期刊文献+
共找到8篇文章
< 1 >
每页显示 20 50 100
HD-YOLO:复杂场景下安全帽佩戴检测算法 被引量:2
1
作者 邱云飞 腰瑞琳 +1 位作者 金海波 张嘉宁 《安全与环境学报》 北大核心 2025年第1期165-174,共10页
针对目标密集、有遮挡的复杂施工场景下安全帽佩戴检测存在漏检、误检的问题,提出了一种基于YOLOv8的HD-YOLO安全帽佩戴检测算法。首先,设计了GRC-C2f结构,使用多分支结构捕获多尺度特征,兼顾训练阶段的特征提取能力和推理阶段的计算效... 针对目标密集、有遮挡的复杂施工场景下安全帽佩戴检测存在漏检、误检的问题,提出了一种基于YOLOv8的HD-YOLO安全帽佩戴检测算法。首先,设计了GRC-C2f结构,使用多分支结构捕获多尺度特征,兼顾训练阶段的特征提取能力和推理阶段的计算效率。其次,设计了DSASF颈部结构,结合动态上采样和多尺度特征融合,精准识别和定位图像中的小目标,以提高检测性能。然后,引入Focal Modulation模块替换原有的快速空间金字塔池化(Spatial Pyramid Pooling-Fast,SPPF)结构模块,捕捉图像中的长距离依赖和上下文信息,聚焦于复杂背景中的目标。最后,采用空间增强注意力模块(Spatially Enhanced Attention Module,SEAM)解决小目标遮挡问题。试验结果表明,HD-YOLO算法在同一数据集上平均精度均值为81.8%,相比原始YOLOv8算法提高了5.0百分点。设计的HD-YOLO算法有效提高了复杂场景中佩戴安全帽的检测精度。 展开更多
关键词 安全社会工程 安全帽检测 YOLOv8算法 GRC-C2f模块 DSASF颈部结构 focal modulation模块 空间增强注意力模块
原文传递
基于YOLOv8的轻量级无人机航拍小目标检测模型
2
作者 陈奕衡 廉佐政 《齐齐哈尔大学学报(自然科学版)》 2025年第3期37-42,共6页
针对无人机携带计算资源受限的问题,提出一种轻量化的YOLOv8小目标检测模型。在主干网络中引入AKConv,通过灵活的卷积核减少不必要的参数;构建新的颈部网络SBN,通过GSConv降低计算开销,并利用双向信息传递机制,提高对不同尺度目标的检... 针对无人机携带计算资源受限的问题,提出一种轻量化的YOLOv8小目标检测模型。在主干网络中引入AKConv,通过灵活的卷积核减少不必要的参数;构建新的颈部网络SBN,通过GSConv降低计算开销,并利用双向信息传递机制,提高对不同尺度目标的检测精度;使用Focal Modulation替代原有的SPPF模块,增强模型在图像关键区域的注意力机制。结果表明,改进后的模型较YOLOv8,参数量降低30%,mAP50提高2.3%,同时相较于目标检测领域常用的一些模型,在轻量化和检测精度上均有更好的表现,证明了改进的有效性。 展开更多
关键词 小目标检测 YOLOv8 AKConv SBN focal modulation
在线阅读 下载PDF
基于YOLOv8改进的EFCW-YOLO高精度木材缺陷检测模型
3
作者 朱一凡 《数字技术与应用》 2025年第6期186-190,共5页
本文提出了一种基于YOLOv8改进的EFCWYOLO模型,针对木材缺陷检测精度低、效率不足等实际问题进行优化。通过引入Focal Modulation Networks模块、ECA注意力机制、C2fCIB模块和Wise-IoU损失函数,显著提升了检测精度和效率。实验结果表明,... 本文提出了一种基于YOLOv8改进的EFCWYOLO模型,针对木材缺陷检测精度低、效率不足等实际问题进行优化。通过引入Focal Modulation Networks模块、ECA注意力机制、C2fCIB模块和Wise-IoU损失函数,显著提升了检测精度和效率。实验结果表明,EFCW-YOLO模型实现了90.3%的mAP50和84.9%的召回率,均优于YOLOv7、YOLOv8n、YOLOv10n和YOLOv11n等主流模型。在泛化性测试实验中,模型进一步取得了97%的mAP50和93.4%的召回率,展现出优异的泛化能力。由此可见,该模型为木材加工行业提供了一种高效、准确的自动化检测方案,提升了生产效率和市场竞争力,具有重要的应用价值,可有效增强企业的生产效能与市场竞争力。 展开更多
关键词 focal modulation Networks YOLOv8
在线阅读 下载PDF
基于YOLOv8-NFMC的带钢表面缺陷检测算法 被引量:9
4
作者 朱成杰 刘乐乐 朱洪波 《国外电子测量技术》 2024年第7期97-104,共8页
针对YOLOv8算法在应用于带钢表面缺陷检测时存在漏检和错检等问题,提出了一种改进YOLOv8算法。针对数据集中的小目标的标签,在原损失CIOU的基础上面加入标准化高斯瓦瑟斯坦距离(normalized Gaussian Wasserstein distance,NWD),提升模... 针对YOLOv8算法在应用于带钢表面缺陷检测时存在漏检和错检等问题,提出了一种改进YOLOv8算法。针对数据集中的小目标的标签,在原损失CIOU的基础上面加入标准化高斯瓦瑟斯坦距离(normalized Gaussian Wasserstein distance,NWD),提升模型对小目标缺陷的检测能力;采用聚焦调制(focal modulation)替换YOLOv8模型的空间池化金字塔(spatial pyramid pooling-fast,SPPF),在轻量化的同时,提高多尺度特征的表达能力;采用移动翻转瓶颈卷积(mobile inverted bottleneck conv,MBConv)替换C2f中的Conv构建新模块C2f-MB,同时使用C2f-MB替换原有的C2f模块,增强特征表达能力和多尺度特征融合能力;在主干部分加入卷积块注意力机制(convolutional block attention module,CBAM)来抑制背景干扰,能更好捕获全局信息,提升了主干部分的特征提取能力。实验结果表明,改进后的YOLOv8算法在计算量下降的同时,mAP@0.5提高了3%,对漏检和错检等问题有明显改善。 展开更多
关键词 带钢表面缺陷 NWD focal modulation MBConv 注意力机制
原文传递
