Method and device for atmospheric correction assisted by aod inversion based on deep learning, and medium

Abstract

The application discloses a kind of based on deep learning inversion AOD auxiliary atmospheric correction method, device and medium, it is related to atmospheric remote sensing and deep learning cross technical field, including: the multiple type training sample set of fusion measured and physical simulation data is constructed, after quality screening and stratification processing, obtain the pre-training sample of adaptation satellite remote sensing data, fine-tuning sample and extreme scene sample;Create deep learning network model with data loss and anti-operator constraint loss weighted fusion as total loss function, adopt cross-satellite transfer learning strategy, in turn complete pre-training, satellite special fine-tuning and extreme scene enhancement training, obtain the training completed aerosol optical depth inversion model;Radiometric calibration is executed to satellite remote sensing original data, cloud detection, geometric correction preprocessing, input model inference calculation, to output result is post-processed and quality classification, obtain the aerosol optical depth inversion result used for auxiliary atmospheric correction.

Images