Fog computing is introduced to relieve the problems triggered by the long distance between the cloud and terminal devices. In this paper, considering the mobility of terminal devices represented as mobile multimedia u...Fog computing is introduced to relieve the problems triggered by the long distance between the cloud and terminal devices. In this paper, considering the mobility of terminal devices represented as mobile multimedia users(MMUs) and the continuity of requests delivered by them, we propose an online resource allocation scheme with respect to deciding the state of servers in fog nodes distributed at different zones on the premise of satisfying the quality of experience(QoE) based on a Stackelberg game. Specifically, a multi-round of a predictably\unpredictably dynamic scheme is derived from a single-round of a static scheme. The optimal allocation schemes are discussed in detail, and related experiments are designed. For simulations, comparing with non-strategy schemes, the performance of the dynamic scheme is better at minimizing the cost used to maintain fog nodes for providing services.展开更多
This paper summarized the previous researches about resin emciency of particleboard in the world, and introduced the computer vision (CV) technique on resin effciency. which has the properties of high measuring speed,...This paper summarized the previous researches about resin emciency of particleboard in the world, and introduced the computer vision (CV) technique on resin effciency. which has the properties of high measuring speed, automatic pattern recognition and low environmental requirement. etc. The theory of the CV technique used for resin effciency in particleboard was studied,along with the handling of resined-particle images and the gathering of relative gray image features. Some quantitative parameters describing the resin efficiency of particleboard were established, and the results indicated that the computer vision method on the resin effciency was much better than others and can control the producing of pafticleboard more effect.展开更多
A key problem that plagues camera self-calibration, namely that the classical self-calibration algorithms are very sensitive to the initial values of the camera intrinsic parameters, is analyzed and a practical soluti...A key problem that plagues camera self-calibration, namely that the classical self-calibration algorithms are very sensitive to the initial values of the camera intrinsic parameters, is analyzed and a practical solution is provided. The effect of the camera intrinsic parameters, mainly the principal point and the skew factor is first discussed. Then a practical method via a controlled motion of the camera is introduced so as to obtain an accurate estimation of these parameters. Feasibility of this approach is illustrated by carrying out comprehensive experiments using synthetic data as well as real image sequences. Unreasonable initial values can often make self-calibration impossible, yet a precise initialization guarantees a better and successful reconstruction. Trying to obtain a more reasonable initialization is worthwhile the effort in camera self-calibration.展开更多
