期刊文献+
共找到122,687篇文章
< 1 2 250 >
每页显示 20 50 100
基于SSD与TPR法对比的中国水域二甲基甲酰胺淡水生物水质基准
1
作者 胡雨晴 吴绅溶 +5 位作者 樊怡利 江浩 王海燕 白英臣 于家琳 王何灵 《环境工程技术学报》 北大核心 2026年第1期370-379,共10页
作为一种典型的污染物,N,N-二甲基甲酰胺(N,N-dimethylformamide,DMF)具有排放量大、生态毒性高、难以降解的特性,但目前国内外尚缺乏相关基准标准研究。基于ECOTOX、Web of Science及中国知网等相关中英文文献和毒理数据库,收集并筛选... 作为一种典型的污染物,N,N-二甲基甲酰胺(N,N-dimethylformamide,DMF)具有排放量大、生态毒性高、难以降解的特性,但目前国内外尚缺乏相关基准标准研究。基于ECOTOX、Web of Science及中国知网等相关中英文文献和毒理数据库,收集并筛选获得222条DMF对中国淡水生物毒性的文献数据,同时补充了DMF对麦穗鱼、黄颡鱼、河蚬、苏式尾鳃蚓、泥鳅、萼花臂尾轮虫、角突臂尾轮虫、莱茵衣藻共8种本土代表性物种的急性、慢性毒性测试数据,用于DMF的淡水生物水质基准推导。DMF水质基准共涉及26种淡水水生生物,涵盖了3个营养级,同时包含了草鱼、鳙鱼等在我国广泛分布的经济物种。其中急性数据涵盖7门15科24个物种,最敏感的物种为柱孢鱼腥藻和多变鱼腥藻;慢性毒性数据涵盖6门10科12个物种,最敏感的物种为模糊网纹溞。采用物种敏感度分布法(SSD)和毒性百分数排序法(TPR)分别计算我国淡水水生生物的DMF水质基准,其中SSD推导出短期水质基准为824.12 mg/L,长期水质基准为14.53 mg/L,TPR推导出短期水质基准为1 081.58 mg/L,长期水质基准为22.48 mg/L,建议采用SSD所推导基准作为DMF的水质标准。 展开更多
关键词 二甲基甲酰胺(DMF) 水质基准 物种敏感度分布法(ssd) 毒性百分数排序法(TPR)
在线阅读 下载PDF
边缘设备端轻量级SSD变电站缺陷检测算法
2
作者 蔡宇翔 陈丽娟 安琪 《郑州大学学报(工学版)》 北大核心 2026年第1期140-146,共7页
针对电力物联网中设备表面缺陷自动化检测难题(如破损、污损及人为违规操作导致的缺陷),提出一种面向边缘计算设备的轻量级SSD检测算法。该算法通过3个关键技术创新实现高效检测。首先,在MobileNetV2的瓶颈结构中引入密集连接机制,动态... 针对电力物联网中设备表面缺陷自动化检测难题(如破损、污损及人为违规操作导致的缺陷),提出一种面向边缘计算设备的轻量级SSD检测算法。该算法通过3个关键技术创新实现高效检测。首先,在MobileNetV2的瓶颈结构中引入密集连接机制,动态增强图像特征表达能力;其次,基于Non-Local注意力机制构建跨层注意力隐式特征金字塔网络(CL-IFPN),通过与MobileNetV2-SSD的深度融合显著提升小缺陷检测能力;最后,通过在卷积层添加特征融合模块并采用QFL函数,强化不同尺度缺陷的预测精度及正负样本训练平衡性。实验结果表明:在公共数据集VOC2007上,所提算法以79.62%的mAP检测精度和36帧/s的检测速度表现优于同类算法;在自建电力器件缺陷数据集上,检测性能进一步提升至95.19%的检测精度和24帧/s的检测速度,充分验证了算法在电力设备缺陷检测场景的实用价值。所提算法为边缘计算环境下的电力物联网设备智能运维提供了有效的技术解决方案。 展开更多
关键词 缺陷检测 MobileNetV2 边缘计算设备 注意力机制 ssd
在线阅读 下载PDF
DPZB+tree:基于ZNS SSD与持久化内存的高效B+树索引设计
3
作者 曹夕 李明杰 +2 位作者 杨朝树 杨程 张润宇 《计算机研究与发展》 北大核心 2026年第3期550-566,共17页
新型分区命名空间固态硬盘(zoned namespace solid state drive,ZNS SSD)有望解决传统块设备固态盘写放大率高、存储密度低、I/O路径复杂等问题,为存储技术的发展创造机遇。B+树作为一种高效的树形索引结构,被广泛应用于各类数据库和文... 新型分区命名空间固态硬盘(zoned namespace solid state drive,ZNS SSD)有望解决传统块设备固态盘写放大率高、存储密度低、I/O路径复杂等问题,为存储技术的发展创造机遇。B+树作为一种高效的树形索引结构,被广泛应用于各类数据库和文件系统中,以支撑大模型高效数据加载、外部知识库构建及结构化元数据管理,从而显著提升训练效率与知识调用性能。然而,由于ZNS SSD的硬件特性不同于传统块设备,直接将B+树部署到ZNS SSD中不仅会导致较高的写放大率,还会引起级联更新,严重影响存储系统的性能。针对以上问题,结合新型持久化内存(persistent memory,PM)提出了一种基于ZNS SSD的B+树索引结构DPZB+tree。首先,DPZB+tree采用DRAM-PM-ZNS SSD混合存储架构,实现冷热数据分离存储;其次,DPZB+tree设计了冷热节点识别策略,以提高存储系统的读写效率;然后,针对PM容量有限的问题,提出了冷热节点动态放置策略,实现冷热数据的自适应迁移;最后,结合硬件特性和局部性原理设计了叶节点分裂及合并操作。DPZB+tree索引方案基于ZNS SSD模拟器和英特尔傲腾PM实现。实验结果表明,在多种工作负载下,相较于LSM-tree,SSDB+tree,DZB+tree,Baseline,DPZB+tree均取得了优异的读写性能以及更低的恢复耗时。 展开更多
关键词 分区命名空间固态硬盘 持久化内存 B+树 索引结构 温度感知
在线阅读 下载PDF
Impact of Well Thickness on Static and Dynamic Behavior of InGaN Light-emitting Diode with Single Quantum Well
4
作者 CHEN Guichu HE Longfei PENG Kun 《发光学报》 北大核心 2026年第2期314-320,共7页
