The identification of hydrocarbons using seismic methods is critical in the prediction of shale oil res-ervoirs.However,delineating shales of high oil saturation is challenging owing to the similarity in the elastic p...The identification of hydrocarbons using seismic methods is critical in the prediction of shale oil res-ervoirs.However,delineating shales of high oil saturation is challenging owing to the similarity in the elastic properties of oil-and water-bearing shales.The complexity of the organic matter properties associated with kerogen and hydrocarbon further complicates the characterization of shale oil reservoirs using seismic methods.Nevertheless,the inelastic shale properties associated with oil saturation can enable the utilization of velocity dispersion for hydrocarbon identification in shales.In this study,a seismic inversion scheme based on the fluid dispersion attribute was proposed for the estimation of hydrocarbon enrichment.In the proposed approach,the conventional frequency-dependent inversion scheme was extended by incorporating the PP-wave reflection coefficient presented in terms of the effective fluid bulk modulus.A rock physics model for shale oil reservoirs was constructed to describe the relationship between hydrocarbon saturation and shale inelasticity.According to the modeling results,the hydrocarbon sensitivity of the frequency-dependent effective fluid bulk modulus is superior to the traditional compressional wave velocity dispersion of shales.Quantitative analysis of the inversion re-sults based on synthetics also reveals that the proposed approach identifies the oil saturation and related hydrocarbon enrichment better than the above-mentioned conventional approach.Meanwhile,in real data applications,actual drilling results validate the superiority of the proposed fluid dispersion attribute as a useful hydrocarbon indicator in shale oil reservoirs.展开更多
Person re-identification(re-id)involves matching a person across nonoverlapping views,with different poses,illuminations and conditions.Visual attributes are understandable semantic information to help improve the iss...Person re-identification(re-id)involves matching a person across nonoverlapping views,with different poses,illuminations and conditions.Visual attributes are understandable semantic information to help improve the issues including illumination changes,viewpoint variations and occlusions.This paper proposes an end-to-end framework of deep learning for attribute-based person re-id.In the feature representation stage of framework,the improved convolutional neural network(CNN)model is designed to leverage the information contained in automatically detected attributes and learned low-dimensional CNN features.Moreover,an attribute classifier is trained on separate data and includes its responses into the training process of our person re-id model.The coupled clusters loss function is used in the training stage of the framework,which enhances the discriminability of both types of features.The combined features are mapped into the Euclidean space.The L2 distance can be used to calculate the distance between any two pedestrians to determine whether they are the same.Extensive experiments validate the superiority and advantages of our proposed framework over state-of-the-art competitors on contemporary challenging person re-id datasets.展开更多
Kinematic semantics is often an important content of a CAD model(it refers to a single part/solid model in this work)in many applications,but it is usually not the belonging of the model,especially for the one retriev...Kinematic semantics is often an important content of a CAD model(it refers to a single part/solid model in this work)in many applications,but it is usually not the belonging of the model,especially for the one retrieved from a common database.Especially,the effective and automatic method to reconstruct the above information for a CAD model is still rare.To address this issue,this paper proposes a smart approach to identify each assembly interface on every CAD model since the assembly interface is the fundamental but key element of reconstructing kinematic semantics.First,as the geometry of an assembly interface is formed by one or more adjacent faces on each model,a face-attributed adjacency graph integrated with face structure fingerprint is proposed.This can describe each CAD model as well as its assembly interfaces uniformly.After that,aided by the above descriptor,an improved graph attention network is developed based on a new dual-level anti-interference filtering mechanism,which makes it have the great potential to identify all representative kinds of assembly interface faces with high accuracy that have various geometric shapes but consistent kinematic semantics.Moreover,based on the abovementioned graph and face-adjacent relationships,each assembly interface on a model can be identified.Finally,experiments on representative CAD models are implemented to verify the effectiveness and characteristics of the proposed approach.The results show that the average assembly-interface-face-identification accuracy of the proposed approach can reach 91.75%,which is about 2%–5%higher than those of the recent-representative graph neural networks.Besides,compared with the state-of-the-art methods,our approach is more suitable to identify the assembly interfaces(with various shapes)for each individual CAD model that has typical kinematic pairs.展开更多
