A fuzzy object-based data processing for high resolution remote sensing image

The present paper deals with the land cover classification of high resolution IKONOS images using multi-feature fusion, which is the traditional optical features analysis combined with texture feature analysis. The study area covers the residential region of the urbanized environment of Beijing, China. Different optical features including grey data in different layers and texture features contains Entropy and ASM were extracted based on GLCM (Grey Level Co-occurrence Matrix) texture feature and used as the distinct feature value in classification procedures. The image analysis process was carried out with the software eCognition. The object classes of buildings, shadow, green land, road and fields will be identified in the image. Results indicate that the proposed approach brings significant improvement of the classification rate based on the different texture feature images of various bands, allowing a better discrimination and mapping of mixed land cover types. Experiments show that the method has a stronger advantage than the traditional pixel-level methods for high resolution remote sensing image change detection.