本发明公开了一种基于协方差矩阵与DCNLSTM模型的大气温度廓线生成方法,包括以下步骤:(1)获取ATMS遥感数据与对应日期的ERA5再分析资料;(2)对ATMS与ERA5数据进行时空匹配处理,进行归一化处理;(3)对ERA5温度廓线数据的平均值进行统计,获得平均值数据集,并利用该数据集计算各气压层间的协方差,获得协方差矩阵即背景协...
本发明属于表情追踪技术,具体的说是一种基于DCNNsLSTM模型的表情追踪方法.本发明考虑到卷积神经网络在图像特征提取上的有效性及长短期记忆网络在处理存在时间序列信息上的优势,提出一种结合DCNN和LSTM的深度神经网络模型用于处理面部68个特征点及眼神,头部特征信息的提取.数据集囊括了老年,青年,少年及中外等不同的人脸...
Aiming at the problem that it is d璁ゝicult to predict the trajectory of ultra-long-range attack targets such as ballistic missiles, a target trajectory prediction method based on long short-term memory (LSTM) network and 1-dimensional convolutional neural network (1DCNN) is ...
Aiming at the problem that it is d让ficult to predict the trajectory of ultra-long-range attack targets such as ballistic missiles, a target trajectory prediction method based on long short-term memory (LSTM) network and 1-dimensional convolutional neural network (1DCNN) is proposed. Firstly, ...