelet Transform,CWT)生成二维时频图,并通过WGAN-GP进行数据增强;其次在视觉几何群网络(Visual Geometry Group Network-16,VGG16)的基础上,引入Ghost模块和全局平均池化(Global Average Pooling,GAP)对其进行轻量化改进;再次,利用卷积注意力模块(Convolutional Block Attention Module,CBAM)和带重启的余弦退火衰减法提高VGG16模型的性能,构建CBAM-VGG16轻量化卷积神经网络模型,将增强后的数据进行预处理后输入到模型中进行训练,建立故障诊断模型;最后采用西储大学轴承数据集进行模型验证和分析。实验结果表明:该方法证实了故障数据不足时进行故障诊断的可行性,缩短了模型的训练时间、诊断时间并缩减了模型的大小和参数量,提高了故障诊断的效率和准确率。

基于改进ResNet34网络的变电站设备巡检图像分类识别的方法

针对变电站设备巡检图像识别领域中存在的图像规模有限和识别准确率低等问题,提出了一种基于改进ResNet34网络的图像分类识别方法。采用Seam Carving算法对图像中的低能量区域进行压缩以保留关键特征;同时使用弹性变换、高斯噪声等6种图像增强技术来增强图像的多样性。将基础ResNet34网络与卷积注意力模块结合,增强模型对设备巡检图像关键特征的提取能力。使用在ImageNet数据集上的预训练模型作为迁移学习的特征提取器来解决样本数量不足的问题。在Adam优化器中引入余弦退火策略来动态调整学习率使改进的ResNet34网络更快收敛至最优解。试验结果表明所提方法比基础ResNet34网络的准确率提升了0.0733,损失率降低了0.2019,为变电站设备巡检图像识别领域提供了一种可靠的解决方案。

改进卷积注意力机制的轻量级检测无人机目标模型

利用深度学习中特征提取的优势,提出一种改进算法,结合改进的卷积注意力模块,将YOLOv5模型骨干网络替换为改进的MobileNetv2轻量化网络,形成I-YOLOv5算法,旨在提高检测精确度和小目标、多目标的检测能力,同时保持实时性.为构建数据集,通过网络搜索和自主录制无人机视频的方式,用Label Img工具完成标注.结果表明,I-YOLOv5算法在检测精度上有显著提升,对小目标和多目标的检测效果更优秀,在视频检测方面表现出色,具有较好的实时性能.通过模型结构优化,使检测模型的大小减少为原来的18.6%,检测速度提升120%.I-YOLOv5算法的平均精度均值达到97.8%.

基于注意力机制和迁移学习的服装分类方法

针对服装图像分类效率低、准确率不高等问题,提出了一种基于注意力机制和迁移学习的服装图像分类方法。主要采用预训练的ResNet50网络模型在服装数据集上进行迁移学习,以降低对数据集的依赖,并减少网络训练时间;通过几何变换和颜色抖动2种数据增强手段处理数据集,提高模型的泛化能力;以ResNet50为基础网络,加入卷积注意力机制模块(convolutional block attention module, CBAM),依次从通道和空间2个维度提高对服装不同区域的关注度,增强了特征表达能力。在CD和IDFashion两类背景干扰信息不同的数据集上进行验证,实验结果表明:所提出的模型能够提取更多服装特征信息,在IDFashion数据集的平均分类准确率为95.60%,分别高于ResNet50、ResNet50+STN和ResNet50+ECA模型6.65%、6.69%、6.62%,一定程度上提高了服装图像分类的准确率和效率。

基于CBAM-CNN的电力系统暂态电压稳定评估

为进一步提高电力系统暂态电压稳定评估模型的特征提取能力和模型在系统拓扑结构发生变化时的适应性,提出一种将改进的卷积神经网络与迁移学习相结合的方法。首先,在卷积神经网络的卷积层后插入卷积块注意力模块,对输入的数据从通道和空间两个独立的维度依次提取特征,提高卷积神经网络对系统暂态电压状态的识别能力。然后,将该模块与微调技术相结合,提高模型在系统拓扑结构改变时的在线更新速度。最后,算例分析验证了所提模型的有效性。

基于残差卷积网络的多传感器融合永磁同步电机故障诊断

作为工业生产与日常生活的常见设备,永磁同步电机的故障诊断研究具有十分重要的意义。以永磁同步电机的匝间短路、退磁、轴承故障为诊断目标,提出一种新型的多传感器特征融合网络(MSFFN),结合多传感器融合技术与卷积神经网络实现永磁同步电机的可靠故障诊断。网络采用2个带有残差模块的卷积神经网络,对输入的电流信号与振动信号并行提取隐藏特征,并设计一种中间特征融合模块(IFFM)有效融合电流和振动的各层隐藏特征,IFFM基于注意力机制对网络中的电流特征与振动特征进行筛选,自适应关注不同信号的内在相关特征,以实现更好的诊断效果。搭建了故障样机测试平台进行数据采集与实验验证,实验结果表明,提出方法具有更高的诊断准确率,同时在叠加了强噪声的条件下,具备更强的抗干扰能力。

面向超分辨率重建的层次间局部特征增强网络

基于卷积神经网络的超分辨率重建模型以单项传播为主,层次越靠后感知信息的能力越微弱,导致层次间局部特征部分丢失,难以实质提升网络的特征表达能力。针对此问题,提出层次间局部特征增强网络。该方法由级联残差模块、层次间特征增强块和特征感知注意力机制组成。级联残差模块通过有效残差连接增加对残差分支信息的利用;层次间特征增强块提取不同深度特征的依赖关系,自适应调整中间层特征权值增强捕获关键信息的能力;特征感知注意力机制采用方向感知和位置判断的方式准确定位和识别感兴趣对象。多项标准数据集的实验结果表明,该方法能改善超分辨率的视觉重建效果,整体性能优于现有方法。

基于端口注意力与通道空间注意力的网络异常流量检测

网络异常流量检测是网络安全保护重要组成部分之一。目前,基于深度学习的异常流量检测方法都是将端口号属性与其他流量属性同等对待,忽略了端口号的重要性。为了提高异常流量检测性能,借鉴注意力思想,提出一个卷积神经网络(CNN)结合端口注意力模块(PAM)和通道空间注意力模块(CBAM)的网络异常流量检测模型。首先,将原始网络流量作为PAM的输入,分离得到端口号属性送入全连接层,得到学习后的端口注意力权重值,并与其他流量属性点乘,输出端口注意力后的流量数据;其次,将流量数据转换成灰度图,利用CNN和CBAM更充分地提取特征图在通道和空间上的信息;最后,使用焦点损失函数解决数据不平衡的问题。所提PAM具有参数量少、即插即用和普遍适用的优点。在CICIDS2017数据集上,所提模型的异常流量检测二分类任务准确率为99.18%,多分类任务准确率为99.07%,对只有少数训练样本的类别也有较高的识别率。