In this dissertation, a new approach to automatic road extraction from high resolution aerial imagery taken over urban areas is developed. In order to deal with the high complexity of this type of scenes, the approach integrates detailed knowledge about roads and their context using explicitly formulated scale-dependent models. The knowledge about how and when certain parts of the road and context model are optimally exploited is expressed by an extraction strategy. It is subdivided into three levels: Context-based data analysis (Level 1) comprises the segmentation of the scene into the urban, rural, and forest area as wells as the analysis of context relations (e.g. the determination of shadow areas and the detection of vehicles). Processing continues in urban areas. To determine salient roads (Level 2) the extraction includes the detection of homogeneous ribbons as preliminary road segments in coarse scale, collinear grouping thin bright road markings in fine scale, and the construction of lanes from groups of road markings and road sides. Then, the lanes are further grouped into larger road objects, so-called road segments. During road network completion (Level 3), road segments detected in overlapping images are fused and gaps in the extraction are iteratively closed by hypothesizing and verifying connections between previously extracted road segments. To this end, local as well as global criteria exploiting the network characteristics are used. The result of network completion are roads that connect junctions or endpoints. A key feature of the presented approach is the incorporation of a scheme for self-diagnosis. With this scheme each hypothesis generated during extraction is internally evaluated so that ist relevance for further processing can be assessed. This facilitates decisions inherently appearing during the extraction process. The results achieved with the approach show that the implemented system is able to extract roads in complex environments, i.e. the extraction is also possible when the appearance of roads is heavily affected by other objects. Based on an external evaluation of the results, the system is validated, and advantages but also remaining deficiencies are discussed.
«
In this dissertation, a new approach to automatic road extraction from high resolution aerial imagery taken over urban areas is developed. In order to deal with the high complexity of this type of scenes, the approach integrates detailed knowledge about roads and their context using explicitly formulated scale-dependent models. The knowledge about how and when certain parts of the road and context model are optimally exploited is expressed by an extraction strategy. It is subdivided into three l...
»