A Hybrid Classification Method for High Spatial Resolution Remote Sensing Image

As the spatial resolution being higher than before, pixel-wise classification method cannot satisfy the demanding of remote sensing image classification. object-based image analysis (OBIA) is introduced into remote sensing image classification. Here, we, first, applied the vector field model (VFM) and phase congruency model to obtain the multiple edge strength. Second, watershed transform is employed to get the image segmentation. Finally, support vector machine (SVM) that is proved to be a stable model to handle high-dimensional data analysis, is used to classify the land cover. Finally, voting principle is used to get the final object-wise land cover classification by combining the pixel-wise classification and image segmentation. The experimental results shows that our proposed method can be used into land cover classification efficiently.