A long-term variation of the inner zone high-energy proton environment at low orbits was investigated by DSTM using the adiabatic approximation of charge particle motion with NASA standard radiation models as referenc...A long-term variation of the inner zone high-energy proton environment at low orbits was investigated by DSTM using the adiabatic approximation of charge particle motion with NASA standard radiation models as reference states. The DST results show that over the past three decades the fluxes of high-energy protons at ～1000 km in the South Atlantic Anomaly (SAA) noticeably increased, the center region of proton SAA apparently moved westward and expanded. Calculations of the L-shell averaged lifetime of high-energy protons indicate that the DST pro-vides a reasonable means for estimation of the secular variation of inner zone proton environ-ment.展开更多
基于深度序列的人体行为识别,一般通过提取特征图来提高识别精度,但这类特征图通常存在时序信息缺失的问题.针对上述问题,本文提出了一种新的深度图序列表示方式,即深度时空图(Depth space time maps,DSTM).DSTM降低了特征图的冗余度,...基于深度序列的人体行为识别,一般通过提取特征图来提高识别精度,但这类特征图通常存在时序信息缺失的问题.针对上述问题,本文提出了一种新的深度图序列表示方式,即深度时空图(Depth space time maps,DSTM).DSTM降低了特征图的冗余度,弥补了时序信息缺失的问题.本文通过融合空间信息占优的深度运动图(Depth motion maps,DMM)与时序信息占优的DSTM,进行高精度的人体行为研究,并提出了多聚点子空间学习(Multi-center subspace learning,MCSL)的多模态数据融合算法.该算法为各类数据构建多个投影聚点,以此增大样本的类间距离,降低了投影目标区域维度.本文在MSR-Action3D数据集和UTD-MHAD数据集上进行人体行为识别.最后实验结果表明,本文方法相较于现有人体行为识别方法有着较高的识别率.展开更多
基金This work was supported by SRFDP,the NSFC project(Grant No.49984002)the Chinese Key Research Project(Grant No.G200000784).
文摘A long-term variation of the inner zone high-energy proton environment at low orbits was investigated by DSTM using the adiabatic approximation of charge particle motion with NASA standard radiation models as reference states. The DST results show that over the past three decades the fluxes of high-energy protons at ～1000 km in the South Atlantic Anomaly (SAA) noticeably increased, the center region of proton SAA apparently moved westward and expanded. Calculations of the L-shell averaged lifetime of high-energy protons indicate that the DST pro-vides a reasonable means for estimation of the secular variation of inner zone proton environ-ment.
文摘基于深度序列的人体行为识别,一般通过提取特征图来提高识别精度,但这类特征图通常存在时序信息缺失的问题.针对上述问题,本文提出了一种新的深度图序列表示方式,即深度时空图(Depth space time maps,DSTM).DSTM降低了特征图的冗余度,弥补了时序信息缺失的问题.本文通过融合空间信息占优的深度运动图(Depth motion maps,DMM)与时序信息占优的DSTM,进行高精度的人体行为研究,并提出了多聚点子空间学习(Multi-center subspace learning,MCSL)的多模态数据融合算法.该算法为各类数据构建多个投影聚点,以此增大样本的类间距离,降低了投影目标区域维度.本文在MSR-Action3D数据集和UTD-MHAD数据集上进行人体行为识别.最后实验结果表明,本文方法相较于现有人体行为识别方法有着较高的识别率.