In this paper,we present a circuit model of single-quantum-well InGaN/GaN light-emitting diodes based on the standard rate equations.Two rate equations describe carrier transport processes occurring in sep-arate confi... In this paper,we present a circuit model of single-quantum-well InGaN/GaN light-emitting diodes based on the standard rate equations.Two rate equations describe carrier transport processes occurring in sep-arate confinement heterostructure and quantum well respectively,and the third equation describes the varied photons in quantum well.By using the presented model,impacts of quantum well thickness on the static and dynamic performances are investigated.Simulated results show that LED with 4 nm well exhibits better lightcurrent(L-I)performance,but LED with 3 nm well presents wider 3 dB modulation bandwidth.It reveals that high carrier density in quantum well is detrimental to the static performance,but beneficial to the dynamic performance. 展开更多
关键词 single quantum well rate equations circuit model L-I performance modulation bandwidth
在线阅读 下载PDF
Advancing pediatric cancer diagnosis: the promise of single-cell liquid biopsy for early detection and surveillance
5
作者 Pavithra Ayyadurai Chinnasamy Ragavendran 《Biomedical Engineering Communications》 2026年第1期13-25,共13页
Pediatric cancers are particularly significant due to their uncommon occurrence in children,driven by a variety of underlying factors.Because of their distinct molecular and genetic makeup,which makes early detection ... Pediatric cancers are particularly significant due to their uncommon occurrence in children,driven by a variety of underlying factors.Because of their distinct molecular and genetic makeup,which makes early detection challenging,they are linked to problems.Diagnostic methods like imaging and tissue biopsy are only effective when the tumor has reached a size that can be identified.The liquid biopsy technique,the least intrusive and most convenient diagnostic method,is the subject of this review.It focuses on the significance of single cell analysis in examining uncommon cancer types.The many biomarkers found in bodily fluids and the cancer types they are linked to in children have been assessed,as has the potential route towards early detection and cancer recurrence forecasting.Combining the single cell liquid biopsy with the newest technologies,such as computational and multi-omics approaches,which have improved the efficiency of processing massive and unique genetic data,appears promising.This article discusses on a number of case reports for uncommon pediatric malignancies,such as Neuroblastoma,Medulloblastoma,Wilms Tumor,Rhabdomyosarcoma,Ewing Sarcoma,and Retinoblastoma,as well as their liquid biopsy profiles.Furthermore,the findings raise ethical questions regarding the therapeutic application of the technology as well as possible difficulties related to clinical translation.The likelihood that this single cell liquid biopsy will be clinically validated and eventually used as a routine diagnostic tool for uncommon pediatric cancers will rise with the realistic approach to sensitivity monitoring,specificity upgrading,and optimization. 展开更多
关键词 pediatric cancer single cell liquid biopsy biomarkers NEUROBLASTOMA MEDULLOBLASTOMA wilms tumor
