Multispectral IKONOS imagery automatic road extraction from high-resolution

An automatic road extraction strategy from pansharpened multispectral (PS-MS) IKONOS imagery is presented in this paper. Roads are modeled as a network of continuous and homogeneous elongated regions with constant width. Road seeds are selected through two steps. First, initial road seeds are generated based on shape and structural analysis through segmentation and grouping. A line detection method based on differential geometry is then used to extract additional road seeds and these are integrated with the initial seeds using graph theory. Missing pieces of the road network are added using a road tracker based on profile matching. Evaluations of the extracted road network using test sites on residential areas are reported. The automated extraction completeness ranges between 70-90% and the correctness ranges between 65-85%. Keywords-automatic road extraction; IKONOS

[1]  George Vosselman,et al.  Road tracing by profile matching and Kaiman filtering , 1995 .

[2]  Carsten Steger,et al.  An Unbiased Detector of Curvilinear Structures , 1998, IEEE Trans. Pattern Anal. Mach. Intell..

[3]  Curt H. Davis,et al.  A genetic image segmentation algorithm with a fuzzy-based evaluation function , 2003, The 12th IEEE International Conference on Fuzzy Systems, 2003. FUZZ '03..

[4]  Curt H. Davis,et al.  A fuzzy classification approach for high-resolution multispectral data over urban areas , 2002, IEEE International Geoscience and Remote Sensing Symposium.

[5]  Christian Heipke,et al.  EMPIRICAL EVALUATION OF AUTOMATICALLY EXTRACTED ROAD AXES , 1998 .