In divertor tokamak plasma, the energetic ion losses of edge plasma are considered to be responsible for the negative radial electric field. In the present paper, a guiding center approximation orbit equation is found...In divertor tokamak plasma, the energetic ion losses of edge plasma are considered to be responsible for the negative radial electric field. In the present paper, a guiding center approximation orbit equation is found by assuming the conservation of three integrals of motion, i.e. the total ion energy E, the magnetic moment # and toroidal angular momentum Pc, and it is used to calculate expediently the ion orbit loss region. The direct ion orbit losses in the initial velocity space near the plasma edge of EAST with SN （single null） divertor configuration are analyzed systematically. The ion loss regions are obtained by solving the guiding center approximation orbit equation of critical ions with the effect of the radial electric field taken into account. Under the influence of plasma current Ip, the type of ions, the toroidal field Bt and the changes of the loss regions are analyzed and calculated accordingly.展开更多
针对可见光-红外跨模态行人重识别中模态差异导致的识别精确率低的问题,提出了一种基于双流结构的跨模态行人重识别关系网络(IVRNBDS)。首先,利用双流结构分别提取可见光模态和红外模态行人图像的特征;然后,将行人图像的特征图水平切分...针对可见光-红外跨模态行人重识别中模态差异导致的识别精确率低的问题,提出了一种基于双流结构的跨模态行人重识别关系网络(IVRNBDS)。首先,利用双流结构分别提取可见光模态和红外模态行人图像的特征;然后,将行人图像的特征图水平切分为6个片段,以提取行人的每个片段的局部特征和其他片段的特征之间的关系,以及行人的核心特征和平均特征之间的关系;最后,在设计损失函数时,引入异质中心三元组损失(HC Loss)函数放松普通三元组损失函数的严格约束,从而使不同模态的图像特征可以更好地映射到同一特征空间中。在公开数据集SYSU-MM01(Sun Yat-Sen University Multi Modal re-identification)和Reg DB(Dongguk Body-based person Recognition)上的实验结果表明,虽然IVRNBDS的计算量略高于当前主流的跨模态行人重识别算法,但所提网络在相似度排名第1(Rank-1)指标和平均精度均值(m AP)指标上都有所提高,提高了跨模态行人重识别算法的识别精确率。展开更多
基金supported by National Natural Science Foundation of China(No.10075048)
文摘In divertor tokamak plasma, the energetic ion losses of edge plasma are considered to be responsible for the negative radial electric field. In the present paper, a guiding center approximation orbit equation is found by assuming the conservation of three integrals of motion, i.e. the total ion energy E, the magnetic moment # and toroidal angular momentum Pc, and it is used to calculate expediently the ion orbit loss region. The direct ion orbit losses in the initial velocity space near the plasma edge of EAST with SN （single null） divertor configuration are analyzed systematically. The ion loss regions are obtained by solving the guiding center approximation orbit equation of critical ions with the effect of the radial electric field taken into account. Under the influence of plasma current Ip, the type of ions, the toroidal field Bt and the changes of the loss regions are analyzed and calculated accordingly.
文摘针对可见光-红外跨模态行人重识别中模态差异导致的识别精确率低的问题,提出了一种基于双流结构的跨模态行人重识别关系网络(IVRNBDS)。首先,利用双流结构分别提取可见光模态和红外模态行人图像的特征;然后,将行人图像的特征图水平切分为6个片段,以提取行人的每个片段的局部特征和其他片段的特征之间的关系,以及行人的核心特征和平均特征之间的关系;最后,在设计损失函数时,引入异质中心三元组损失(HC Loss)函数放松普通三元组损失函数的严格约束,从而使不同模态的图像特征可以更好地映射到同一特征空间中。在公开数据集SYSU-MM01(Sun Yat-Sen University Multi Modal re-identification)和Reg DB(Dongguk Body-based person Recognition)上的实验结果表明,虽然IVRNBDS的计算量略高于当前主流的跨模态行人重识别算法,但所提网络在相似度排名第1(Rank-1)指标和平均精度均值(m AP)指标上都有所提高,提高了跨模态行人重识别算法的识别精确率。
