Dual-Path Vision Transformer用于急性缺血性脑卒中辅助诊断

急性缺血性脑卒中是由于脑组织血液供应障碍导致的脑功能障碍,数字减影脑血管造影(DSA)是诊断脑血管疾病的金标准。基于患者的正面和侧面DSA图像,对急性缺血性脑卒中的治疗效果进行分级评估,构建基于Vision Transformer的双路径图像分类智能模型DPVF。为了提高辅助诊断速度,基于EdgeViT的轻量化设计思想进行了模型的构建;为了使模型保持轻量化的同时具有较高的精度,提出空间-通道自注意力模块,促进Transformer模型捕获更全面的特征信息,提高模型的表达能力;此外,对于DPVF的两分支的特征融合,构建交叉注意力模块对两分支输出进行交叉融合,促使模型提取更丰富的特征,从而提高模型表现。实验结果显示DPVF在测试集上的准确率达98.5%,满足实际需求。

SVMD-PE-BP-Transformer短期光伏功率预测

考虑到光伏功率受气象因素变化影响而波动性大难以预测的问题,将逐次变分模态分解SVMD-排列熵PE与BPTransformer相结合,给出了一种组合预测方法,以下简称SPBT模型。在去除非相关因子的基础上,利用SOM聚类方法,对全年光伏数据进行3种类型的分类;针对光伏发电初始时序中所蕴含的重要信息,利用SVMD自适应K值的方法,对其进行分解。再利用PE方法计算各个子序列的熵值,即序列的起伏复杂程度,根据熵的大小,对频率接近的成分进行重构,将其分为两个区间:复杂度低的部分和复杂度高的部分。最后利用BP网络与Transformer分别对其进行预测,并对预测输出进行综合处理。该文以江苏省一光伏电站观测的气象与功率数据为例,通过比较试验验证了该模型的优势,该模型具有较低的预测误差,有助于提高预测精度。

A new insight into LPSO phase transformation and mechanical properties uniformity of large-scale Mg-Gd-Y-Zn-Zr alloy prepared by multi-pass friction stir processing

A large-scale fine-grained Mg-Gd-Y-Zn-Zr alloy plate with high strength and ductility was successfully prepared by multi-pass friction stir processing(MFSP)technology in this work.The structure of grains and long period stacking ordered(LPSO)phase were characterized,and the mechanical properties uniformity was investigated.Moreover,a quantitative relationship between the microstructure and tensile yield strength was established.The results showed that the grains in the processed zone(PZ)and interfacial zone(IZ)were refined from 50μm to 3μm and 4μm,respectively,and numerous original LPSO phases were broken.In IZ,some block-shaped 18R LPSO phases were transformed into needle-like 14H LPSO phases due to stacking faults and the short-range diffusion of solute atoms.The severe shear deformation in the form of kinetic energy caused profuse stacking fault to be generated and move rapidly,greatly increasing the transformation rate of LPSO phase.After MFSP,the ultimate tensile strength,yield strength and elongation to failure of the large-scale plate were 367 MPa,305 MPa and 18.0% respectively.Grain refinement and LPSO phase strengthening were the major strengthening mechanisms for the MFSP sample.In particularly,the strength of IZ was comparable to that of PZ because the strength contribution of the 14H LPSO phase offsets the lack of grain refinement strengthening in IZ.This result opposes the widely accepted notion that IZ is a weak region in MFSP-prepared large-scale fine-grained plate.

考虑多维影响因素的改进Transformer-PSO短期电价预测方法

随着多元化电力市场的建设,电价影响因素日益增加,市场环境变化也更加剧烈.为提高市场短期电价的预测精度,提出一种考虑多种电价影响因素的改进Transformer-粒子群优化(PSO)算法短期电价预测方法.首先,在考虑历史电价、负荷的基础上进一步分析电价形成的相关因素,利用自相关函数分析电价的多周期特性并在此基础上调整输入序列,克服了仅采用历史数据以及经验调整输入序列导致预测精度受限的问题.其次,结合长短期记忆(LSTM)、自注意力机制与多层注意力机制并采用多输入结构建立改进Transformer模型,进一步提升LSTM模型捕获不同时间步信息间的长短期依赖关系的能力,克服LSTM的信息利用瓶颈,适应包括历史电价及多种电价成因的复杂多序列输入.此外,还利用PSO智能算法搜索模型不同学习阶段的最佳学习率克服手动调整学习率的局限性.最后,采用PJM市场电价进行算例分析,结果表明所提短期电价预测模型能应用于电价影响因素多、变化剧烈的市场环境,并有效提升短期电价预测精度.

