基于Point Transformer v2的点云枝叶分离方法研究

准确高效的点云枝叶分离对精确计算森林树木的垂直参数至关重要。然而,当前的研究方法计算成本高,且依赖先验知识导致泛化能力不足。针对以上问题,文章提出利用基于点特征的Transformer网络进行自动化的森林场景三维点云的枝叶分离研究。该方法使用Point Transformer v2网络,首先利用网格编码模块提取可学习的局部结构关系,保留点云的几何拓扑结构;其次使用分组注意力实现多通道联合学习,降低特征的冗余度,提高计算的效率;最后构建了基于点的Transformer网络实现高精度森林树木三维点云语义分割,降低了对于先验知识的需求。使用地基激光扫描仪获取的加拿大和芬兰7个不同树种样地的三维点云数据,进行枝叶分离实验和精度评价。实验结果表明,网络整体精度(OA)为94.42%,mIoU为78.89%,能够适应不同树种、不同点云密度的森林场景的枝叶分离。

基于YOLO V5和Transformer的实时手语识别和翻译方法研究

针对手语识别和翻译的需求,文中首先结合OpenCV和Mediapipe hands模型,制作了一个包含多种手势、不同角度、明暗区域、简单和复杂背景等特定环境下的手语数据集,该数据集共计3000张图像。其次,为了实现更准确和实时的手语识别和翻译,选用YOLO V5模型进行了手语识别检测。在测试集上,平均准确率mAP(0.5)达99.1%。然后,结合Transformer模型进行手语孤立词连接,实现手语句子的生成。这些技术的应用能为听障人士提供更智能、便捷的交流方式,将在未来得到更广泛的应用。

基于S-YOLO V5和Vision Transformer的视频内容描述算法

视频内容描述的自动生成是结合计算机视觉和自然语言处理等相关技术提出的一种新型交叉学习任务。针对当前视频内容生成描述模型可读性不佳的问题,本研究提出一种基于S-YOLO V5和Vison Transformer(ViT)的视频内容描述算法。首先,基于神经网络模型KATNA提取关键帧,以最少帧数进行模型训练;其次,利用S-YOLO V5模型提取视频帧中的语义信息,并结合预训练ResNet101模型和预训练C3D模型提取视频静态视觉特征和动态视觉特征,并对两种模态特征进行融合;然后,基于ViT结构的强大长距离编码能力,构建模型编码器对融合特征进行长距离依赖编码;最后,将编码器的输出作为LSTM解码器的输入,依次输出预测词,生成最终的自然语言描述。通过在MSR-VTT数据集上进行测试,本研究模型的BLEU-4、METEOR、ROUGEL和CIDEr分别为42.9、28.8、62.4和51.4;在MSVD数据集上进行测试,本研究模型的BLEU-4、METEOR、ROUGEL和CIDEr分别为56.8、37.6、74.5以及98.5。与当前主流模型相比,本研究模型在多项评价指标上表现优异。

退火工艺对Ti-Ni-V形状记忆合金相变和形变行为的影响

利用差示扫描量热仪和拉伸试验研究了400℃×10~120 min和500℃×10~120 min退火态Ti-50.8Ni-0.5V合金的相变行为、拉伸性能和形状记忆行为。结果表明:随退火温度和时间变化,合金的相变类型、R相变温度T_(R)和热滞ΔT_(R)、马氏体相变温度T_(M)和热滞ΔT_(M)、抗拉强度R_(m)、伸长率A、应力诱发马氏体相变临界应力σ_(M)、加载-卸载残余应变ε_(R)和能耗W_(D)的变化规律如下:400和500℃退火态Ti-50.8Ni-0.5V合金皆发生B2→R→B19’/B19’→R→B2(B2-母相,CsCl型结构;R-R相,菱方结构;B19’-马氏体相,单斜结构)型相变;T_(R)^(400℃)>T_(R)^(500℃),T_(M)^(500℃)>T_(M)^(400℃),ΔT_(M)^(400℃)>ΔT_(M)^(500℃)≥ΔT_(R)^(400℃)≈ΔT_(R)^(500℃)≈4℃;R_(m)^(400℃)>R_(m)^(500℃),A^(500℃)>A^(400℃);400℃退火态合金的超弹性(SE)稳定性高于500℃退火态合金。随退火时间延长,T_(R)^(400℃)、T_(R)^(500℃)、T_(M)^(400℃)和T_(M)^(500℃)升高,ΔT_(M)^(400℃)和ΔT_(M)^(500℃)减小,ΔT_(R)^(400℃)和ΔT_(R)^(500℃)在3.5~4.6℃之间变化;400℃退火态合金保持SE,500℃退火态合金由SE向SE+SME(形状记忆效应)转变;σ_(M)^(400℃)和σ_(M)^(500℃)降低;W_(D)^(400℃)和W_(D)^(500℃)增加;ε_(R)^(500℃)先降后升;ε_(R)^(400℃)在0.64%~1.36%之间变化。要使Ti-50.8Ni-0.5V合金获得较高的T_(R),可对其进行400℃退火;要使该合金获得较高的T_(M),可对其进行500℃退火;要使该合金在室温下获得超弹性,可对其进行400℃×10~120 min或500℃×10~60 min退火。

