A Refined Algorithm ( GA-Powell ) for Camera Parameters Estimation and 3 D Modeling Based on Silhouettes

In this paper, a hybrid method is applied to recover parameters and motion of camera from a set of silhouettes of an object taken under circular motion. Camera parameters can be obtained by maximizing the total coherence between all silhouettes. Two optimization methods, the Powell optimizer (PO) and the Genetic algorithms (GA), are applied to maximize the silhouette coherence and their performances are compared for several experiments. To take advantage of the strengths of the two methods, we developed a hybrid method that combines the genetic algorithm and the Powell optimizer to improve the performances in term of convergence speed and accuracy. The recovered parameters are used for 3D image-based modeling to obtain high fidelity 3D reconstruction.