Disordered Structure and Reversible Phase Transformation from K-Birnessite to Zn-Buserite Enable High-Performance Aqueous Zinc-lon Batteries

The layeredδ-MnO_(2)(dMO)is an excellent cathode material for rechargeable aqueous zinc-ion batteries owing to its large interlayer distance(~0.7 nm),high capacity,and low cost;however,such cathodes suffer from structural degradation during the long-term cycling process,leading to capacity fading.In this study,a Co-doped dMO composite with reduced graphene oxide(GC-dMO)is developed using a simple cost-effective hydrothermal method.The degree of disorderness increases owing to the hetero-atom doping and graphene oxide composites.It is demonstrated that layered dMO and GC-dMO undergo a structural transition from K-birnessite to the Zn-buserite phase upon the first discharge,which enhances the intercalation of Zn^(2+)ions,H_(2)O molecules in the layered structure.The GC-dMO cathode exhibits an excellent capacity of 302 mAh g^(-1)at a current density of 100 mAg^(-1)after 100 cycles as compared with the dMO cathode(159 mAhg^(-1)).The excellent electrochemical performance of the GC-dMO cathode owing to Co-doping and graphene oxide sheets enhances the interlayer gap and disorderness,and maintains structural stability,which facilitates the easy reverse intercalation and de-intercalation of Zn^(2+)ions and H_(2)O molecules.Therefore,GC-dMO is a promising cathode material for large-scale aqueous ZIBs.

Triple-path feature transform network for ring-array photoacoustic tomography image reconstruction

Photoacoustic imaging(PAI)is a noninvasive emerging imaging method based on the photoacoustic effect,which provides necessary assistance for medical diagnosis.It has the characteristics of large imaging depth and high contrast.However,limited by the equipment cost and reconstruction time requirements,the existing PAI systems distributed with annular array transducers are difficult to take into account both the image quality and the imaging speed.In this paper,a triple-path feature transform network(TFT-Net)for ring-array photoacoustic tomography is proposed to enhance the imaging quality from limited-view and sparse measurement data.Specifically,the network combines the raw photoacoustic pressure signals and conventional linear reconstruction images as input data,and takes the photoacoustic physical model as a prior information to guide the reconstruction process.In addition,to enhance the ability of extracting signal features,the residual block and squeeze and excitation block are introduced into the TFT-Net.For further efficient reconstruction,the final output of photoacoustic signals uses‘filter-then-upsample’operation with a pixel-shuffle multiplexer and a max out module.Experiment results on simulated and in-vivo data demonstrate that the constructed TFT-Net can restore the target boundary clearly,reduce background noise,and realize fast and high-quality photoacoustic image reconstruction of limited view with sparse sampling.

Major-Countries Competition and Pivot to the Asia-Pacific:the Strategic Transformation and Influence of“Global NATO”

As a product of the Cold War,NATO did not disintegrate with the end of it,but rather,it has sought to become“global NATO”through continuous expansion and transformation,in order to hold wider sway in international security.In its 75-year history,NATO has made numerous moves including adopting confrontational security thinking,building exclusive alliances,conducting humanitarian intervention and crisis management,expanding eastward and northward and involving in the Asia-Pacific.

Efficient single-pixel imaging encrypted transmission based on 3D Arnold transformation

Single-pixel imaging(SPI)can transform 2D or 3D image data into 1D light signals,which offers promising prospects for image compression and transmission.However,during data communication these light signals in public channels will easily draw the attention of eavesdroppers.Here,we introduce an efficient encryption method for SPI data transmission that uses the 3D Arnold transformation to directly disrupt 1D single-pixel light signals and utilizes the elliptic curve encryption algorithm for key transmission.This encryption scheme immediately employs Hadamard patterns to illuminate the scene and then utilizes the 3D Arnold transformation to permutate the 1D light signal of single-pixel detection.Then the transformation parameters serve as the secret key,while the security of key exchange is guaranteed by an elliptic curve-based key exchange mechanism.Compared with existing encryption schemes,both computer simulations and optical experiments have been conducted to demonstrate that the proposed technique not only enhances the security of encryption but also eliminates the need for complicated pattern scrambling rules.Additionally,this approach solves the problem of secure key transmission,thus ensuring the security of information and the quality of the decrypted images.

