Extraction automatic or semi-automatic of primary object such as roads and buildings from remotely sensed imagery can meet the requirement of mapping of imagery, acquiring and updating automatically data of GIS , and also have important significance for computer vision and image understanding etc. . The spectral information of high-resolution remotely sensed imagery is insufficient ,and phenomena of different spectrum with same object or same spectrum with different object is obvious . But the spatial resolution is relative high and the geometrical structure of object is clear , so we can use spectral feature with geometrical feature to extract information of object . In this paper ,we utilize the imagery with greater 1 meter resolution which is so called very high resolution as experimental data and do the research of ribbon road and simple building roof , and we talk about the process and method of extraction from high-resolution remotely sensed imagery :the representation of road and building in high-resolution image is analyzed ;a method of semi-automatic extraction for ribbon road based on angular texture signature and profile matching is promoted ;a algorithm for road extraction based on Snakes model is compared and analyzed ;the method of object-oriented image analysis and classification is introduced and we utilize this method to extract simple building roof semi-automatically ;the algorithm of road extraction is implemented as prototype system ;finally , the practicability of above-mentioned method are evaluated .The research in experimental area indicate that above-mentioned semi-automatic method is operable , and they have applicable prospects to some extents. |