Complex Scene Classification of High Resolution Remote Sensing Images Based on DCNN Model

摘要

Aiming at the problem that the traditional scene classification methods are not accurate to the semantic description of high resolution remote sensing images, a method based on deep convolutional neural network (DCNN) is proposed. It achieves an accuracy of 93% on the dataset (whu-6) made by myself, which significantly improves the classification accuracy compared with the traditional scene classification methods.