暂未订购
Single electron capture in low-and intermediate-energy collisions of Si^(3,4+)with He
6
作者 Yingzhou Li Yadong Liu +3 位作者 Yueying Qi Ling Liu Yizhi Qu Jianguo Wang 《Chinese Physics B》 2026年第1期150-157,共8页
The single electron capture processes in Si^(3,4+)+He collisions have been investigated theoretically employing the two-center atomic orbital close-coupling method in the energy range 0.01-100 keV/u.Total and state-se... The single electron capture processes in Si^(3,4+)+He collisions have been investigated theoretically employing the two-center atomic orbital close-coupling method in the energy range 0.01-100 keV/u.Total and state-selective electron capture cross sections for the dominant and subdominant reaction channels are calculated and compared with the available experimental and theoretical data.For the total charge transfer cross sections,the present results show better agreements with the available experimental data than the other theoretical ones in the overlapping energy region for both collision systems.For the state-selective cross sections,the present results for 3s and 3p states are in general agreement with the previous MOCC results in the low energy region for both collision systems.Furthermore,the cross sections for electron captured to the 3d,4l and 5l(l=0,1,...,n-1)states of Si^(2+)and Si^(3+)ions are first provided in a broad energy region in our work.These results are useful for the investigations in astrophysics.The datasets presented in this paper,including the total and state-selective electron capture cross sections of Si^(3,4+)+He collisions in 0.01-100 ke V/u,are openly available at https://doi.org/10.57760/sciencedb.j00113.00257. 展开更多
关键词 atomic orbital close-coupling single electron capture cross section
原文传递
Characterization of ultrahigh-strain-rate compressive behaviors in single 10-μm scale fibers using a micro-scale Hopkinson bar method
7
作者 Liang Ma Lingxin Hu +9 位作者 Haoxiang Wang Yichao Yuan Jian Wei Xiaoxin Zhao Kunkun Zeng Yuze Zhao Zhiyin Zhao Jiagui Liu Shizhao Chen Jinling Gao 《Defence Technology(防务技术)》 2026年第2期270-281,共12页
High-performance fiber fabrics and composites experienced transverse compression deformation at ultrahigh strain rates near the impact point when subjected to high-velocity impacts,which significantly affected their b... High-performance fiber fabrics and composites experienced transverse compression deformation at ultrahigh strain rates near the impact point when subjected to high-velocity impacts,which significantly affected their ballistic limits.In this paper,a fiber-scale experimental method for characterizing ultrahigh strain-rate transverse compression behavior was proposed.To begin with,in order to measure the extremely low stress and strain in small specimens,the conventional Hopkinson bar was reduced to the hundred-micron scale,thereby achieving wave impedance matching with single