Mucosa color and size may indicate malignant transformation of chicken skin mucosa-positive colorectal neoplastic polyps

BACKGROUND Lipid metabolism reprogramming is suspected to exist in pre-cancerous lesions,including colorectal adenoma.Screening colonoscopy frequently reveals chicken skin mucosa(CSM;white or yellow-white speckled mucosa)surrounding colo-rectal polyps,caused by macrophages engulfing and accumulating the lipids decomposed by colon cells or adjacent tumors.CSM-positive colorectal polyps are associated with various diseases;however,their prognosis varies greatly.Cold snare polypectomy is commonly used to resect lesions up to 10 to 15 mm in diameter without signs of submucosal invasion but is controversial for CSM-positive colorectal polyps.Improved imaging is required to diagnose and treat CSM-positive colorectal polyps.METHODS This retrospective cohort study included 177 patients with CSM-positive colorectal polyps diagnosed using endoscopy.All patient-related information was extracted from the Goldisc soft-clinic DICOM system or electronic medical record system.Based on the pathological results,patients were classified as non-neoplastic polyps(five juvenile polyps),neoplastic polyps,non-invasive high-grade neoplasia(NHGN),or submucosal invasive carcinoma(SM stage cancer).We analyzed and compared the clinical features,suspected risk factors for malignant transformation of neoplastic polyps,and early infiltration of sub-mucosal carcinoma.RESULTS The diameters of NHGN and SM polyps were much smaller than those of neoplastic polyps.Most NHGN polyps had a deeper red mucosal color.On logistic regression analyses,diameter and deeper red mucosal color were independent risk factors for malignant transformation of neoplastic polyps.Type 1 CSM was more common in high-grade intraepithelial neoplasia and SM;type 2 CSM was more common in neoplastic polyps.Logistic regression analyses revealed no significant differences in the malignant transformation of neoplastic polyps or early submucosal invasion of CSM-positive colorectal cancer.Changes in the CSM mucosa surrounding neoplastic polyps and submucosal invasion of colorectal cancer disappeared within 12 months.No tumor recurrence was found during either partial or complete endoscopic resection of the CSM.CONCLUSION CSM-positive colorectal polyps>1 cm in diameter or with deeper red mucosa may be related to NHGN.Resection of CSM surrounding colorectal adenomas did not affect tumor recurrence.

Online Capacitor Voltage Transformer Measurement Error State Evaluation Method Based on In-Phase Relationship and Abnormal Point Detection

The assessment of the measurement error status of online Capacitor Voltage Transformers (CVT) within the power grid is of profound significance to the equitable trade of electric energy and the secure operation of the power grid. This paper advances an online CVT error state evaluation method, anchored in the in-phase relationship and outlier detection. Initially, this method leverages the in-phase relationship to obviate the influence of primary side fluctuations in the grid on assessment accuracy. Subsequently, Principal Component Analysis (PCA) is employed to meticulously disentangle the error change information inherent in the CVT from the measured values and to compute statistics that delineate the error state. Finally, the Local Outlier Factor (LOF) is deployed to discern outliers in the statistics, with thresholds serving to appraise the CVT error state. Experimental results incontrovertibly demonstrate the efficacy of this method, showcasing its prowess in effecting online tracking of CVT error changes and conducting error state assessments. The discernible enhancements in reliability, accuracy, and sensitivity are manifest, with the assessment accuracy reaching an exemplary 0.01%.

