The summer and winter circulations in the South China Sea (SCS) including the surface elevation and water temperature are simulated using the model described by Cai and Li (1996) with the monthly mean wind stress and ...The summer and winter circulations in the South China Sea (SCS) including the surface elevation and water temperature are simulated using the model described by Cai and Li (1996) with the monthly mean wind stress and air temperature field at the 1000 mb level from the European Centre for Medium-Range Weather Forecasts as inputs. The boundary conditions at Bashi Channel and Taiwan Strait are taken from the simulation results of the Kuroshio using the same numerical model with a grid size of 0.5°×0.5° and the results of Cai and Li (1996) as boundary conditions. The computational domain for the present paper is between 100°E and 123°E and between 4.5°N and 27°N. The horizontal resolution is 0.25°×0.25° and the vertical variations of the velocity components are resolved by 6 layers The computed steady flow, temperature and elevation fields are consistent with the corresponding fields observed. In particular, the temperature and elevation fields of the South China Sea Warm Current (SCSWC) have been successfully simulated. The paths of the branch of the Kuroshio entering the South China Sea (SCSBK) through Bashi Channel in winter and summer are discussed It is found that the SCSBK flows southward to the southern SCS from the coast of the Guangdong Province. A portion of the SCSBK returns to the Bashi Channel and subdivides again in deep waters in winter with a branch flows to the south along the coast of the Philippines instead of flowing back to the Pacific In addition, our results confirm the existence of a eastward current to the northeast of Dongsha in summer with the Kuroshio as its source as suggested by Huang et al. Since the value of the eddy viscosity adopted for the simulation of the Kuroshio is on the high side, resulting in a weaker west boundary current in the western Pacific as the boundary conditions for the present simulations, some deviations from the actual situations are expected although the results are in general consistent with observations.展开更多
针对卷积神经网络在高光谱图像特征提取和分类的过程中,存在空谱特征提取不充分以及网络层数太多引起的参数量大、计算复杂的问题,提出快速三维卷积神经网络(3D-CNN)结合深度可分离卷积(DSC)的轻量型卷积模型。该方法首先利用增量主成...针对卷积神经网络在高光谱图像特征提取和分类的过程中,存在空谱特征提取不充分以及网络层数太多引起的参数量大、计算复杂的问题,提出快速三维卷积神经网络(3D-CNN)结合深度可分离卷积(DSC)的轻量型卷积模型。该方法首先利用增量主成分分析(IPCA)对输入的数据进行降维预处理;其次将输入模型的像素分割成小的重叠的三维小卷积块,在分割的小块上基于中心像素形成地面标签,利用三维核函数进行卷积处理,形成连续的三维特征图,保留空谱特征。用3D-CNN同时提取空谱特征,然后在三维卷积中加入深度可分离卷积对空间特征再次提取,丰富空谱特征的同时减少参数量,从而减少计算时间,分类精度也有所提高。所提模型在Indian Pines、Salinas Scene和University of Pavia公开数据集上验证,并且同其他经典的分类方法进行比较。实验结果表明,该方法不仅能大幅度节省可学习的参数,降低模型复杂度,而且表现出较好的分类性能,其中总体精度(OA)、平均分类精度(AA)和Kappa系数均可达99%以上。展开更多
文摘The summer and winter circulations in the South China Sea (SCS) including the surface elevation and water temperature are simulated using the model described by Cai and Li (1996) with the monthly mean wind stress and air temperature field at the 1000 mb level from the European Centre for Medium-Range Weather Forecasts as inputs. The boundary conditions at Bashi Channel and Taiwan Strait are taken from the simulation results of the Kuroshio using the same numerical model with a grid size of 0.5°×0.5° and the results of Cai and Li (1996) as boundary conditions. The computational domain for the present paper is between 100°E and 123°E and between 4.5°N and 27°N. The horizontal resolution is 0.25°×0.25° and the vertical variations of the velocity components are resolved by 6 layers The computed steady flow, temperature and elevation fields are consistent with the corresponding fields observed. In particular, the temperature and elevation fields of the South China Sea Warm Current (SCSWC) have been successfully simulated. The paths of the branch of the Kuroshio entering the South China Sea (SCSBK) through Bashi Channel in winter and summer are discussed It is found that the SCSBK flows southward to the southern SCS from the coast of the Guangdong Province. A portion of the SCSBK returns to the Bashi Channel and subdivides again in deep waters in winter with a branch flows to the south along the coast of the Philippines instead of flowing back to the Pacific In addition, our results confirm the existence of a eastward current to the northeast of Dongsha in summer with the Kuroshio as its source as suggested by Huang et al. Since the value of the eddy viscosity adopted for the simulation of the Kuroshio is on the high side, resulting in a weaker west boundary current in the western Pacific as the boundary conditions for the present simulations, some deviations from the actual situations are expected although the results are in general consistent with observations.
文摘针对卷积神经网络在高光谱图像特征提取和分类的过程中,存在空谱特征提取不充分以及网络层数太多引起的参数量大、计算复杂的问题,提出快速三维卷积神经网络(3D-CNN)结合深度可分离卷积(DSC)的轻量型卷积模型。该方法首先利用增量主成分分析(IPCA)对输入的数据进行降维预处理;其次将输入模型的像素分割成小的重叠的三维小卷积块,在分割的小块上基于中心像素形成地面标签,利用三维核函数进行卷积处理,形成连续的三维特征图,保留空谱特征。用3D-CNN同时提取空谱特征,然后在三维卷积中加入深度可分离卷积对空间特征再次提取,丰富空谱特征的同时减少参数量,从而减少计算时间,分类精度也有所提高。所提模型在Indian Pines、Salinas Scene和University of Pavia公开数据集上验证,并且同其他经典的分类方法进行比较。实验结果表明,该方法不仅能大幅度节省可学习的参数,降低模型复杂度,而且表现出较好的分类性能,其中总体精度(OA)、平均分类精度(AA)和Kappa系数均可达99%以上。