fibers.In addition,tangential and normal laser Doppler velocimetry(LDV)methods were employed to realize non-contact,high-precision,and high-speed axial velocity measurements of micron-scale incident and transmission bars,respectively.Meanwhile,a microscopic observation system was used to facilitate the installation of miniature fiber samples.The experimental setup and procedures were introduced,and the system accuracy was verified through sample-free loading tests based on one-dimensional stress wave propagation theory.Dynamic compression experiments on Graphene-UHMWPE fibers were carried out,followed by post-compression microstructural characterization via scanning electron microscopy(SEM).Results demonstrated that successful mechanical characterization was achieved at strain rates exceeding 105,an order of magnitude higher than the previously reported maximum rates.Furthermore,during the loading process,the fibers underwent uniform compression deformation while exhibiting pronounced strain-rate effects.This method offers a novel approach for dynamic mechanical characterization of microscale single fibers,enabling the development of comprehensive strain-ratedependent material models to guide the design of advanced composites and high-performance fibers. 展开更多
关键词 single fiber Transverse compression Ultrahigh strain rate Microscale Hopkinson bar Laser Doppler velocimetry
在线阅读 下载PDF
A single-cell landscape of the regenerating spinal cord of zebrafish
8
作者 Lei Yao Xinyi Cai +5 位作者 Saishuai Yang Yixing Song Lingyan Xing Guicai Li Zhiming Cui Jiajia Chen 《Neural Regeneration Research》 2026年第2期780-789,共10页
Unlike mammals,zebrafish possess a remarkable ability to regenerate their spinal cord after injury,making them an ideal vertebrate model for studying regeneration.While previous research has identified key cell types ... Unlike mammals,zebrafish possess a remarkable ability to regenerate their spinal cord after injury,making them an ideal vertebrate model for studying regeneration.While previous research has identified key cell types involved in this process,the underlying molecular and cellular mechanisms remain largely unexplored.In this study,we used single-cell RNA sequencing to profile distinct cell populations at different stages of spinal cord injury in zebrafish.Our analysis revealed that multiple subpopulations of neurons showed persistent activation of genes associated with axonal regeneration post injury,while molecular signals promoting growth cone collapse were inhibited.Radial glial cells exhibited significant proliferation and differentiation potential post injury,indicating their intrinsic roles in promoting neurogenesis and axonal regeneration,respectively.Additionally,we found that inflammatory factors rapidly decreased in the early stages following spinal cord injury,creating a microenvironment permissive for tissue repair and regeneration.Furthermore,oligodendrocytes lost maturity markers