Height measurement precision improvement method based on conditional feature modulation of deep learning error compensation

Abstract

The present application relates to the cross technical field of satellite altimetry, ocean surveying and mapping, and the like, and in particular to a method for improving altimetry precision based on conditional feature modulation deep learning error compensation, comprising: obtaining satellite-borne GNSS-R observation data and auxiliary parameter data, and pre-processing and quality control processing the observation data; obtaining an initial sea surface height estimate value, and calculating the difference between the initial sea surface height estimate value and a reference sea surface height model as an error label; constructing a conditional feature modulation deep learning (CFMDL) error compensation model, and training the CFMDL error compensation model using the processed data and the error label; and using the trained CFMDL error compensation model to perform error compensation on the sea surface height inversion result of a sea area to be measured, to obtain a high-precision sea surface height value. Compared with the prior art, the method is superior to a comparative model in all evaluation indicators, effectively corrects satellite-borne GNSS-R sea surface height inversion errors, and improves sea surface height inversion precision.

Images