Treatable focal region modulated by double excitation signal superimposition to realize platform temperature distribution during transcranial brain tumor therapy with high-intensity focused ultrasound 被引量:2
5
作者 Shi-Hui Chang Rui Cao +4 位作者 Ya-Bin Zhang Pei-Guo Wang Shi-Jing Wu Yu-Han Qian Xi-Qi Jian 《Chinese Physics B》 SCIE EI CAS CSCD 2018年第7期585-594,共10页
Recently, the phase compensation technique has allowed the ultrasound to propagate through the skull and focus into the brain. However, the temperature evolution during treatment is hard to control to achieve effectiv... Recently, the phase compensation technique has allowed the ultrasound to propagate through the skull and focus into the brain. However, the temperature evolution during treatment is hard to control to achieve effective treatment and avoid over-high temperature. Proposed in this paper is a method to modulate the temperature distribution in the focal region. It superimposes two signals which focus on two preset different targets with a certain distance. Then the temperature distribution is modulated by changing triggering time delay and amplitudes of the two signals. The simulation model is established based on an 82-element transducer and computed tomography (CT) data of a volunteer's head. A finite- difference time-domain (FDTD) method is used to calculate the temperature distributions. The results show that when the distances between the two targets respectively are 7.5-12.5 mm on the acoustic axis and 2.0-3.0 mm in the direction perpendicular to the acoustic axis, a focal region with a uniform temperature distribution (64-65 ℃) can be created. Moreover, the volume of the focal region formed by one irradiation can be adjusted (26.8-266.7 mm3) along with the uniform temperature distribution. This method may ensure the safety and efficacy of HIFU brain tumor therapy. 展开更多
关键词 high-intensity focused ultrasound (HIFU) transcranial therapy double excitation signal superim- position temperature modulation of focal region
原文传递
Creation of Multiple Subwavelength Focal Spot Segments Using Phase Modulated Radially Polarized Multi Gaussian Beam
6
作者 K.Prabakaran K.B.Rajesh +4 位作者 S.Sumathira M.D.Bharathi R.Hemamalini A.M.Musthafa V.Aroulmoji 《Chinese Physics Letters》 SCIE CAS CSCD 2016年第9期48-51,共4页
Based on the vector diffraction theory, the effect of complex phase filters on intensity distribution of a radially polarized multi Gaussian beam in the focal region of high NA lens is theoretically investigated. It i... Based on the vector diffraction theory, the effect of complex phase filters on intensity distribution of a radially polarized multi Gaussian beam in the focal region of high NA lens is theoretically