改进Deeplabv3+的双注意力融合作物分类方法

近年来,卷积神经网络(convolutional neural networks,CNN)在农作物分类研究中不断取得新进展,但在建模长期依赖关系方面表现出一定的局限性,对农作物全局特征的捕获存在不足。针对以上问题,将Transformer引入Deeplab v3+模型,提出了一种用于无人机影像农作物分类的并行分支结构--DeepTrans(Deeplab v3+with Trans-former)模型。DeepTrans以一种并行的方式将Transformer和CNN结合在一起,利于全局特征与局部特征的有效捕获。通过引入Transformer来增强图像中信息的远距离依赖关系,提高了作物全局信息的提取能力;加入通道注意力机制和空间注意力机制加强Transformer对通道信息的敏感度及ASPP(atrous spatial pyramid pooling)对作物空间信息捕获能力。实验结果表明,DeepTrans模型在MIoU指标上可达0.812,相较于Deeplab v3+模型提高了3.9%,该模型在五类作物的分类中精度均有提升,对于容易错分的甘蔗、玉米和香蕉三种作物,其IoU分别提高了2.9%、4.7%、13%。由此可见,DeepTrans模型在农作物分类图像的内部填充和全局预测方面有着更好的分割效果,有助于更准确地监测农田作物的种植结构及规模。

Study Hybrid Compensation Cophase Traction Power Supply System Schemes for High-Speed and Heavy Haul Electrical Railway with V Connection Transformer

In order to solve negative phase sequence problem of V connection transformer in the high speed and heavy haul electrical railway of China, the hybrid compensative co-phase traction power supply system which based on passive and active compensation is proposed. Firstly, There construction and capacity distribution are analyzed, and the compensation current of active equipment is gave;Second, the feature of the hybrid compensative schemes are discussed. In the end, the related simulation results have confirmed the effectiveness of the compensation schemes in this paper.

基于改进YOLOv7的织物疵点小目标检测算法

针对织物疵点种类多、尺度变化大、小目标易漏检等问题,提出了一种基于YOLOv7的改进算法(YOLOv7⁃ESL)。首先,将注意力机制ECA⁃Net融入到Neck层中取代CBS模块,在少量增加计算成本的情况下显著提高了检测精度。其次,设计了专用探测头检测小目标,充分利用网络的浅层特征信息,使模型能够有效地检测多尺度的目标。最后,在特征加强部分增加Swin Transformer V2 Block,能够捕捉全局和局部之间的丰富关系,提高模型检测小目标缺陷的能力。试验结果表明:YOLOv7⁃ESL算法精确率为97.7%,召回率为90.3%,平均精度均值为94.9%,FPS为29.9帧/s;与原始YOLOv7模型相比,分别提高了3.8个百分点、3.4个百分点、3.3个百分点、3.1帧/s,可满足纺织工业领域的应用要求。

