METHODS FOR WATER DEPTH INVERSION USING STOCHASTIC OPTIMAL RATIO LOGARITHMIC MODEL IN MULTI-SPECTRAL REMOTE SENSING

The present disclosure relates to a method for multispectral remote sensing stochastic optimal ratio-logarithmic bathymetry inversion. The method addresses an issue that a water body signal is typically below 5%. The method comprehensively considers influences of a solar spectrum, an upward total scattering, a solar zenith angle, a total global transmittance, a direct irradiance, and a total spherical albedo. The method constructs an atmospheric coupled radiometric calibration correction model to correct atmospheric and water surface reflectance. The method corrects a tide height through a divided difference interpolation tide height correction model. The method accurately extracts a water body range by combining an FNDWI normalized difference water index. The method effectively extracts sub-datasets through a stochastic optimal model. The method optimizes model parameters of a ratio-logarithmic model using an extreme gradient boosting manner. The method obtains an optimized optimal ratio-logarithmic model for bathymetry inversion to achieve bathymetry inversion. A comparison is made between inversion results of the optimized optimal ratio-logarithmic model for bathymetry inversion and inversion results of a Stumpf model. The Raccuracy of the bathymetry inversion by the optimal ratio-logarithmic model for bathymetry inversion is improved by 0.592, and an RMSE is improved by 0.841 m.