Abstract: The analysis of the remotely sensed images evolves in three stages:physical classification,semantic extraction,and information recognition.This paper proposes a new method for integrating the three stages into one common platform.Through this method,semantic extraction and information recognition can be synthetically performed with the classification procedure in the image processing.The mechanism analyzing the relationship in a multi-resolution hierarchy benefits the better understanding to the image objects,in both semantic and physical ways.Consequently,this method can be used for integration of remote sensing and GIS at semantic level.