Recognition of 3d Settlement Structure for Generalization

ii Abstract When a given 3D city model, consisting of mainly buildings and roads, is viewed at reduced scale, its objects not only become smaller but tend to conflict due to small area available. Generalization plays an important role to overcome these effects and helps in preserving the required legibility. Various operations are performed during generalization on these objects. One of the primary requirements for generalization is structure recognition. It not only involves structure recognition of individual object but various objects in neighborhood as well and includes many spatial relations among them. Although importance of structure recognition for 2D generalization has been reported in literature but little has been studied and reported for 3D structure recognition. This work aims at recognizing 3D settlement structures for automatic generalization, an innovative extension to 2D. The recognition procedure has been divided into three levels namely micro, meso and macro and is based upon individual buildings, buildings in neighborhood and buildings at cluster level having similar properties such as settlement blocks as well as psychophysically perceived groups. Any of these three levels of structure recognition demands that comprehensive information about the buildings should be known a-priori. Therefore first of all different buildings are recognized from the data available using a bottom-up approach. It starts with recognizing ground plans of buildings and which in turn, along with other information, are used to recognize different roof types and finally entire buildings are recognized in a similar way. After building recognition, their structure description has been studied in detail, which gives rise to various measurable parameters of individual as well as buildings in neighborhood. These parameters not only characterize individual building but also many spatial relations among them. Structure recognition at clustered level is studied next and it involves the recognition of group of buildings as a whole. The human visual system can detect many clusters of patterns and significant arrangements of image elements. Perceptual grouping refers to the human visual ability to extract significant image relations from lower-level primitive image features without any knowledge of the image content and group them to obtain meaningful higher-level structure. Various perceptual grouping principles have been applied to identify these clusters of groups. After a comprehensive study of structure recognition, their findings are then applied to 3D generalization. Among the various generalization algorithms such as aggregation, displacement, simplification, exaggeration, typification, aggregation is chosen here as it almost uses most of …