Conventional electron and optical microscopy techniques require the sample to be sectioned, polished or etched to expose the internal surfaces for imaging. However, such sample preparation techniques have traditionall...Conventional electron and optical microscopy techniques require the sample to be sectioned, polished or etched to expose the internal surfaces for imaging. However, such sample preparation techniques have traditionally prevented the observation of the same sample over time, under realistic three-dimensional geometries and in an environment representative of real-world operating conditions. X-ray microscopy (XRM) is a rapidly emerging technique that enables non-destructive evaluation of buried structures within hard to soft materials in 3D, requiring little to no sample preparation. Furthermore in situ and 4D quantification of microstructural evolution under controlled environment as a function of time, temperature, chemistry or stress can be done repeatable on the same sample, using practical specimen sizes ranging from tens of microns to several cm diameter, with achievable imaging resolution from submicron to 50 nm. Many of these studies were reported using XRM in synchrotron beamlines. These include crack propagation on composite and construction materials; corrosion studies; microstructural changes during the setting of cement; flow studies within porous media to mention but a few.展开更多
针对传统D-S证据理论中基于识别率和误识率构造的基本概率赋值函数(Basic Probability Assignment,BPA)没有考虑训练样本分布的缺点,提出了一种将整体错误率分配给除了正确判别命题以外各个焦元的BPA构造新方法.针对传统D-S证据理论中...针对传统D-S证据理论中基于识别率和误识率构造的基本概率赋值函数(Basic Probability Assignment,BPA)没有考虑训练样本分布的缺点,提出了一种将整体错误率分配给除了正确判别命题以外各个焦元的BPA构造新方法.针对传统D-S证据理论中所采用的基于正交和运算的合成规则不能融合矛盾证据的缺陷,提出一种能融合矛盾证据的大概率赋值法.在此改进D-S证据理论的基础上,给出了两分类器决策层融合流程和多分类器决策层融合系统.在ORL和Yale数据库上的实验结果表明,对几种典型分类器的决策层融合提高了系统人脸识别的正确率,且改进D-S证据理论比传统D-S和投票融合方法的正确率更高.展开更多
文摘Conventional electron and optical microscopy techniques require the sample to be sectioned, polished or etched to expose the internal surfaces for imaging. However, such sample preparation techniques have traditionally prevented the observation of the same sample over time, under realistic three-dimensional geometries and in an environment representative of real-world operating conditions. X-ray microscopy (XRM) is a rapidly emerging technique that enables non-destructive evaluation of buried structures within hard to soft materials in 3D, requiring little to no sample preparation. Furthermore in situ and 4D quantification of microstructural evolution under controlled environment as a function of time, temperature, chemistry or stress can be done repeatable on the same sample, using practical specimen sizes ranging from tens of microns to several cm diameter, with achievable imaging resolution from submicron to 50 nm. Many of these studies were reported using XRM in synchrotron beamlines. These include crack propagation on composite and construction materials; corrosion studies; microstructural changes during the setting of cement; flow studies within porous media to mention but a few.
文摘针对传统D-S证据理论中基于识别率和误识率构造的基本概率赋值函数(Basic Probability Assignment,BPA)没有考虑训练样本分布的缺点,提出了一种将整体错误率分配给除了正确判别命题以外各个焦元的BPA构造新方法.针对传统D-S证据理论中所采用的基于正交和运算的合成规则不能融合矛盾证据的缺陷,提出一种能融合矛盾证据的大概率赋值法.在此改进D-S证据理论的基础上,给出了两分类器决策层融合流程和多分类器决策层融合系统.在ORL和Yale数据库上的实验结果表明,对几种典型分类器的决策层融合提高了系统人脸识别的正确率,且改进D-S证据理论比传统D-S和投票融合方法的正确率更高.