The working processes, machining devices and tools, cutting amount, consumption of materials, productivity and quality of products are directly affected by wood surface roughness. This paper gives an extensive review ...The working processes, machining devices and tools, cutting amount, consumption of materials, productivity and quality of products are directly affected by wood surface roughness. This paper gives an extensive review of methods used previously to measure wood surface roughness, and concludes that computer vision is the most suitable technique. The preliminary study shows that computer vision method has the advantages of a noncontact, three-dimensional measurement, high speed and well correlates with stylus tracing method. This method can be used in classification and in-time measurement of wood surface roughness after being improved.展开更多
The emergence of the Internet-of-Things is anticipated to create a vast market for what are known as smart edge devices,opening numerous opportunities across countless domains,including personalized healthcare and adv...The emergence of the Internet-of-Things is anticipated to create a vast market for what are known as smart edge devices,opening numerous opportunities across countless domains,including personalized healthcare and advanced robotics.Leveraging 3D integration,edge devices can achieve unprecedented miniaturization while simultaneously boosting processing power and minimizing energy consumption.Here,we demonstrate a back-end-of-line compatible optoelectronic synapse with a transfer learning method on health care applications,including electroencephalogram(EEG)-based seizure prediction,electromyography(EMG)-based gesture recognition,and electrocardiogram(ECG)-based arrhythmia detection.With experiments on three biomedical datasets,we observe the classification accuracy improvement for the pretrained model with 2.93%on EEG,4.90%on ECG,and 7.92%on EMG,respectively.The optical programming property of the device enables an ultralow power(2.8×10^(-13) J)fine-tuning process and offers solutions for patient-specific issues in edge computing scenarios.Moreover,the device exhibits impressive light-sensitive characteristics that enable a range of light-triggered synaptic functions,making it promising for neuromorphic vision application.To display the benefits of these intricate synaptic properties,a 5×5 optoelectronic synapse array is developed,effectively simulating human visual perception and memory functions.The proposed flexible optoelectronic synapse holds immense potential for advancing the fields of neuromorphic physiological signal processing and artificial visual systems in wearable applications.展开更多
由于镀金回转体工件的特殊几何特征和尺寸限制,快速准确地获取其全表面图像存在困难。本文提出了一种基于自适应亮度校正的全表面成像方法。首先,为了恢复低亮度区域信息,提出一种自适应调整图像亮度的校正算法,在全局亮度映射预调整后...由于镀金回转体工件的特殊几何特征和尺寸限制,快速准确地获取其全表面图像存在困难。本文提出了一种基于自适应亮度校正的全表面成像方法。首先,为了恢复低亮度区域信息,提出一种自适应调整图像亮度的校正算法,在全局亮度映射预调整后,使用导向滤波器替代传统的高斯滤波器进行图像局部对比度多尺度增强,同时保护划痕和边缘等特征。其次,设计了一种基于ROI(Region of Interest)自适应裁切的图像拼接方法,通过HSV颜色空间下阈值分割和单应矩阵估计提取有效区域,降低图像拼接时由曲面投影失真和视差引起的干扰,并提高算法的运行速度。实验结果表明:本文的亮度校正算法能改善图像特征亮度不一致情况,使得图像配准平均反向投影误差降低约50%,多图像拼接算法速度达1.25幅/s。相比Autostitch、LPC等经典算法,本文算法在精度和效率上都具有明显优势,适用于工业环境中回转体工件的全表面图像获取及缺陷检测。展开更多
中国山水画风格迁移的目标是在保持原有山水真实场景图像内容的前提下,引入传统中国画作特征,以生成具有中国山水画艺术特征的图像。近年,由于深度学习的快速发展,卷积神经网络(CNN)和对抗生成网络(GAN)几乎主导了包括风格迁移在内的大...中国山水画风格迁移的目标是在保持原有山水真实场景图像内容的前提下,引入传统中国画作特征,以生成具有中国山水画艺术特征的图像。近年,由于深度学习的快速发展,卷积神经网络(CNN)和对抗生成网络(GAN)几乎主导了包括风格迁移在内的大部分图像生成任务,但也存在一些问题,如真实场景在风格迁移过程中易丢失语义,GAN网络训练出现模型坍塌,CNN风格迁移方法出现棋盘效应等。视觉Transformer模型为图像处理任务提供了新的解决方案,但训练需大量数据且计算复杂。为了解决生成中国画过程中由上述因素引起的图像质量低及细节特征丢失等问题,本文提出一种能基于细节特征提取融合的中国山水画风格迁移网络,即SSTR(swin style transfer transformer)。该网络在StyTr^(2)网络的基础上,引入了Swin–Transformer模型,利用视觉Transformer的强语义性保留山水场景的特征;同时利用Swin–Transformer模型的分层体系结构及滑窗操作计算注意力机制,提取更多的山水画艺术风格细节,同时降低模型训练复杂度;最后,引入一个CNN解码器细化生成目标图像。本文利用公开视觉数据集COCO 2014与公开山水画数据集进行训练、验证与测试,并将结果与基线方法进行比较。结果表明,SSTR在处理中国山水画风格迁移任务中,风格损失和内容损失分别为1.35和1.88,在风格损失上优于StyTr^(2),表现出了优异的特征提取能力和图像生成能力。展开更多