The boundary identification and quantitative thickness prediction of channel sand bodies are always difficult in seismic exploration.We present a new method for boundary identification and quantitative thickness predi...The boundary identification and quantitative thickness prediction of channel sand bodies are always difficult in seismic exploration.We present a new method for boundary identification and quantitative thickness prediction of channel sand bodies based on seismic peak attributes in the frequency domain.Using seismic forward modeling of a typical thin channel sand body,a new seismic attribute-the ratio of peak frequency to amplitude was constructed.Theoretical study demonstrated that seismic peak frequency is sensitive to the thickness of the channel sand bodies,while the amplitude attribute is sensitive to the strata lithology.The ratio of the two attributes can highlight the boundaries of the channel sand body.Moreover,the thickness of the thin channel sand bodies can be determined using the relationship between seismic peak frequency and thin layer thickness.Practical applications have demonstrated that the seismic peak frequency attribute can depict the horizontal distribution characteristics of channels very well.The ratio of peak frequency to amplitude attribute can improve the identification ability of channel sand body boundaries.Quantitative prediction and boundary identification of channel sand bodies with seismic peak attributes in the frequency domain are feasible.展开更多
针对目前RFID(Radio Frequency Identification,射频识别技术)系统安全分析中忽略攻击事件对系统安全状态动态影响的问题,为了有效实现RFID系统的安全风险评估,文章提出了一种基于贝叶斯攻击图的RFID系统安全评估模型。该模型首先通过对...针对目前RFID(Radio Frequency Identification,射频识别技术)系统安全分析中忽略攻击事件对系统安全状态动态影响的问题,为了有效实现RFID系统的安全风险评估,文章提出了一种基于贝叶斯攻击图的RFID系统安全评估模型。该模型首先通过对RFID系统结构、所用协议进行分析确定系统的脆弱性漏洞及其依赖关系,建立攻击图。针对攻击图模型只能进行定性分析的问题,构建出相应的攻击图模型结构后可以结合贝叶斯理论对其进行量化。依据漏洞的利用难易度和影响程度建立RFID漏洞量化评价指标,计算出对应的原子攻击概率,然后以条件转移概率的形式将攻击节点与RFID系统的安全属性节点联系在一起,不仅能推断攻击者能够成功到达各个属性节点的风险概率,而且能够依据攻击者的不同行为动态展示系统风险状况的变化,实现评估不同状态下目标RFID系统的整体风险状况。实验表明,所提模型可以有效地计算出RFID系统整体的风险概率,为后续实施对应的安全策略提供理论依据。展开更多
跨模态行人重识别研究的重难点主要来自于行人图像之间巨大的模态差异和模态内差异。针对这些问题,提出一种结合多尺度特征与混淆学习的网络结构。为实现高效的特征提取、缩小模态内差异,将网络设计为多尺度特征互补的形式,分别学习行...跨模态行人重识别研究的重难点主要来自于行人图像之间巨大的模态差异和模态内差异。针对这些问题,提出一种结合多尺度特征与混淆学习的网络结构。为实现高效的特征提取、缩小模态内差异,将网络设计为多尺度特征互补的形式,分别学习行人的局部细化特征与全局粗糙特征,从细粒度和粗粒度两方面来增强网络的特征表达能力。利用混淆学习策略,模糊网络的模态识别反馈,挖掘稳定且有效的模态无关属性应对模态差异,来提高特征对模态变化的鲁棒性。在大规模数据集SYSU-MM01的全搜索模式下该算法首位击中率和平均精度(mean average precision,mAP)的结果分别为76.69%和72.45%,在RegDB数据集的可见光到红外模式下该算法首位击中率和mAP的结果分别为94.62%和94.60%,优于现有的主要方法,验证了所提方法的有效性。展开更多
基金supported by the National Natural Science Foundation of China(Grant numbers 42074153 and 42274160)the Open Research Fund of SINOPEC Key Laboratory of Geophysics(Grant number 33550006-20-ZC0699-0006).
文摘The identification of hydrocarbons using seismic methods is critical in the prediction of shale oil res-ervoirs.However,delineating shales of high oil saturation is challenging owing to the similarity in the elastic properties of oil-and water-bearing shales.The complexity of the organic matter properties associated with kerogen and hydrocarbon further complicates the characterization of shale oil reservoirs using seismic methods.Nevertheless,the inelastic shale properties associated with oil saturation can enable the utilization of velocity dispersion for hydrocarbon identification in shales.In this study,a seismic inversion scheme based on the fluid dispersion attribute was proposed for the estimation of hydrocarbon enrichment.In the proposed approach,the conventional frequency-dependent inversion scheme was extended by incorporating the PP-wave reflection coefficient presented in terms of the effective fluid bulk modulus.A rock physics model for shale oil reservoirs was constructed to describe the relationship between hydrocarbon saturation and shale inelasticity.According to the modeling results,the hydrocarbon sensitivity of the frequency-dependent effective fluid bulk modulus is superior to the traditional compressional wave velocity dispersion of shales.Quantitative analysis of the inversion re-sults based on synthetics also reveals that the proposed approach identifies the oil saturation and related hydrocarbon enrichment better than the above-mentioned conventional approach.Meanwhile,in real data applications,actual drilling results validate the superiority of the proposed fluid dispersion attribute as a useful hydrocarbon indicator in shale oil reservoirs.
基金supported by the National Natural Science Foundation of China(6147115461876057)the Fundamental Research Funds for Central Universities(JZ2018YYPY0287)
文摘Person re-identification(re-id)involves matching a person across nonoverlapping views,with different poses,illuminations and conditions.Visual attributes are understandable semantic information to help improve the issues including illumination changes,viewpoint variations and occlusions.This paper proposes an end-to-end framework of deep learning for attribute-based person re-id.In the feature representation stage of framework,the improved convolutional neural network(CNN)model is designed to leverage the information contained in automatically detected attributes and learned low-dimensional CNN features.Moreover,an attribute classifier is trained on separate data and includes its responses into the training process of our person re-id model.The coupled clusters loss function is used in the training stage of the framework,which enhances the discriminability of both types of features.The combined features are mapped into the Euclidean space.The L2 distance can be used to calculate the distance between any two pedestrians to determine whether they are the same.Extensive experiments validate the superiority and advantages of our proposed framework over state-of-the-art competitors on contemporary challenging person re-id datasets.
