Combination of UAV Images and DSM for Land Cover Classification Using Convolutional Neural Network

Machine learning applies predominantly to the classification of the satellite images, aerial photo, unmanned aerial vehicle (UAV) data, point clouds with considerable achievements. However, the dynamic and complex structures of land surface prevent accurate land cover segregation through built-in models, and there is a crucial need to investigate novel ones. This study integrates Catboost into a Convolutional neural network for land cover classification from UAV images, with a case study in Hanoi. The combination of these images and Digital surface model to form the input datasets. The results show that the overall accuracy reaches 91,5%, which is relatively higher than other comparing methods. The proposal model can be used as an alternative method for land cover classification.