Organization and Representation of Objects in Multi-source Remote Sensing Image Classification

Compared with the traditional remote sensing image processing system based on pixel, there are some unique characters in the object-based classification: various operations on objects; variety of objects features; complicated relationships among objects; highdimensional feature set. All of these increase the difficulty of the organization and representation of objects. Therefore, the effective organization and representation of objects plays a key role in this system. This paper proposes a synthetic dynamic representation method of objects. The characteristic of the method lies in: the objects are represented in both the vector and raster format; the highdimensional object features are dynamically recorded; the complicated relations among objects are recorded based on hierarchal object network. This method considers the characters of the object-oriented multi-source remote sensing imagery classification system. It effectively solves the problems of object representation and meets the demands of the multi-source remote sensing imagery classification.