Functional Area Classification Based On Superpixel In High-Resolution Remote Sensing Image

With the development of urbanization in China,the planning,construction and management of urban functional areas play an important role in urban development.As urban functional area contains different types of object,and different urban functional areas may contains the same type of object,in the high resolution remote sensing images,the spectral heterogeneity in urban functional areas and the spectral overlap of urban functional areas.Therefore,the traditional classification method based on the consistency of the spectral measure of pixel can not realize the classification of urban functional areas.To classify urban functional areas in high resolution remote sensing image,a semi-automatic method based on superpixel is proposed.First of all,Simple Linear Iterative Clustering(SLIC)is used to segment a given image and a set of superpixels can be obtained.The superpixels,which can provide effective information for functional area classification,have similar features and keep the boundaries of functional area of the image.Then the color histogram of each superpixel is computed,the similarity within superpixels is represented by Bhattacharyya coefficient and the principle of maximal similarity within neighborhood superpixel is defined.Further,the samples are chosen in each functional area,and the sclassification of the functional areas is implemented by merging superpixels according to the samples and the defined principle.To test the proposed method,experiments are carried on with WorldView-2 images,the experimental results show that the proposed method can classify the functional areas of high resolution remote sensing image well,thus it is demonstrated that the proposed method is a feasible and effective method for urban functional area classification.