Land-use Classification with Remote Sensing Image Based on Stacked Autoencoder

Focused on the issue that conventional land-use classification methods can't reach better performance, a new remote sensing image classification method based on Stacked Autoencoder inspired by deep learning was proposed. Firstly, the deep network model was built through the stacked layers of Autoencoder, then the unsupervised Greedy layer-wise training algorithm was used to train each layer in turn for more robust expressing, characteristics were learnt supervised by Back Propagation neural network and the whole net was optimized by using error back propagation. Finally, GF-1 remote sensing data were used for evaluation and the total accuracy and kappa accuracy which were higher than that of Support Vector Machine and Back Propagation neural network reached 95.5% and 95.3% respectively. The experiment results show that the proposed method can effectively improve the accuracy of land cover classification.