An iterative approach to object-oriented classification of remotely sensed image

While spectral analysis of images has yielded satisfactory results, they may not be enough to extract features from high spatial resolution satellite data. A novel iterative classification approach is proposed, which based on object-oriented technologies. This approach makes use of both pixel spectral features and object features through an iterative model to improve the accuracy of classification. Some adaptive techniques are used for each iteration, which are leading class selection, advanced object-based feature calculating, feature selection based on mutual information. The experiment of land cover classification on SPOT5 image and aerial photography shows this approach produces higher classification accuracy than normal object-oriented classification.