This paper re-examines the status quo of China's territorial sovereignty security from a strategic perspective.Territory is the most basic physical where-about for a country to express national sovereignty.It is t...This paper re-examines the status quo of China's territorial sovereignty security from a strategic perspective.Territory is the most basic physical where-about for a country to express national sovereignty.It is the territory that the national sovereignty derives its physical basis from.In the context of globalization,however, the traditional national territorial sovereignty space is continually squeezed and constrained,but the national territorial sovereignty is still the cornerstone of the international relations.And the inviolability of national territorial sovereignty is still the most important principle of modern international law.In this regard,the primary goal of Chinese national security today remains as a goal to safeguard the unification,integrity and security of China's territorial sovereignty.At present,the integrity,unification and security of Chinese national territorial sovereignty have yet to be achieved on the strategic level.As one of China's basic national conditions,it should be taken seriously and paid adequate strategic attentions.展开更多
目的在行人再识别中,行人朝向变化会导致表观变化,进而导致关联错误。现有方法通过朝向表示学习和基于朝向的损失函数来改善这一问题。然而,大多数朝向表示学习方法主要以嵌入朝向标签为主,并没有显式的向模型传达行人姿态的空间结构,...目的在行人再识别中,行人朝向变化会导致表观变化,进而导致关联错误。现有方法通过朝向表示学习和基于朝向的损失函数来改善这一问题。然而,大多数朝向表示学习方法主要以嵌入朝向标签为主,并没有显式的向模型传达行人姿态的空间结构,从而减弱了模型对朝向的感知能力。此外,基于朝向的损失函数通常对相同身份的行人进行朝向聚类,忽略了由表观相似且朝向相同的负样本造成的错误关联的问题。方法为了应对这些挑战,提出了面向行人再识别的朝向感知特征学习。首先,提出了基于人体姿态的朝向特征学习,它能够显式地捕捉人体姿态的空间结构。其次,提出的朝向自适应的三元组损失主动增大表观相似且相同朝向行人之间的间隔,进而将它们分离。结果本文方法在大规模的行人再识别公开数据集MSMT17(multi-scene multi-time person ReID dataset)、Market1501等上进行测试。其中,在MSMT17数据集上,相比于性能第2的UniHCP(unified model for human-centric perceptions)模型,Rank1和mAP值分别提高了1.7%和1.3%;同时,在MSMT17数据集上的消融实验结果证明本文提出的算法有效改善了行人再识别的关联效果。结论本文方法能够有效处理上述挑战导致的行人再识别系统中关联效果变差的问题。展开更多
现有行人重识别技术主要关注水平视角下的图像。在例如无人超市这类特定场景下,摄像头以俯视角度拍摄,仅能获得有限的行人信息。针对此问题,将多模态视觉Transformer应用于俯视图行人重识别任务,利用俯视数据集中额外的深度模态来提高...现有行人重识别技术主要关注水平视角下的图像。在例如无人超市这类特定场景下,摄像头以俯视角度拍摄,仅能获得有限的行人信息。针对此问题,将多模态视觉Transformer应用于俯视图行人重识别任务,利用俯视数据集中额外的深度模态来提高俯视图的检索精度。具体而言,提出一种基于RGB(red,green,blue)与深度多模态视觉Transformer的特征提取方法,利用双流网络提取数据集的深度信息,自集成多个自注意力层的特征输出,以此作为最终的图像特征,并对损失函数进行改进,从而提高了模型的检索效果。通过在俯视图数据集TVPR(top-view person re-identification)和TVPR2上开展实验,结果表明:所提方法能有效提升检索效果,且超过了几种先进的俯视图行人重识别方法。展开更多
文摘This paper re-examines the status quo of China's territorial sovereignty security from a strategic perspective.Territory is the most basic physical where-about for a country to express national sovereignty.It is the territory that the national sovereignty derives its physical basis from.In the context of globalization,however, the traditional national territorial sovereignty space is continually squeezed and constrained,but the national territorial sovereignty is still the cornerstone of the international relations.And the inviolability of national territorial sovereignty is still the most important principle of modern international law.In this regard,the primary goal of Chinese national security today remains as a goal to safeguard the unification,integrity and security of China's territorial sovereignty.At present,the integrity,unification and security of Chinese national territorial sovereignty have yet to be achieved on the strategic level.As one of China's basic national conditions,it should be taken seriously and paid adequate strategic attentions.
文摘目的在行人再识别中,行人朝向变化会导致表观变化,进而导致关联错误。现有方法通过朝向表示学习和基于朝向的损失函数来改善这一问题。然而,大多数朝向表示学习方法主要以嵌入朝向标签为主,并没有显式的向模型传达行人姿态的空间结构,从而减弱了模型对朝向的感知能力。此外,基于朝向的损失函数通常对相同身份的行人进行朝向聚类,忽略了由表观相似且朝向相同的负样本造成的错误关联的问题。方法为了应对这些挑战,提出了面向行人再识别的朝向感知特征学习。首先,提出了基于人体姿态的朝向特征学习,它能够显式地捕捉人体姿态的空间结构。其次,提出的朝向自适应的三元组损失主动增大表观相似且相同朝向行人之间的间隔,进而将它们分离。结果本文方法在大规模的行人再识别公开数据集MSMT17(multi-scene multi-time person ReID dataset)、Market1501等上进行测试。其中,在MSMT17数据集上,相比于性能第2的UniHCP(unified model for human-centric perceptions)模型,Rank1和mAP值分别提高了1.7%和1.3%;同时,在MSMT17数据集上的消融实验结果证明本文提出的算法有效改善了行人再识别的关联效果。结论本文方法能够有效处理上述挑战导致的行人再识别系统中关联效果变差的问题。
文摘现有行人重识别技术主要关注水平视角下的图像。在例如无人超市这类特定场景下,摄像头以俯视角度拍摄,仅能获得有限的行人信息。针对此问题,将多模态视觉Transformer应用于俯视图行人重识别任务,利用俯视数据集中额外的深度模态来提高俯视图的检索精度。具体而言,提出一种基于RGB(red,green,blue)与深度多模态视觉Transformer的特征提取方法,利用双流网络提取数据集的深度信息,自集成多个自注意力层的特征输出,以此作为最终的图像特征,并对损失函数进行改进,从而提高了模型的检索效果。通过在俯视图数据集TVPR(top-view person re-identification)和TVPR2上开展实验,结果表明:所提方法能有效提升检索效果,且超过了几种先进的俯视图行人重识别方法。
