Method of 3D mesh reconstruction from sequences of calibrated images

In this paper we propose a technical mesh of an unknown 3D scene from two or more images taken by cameras having the constant parameters based on genetic algorithms. The present approach is based on the formulation of a nonlinear cost function from the determination of a relationship between points of the image planes and all parameters of the cameras. The minimization of this function by a genetic approach enables us to simultaneously estimate the intrinsic and extrinsic parameters of the cameras. These parameters are used with matching to determine the 3D point clouds. The mesh generation is performed by Delaunay triangulation. The strong point of this approach can be clearly seen in the minimization of the constraints of a self-calibration system: The use a genetic algorithm to accelerate the speed of convergence and avoid local minima to obtain a good estimation of the camera parameters and the use of a 3D scene reduces the planarity constraints. Our study is performed on real data to demonstrate the validity and performance of the presented approach in terms of accuracy, simplicity, stability, and convergence.