基于改进YOLO v5的苹果叶部病害检测

针对YOLO v5检测模型存在的漏检率和误检率问题,改进目标检测技术,提升苹果叶部病害早期发现及定位的准确性和速度,从而减少经济损失。先采用加权双向特征金字塔网络(BiFPN)特征融合方法,有效改善PANet对多尺度特征融合的不良影响,并引入Transformer机制,有效改进原始网络结构,使其专注于有用的信息,并增强特征融合效果。再使用ATCSP模块和自上而下的特征融合方法,增强模型对多尺度疾病的检测效果,并将强大的语义信息传达至模型底层,进一步提高检测精度。使用数据集由实验室采集的苹果树叶样本构成,包含3 331张标记图像的矩形位置来标记病害。由于图像亮度分布不均匀,采用直方图均衡化和改进的直方图均衡化处理,使图像对比度得到增强,大幅降低后续图像特征提取的计算量。在训练和测试过程中,还对原始病害图像进行旋转、随机亮度增强、随机色度增强、随机对比度增强和锐化等数据增强操作。结果表明,改进的YOLO v5检测算法可以显著提高苹果叶部病害检测的精度,对比原始算法,平均精度mAP@0.5提高20.8%。改进YOLO v5苹果叶部病害检测算法能够及时发现和定位苹果叶部病害,进而为深度学习技术在农业病害监测中的广泛应用提供技术支撑。

FVIT-YOLO v8:基于多尺度融合注意机制的改进YOLO v8小目标检测

本文研究了遥感与无人机航拍图像中的小目标检测问题。由于这类图像存在目标尺度小、目标分布密集、背景复杂等特点,使得特征提取困难。目前针对小目标检测的算法,为了提升精度,大多忽略了参数量与推理速度的影响,这使得算法缺乏实用性。针对上述问题,本文提出了一种基于轻量化的多尺度融合注意机制的改进YOLO v8小目标检测算法。算法首先在YOLO v8的FPN结构中加入F算子,设计了多尺度特征的加权融合;然后在网络预测层剔除了P4、P5预测层,加入P2层用于小目标的预测;最后对轻量化自注意力机制进行图像输入网格化分割整合改进,并用它替换了FPN中的C2f模块,使得算法具有更好的全局感知能力,并大幅降低了参数量。与YOLO v8s相比,本文算法在DOTA数据集上的mAP提升了4.4%,网络参数量下降了52%,FPS达到了46帧/s。在VisDrone数据集中,本算法在精度上提升了8.2%。

响沙湾站接触网V停作业穿越电流问题探讨

对包西铁路响沙湾站接触网V停作业穿越电流问题进行分析,从穿越电流产生的原因入手,对穿越电流的路径、穿越电流数值理论进行分析,提出了改善穿越电流问题的措施与建议,并通过运营实践验证,取得了一定效果。

Unveiling the therapeutic potential:KBU2046 halts triple-negative breast cancer cell migration by constricting TGF-β1 activation in vitro

Background:Triple-negative breast cancer(TNBC)is a heterogeneous,recurring cancer characterized by a high rate of metastasis,poor prognosis,and lack of efficient therapies.KBU2046,a small molecule inhibitor,can inhibit cell motility in malignant tumors,including breast cancer.However,the specific targets and the corresponding mechanism of its function remain unclear.Methods:In this study,we employed(3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H tetrazolium)(MTS)assay and transwell assay to investigate the impact of KBU2046 on the proliferation and migration of TNBC cells in vitro.RNA-Seq was used to explore the targets of KBU2046 that inhibit the motility of TNBC.Finally,confirmed the predicted important signaling pathways through RT-qPCR and western blotting.Results:In this study,we found that KBU2046 functioned as a novel transforming growth factor-β(TGF-β1)inhibitor,effectively suppressing tumor cell motility in vitro.Mechanistically,it directly down-regulated leucine-rich repeat-containing 8 family,member E(LRRC8E),latent TGFβ-binding protein 3(LTBP3),dynein light chain 1(DNAL1),and MAF family of bZIP transcription factors(MAFF)genes,along with reduced protein expression of the integrin family.Additionally,KBU2046 decreased phosphorylation levels of Raf and ERK.This deactivation of the ERK signaling pathway impeded cancer invasion and metastasis.Conclusions:In summary,these findings advocate for the utilization of TGF-β1 as a diagnostic and prognostic biomarker and as a therapeutic target in TNBC.Furthermore,our data underscore the potential of KBU2046 as a novel therapeutic strategy for combating cancer metastasis.