while exhibiting increased proliferation following injury.These findings demonstrated that the rapid and orderly regulation of inflammation,as well as the efficient proliferation and redifferentiation of new neurons and glial cells,enabled zebrafish to reconstruct the spinal cord.This research provides new insights into the cellular transitions and molecular programs that drive spinal cord regeneration,offering promising avenues for future research and therapeutic strategies. 展开更多
关键词 dividing oligodendrocyte macrophage MICROGLIA neuron proliferating oligodendrocyte radial glia single cell sequencing spinal cord regeneration transcriptome ZEBRAFISH
暂未订购
Doping dependence of resistivity,upper critical field and its anisotropy in overdoped Ba_(1-x)K_(x)Fe_(2)As_(2)(x=0.6-1)single crystals
9
作者 Ke Shi Wenshan Hong +10 位作者 Yang Li Minjie Zhang Yongqi Han Yu Zhao Jiating Wu Ze Wang Langsheng Ling Chuanying Xi Li Pi Huiqian Luo Zhaosheng Wang 《Chinese Physics B》 2026年第1期272-278,共7页
Temperature-dependent resistivity,upper critical field H_(c2)and its anisotropy in overdoped superconducting Ba_(1-x)K_x Fe_2As_2(x=0.6-1)single crystals have been measured in steady magnetic fields up to 44 T and low... Temperature-dependent resistivity,upper critical field H_(c2)and its anisotropy in overdoped superconducting Ba_(1-x)K_x Fe_2As_2(x=0.6-1)single crystals have been measured in steady magnetic fields up to 44 T and low temperatures down to 0.4 K.Analysis using both the quadratic term and power-law fitting demonstrates that the in-plane resistivityρ_(ab)(T)progressively approaches the Fermi-liquid T~2behavior with increasing K doping and reaches a saturation plateau at x≈0.8.The temperature dependence of both H_(c2)^(ab)and H^(c)_(c2)follows the Werthamer-Helfand-Hohenberg model,incorporating orbital and spin paramagnetic effects.For x≤0.8,the orbital effect dominates for H ab,while the Pauli paramagnetic effect prevails for H c.For x>0.8,the Pauli paramagnetic effect becomes dominant in both crystallographic directions.The anisotropy of H_(c2)(0)exhibits a discontinuity in its dependence on K doping concentration with a significant enhancement at x=0.8 and a maximum at x=0.9.These experimental results indicate that the electron correlation effect is enhanced in the heavily overdoped Ba_(1-x)K_(x)Fe_(2)As_(2)system where the underlying symmetries are broken due to the Fermi surface reconstruction before x=0.9. 展开更多
关键词 BaK122 single crystals high magnetic fields upper critical field H_(c2) MAGNETORESISTANCE
原文传递
Utilizing Single-cell and Spatial RNA-seq databasE for Alzheimer’s Disease(ssREAD)in hypothesis-driven queries
10
作者 Diana Acosta Cankun Wang +1 位作者 Qin Ma Hongjun Fu 《Neural Regeneration Research》 2026年第2期677-678,共2页
