With the growing demand for shallow sea surveying and mapping in China, this paper uses four-bands high resolution WorldView-2 images to conduct satellite-derived bathymetry experiments in the Ganquan Island and the Nanwan area in Taiwan. The random forest satellite-derived bathymetry model is constructed by using the random forest algorithm, and the accuracy is compared with three classic satellite-derived bathymetry model. The results show that the bathymetry accuracy of the random forest model is optimal in Ganquan Island and Nanwan area, with RMSE are 0.85m and 1.59m, and MRE are 8% and 12%, respectively. |