基于RoBERTa和图增强Transformer的序列推荐方法

自推荐系统出现以来,有限的数据信息就一直制约着推荐算法的进一步发展。为降低数据稀疏性的影响,增强非评分数据的利用率,基于神经网络的文本推荐模型相继被提出,但主流的卷积或循环神经网络在文本语义理解和长距离关系捕捉方面存在明显劣势。为了更好地挖掘用户与商品之间的深层潜在特征,进一步提高推荐质量,提出一种基于Ro BERTa和图增强Transformer的序列推荐(RGT)模型。引入评论文本数据,首先利用预训练的Ro BERTa模型捕获评论文本中的字词语义特征,初步建模用户的个性化兴趣,然后根据用户与商品的历史交互信息,构建具有时序特性的商品关联图注意力机制网络模型,通过图增强Transformer的方法将图模型学习到的各个商品的特征表示以序列的形式输入Transformer编码层,最后将得到的输出向量与之前捕获的语义表征以及计算得到的商品关联图的全图表征输入全连接层,以捕获用户全局的兴趣偏好,实现用户对商品的预测评分。在3组真实亚马逊公开数据集上的实验结果表明,与Deep FM、Conv MF等经典文本推荐模型相比,RGT模型在均方根误差(RMSE)和平均绝对误差(MAE)2种指标上有显著提升,相较于最优对比模型最高分别提升4.7%和5.3%。

基于Depth-wise卷积和视觉Transformer的图像分类模型

图像分类作为一种常见的视觉识别任务,有着广阔的应用场景。在处理图像分类问题时,传统的方法通常使用卷积神经网络,然而,卷积网络的感受野有限,难以建模图像的全局关系表示,导致分类精度低,难以处理复杂多样的图像数据。为了对全局关系进行建模,一些研究者将Transformer应用于图像分类任务,但为了满足Transformer的序列化和并行化要求,需要将图像分割成大小相等、互不重叠的图像块,破坏了相邻图像数据块之间的局部信息。此外,由于Transformer具有较少的先验知识,模型往往需要在大规模数据集上进行预训练,因此计算复杂度较高。为了同时建模图像相邻块之间的局部信息并充分利用图像的全局信息,提出了一种基于Depth-wise卷积的视觉Transformer(Efficient Pyramid Vision Transformer,EPVT)模型。EPVT模型可以实现以较低的计算成本提取相邻图像块之间的局部和全局信息。EPVT模型主要包含3个关键组件:局部感知模块(Local Perceptron Module,LPM)、空间信息融合模块(Spatial Information Fusion,SIF)和“+卷积前馈神经网络(Convolution Feed-forward Network,CFFN)。LPM模块用于捕获图像的局部相关性;SIF模块用于融合相邻图像块之间的局部信息,并利用不同图像块之间的远距离依赖关系,提升模型的特征表达能力,使模型学习到输出特征在不同维度下的语义信息;CFFN模块用于编码位置信息和重塑张量。在图像分类数据集ImageNet-1K上,所提模型优于现有的同等规模的视觉Transformer分类模型,取得了82.6%的分类准确度,证明了该模型在大规模数据集上具有竞争力。

基于Transformer和自适应特征融合的矿井低照度图像亮度提升和细节增强方法

高质量矿井影像为矿山安全生产提供保障,也有利于提高后续图像分析技术的性能。矿井影像受低照度环境的影响,易出现亮度低,照度不均,颜色失真,细节信息丢失严重等问题。针对上述问题,提出一种基于Transformer和自适应特征融合的矿井低照度图像亮度提升和细节增强方法。基于生成对抗思想搭建生成对抗式主体模型框架,使用目标图像域而非单一参考图像驱动判别器监督生成器的训练,实现对低照度图像的充分增强;基于特征表示学习理论搭建特征编码器,将图像解耦为亮度分量和反射分量,避免图像增强过程中亮度与颜色特征相互影响从而导致颜色失真问题;设计CEM-Transformer Encoder通过捕获全局上下文关系和提取局部区域特征,能够充分提升整体图像亮度并消除局部区域照度不均;在反射分量增强过程中,使用结合CEM-Cross-Transformer Encoder的跳跃连接将低级特征与深层网络处特征进行自适应融合,能够有效避免细节特征丢失,并在编码网络中添加ECA-Net,提高浅层网络的特征提取效率。制作矿井低照度图像数据集为矿井低照度图像增强任务提供数据资源。试验显示,在矿井低照度图像数据集和公共数据集中,与5种先进的低照度图像增强算法相比,该算法增强图像的质量指标PSNR、SSIM、VIF平均提高了16.564%,10.998%,16.226%和14.438%,10.888%,14.948%,证明该算法能够有效提升整体图像亮度,消除照度不均,避免颜色失真和细节丢失,实现矿井低照度图像增强。