为实现南极磷虾粉中虾青素含量的快速检测,借助计算机视觉和卷积神经网络建立了一种虾粉虾青素含量的测定方法。以70个南极磷虾粉样本,通过高效液相色谱法测定虾青素含量,计算机视觉系统采集图像,将虾青素含量与图像对应组成数据集并对...为实现南极磷虾粉中虾青素含量的快速检测,借助计算机视觉和卷积神经网络建立了一种虾粉虾青素含量的测定方法。以70个南极磷虾粉样本,通过高效液相色谱法测定虾青素含量,计算机视觉系统采集图像,将虾青素含量与图像对应组成数据集并对数据集进行数据增强;使用TensorFlow学习框架构建模型,使用5折交叉验证进行模型调参及评估并选出最优参数模型;随机划分数据集对最优参数模型进行评估,最后随机挑选数据集中的30张图像进行模型验证。结果显示经过交叉验证后的最优参数模型的均方根误差(Root Mean Square Error,RMSE)为3.59;模型评估阶段,模型重复运行3次,测试集的决定系数(Coefficient of Determination,R2)、均绝对误差(Mean Absolute Error,MAE)、均方误差(Mean Square Error,MSE)、RMSE的平均值分别为0.9626、1.49、4.22、2.05。模型验证阶段,模型预测虾青素含量的相对误差介于0.10%~6.46%之间,预测结果与观测值之间偏差较小。因此,该虾青素含量预测模型能够较准确地预测虾青素含量,进而实现虾粉虾青素含量的快速无损检测。展开更多
【目的】绿化资源配置是城市公共空间优化的重要环节之一,对居民生活质量的提升有着积极的作用。城市街道绿化泛类结构(urban street greening general structure,USGGS)能够反映街道绿化在行人视觉环境中的整体特征,研究USGGS聚类对于...【目的】绿化资源配置是城市公共空间优化的重要环节之一,对居民生活质量的提升有着积极的作用。城市街道绿化泛类结构(urban street greening general structure,USGGS)能够反映街道绿化在行人视觉环境中的整体特征,研究USGGS聚类对于物质空间要素数量以及物质空间形态的改变,能够有效探究街道绿化对行人视觉感知水平的影响。【方法】采用百度街景数据,利用DeepLabV3+神经网络模型,对天津市市内六区街道的物质空间要素进行分割,使用ArcGIS软件对空间分布特征进行可视化处理,结合数理统计分析结果,探讨USGGS与行人视觉感知之间的关系。【结果】USGGS聚类呈现向心聚集型的空间分布特征,城市主干道及快速路的行人视觉感知空间分布特征较为同质化,空间异质化现象集中出现在街道断面狭窄的生活型街道以及商业型街道。不同聚类的USGGS不仅对行人视觉感知有不同程度的影响,也与场所属性以及绿化空间位置密切相关。【结论】提升城市街道环境质量需要考虑行人视觉感知水平。合理的USGGS配置以及适当的种植点位能够更好地适应周围场所的属性,促进城市公共空间与城市街道绿化的有机融合,助推城市更新工作的精细化管理,提升城市人居环境质量。展开更多
基金supported by the National Natural Science Foundation of China under grant No. 61501080, 61572095, 61871064, and 61877007
文摘Fog computing is introduced to relieve the problems triggered by the long distance between the cloud and terminal devices. In this paper, considering the mobility of terminal devices represented as mobile multimedia users(MMUs) and the continuity of requests delivered by them, we propose an online resource allocation scheme with respect to deciding the state of servers in fog nodes distributed at different zones on the premise of satisfying the quality of experience(QoE) based on a Stackelberg game. Specifically, a multi-round of a predictably\unpredictably dynamic scheme is derived from a single-round of a static scheme. The optimal allocation schemes are discussed in detail, and related experiments are designed. For simulations, comparing with non-strategy schemes, the performance of the dynamic scheme is better at minimizing the cost used to maintain fog nodes for providing services.
文摘This paper summarized the previous researches about resin emciency of particleboard in the world, and introduced the computer vision (CV) technique on resin effciency. which has the properties of high measuring speed, automatic pattern recognition and low environmental requirement. etc. The theory of the CV technique used for resin effciency in particleboard was studied,along with the handling of resined-particle images and the gathering of relative gray image features. Some quantitative parameters describing the resin efficiency of particleboard were established, and the results indicated that the computer vision method on the resin effciency was much better than others and can control the producing of pafticleboard more effect.
