| Road information is an important data source of basic geographic database. Research on intelligent, automated road extraction is significant for studies on identification and location of objects from remote sensing image, mass GIS data acquisition and the update of electronic maps and spatial databases. In recent years, mass high resolution remote sensed images have been introduced into road extraction, but although these images could provide more detail information, owing to the limit of high complexity of target recognition, the current artificial, man-machine interactive feature interpretation is prevailed, which is heavy burden for users and what's more, low efficient. Thus, study of road extraction algorithm from high resolution remote sensed images has important theoretical and practical significance.Mathematical morphology (MM) developed since 1960s can simplify image structure, keep image features, remove redundant structure, and its operations are simple, flexible and fast. In this thesis, base on MM strategy, combination with image threshold segment method, a road extraction strategy was proposed. First, according to the high resolution remote sensing image road characteristics, roads are classified into four types including linear, curvilinear, cross and fault.; Second, based on MM theory and technology, reasonable road segmentation strategy was developed and then using the road cross block technology, remote sensed image can be segmented into several sub images which contains only one type road. After that,road information is extracted and segmented by corresponding strategy; At last, road information from sub images can be merged to a whole road network, and road skeleton can also be got by MM thin and trim processing.Results show that road information extraction method from high resolution remote sensing image developed in this thesis based on MM strategy can perform well although it is not perfect for all situations. |
PDF Full Text Request