Classification using semantic feature and machine learning: Land-use case application
Hela Elmannai, Abeer Dhafer AlGarni
Land cover classification has interested recent works especially for deforestation, urban are monitoring and agricultural land use. Traditional classification approaches have limited accuracy especially for non-heterogeneous land cover. Thus, using machine may improve the classification accuracy. The presented paper deals with the land-use scene recognition on very high-resolution remote sensing imagery. We proposed a new framework based on semantic features, handcrafted features and machine learning classifiers decisions. The method starts by semantic feature extraction using a convolutional neural network. Handcraft features are also extracted based on color and multi-resolution characteristics. Then, the classification stage is processed by three learning machine algorithms. The final classification result performed by majority vote algorithm. The idea behind is to take advantages from semantic features and handcrafted features. The second scope is to use the decision fusion to enhance the classification result. Experimentation results show that the proposed method provides good accuracy and trustable tool for land use image identification.