文摘A key problem that plagues camera self-calibration, namely that the classical self-calibration algorithms are very sensitive to the initial values of the camera intrinsic parameters, is analyzed and a practical solution is provided. The effect of the camera intrinsic parameters, mainly the principal point and the skew factor is first discussed. Then a practical method via a controlled motion of the camera is introduced so as to obtain an accurate estimation of these parameters. Feasibility of this approach is illustrated by carrying out comprehensive experiments using synthetic data as well as real image sequences. Unreasonable initial values can often make self-calibration impossible, yet a precise initialization guarantees a better and successful reconstruction. Trying to obtain a more reasonable initialization is worthwhile the effort in camera self-calibration.
文摘The working processes, machining devices and tools, cutting amount, consumption of materials, productivity and quality of products are directly affected by wood surface roughness. This paper gives an extensive review of methods used previously to measure wood surface roughness, and concludes that computer vision is the most suitable technique. The preliminary study shows that computer vision method has the advantages of a noncontact, three-dimensional measurement, high speed and well correlates with stylus tracing method. This method can be used in classification and in-time measurement of wood surface roughness after being improved.
基金financial support by the Semiconductor Initiative at the King Abdullah University of Science and Technologysupported by King Abdullah University of Science and Technology(KAUST)Research Funding(KRF)under Award No.ORA-2022-5314.
文摘The emergence of the Internet-of-Things is anticipated to create a vast market for what are known as smart edge devices,opening numerous opportunities across countless domains,including personalized healthcare and advanced robotics.Leveraging 3D integration,edge devices can achieve unprecedented miniaturization while simultaneously boosting processing power and minimizing energy consumption.Here,we demonstrate a back-end-of-line compatible optoelectronic synapse with a transfer learning method on health care applications,including electroencephalogram(EEG)-based seizure prediction,electromyography(EMG)-based gesture recognition,and electrocardiogram(ECG)-based arrhythmia detection.With experiments on three biomedical datasets,we observe the classification accuracy improvement for the pretrained model with 2.93%on EEG,4.90%on ECG,and 7.92%on EMG,respectively.The optical programming property of the device enables an ultralow power(2.8×10^(-13) J)fine-tuning process and offers solutions for patient-specific issues in edge computing scenarios.Moreover,the device exhibits impressive light-sensitive characteristics that enable a range of light-triggered synaptic functions,making it promising for neuromorphic vision application.To display the benefits of these intricate synaptic properties,a 5×5 optoelectronic synapse array is developed,effectively simulating human visual perception and memory functions.The proposed flexible optoelectronic synapse holds immense potential for advancing the fields of neuromorphic physiological signal processing and artificial visual systems in wearable applications.