基金supported by the National Natural Science Foundation of China[61702147]the Zhejiang Provincial Science and Technology Program in China[2021C03137].
文摘Kinematic semantics is often an important content of a CAD model(it refers to a single part/solid model in this work)in many applications,but it is usually not the belonging of the model,especially for the one retrieved from a common database.Especially,the effective and automatic method to reconstruct the above information for a CAD model is still rare.To address this issue,this paper proposes a smart approach to identify each assembly interface on every CAD model since the assembly interface is the fundamental but key element of reconstructing kinematic semantics.First,as the geometry of an assembly interface is formed by one or more adjacent faces on each model,a face-attributed adjacency graph integrated with face structure fingerprint is proposed.This can describe each CAD model as well as its assembly interfaces uniformly.After that,aided by the above descriptor,an improved graph attention network is developed based on a new dual-level anti-interference filtering mechanism,which makes it have the great potential to identify all representative kinds of assembly interface faces with high accuracy that have various geometric shapes but consistent kinematic semantics.Moreover,based on the abovementioned graph and face-adjacent relationships,each assembly interface on a model can be identified.Finally,experiments on representative CAD models are implemented to verify the effectiveness and characteristics of the proposed approach.The results show that the average assembly-interface-face-identification accuracy of the proposed approach can reach 91.75%,which is about 2%–5%higher than those of the recent-representative graph neural networks.Besides,compared with the state-of-the-art methods,our approach is more suitable to identify the assembly interfaces(with various shapes)for each individual CAD model that has typical kinematic pairs.
基金supported by National Key Science and Technology Special Projects (Grant No.2008ZX05000-004)CNPC Key S and T Special Projects (Grant No.2008E-0610-10)
文摘The boundary identification and quantitative thickness prediction of channel sand bodies are always difficult in seismic exploration.We present a new method for boundary identification and quantitative thickness prediction of channel sand bodies based on seismic peak attributes in the frequency domain.Using seismic forward modeling of a typical thin channel sand body,a new seismic attribute-the ratio of peak frequency to amplitude was constructed.Theoretical study demonstrated that seismic peak frequency is sensitive to the thickness of the channel sand bodies,while the amplitude attribute is sensitive to the strata lithology.The ratio of the two attributes can highlight the boundaries of the channel sand body.Moreover,the thickness of the thin channel sand bodies can be determined using the relationship between seismic peak frequency and thin layer thickness.Practical applications have demonstrated that the seismic peak frequency attribute can depict the horizontal distribution characteristics of channels very well.The ratio of peak frequency to amplitude attribute can improve the identification ability of channel sand body boundaries.Quantitative prediction and boundary identification of channel sand bodies with seismic peak attributes in the frequency domain are feasible.
文摘针对目前RFID(Radio Frequency Identification,射频识别技术)系统安全分析中忽略攻击事件对系统安全状态动态影响的问题,为了有效实现RFID系统的安全风险评估,文章提出了一种基于贝叶斯攻击图的RFID系统安全评估模型。该模型首先通过对RFID系统结构、所用协议进行分析确定系统的脆弱性漏洞及其依赖关系,建立攻击图。针对攻击图模型只能进行定性分析的问题,构建出相应的攻击图模型结构后可以结合贝叶斯理论对其进行量化。依据漏洞的利用难易度和影响程度建立RFID漏洞量化评价指标,计算出对应的原子攻击概率,然后以条件转移概率的形式将攻击节点与RFID系统的安全属性节点联系在一起,不仅能推断攻击者能够成功到达各个属性节点的风险概率,而且能够依据攻击者的不同行为动态展示系统风险状况的变化,实现评估不同状态下目标RFID系统的整体风险状况。实验表明,所提模型可以有效地计算出RFID系统整体的风险概率,为后续实施对应的安全策略提供理论依据。
文摘跨模态行人重识别研究的重难点主要来自于行人图像之间巨大的模态差异和模态内差异。针对这些问题,提出一种结合多尺度特征与混淆学习的网络结构。为实现高效的特征提取、缩小模态内差异,将网络设计为多尺度特征互补的形式,分别学习行人的局部细化特征与全局粗糙特征,从细粒度和粗粒度两方面来增强网络的特征表达能力。利用混淆学习策略,模糊网络的模态识别反馈,挖掘稳定且有效的模态无关属性应对模态差异,来提高特征对模态变化的鲁棒性。在大规模数据集SYSU-MM01的全搜索模式下该算法首位击中率和平均精度(mean average precision,mAP)的结果分别为76.69%和72.45%,在RegDB数据集的可见光到红外模式下该算法首位击中率和mAP的结果分别为94.62%和94.60%,优于现有的主要方法,验证了所提方法的有效性。