|
| 基于深度全连接网络Himawari-8卫星气溶胶反演研究 |
| Research on Aerosol Optical Depth Retrieval of Himawari 8 Data Based on Deep Neural Networks |
| |
| DOI： |
| 中文关键词: 气溶胶光学厚度 深度全连接网络 葵花8 遥感 |
| 英文关键词: Aerosol optical depth（AOD） Deep neural networks Himawari 8 Remote sensing |
| 基金项目:国家自然科学基金资助项目（41604028）；安徽省科技重大专项基金资助项目（18030801111）； 安徽省自然科学基金资助项目（1708085QD83） |
|
| 摘要点击次数: 633 |
| 全文下载次数: 57 |
| 中文摘要: |
| 利用葵花8（Himawari 8，H8） 16个波段数据、卫星、太阳角度数据和深度学习技术，提出一种基于深度全连接网络（Deep Neural Networks，DNN）模型的AOD遥感反演方法（Himawari DNN）。该方法直接建立H8影像本身与AERONET站点AOD数据间的关系，可避免传统AOD遥感反演方法中复杂过程，得到精度较高的反演结果。通过有效数据对所构建的模型做精度测试，同时将反演结果和实测数据对比分析，结果表明，模型反演结果与研究区内所有站点的观测值均具有较高的一致性（R2均＞089）。可见，应用DNN对H8气象静止卫星开展AOD反演具有一定的可行性。 |
| 英文摘要: |
| This paper used 16 bands of data， satellite， sun angle data of Himawari 8(H8) and deep learning technology to establish an AOD remote sensing inversion method based on Deep Neural Networks (DNN) model. This method could build directly a relationship between H8 images and AOD data from AERONET sites， avoid the complicated process in traditional AOD remote sensing retrieval method， and obtain highly precise inversion results. The accuracy of the model was tested by valid data. Meanwhile， comparing the inversion results with measured data， it showed that the retrieval results by the model were consistent with the data from all the sites in the study area(all of R2＞ 0.89）. Therefore， it was feasible to use Deep Neural Networks model to retrieve the aerosol optical depth of H8 meteorological geostationary satellite. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