V弯试验设备的自动化改造及效率提升

V弯试验是评估镀锌板性能的重要方法,但传统设备依靠人工操作,效率低且存在安全隐患。因此,文章提出一种V弯试验设备的自动化改造方案,通过集成机械手、优化夹具设计、升级控制系统等,实现装夹、弯曲、压平等全流程自动化。结果显示,改造后的设备单件效率提升30%~50%,角度重复精度改善64%~66%,在提质增效方面效果显著。

基于改进YOLO v7的农田复杂环境下害虫识别算法研究

为使巡检机器人能够对体积小且密集、形态多变、数量多且分布不均的害虫进行高效精准识别,提出了一种基于改进YOLO v7的害虫识别方法。该方法将CSP Bottleneck与基于移位窗口Transformer(Swin Transformer)自注意力机制相结合,提高了模型获取密集害虫目标位置信息的能力;在路径聚合部分增加第4检测支路,提高模型对小目标的检测性能;将卷积注意力模块(CBAM)集成到YOLO v7模型中,使模型更加关注害虫区域,抑制背景等一般特征信息,提高被遮挡害虫的识别精确率;使用Focal EIoU Loss损失函数减少正负样本不平衡对检测结果的影响,提高识别精度。采用基于实际农田环境建立的数据集的实验结果表明,改进后算法的精确率、召回率及平均精度均值分别为91.6%、82.9%和88.2%,较原模型提升2.5、1.2、3个百分点。与其它主流模型的对比实验结果表明,本文方法对害虫的实际检测效果更优,对解决农田复杂环境下害虫的精准识别问题具有参考价值。

时效态Ti-50.8Ni-0.5V合金的组织、相变和形状记忆行为

为了揭示时效温度和时间对Ti-50.8Ni-0.5V形状记忆合金超弹性(SE)和形状记忆效应(SME)的影响规律,本文用透射电子显微技术、示差扫描量热仪和拉伸试验分别研究了经300℃、400℃、500℃分别时效0.5 h、1 h、5 h、10 h、20 h、50 h后Ti-50.8Ni-0.5V合金的显微组织特征、相变行为和形状记忆行为。结果表明:300~500℃时效态Ti-50.8Ni-0.5V合金的显微组织由基体和Ti3Ni4析出相组成,其中Ti3Ni4析出相呈透镜状,分布在基体的晶内和晶界。随着时效温度升高和时效时间的延长,Ti3Ni4析出相尺寸增大,密度减小。在冷却/加热过程中,300℃和400℃时效态合金发生B2→R→B19′/B19′→R→B2(B2—母相,CsCl型结构;R—R相,菱方结构;B19′—马氏体,单斜结构)型相变,500℃时效态合金发生B2→R→B19′/B19′→B2型相变。随着时效时间的延长,合金的R相变温度TR和马氏体相变温度TM增加,马氏体相变热滞ΔTM减小,R相变热滞ΔTR变化不大,其中TR400℃>TR300℃>TR500℃,TM500℃>TM400℃>TM300℃,ΔTM400℃>ΔTM300℃>ΔTM500℃,ΔTR300℃≈ΔTR400℃≈4℃。室温下,(300℃,0.5~50 h)、(400℃,0.5~10 h)和(500℃,0.5~1 h)时效态Ti-50.8Ni-0.5V合金呈SE特性,(400℃,10~50 h)和(500℃,1~50 h)时效态合金呈SE+SME特性。要使Ti-50.8Ni-0.5V合金在室温下获得优异的SE特性,可对其进行(300℃,0.5~50 h)或(400℃,0.5~10 h)或500℃/0.5~1 h时效处理。

Site occupation evolution of alloying elements in L1_2 phase during phase transformation in Ni_(75)Al_(7.5)V_(17.5)

Correlation between site occupation evolution of alloying elements in L12 phase and growth of DO22 phase in Ni75Al7.5V17.5 was studied using microscopic phase field model. The results demonstrate that the growing process of DO22 phase can be divided into two stages. At the early stage, composition in the centre part of L12 phase almost remains unchanged, and the nucleation and growth of DO22 phase is controlled by the decrease of interface between L12 phases. At the late stage, part of V for growth of DO22 phase is supplied from the centre part of L12 phase and mainly comes from Al sublattice, the excess Ni spared from the decreasing L12 phase migrates into the centre part of L12 phase and occupies the Ni sublattices exclusively, while the excess Al mainly occupies the Al sublattice. At the late stage, the growth of DO22 phase is controlled by the evolution of antisite atoms and ternary additions in the centre part of L12 phase.