Alzheimer’s disease(AD)is the most common form of dementia.In addition to the lack of effective treatments,there are limitations in diagnostic capabilities.The complexity of AD itself,together with a variety of other... Alzheimer’s disease(AD)is the most common form of dementia.In addition to the lack of effective treatments,there are limitations in diagnostic capabilities.The complexity of AD itself,together with a variety of other diseases often observed in a patient’s history in addition to their AD diagnosis,make deciphering the molecular mechanisms that underlie AD,even more important.Large datasets of single-cell RNA sequencing,single-nucleus RNA-sequencing(snRNA-seq),and spatial transcriptomics(ST)have become essential in guiding and supporting new investigations into the cellular and regional susceptibility of AD.However,with unique technology,software,and larger databases emerging;a lack of integration of these data can contribute to ineffective use of valuable knowledge.Importantly,there was no specialized database that concentrates on ST in AD that offers comprehensive differential analyses under various conditions,such as sex-specific,region-specific,and comparisons between AD and control groups until the new Single-cell and Spatial RNA-seq databasE for Alzheimer’s Disease(ssREAD)database(Wang et al.,2024)was introduced to meet the scientific community’s growing demand for comprehensive,integrated,and accessible data analysis. 展开更多
关键词 sex specific alzheimer s disease ad deciphering molecular mechanisms spatial transcriptomics ssread spatial transcriptomics st Alzheimers disease single cell RNA seq
暂未订购
Broadband Telecom Single-Photon Emissions from InAs/InP Quantum Dots Grown by MOVPE Droplet Epitaxy
11
作者 Shichen Zhang Li Liu +6 位作者 Kai Guo Xingli Mu Yuanfei Gao Junqi Liu Fengqi Liu Quanyong Lu Zhiliang Yuan 《Chinese Physics Letters》 2026年第1期37-43,共7页
The development of quantum materials for single-photon emission is crucial for the advancement of quantum information technology.Although significant advancements have been witnessed in recent years for single-photon ... The development of quantum materials for single-photon emission is crucial for the advancement of quantum information technology.Although significant advancements have been witnessed in recent years for single-photon sources in the near-infrared band(λ∼700–1000 nm),several challenges have yet to be addressed for ideal single-photon emission at the telecommunication band.In this study,we present a droplet-epitaxy strategy for O-band to C-band single-photon source-based semiconductor quantum dots(QDs)using metal-organic vaporphase epitaxy(MOVPE).By investigating the growth conditions of the epitaxial process,we have successfully synthesized InAs/InP QDs with narrow emission lines spanning a broad spectral range of λ∼1200–1600 nm.The morphological and optical properties of the samples were characterized using atomic force microscopy and microphotoluminescence spectroscopy.The recorded single-photon purity of a plain QD structure reaches g^((2))(0)=0.16,with a radiative recombination lifetime as short as 1.5 ns.This work provides a crucial platform for future research on integrated microcavity enhancement techniques and coupled QDs with other quantum photonics in the telecom bands,offering significant prospects for quantum network applications. 展开更多
