Road extraction in suburban areas by region-based road subgraph extraction and evaluation

In this paper, a road extraction approach for suburban areas from high resolution CIR images is presented. The approach is region-based: the image is first segmented using the normalized cuts algorithm, then the initial segments are grouped to form segments, and road parts are extracted from these segments. Ideally roads in the image correspond to only one extracted road part, but they are often covered by several road parts with gaps between them. In order to combine these road parts, neighbouring road parts are connected to a road subgraph if there is evidence that they belong to the same road, such as similar direction and smooth continuation. This process allows several branches in the subgraph which is why another step follows to evaluate the subgraphs and divide them at gaps which show weak connections. The subgraph evaluation step is the focus of this paper. Linear programming is used for the subgraph evaluation after gap weights are determined. Two ways of determining gap weights are discussed, one using criteria which describe the properties of the road parts and their interrelations, and one using context objects (vehicles, trees, vegetation) in the gaps. The determination of the gap weights and the division of the road subgraphs is shown with an example.