Land use classification of remote sensing images based on convolution neural network

Abstract

In order to further improve the accuracy of land use classification, this paper uses UC Merced land use data set to fine-tune the parameters of CaffeNet, VGG-S, and VGG-F CNN models. Then, the fine-tuned network is used as the feature extractor, and the extracted full connection layer output features are cascaded as the final expression of the image. Finally, the cascaded features are input into the mcODM classifier to obtain the classification results. The results showed that the overall classification accuracy of the multi-structure CNN feature cascade method in UC Merced landuse dataset reached 97.55%, indicating an improvement between 2 and 5% compared with the single CNN model, and the classification accuracy after fine-tuning was improved in the range of 3–5%. In conclusion, this method can effectively improve the expression of features in scene level classification and improve classification performance.