关键词 development quantum materials broadband telecom single photon emissions MOVPE droplet epitaxy InAs InP quantum dots microcavity enhancement quantum dots qds using information technologyalthough
原文传递
基于SSD算法的煤矿运输车辆智能识别研究
12
作者 李鑫 景若光 彭飞 《机械设计与制造工程》 2026年第2期133-137,共5页
针对传统车辆识别算法难以适应复杂环境,易出现错误识别与识别迟钝等问题,在单步多框目标检测算法的基础上,提出了一种多尺度特征融合模型,旨在提升算法对煤矿环境的适应性及车辆识别的整体性能。通过对比测试实验结果发现,该模型的召... 针对传统车辆识别算法难以适应复杂环境,易出现错误识别与识别迟钝等问题,在单步多框目标检测算法的基础上,提出了一种多尺度特征融合模型,旨在提升算法对煤矿环境的适应性及车辆识别的整体性能。通过对比测试实验结果发现,该模型的召回率为91.1%,F1值为0.938,准确率为94.2%。此外,在进一步的测试中还发现,模型的平均识别时间为0.04 s,平均正确率达94.5%,且能良好识别复杂车辆场景。由此表明,该模型能良好地应用于煤矿运输车辆的智能识别。 展开更多
关键词 车辆识别 单步多框目标检测 多尺度特征融合 智能识别 改进
在线阅读 下载PDF
基于MSG-SSD的复合绝缘子憎水性等级智能识别方法 被引量:2
13
作者 陈伟华 马士博 +1 位作者 闫孝姮 李健华 《电子测量与仪器学报》 北大核心 2025年第1期234-243,共10页
复合绝缘子憎水性等级的检测是电力系统巡检中的重要环节,针对现有方法存在检测效率低、实时性差及模型结构复杂的问题,提出一种基于MSG-SSD的复合绝缘子憎水性等级智能识别方法。首先,检测模型以SSD算法为基准,采用轻量级MobileNetV2... 复合绝缘子憎水性等级的检测是电力系统巡检中的重要环节,针对现有方法存在检测效率低、实时性差及模型结构复杂的问题,提出一种基于MSG-SSD的复合绝缘子憎水性等级智能识别方法。首先,检测模型以SSD算法为基准,采用轻量级MobileNetV2作为主干网络,在提升模型检测速度的同时实现网络的轻量化;其次,为增强对水迹特征的提取能力,构建高分辨率特征融合模块Sim-HRFPN,在特征融合的同时保留高分辨率的特征,以弥补因轻量化造成的精度损失;最后,为进一步提高模型的计算效率,将GhostConv替换额外预测特征层的传统卷积,在保持模型高性能的同时,减轻了计算负担。实验结果表明,相较于SSD,MSG-SSD的检测速度和检测精度分别提高48.17%和4.89%,计算量和参数量分别减少97.63%和82.99%。由此可知,改进模型不仅能精准识别和快速定位复合绝缘子的憎水性等级,而且满足边缘巡检设备轻量化部署的需求,为电力系统中复合绝缘子运行状态的智能检测提供了一种行之有效的方法。 展开更多
关键词 复合绝缘子 憎水性检测 智能识别 ssd算法 轻量化 特征融合
原文传递
基于改进SSD算法的地铁场景小行人目标检测 被引量:3
14
作者 张秀再 邱野 沈涛 《计算机研究与发展》 北大核心 2025年第2期397-407,共11页
在地铁场景中,小行人目标由于分辨率低,包含特征信息较少,现阶段目标检测器对此类目标的检测仍具有挑战性.SSD目标检测算法利用金字塔网络的多尺度检测头,能一定程度提高行人目标检测性能,但将其应用于地铁等复杂环境中实现小行人目标... 在地铁场景中,小行人目标由于分辨率低,包含特征信息较少,现阶段目标检测器对此类目标的检测仍具有挑战性.SSD目标检测算法利用金字塔网络的多尺度检测头,能一定程度提高行人目标检测性能,但将其应用于地铁等复杂环境中实现小行人目标检测仍具有一定局限性.针对上述问题,提出一种改进SSD算法以加强地铁场景中小行人目标检测效果.通过构建地铁场景行人目标数据集,标注相应标签,同时进行数据预处理操作;在特征提取网络中加入金字塔特征加强模块,将多分支残差单元、亚像素卷积和特征金字塔相结合获得图像多尺度、多感受野融合特征;利用上下文信息融合模块将图像低层特征与上下文特征相融合,生成扩展特征层用于检测小行人目标;设计一种基于Anchor-free的动态正负样本分配策略,为小行人目标生成最优正样本.实验结果表明:提出的改进SSD算法能有效提高地铁场景小行人目标检测性能,对遮挡严重的小行人目标检测,效果提升更为明显. 展开更多
关键词 小行人目标检测 ssd算法 注意力机制 亚像素卷积 多尺度特征融合
在线阅读 下载PDF
RMSFF-SSD:基于重参数化与多尺度特征融合的遥感图像目标检测模型
15
作者 陈海燕 马舒豪 张振霄 《计算机科学》 北大核心 2025年第S2期356-362,共7页
遥感图像目标检测在国土资源调查、灾害监测、军事侦察等领域具有广泛的应用。针对SSD(Single Shot MultiBox Detector)模型在遥感图像目标检测时难以有效提取小目标的特征,从而对小目标检测不利的问题,文中提出了一种基于重参数化与多... 遥感图像目标检测在国土资源调查、灾害监测、军事侦察等领域具有广泛的应用。针对SSD(Single Shot MultiBox Detector)模型在遥感图像目标检测时难以有效提取小目标的特征,从而对小目标检测不利的问题,文中提出了一种基于重参数化与多尺度特征融合的RMSFF-SSD(Reparameterization Multi-Scale Feature Fusion SSD)遥感图像目标检测模型,该模型在SSD模型的基础上进行改进。首先,对SSD的骨干特征提取网络中的卷积层使用具有重参数化性质的卷积来提取特征,同时在重参数化卷积中引入SE注意力机制,以捕获通道之间的依赖关系并抑制无用的特征;其次,将特征提取网络中提取到的特征用多级特征融合的方式对全局信息与局部细节信息进行融合,来进一步增强目标的特征;最后,将融合后所获得的6个不同尺度的特征图用于目标检测。在NWPU VHR-10数据集上进行目标检测实验,实验结果表明,所提出的RMSFF-SSD512目标检测模型平均精度为89.7%,显著高于DSSD(78.7%)模型、FSSD(86.7%)模型、FPN(68.9%)模型、Faster R-CNN(44.2%)模型和YOLOv5(83.7%)模型。 展开更多
