The Research on Suburb Residential Areas Extraction of High Spatial Resolution Remotely Sensed Imagery Based on Texture Features

Traditional extraction algorithms of residential areas extraction in suburb environment do not give the desired result due to large within-class spectral variations and between-class spectral confusions that characterize the high spatial resolution remotely sensed data. The Subgraph Method was introduced for objects extraction from high spatial resolution remotely sensed imagery. Reasonable expression of texture features was achieved by integrating the multi-scale multi-directional capabilities of Contourlet transform and the statistical capacity of GLCM,which can correctly describe the features of remote sensing images objects. Weights to show distinguishing ability of each feature was calculated by the Multiple kernel learning( MKL) methods. A case study taking QucikBird imagery as sample data,proves the effectiveness of the innovative method adopted in this research.