investigated. It is observed that a properly designed multi belt complex phase filter can generate subwavelength novel focal patterns including splitting of focal spots and generation of multiple focal spot segments such as eight, six and four focal spots along the optical axis are obtained. We expect that such an investigation is useful for optical manipulation and material processing, multiple high refractive index particle trapping technologies. 展开更多
关键词 of for Creation of Multiple Subwavelength focal Spot Segments Using Phase Modulated Radially Polarized Multi Gaussian Beam on is in
原文传递
基于YOLOv8n改进的路面裂纹检测
7
作者 侯传康 戚可文 +4 位作者 刘海龙 李勇 卓令军 董磊 张林 《现代计算机》 2024年第23期20-25,共6页
路面裂缝检测是保障道路安全、实现道路损伤及时修复的一项重要任务。针对现有路面检测存在的检测精度低、定位不准等问题,提出一种基于YOLOv8改进的路面裂纹检测算法YOLOv8⁃pavement。首先在模型训练时对数据集进行离线数据增强,提高... 路面裂缝检测是保障道路安全、实现道路损伤及时修复的一项重要任务。针对现有路面检测存在的检测精度低、定位不准等问题,提出一种基于YOLOv8改进的路面裂纹检测算法YOLOv8⁃pavement。首先在模型训练时对数据集进行离线数据增强,提高模型的泛化能力,其次在骨干网络末端添加Focal Modulation(FM)模块来捕捉图像中长距离依赖和上下文信息以适应裂纹对象的大跨度和细长特征。最后在颈部网络中使用CSPStage(CS)模块,提高特征表达性能,减小模型的参数量和计算量。实验证明,与初始YOLOv8n模型相比mAP50提高了1.2个百分点,而模型的参数量和计算量分别降低3个百分点和4.9个百分点,该算法具有良好的检测效果。 展开更多
关键词 路面裂缝 YOLOv8n focal modulation CSPStage
在线阅读 下载PDF
Advanced optical microscopy methods for in vivo imaging of sub-cellular structures in thick biological tissuesl
8
作者 Nanguang Chen Shakil Rehman Colin J.R.Sheppard 《Journal of Innovative Optical Health Sciences》 SCIE EI CAS 2014年第5期64-90,共27页
Optical microscopy has become an indispensable tool for visualizing sub-cellular structures andbiological processes.However,scattering in biological tissues is a major obstacle that preventshigh-resolution images from... Optical microscopy has become an indispensable tool for visualizing sub-cellular structures andbiological processes.However,scattering in biological tissues is a major obstacle that preventshigh-resolution images from being obtained from deep regions of tissue.We review commontechniques,such as multiphoton microscopy(MPM)and optical coherence microscopy(OCM),for diffraction limited imaging beyond an imaging depth of 0.5 mm.Novel implementations havebeen emerging in recent years giving higher imaging speed,deeper penetration,and better imagequality.Focal modulation microscopy(FMM)is a novel method that combines confocal spatialfltering with focal modulation to reject out-of-focus background.FMM has demonstrated animaging depth comparable to those of MPM and OCM,near-real-time image acquisition,and thecapability for multiple contrast mechanisms. 展开更多
关键词 Confocal microscopy multiphoton microscopy(MPM) optical gating optical coherence tomography(OCT) optical coherence microscopy(OCM) focal modulation microscopy(FMM).
原文传递
上一页 1 下一页 到第
使用帮助 返回顶部