The study of aerosol optical depth retrieval is of great significance for monitoring and assessing the quality of the atmospheric environment.Aiming at the low accuracy of the existing aerosol optical depth retrieval,the aerosol optical depth data measured at the AERONET site were temporally and spatially matched with the preprocessed Himawari-8 image data to establish a sample data set so as to build an aerosol optical depth retrieval model.The accuracy of the deep belief network retrieval model was verified by comparing and analyzing the deep neural network model with the deep belief network model,and the feasibility of the deep belief network retrieval of aerosol optical depth was verified by using the measured aerosol optical depth data from the AERONET site and the MCD19 A2 aerosol optical depth data as the validation data.The results show that the aerosol optical depth based on deep belief network retrieval has high accuracy,good fitting,and a small error.
关键词
深度置信网络/受限玻尔兹曼机/气溶胶光学厚度/Himawari-8影像/反演
Key words
deep belief network/restricted Boltzmann machine/aerosol optical depth/Himawari-8 image/retrieval
维普
万方数据