Site occupation evolution of alloying elements in Ni_3 V phase during phase transformation in Ni_(75)Al_(4.2)V_(20.8)

Based on the microscopic phase-field model, the correlation between site occupation evolution of alloying elements in Ni3V-DO22 phase and growth of Ni3Al-L12 phase was studied during the phase transformation of Ni75Al4.2V20.8. The results demonstrate that the growth of L12 phase can be divided into two stages： at the early stage, the composition of alloying elements in DO22 phase almost remains unchanged; at the late stage, the compositions of Ni and Al decrease while V increases in DO22 phase. Part of alloying elements for L12 phase growth are supplied from the site occupation evolution of alloying elements on three kinds of sublattices in DO22 phase. Ni is mainly supplied from V sublattice, and part of Al is supplied from NiⅠ and V sites at the centre of DO22 phase. The excessive V from the decreasing DO22 phase migrates into the centre of DO22 phase and mainly occupies V and NiII sites. It is the site occupation evolution of antisite atoms and ternary additions in DO22 phase that controls the growth rate of L12 phase at the late stage.

V微合金化对高强弹簧钢静态相变和拉伸性能的影响

针对55SiCr钢,利用热力学软件计算了V微合金化对实验钢析出的影响,根据拉伸性能确定了较优的V添加量(质量分数),分析了较优的V质量分数对55SiCr钢静态相变的影响,同时利用金相和SEM表征了不同冷速下实验钢的显微组织特征.结果表明:V质量分数的增加,提高了MC相的析出量和析出温度,抑制了M_(3)C,M_(7)C_(3)和M_(3)C_(2)碳化物的析出;V的质量分数为0.15%时实验钢性能最优,此时其抗拉强度和屈服强度分别提高180,200 MPa,断面收缩率超过40%,总伸长率超过12%;V微合金化使实验钢的相变温度A_(c3)和A_(c1)分别降至805,745℃,使马氏体相变温度降至255℃.

融合CNN与时序Transformer的动态手势识别

针对传统手势识别方法没有综合考虑手势的全局空间、局部空间、时序等特征信息,提取的特征通常很难全面表征手势之间的区别,提出了一种卷积神经网络(CNN)和Transformer网络相结合的网络结构。首先对输入视频序列的每一帧使用轻量化MobileNet V3卷积神经网络来提取空间特征信息,再将输出经过pathch embedding后加上时序嵌入序列,输入到Transformer模型中利用注意力机制来提取手势的全局注意力特征和时序特征。并在两个公开数据集DHG-14/28和VIVA上进行了实验,与经典方法相比,平均识别精度分别提升了2.38%、1.87%和3.74%。实验结果表明,提出的方法能够准确地提取动态手势序列的特征并表征手势类别。

Effect of V and V-N Microalloying on Deformation-Induced Ferrite Transformation in Low Carbon Steels

Deformation-induced ferrite transformation （DIFT） has been proved to be an effective approach to refine ferrite grains. This paper shows that the ferrite grains can further be refined through combination of DIFT and V or V-N microalloying. Vanadium dissolved in γ matrix restrains DIFT. During deformation, vanadium carbonitrides rapidly precipitate due to strain-induced precipitation, which causes decrease in vanadium dissolved in matrix and indirectly accelerates DIFT. Under heavy deformation, deformation induced ferrite （DIF） grains in V microalloyed steel were finer than those in V free steel. The more V added to steel, the finer DIF grains obtained. Moreover, the addition of N to V microalloyed steels can remarkably accelerate precipitation of V, and then promote DIFT. However, DIF grains in V-N microalloyed steel easily coarsen.

Effect of hot deformation on α→β phase transformation in 47Zr-45Ti-5Al-3V alloy

The effect of strain rate and deformation temperature on theα→βphase transformation in 47Zr-45Ti-5Al-3V alloy with an initial widmanstattenαstructure was investigated.At the deformation temperature of 550°C,the volume fraction ofαphase decreased with increasing strain rate.At 600 and 650°C,the volume fraction ofαphase firstly increased to a maximum value with increasing strain rate from 1×10-3 to 1×10-2 s-1,and then decreased.At 700°C,the microstructure consisted of singleβphase.At a given strain rate,the volume fraction ofαphase decreased with increasing deformation temperature.With decreasing strain rate and increasing deformation temperature,the volume fraction and size of globularαphase increased.At 650°C and 1×10-3 s-1,the lamellarαphase was fully globularized.The variation in the volume fraction and morphology ofαphase with strain rate and deformation temperature significantly affected the hardness of 47Zr-45Ti-5Al-3V alloy.