Model-Based Optimization : An Approach to Fast ,

Model-Based Optimization (MBO) is a paradigm in which an objective function is used to express both geometric and photometric constraints on features of interest. A parametric model of a feature, such as a road, a building, a river or the underlying terrain, is extracted from one or more images by automatically adjusting the model's state variables until a minimum value of the objective function is obtained. The optimization procedure yields a description that simultaneously satis es (or nearly satis es) all constraints, and, as a result, is likely to be a good model of the feature. Using this approach, a rough initial sketch of a 3-D object can automatically be re ned, resulting in an accurate model. The paper presents experimental results on road delineation, showing the speedup obtained when compared to totally manual road delineation. Furthermore, because objects are all modeled in the same fashion, we can re ne the models simultaneously and enforce geometric and semantic constraints between objects, thus increasing The research reported here was funded by the Advanced Research Projects Agency and the U.S. Army Topographic Engineering Center under contracts DACA7692-C-034 and DACA76-95-C-0010. not only the accuracy but also the consistency of the reconstruction. We believe that MBO capabilities will prove indispensable to automating the generation of complex object databases from imagery, such as the ones required for realistic simulations or intelligence analysis.