文摘由于镀金回转体工件的特殊几何特征和尺寸限制,快速准确地获取其全表面图像存在困难。本文提出了一种基于自适应亮度校正的全表面成像方法。首先,为了恢复低亮度区域信息,提出一种自适应调整图像亮度的校正算法,在全局亮度映射预调整后,使用导向滤波器替代传统的高斯滤波器进行图像局部对比度多尺度增强,同时保护划痕和边缘等特征。其次,设计了一种基于ROI(Region of Interest)自适应裁切的图像拼接方法,通过HSV颜色空间下阈值分割和单应矩阵估计提取有效区域,降低图像拼接时由曲面投影失真和视差引起的干扰,并提高算法的运行速度。实验结果表明:本文的亮度校正算法能改善图像特征亮度不一致情况,使得图像配准平均反向投影误差降低约50%,多图像拼接算法速度达1.25幅/s。相比Autostitch、LPC等经典算法,本文算法在精度和效率上都具有明显优势,适用于工业环境中回转体工件的全表面图像获取及缺陷检测。
文摘中国山水画风格迁移的目标是在保持原有山水真实场景图像内容的前提下,引入传统中国画作特征,以生成具有中国山水画艺术特征的图像。近年,由于深度学习的快速发展,卷积神经网络(CNN)和对抗生成网络(GAN)几乎主导了包括风格迁移在内的大部分图像生成任务,但也存在一些问题,如真实场景在风格迁移过程中易丢失语义,GAN网络训练出现模型坍塌,CNN风格迁移方法出现棋盘效应等。视觉Transformer模型为图像处理任务提供了新的解决方案,但训练需大量数据且计算复杂。为了解决生成中国画过程中由上述因素引起的图像质量低及细节特征丢失等问题,本文提出一种能基于细节特征提取融合的中国山水画风格迁移网络,即SSTR(swin style transfer transformer)。该网络在StyTr^(2)网络的基础上,引入了Swin–Transformer模型,利用视觉Transformer的强语义性保留山水场景的特征;同时利用Swin–Transformer模型的分层体系结构及滑窗操作计算注意力机制,提取更多的山水画艺术风格细节,同时降低模型训练复杂度;最后,引入一个CNN解码器细化生成目标图像。本文利用公开视觉数据集COCO 2014与公开山水画数据集进行训练、验证与测试,并将结果与基线方法进行比较。结果表明,SSTR在处理中国山水画风格迁移任务中,风格损失和内容损失分别为1.35和1.88,在风格损失上优于StyTr^(2),表现出了优异的特征提取能力和图像生成能力。
文摘为实现南极磷虾粉中虾青素含量的快速检测,借助计算机视觉和卷积神经网络建立了一种虾粉虾青素含量的测定方法。以70个南极磷虾粉样本,通过高效液相色谱法测定虾青素含量,计算机视觉系统采集图像,将虾青素含量与图像对应组成数据集并对数据集进行数据增强;使用TensorFlow学习框架构建模型,使用5折交叉验证进行模型调参及评估并选出最优参数模型;随机划分数据集对最优参数模型进行评估,最后随机挑选数据集中的30张图像进行模型验证。结果显示经过交叉验证后的最优参数模型的均方根误差(Root Mean Square Error,RMSE)为3.59;模型评估阶段,模型重复运行3次,测试集的决定系数(Coefficient of Determination,R2)、均绝对误差(Mean Absolute Error,MAE)、均方误差(Mean Square Error,MSE)、RMSE的平均值分别为0.9626、1.49、4.22、2.05。模型验证阶段,模型预测虾青素含量的相对误差介于0.10%~6.46%之间,预测结果与观测值之间偏差较小。因此,该虾青素含量预测模型能够较准确地预测虾青素含量,进而实现虾粉虾青素含量的快速无损检测。
文摘【目的】绿化资源配置是城市公共空间优化的重要环节之一,对居民生活质量的提升有着积极的作用。城市街道绿化泛类结构(urban street greening general structure,USGGS)能够反映街道绿化在行人视觉环境中的整体特征,研究USGGS聚类对于物质空间要素数量以及物质空间形态的改变,能够有效探究街道绿化对行人视觉感知水平的影响。【方法】采用百度街景数据,利用DeepLabV3+神经网络模型,对天津市市内六区街道的物质空间要素进行分割,使用ArcGIS软件对空间分布特征进行可视化处理,结合数理统计分析结果,探讨USGGS与行人视觉感知之间的关系。【结果】USGGS聚类呈现向心聚集型的空间分布特征,城市主干道及快速路的行人视觉感知空间分布特征较为同质化,空间异质化现象集中出现在街道断面狭窄的生活型街道以及商业型街道。不同聚类的USGGS不仅对行人视觉感知有不同程度的影响,也与场所属性以及绿化空间位置密切相关。【结论】提升城市街道环境质量需要考虑行人视觉感知水平。合理的USGGS配置以及适当的种植点位能够更好地适应周围场所的属性,促进城市公共空间与城市街道绿化的有机融合,助推城市更新工作的精细化管理,提升城市人居环境质量。