关键词 重参数化 特征融合 遥感检测 ssd SE注意力机制
在线阅读 下载PDF
基于改进SSD模型的环焊缝DR图像缺陷检测
16
作者 张宽 王树强 《化工装备技术》 2025年第5期33-37,共5页
针对环焊缝数字化射线成像(DR)图像缺陷快速准确的检测,进而提高效率并减少人力消耗等问题,选用VGG16作为主干网络,以SSD(Single Shot MultiBox Detector)为目标检测基础框架,引入了一种能够有效实现局部跨通道交互且无需降维的注意力... 针对环焊缝数字化射线成像(DR)图像缺陷快速准确的检测,进而提高效率并减少人力消耗等问题,选用VGG16作为主干网络,以SSD(Single Shot MultiBox Detector)为目标检测基础框架,引入了一种能够有效实现局部跨通道交互且无需降维的注意力机制——ECA(Efficient Channel Attention)模块,构建了一种新型ECA-SSD模型用于缺陷识别,实现对焊缝缺陷的精准分类。结果显示,改进后的网络模型在环焊缝DR图像数据集上准确率达到90.81%,召回率达到93.23%,性能优于传统的焊缝缺陷检测算法,能够实现快速高效的缺陷识别。 展开更多
关键词 焊缝缺陷检测 目标检测 ssd模型 ECA
在线阅读 下载PDF
SSD(Single Shot MultiBox Detector)目标检测算法的研究与改进 被引量:33
17
作者 汪宋 费树岷 《工业控制计算机》 2019年第4期103-105,共3页
目标检测是图像处理的重要领域,SSD算法是基于深度学习下对物体进行检测。首先对SSD检测算法进行了深入研究,然后使用MxNet框架对SSD算法进行了实验,并对训练过程m AP值以及分类准确率进行了分析。由于SSD算法对小物体的检测准确率并没... 目标检测是图像处理的重要领域,SSD算法是基于深度学习下对物体进行检测。首先对SSD检测算法进行了深入研究,然后使用MxNet框架对SSD算法进行了实验,并对训练过程m AP值以及分类准确率进行了分析。由于SSD算法对小物体的检测准确率并没有那么理想,通过在预测层加入反卷积和对预测框长宽比的调整,使得改进后物体检测的平均准确率得到了可观的提高,并且提高了对小物体的检测准确率。 展开更多
关键词 目标检测 ssd 深度学习 MxNet
在线阅读 下载PDF
基于卷积神经网络轻量化的改进SSD异纤检测方法 被引量:4
18
作者 胡胜 王紫悦 +3 位作者 张守京 李博豪 赵小惠 刘文慧 《计算机集成制造系统》 北大核心 2025年第1期171-181,共11页
精准检测棉花中混杂的小型异纤是保障纱线与织物质量的基础和关键。针对现有算法在棉花小型异纤检测中存在的漏检率高、网络结构复杂等问题,提出一种基于卷积神经网络轻量化的改进单步多框检测器(SSD)的棉花异纤检测方法。首先,通过引... 精准检测棉花中混杂的小型异纤是保障纱线与织物质量的基础和关键。针对现有算法在棉花小型异纤检测中存在的漏检率高、网络结构复杂等问题,提出一种基于卷积神经网络轻量化的改进单步多框检测器(SSD)的棉花异纤检测方法。首先,通过引入深度可分离卷积、倒残差结构等创新性设计,将SSD算法中原有骨干特征提取网络VGGNet16替换为MobileNetv2网络;然后,对于SSD算法中生成的候选框尺寸与棉花异纤大小不匹配导致棉花背景占比过高,从而引起正负样本不均衡的问题,采用K-means++算法对棉花异纤尺寸进行聚类分析,根据聚类结果修正候选框尺寸。通过算例进行验证,结果显示所提方法在实现模型轻量化的同时有效提升了异纤检测效果和计算效率。 展开更多
关键词 异纤检测 改进ssd 卷积神经网络 K-means++聚类 轻量化
在线阅读 下载PDF
基于VMD-SSDEO的全并联AT牵引网故障定位 被引量:1
19
作者 李泽文 冯译萱 +4 位作者 夏翊翔 张一鸣 刘国胜 张灵芝 罗姗姗 《电力科学与技术学报》 北大核心 2025年第1期55-66,共12页
针对全并联自耦变压器(auto-transformer,AT)牵引网故障点反射波波头微弱难辨识,且线路并联结构导致故障行波折反射复杂而造成定位困难的问题,提出基于变分模态分解(variational mode decomposition,VMD)和改进能量算子的牵引网故障行... 针对全并联自耦变压器(auto-transformer,AT)牵引网故障点反射波波头微弱难辨识,且线路并联结构导致故障行波折反射复杂而造成定位困难的问题,提出基于变分模态分解(variational mode decomposition,VMD)和改进能量算子的牵引网故障行波单端定位方法。首先,深入分析牵引网故障行波的传输特性,研究全并联结构对行波折反射的影响,确定不同故障类型和潮流特征,以此为依据提取故障特征量,将波头辨识转化为能量突变值提取;随后,利用VMD去噪提取电压行波真实分量,再针对第二反射波波头微弱难标定问题,在滑动时间窗口(sliding time window,STW)下,结合对称差分能量算子(symmetrical differencing energy operator,SDEO)构造故障信号的二次瞬时能量谱,效果良好。仿真结果表明:所提方法抗过渡电阻能力强,能够反应不同工况下故障牵引网电磁能量的变化,具有较高的定位精度。 展开更多
关键词 全并联AT牵引网 故障测距 行波单端定位 对称差分能量算子
在线阅读 下载PDF
基于改进型SSD算法的钢缆表面缺陷检测方法 被引量:1
20
作者 姜冲 莫文洁 +2 位作者 冯上榜 杨永杰 许鹏 《电子设计工程》 2025年第7期66-71,共6页
针对在框绞过程中钢缆表面会产生缺陷的问题,采用了改进型SSD算法的方法。具体措施为:使用ResNet18作为SSD的骨干算法来提取特征信息;通过萤火虫优化算法和K-Means算法优化先验框,提高匹配精度;引入单向特征融合模块和改进的CBAM注意力... 针对在框绞过程中钢缆表面会产生缺陷的问题,采用了改进型SSD算法的方法。具体措施为:使用ResNet18作为SSD的骨干算法来提取特征信息;通过萤火虫优化算法和K-Means算法优化先验框,提高匹配精度;引入单向特征融合模块和改进的CBAM注意力模块,提高了检测的精度;将Focalloss作为损失函数,减少训练过程中负样本的权值。结合钢缆缺陷数据集进行训练实验,得出改进型SSD算法平均精度为80.3%,相对于传统的SSD-VGG模型提升9%,FPS保持在63.3。在检测精度和检测速度方面上,改进型SSD算法能满足实际需求。 展开更多
关键词 钢缆缺陷 注意力机制 目标检测 ssd算法 特征融合
在线阅读 下载PDF
上一页 1 2 250 下一页 到第
使用帮助 返回顶部