Three algorithms（the Charge Comparison Method,n-γ Model Analysis and the Centroid Algorithm）have been revised to improve their accuracy and broaden the scope of applications to real-time digital n-γ discrimination...Three algorithms（the Charge Comparison Method,n-γ Model Analysis and the Centroid Algorithm）have been revised to improve their accuracy and broaden the scope of applications to real-time digital n-γ discrimination.To evaluate the feasibility of the revised algorithms,a comparison between the improved and original versions of each is presented.To select an optimal real-time discrimination algorithm from these six algorithms（improved and original）,the figure-of-merit（FOM）,Peak-Threshold Ratio（PTR）,Error Probability（EP） and Simulation Time（ST） for each were calculated to obtain a quantitatively comprehensive assessment of their performance.The results demonstrate that the improved algorithms have a higher accuracy,with an average improvement of 10%in FOM,95%in PTR and 25%in EP,but all the STs are increased.Finally,the Adjustable Centroid Algorithm（ACA） is selected as the optimal algorithm for real-time digital n-γ discrimination.展开更多
二维方法用于图像矩阵特征提取,虽然速度快,但影响了分类速度。针对二维线性鉴别分析(Two-Dimensional Linear Discriminant Analysis,2DLDA)的特点,研究了一种基于图像分块的改进Fisher人脸识别算法,该算法首先对人脸图像进行压缩降维...二维方法用于图像矩阵特征提取,虽然速度快,但影响了分类速度。针对二维线性鉴别分析(Two-Dimensional Linear Discriminant Analysis,2DLDA)的特点,研究了一种基于图像分块的改进Fisher人脸识别算法,该算法首先对人脸图像进行压缩降维处理,得到相应的特征矩阵,然后利用改进Fisher算法对特征矩阵进行类间离散度矩阵和类内离散度矩阵的计算,该算法充分考虑了类别信息,避免了传统Fisher算法造成的小样本问题,有效提高了分类速度。基于ORL(Olivetti Research Laboratory)与Yale人脸数据库的实验结果证明了该算法的有效性。展开更多
基于冻融土的微波辐射特征,在HUT(Helsinki University of Technology)积雪辐射模型的基础上,引入新的冻土介电常数模型计算冻/融土的介电常数,利用高级积分方程模型(Advanced Integrated Emission Model,AIEM)计算地表发射率,改进了寒...基于冻融土的微波辐射特征,在HUT(Helsinki University of Technology)积雪辐射模型的基础上,引入新的冻土介电常数模型计算冻/融土的介电常数,利用高级积分方程模型(Advanced Integrated Emission Model,AIEM)计算地表发射率,改进了寒区复杂地表微波辐射模型和冻融状态判别式算法。采用AMSR^2(The Advanced Microwave Scanning Radiometer 2)被动微波辐射计亮温数据和地基微波辐射计观测数据进行了地表冻融状态判别式算法精度的验证与比较。结果显示:改进后的判别式算法对冻土的判识精度有明显提升,总体判识精度在82%以上,是一种较可靠的判别模式。展开更多
文摘Three algorithms（the Charge Comparison Method,n-γ Model Analysis and the Centroid Algorithm）have been revised to improve their accuracy and broaden the scope of applications to real-time digital n-γ discrimination.To evaluate the feasibility of the revised algorithms,a comparison between the improved and original versions of each is presented.To select an optimal real-time discrimination algorithm from these six algorithms（improved and original）,the figure-of-merit（FOM）,Peak-Threshold Ratio（PTR）,Error Probability（EP） and Simulation Time（ST） for each were calculated to obtain a quantitatively comprehensive assessment of their performance.The results demonstrate that the improved algorithms have a higher accuracy,with an average improvement of 10%in FOM,95%in PTR and 25%in EP,but all the STs are increased.Finally,the Adjustable Centroid Algorithm（ACA） is selected as the optimal algorithm for real-time digital n-γ discrimination.
文摘二维方法用于图像矩阵特征提取,虽然速度快,但影响了分类速度。针对二维线性鉴别分析(Two-Dimensional Linear Discriminant Analysis,2DLDA)的特点,研究了一种基于图像分块的改进Fisher人脸识别算法,该算法首先对人脸图像进行压缩降维处理,得到相应的特征矩阵,然后利用改进Fisher算法对特征矩阵进行类间离散度矩阵和类内离散度矩阵的计算,该算法充分考虑了类别信息,避免了传统Fisher算法造成的小样本问题,有效提高了分类速度。基于ORL(Olivetti Research Laboratory)与Yale人脸数据库的实验结果证明了该算法的有效性。