CNN-Transformer特征融合多目标跟踪算法

在卷积神经网络(CNN)中,卷积运算能高效地提取目标的局部特征,却难以捕获全局表示;而在视觉Transformer中,注意力机制可以捕获长距离的特征依赖,但会忽略局部特征细节。针对以上问题,提出一种基于CNN-Transformer双分支主干网络进行特征提取和融合的多目标跟踪算法CTMOT(CNN-transformer multi-object tracking)。使用基于CNN和Transformer双分支并行的主干网络分别提取图像的局部和全局特征。使用双向桥接模块(two-way braidge module,TBM)对两种特征进行充分融合。将融合后的特征输入两组并行的解码器进行处理。将解码器输出的检测框和跟踪框进行匹配,完成多目标跟踪任务。在多目标跟踪数据集MOT17、MOT20、KITTI以及UADETRAC上进行评估,CTMOT算法的MOTP和IDs指标在四个数据集上均达到了SOTA效果,MOTA指标分别达到了76.4%、66.3%、92.36%和88.57%,在MOT数据集上与SOTA方法效果相当,在KITTI数据集上达到SOTA效果。由于同时完成目标检测和关联,能够端到端进行目标跟踪,跟踪速度可达35 FPS,表明CTMOT算法在跟踪的实时性和准确性上达到了较好的平衡,具有较大潜力。

基于残差U-Net和自注意力Transformer编码器的磁场预测方法

利用有限元方法对几何结构复杂的电机和变压器进行磁场分析,存在仿真时间长且无法复用的问题。因此,该文提出一种基于残差U-Net和自注意力Transformer编码器的磁场预测方法。首先建立永磁同步电机(PMSM)和非晶合金变压器(AMT)有限元模型,得到深度学习训练所需的数据集;然后将Transformer模块与U-Net模型结合,并引入短残差机制建立ResUnet-Transformer模型,通过预测图像的像素实现磁场预测;最后通过Targeted Dropout算法和动态学习率调整策略对模型进行优化,解决拟合问题并提高预测精度。计算实例证明,ResUnet-Transformer模型在PMSM和AMT数据集上测试集的平均绝对百分比误差(MAPE)均小于1%,且仅需500组样本。该文提出的磁场预测方法能减少实际工况和多工况下精细模拟和拓扑优化的时间和资源消耗,亦是虚拟传感器乃至数字孪生的关键实现方法之一。

融合卷积注意力和Transformer架构的行人重识别方法

行人重识别技术是智能安防系统中的重要方法之一,为构建一个适用各种复杂场景的行人重识别模型,基于现有的卷积神经网络和Transformer模型,提出一种融合卷积注意力和Transformer(FCAT)架构的方法,以增强Transformer对局部细节信息的关注。所提方法主要将卷积空间注意力和通道注意力嵌入Transformer架构中,分别加强对图像中重要区域的关注和对重要通道特征的关注,以进一步提高Transformer架构对局部细节特征的提取能力。在3个公开行人重识别数据集上的对比消融实验证明,所提方法在非遮挡数据集上取得了与现有方法相当的结果,在遮挡数据集上的性能得到显著提升。所提方法更加轻量化,在不增加额外计算量和模型参数的情况下,推理速度得到了提升。

