Image analysis, in the field of cartography, is not limited only to object recognition,udbut also includes the precise computation of object geometric shapes . Ourudapproach, for automatic extraction ofroad networkfor use in cartography, involvesudtwo distinct steps. Step one, recognition, extracts a topologically correct exhaustiveudgraph of the network and step two, geometric shape extraction, computes alludcenterlines for the complete road network with good quality results both inudaccuracy and in cartographic representation . For object recognition, we haveuddeveloped a road following algorithm based on the minimization of a cost functionudwhich evaluates the homogeneity . For geometric shape computation, we proposeuda method based on the calculation of a well balanced position of road sectionsudwith respect to a set of constraints : internal constraints - derived from a knowledgeudof road-shape characteristics - and external constraints-which force theudconnection between sets of well known points.
展开▼
