Remote sensing of water depth in subsidence area based on artificial neural networks

To measure the remote sensing of water depth in subsidence area,the model based on BP neural network is proposed.After geometric calibration,atmospheric correction and subsidence area extraction,the pixels reflectivity is exported.In order to find the relation between actual water depth and pixels reflectivity,the pixels reflectivity are matched to control points.The depth of 2 m is the threshold of the model which corresponds to actually measured water depth less than 2 m and water depth from 2 m to 6 m.The model is applied to measure water depth in subsidence area of Huainan.It is demonstrated that the mean absolute error is 0.166 3 m and the mean relative error is 13.29%,when the actually measured water depth is less than 2 m.The mean absolute error is 0.578 6 m and the mean relative error is 15.20%,when the actually measured water depth is in range of 2 m to 6 m.