基于Transformer和动态3D卷积的多源遥感图像分类

多源遥感数据具有互补性和协同性,近年来,基于深度学习的方法已经在多源遥感图像分类中取得了一定进展,但当前方法仍面临关键难题,如多源遥感图像特征表达不一致,融合困难,基于静态推理范式的神经网络缺乏对不同类别地物的适应性。为解决上述问题,提出了基于跨模态Transformer和多尺度动态3D卷积的多源遥感图像分类模型。为提高多源特征表达的一致性,设计了基于Transformer的融合模块,借助其强大的注意力建模能力挖掘高光谱和LiDAR数据特征之间的相互作用;为提高特征提取方法对不同地物类别的适应性,设计了多尺度动态3D卷积模块,将输入特征的多尺度信息融入卷积核的调制,提高卷积操作对不同地物的适应性。采用多源遥感数据集Houston和Trento对所提方法进行验证,实验结果表明:所提方法在Houston和Trento数据集上总体准确率分别达到94.60%和98.21%,相比MGA-MFN等主流方法,总体准确率分别至少提升0.97%和0.25%,验证了所提方法可有效提升多源遥感图像分类的准确率。

考虑特征重组与改进Transformer的风电功率短期日前预测方法

短期日前风电功率预测对电力系统调度计划制定有重要意义,该文为提高风电功率预测的准确性,提出了一种基于Transformer的预测模型Powerformer。模型通过因果注意力机制挖掘序列的时序依赖;通过去平稳化模块优化因果注意力以提高数据本身的可预测性;通过设计趋势增强和周期增强模块提高模型的预测能力;通过改进解码器的多头注意力层,使模型提取周期特征和趋势特征。该文首先对风电数据进行预处理,采用完全自适应噪声集合经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)将风电数据序列分解为不同频率的本征模态函数并计算其样本熵,使得风电功率序列重组为周期序列和趋势序列,然后将序列输入到Powerformer模型,实现对风电功率短期日前准确预测。结果表明,虽然训练时间长于已有预测模型,但Poweformer模型预测精度得到提升;同时,消融实验结果验证了模型各模块的必要性和有效性,具有一定的应用价值。

基于遥感多参数和CNN-Transformer的冬小麦单产估测

为了提高冬小麦单产估测精度,改善估产模型存在的高产低估和低产高估等现象,以陕西省关中平原为研究区域,选取旬尺度条件植被温度指数(VTCI)、叶面积指数(LAI)和光合有效辐射吸收比率(FPAR)为遥感特征参数,结合卷积神经网络(CNN)局部特征提取能力和基于自注意力机制的Transformer网络的全局信息提取能力,构建CNN-Transformer深度学习模型,用于估测关中平原冬小麦产量。与Transformer模型(R^(2)为0.64,RMSE为465.40 kg/hm^(2),MAPE为8.04%)相比,CNN-Transformer模型具有更高的冬小麦单产估测精度(R^(2)为0.70,RMSE为420.39 kg/hm^(2),MAPE为7.65%),能够从遥感多参数中提取更多与产量相关的信息,且对于Transformer模型存在的高产低估和低产高估现象均有所改善。基于5折交叉验证法和留一法进一步验证了CNN-Transformer模型的鲁棒性和泛化能力。此外,基于CNN-Transformer模型捕获冬小麦生长过程的累积效应,分析逐步累积旬尺度输入参数对产量估测的影响,评估模型对于冬小麦不同生长阶段的累积过程的表征能力。结果表明,模型能有效捕捉冬小麦生长的关键时期,3月下旬至5月上旬是冬小麦生长的关键时期。

基于Transformer的多尺度遥感语义分割网络

为了提升遥感图像语义分割效果,本文针对分割目标类间方差小、类内方差大的特点,从全局上下文信息和多尺度语义特征2个关键点提出一种基于Transformer的多尺度遥感语义分割网络(muliti-scale Transformer network,MSTNet)。其由编码器和解码器2个部分组成,编码器包含基于Transformer改进的视觉注意网络(visual attention network,VAN)主干和基于空洞空间金字塔池化(atrous spatial pyramid pooling, ASPP)结构改进的多尺度语义特征提取模块(multi-scale semantic feature extraction module, MSFEM)。解码器采用轻量级多层感知器(multi-layer perception,MLP)配合编码器设计,充分分析所提取的包含全局上下文信息和多尺度表示的语义特征。MSTNet在2个高分辨率遥感语义分割数据集ISPRS Potsdam和LoveDA上进行验证,平均交并比(mIoU)分别达到79.50%和54.12%,平均F1-score(m F1)分别达到87.46%和69.34%,实验结果验证了本文所提方法有效提升了遥感图像语义分割的效果。