High-repetition-rate femtosecond lasers enable the precise production of nanofoam from a wide range of materials. Here, the laser-based fabrication of nanofoam from silicon, borosilicate glass, sodalime glass, gallium...High-repetition-rate femtosecond lasers enable the precise production of nanofoam from a wide range of materials. Here, the laser-based fabrication of nanofoam from silicon, borosilicate glass, sodalime glass, gallium lanthanum sulphide and lithium niobate is demonstrated, where the pore size of the nanofoam is shown to depend strongly on the material used, such that the pore width and nanofibre width appear to increase with density and thermal expansion coefficient of the material. In addition, the patterning of nanofoam on a glass slide, with fabricated pattern pixel resolution of ~35 μm, is demonstrated.展开更多
The re-emitted images of the frame camera indicated that the high-Z （Bi） capsule deviated about 29 μm from the center of the hohlraum in experiments at the Shenguang-II （SG-II） laser facility; however, investigat...The re-emitted images of the frame camera indicated that the high-Z （Bi） capsule deviated about 29 μm from the center of the hohlraum in experiments at the Shenguang-II （SG-II） laser facility; however, investigations on this issue have seldom been performed. The influence of three dimensional offsets of a capsule on its radiation asymmetry in inertial confinement fusion （ICF） will be analyzed in this paper. Simulations demonstrate that the axial offset of 100 μm of a capsule from the center of the hohlraum brings an additional 3.5% radiation drive asymmetry and 6.5% P1 asymmetry （Legendre odd model） on the capsule in the SG-II laser facility, and the offset must be within 25 μm if the P1 asymmetry is restricted to below 2%.展开更多
卷烟激光喷码识别是烟草稽查工作的重要手段.本文提出一种基于双态非对称网络的烟码识别方法,针对畸变烟码训练样本不足导致模型泛化能力弱的问题,设计非线性局部增强方法(nonlinear local augmentation,NLA),通过在烟码图像边缘设置可...卷烟激光喷码识别是烟草稽查工作的重要手段.本文提出一种基于双态非对称网络的烟码识别方法,针对畸变烟码训练样本不足导致模型泛化能力弱的问题,设计非线性局部增强方法(nonlinear local augmentation,NLA),通过在烟码图像边缘设置可控基准点进行空间变换,生成有效畸变训练样本以增强模型泛化能力;针对烟码与背景图案特征相似导致识别精度低的问题,提出双态非对称网络(dual-state asymmetric network,DSANet),将CRNN的卷积层划分为训练模式和部署模式,训练模式通过引入非对称卷积优化特征权重分布,增强模型关键特征提取能力;为保证实时性,部署模式设计BN融合和分支融合方法,通过计算融合权重并初始化卷积核,将卷积层等效转换回原始网络结构,降低用户端推理时间;最后,在循环层中引入自注意力机制,通过动态调整序列特征权重,进一步加强模型对烟码特征的提取能力.通过对比实验,该方法具有更高的识别精度和速度,其识别精度达到87.34%.展开更多
文摘High-repetition-rate femtosecond lasers enable the precise production of nanofoam from a wide range of materials. Here, the laser-based fabrication of nanofoam from silicon, borosilicate glass, sodalime glass, gallium lanthanum sulphide and lithium niobate is demonstrated, where the pore size of the nanofoam is shown to depend strongly on the material used, such that the pore width and nanofibre width appear to increase with density and thermal expansion coefficient of the material. In addition, the patterning of nanofoam on a glass slide, with fabricated pattern pixel resolution of ~35 μm, is demonstrated.
基金supported by Science and Technology on Plasma Physics Laboratory of China(Nos.9140C680104140C68287,9140C680104130C68241)in part by National Natural Science Foundation of China(Nos.11475154,51375185,U1430124,11435011,11305160)
文摘The re-emitted images of the frame camera indicated that the high-Z （Bi） capsule deviated about 29 μm from the center of the hohlraum in experiments at the Shenguang-II （SG-II） laser facility; however, investigations on this issue have seldom been performed. The influence of three dimensional offsets of a capsule on its radiation asymmetry in inertial confinement fusion （ICF） will be analyzed in this paper. Simulations demonstrate that the axial offset of 100 μm of a capsule from the center of the hohlraum brings an additional 3.5% radiation drive asymmetry and 6.5% P1 asymmetry （Legendre odd model） on the capsule in the SG-II laser facility, and the offset must be within 25 μm if the P1 asymmetry is restricted to below 2%.
文摘卷烟激光喷码识别是烟草稽查工作的重要手段.本文提出一种基于双态非对称网络的烟码识别方法,针对畸变烟码训练样本不足导致模型泛化能力弱的问题,设计非线性局部增强方法(nonlinear local augmentation,NLA),通过在烟码图像边缘设置可控基准点进行空间变换,生成有效畸变训练样本以增强模型泛化能力;针对烟码与背景图案特征相似导致识别精度低的问题,提出双态非对称网络(dual-state asymmetric network,DSANet),将CRNN的卷积层划分为训练模式和部署模式,训练模式通过引入非对称卷积优化特征权重分布,增强模型关键特征提取能力;为保证实时性,部署模式设计BN融合和分支融合方法,通过计算融合权重并初始化卷积核,将卷积层等效转换回原始网络结构,降低用户端推理时间;最后,在循环层中引入自注意力机制,通过动态调整序列特征权重,进一步加强模型对烟码特征的提取能力.通过对比实验,该方法具有更高的识别精度和速度,其识别精度达到87.34%.
