Estimation des Deplacements a Partir de Deux Scenes 3D Reconstruites Par UNSysteme Stereoscopique (Estimation of Displacements from Two 3D Frames Obtained from Stereo)

摘要

A method for estimating 3D (three-dimensional) displacements from two stereoframes based on the hypothesize and verify paradigm which is used to match 3D line segments between the two frames is presented. In order to produce the complexity of the method, objects are assumed to be rigid. A set of complete rigidity constraints for 3D line segments is formulated and the uncertainty of measurements are integrated into this formulation. The hypothesize and verify stages of the method use an extended Kalman filter to produce estimates of the displacements and of their uncertainty. In the experimental sections, the algorithm is shown to work on indoor and natural scenes. An extension to the case where several mobile objects are present is shown. The method is quite robust, fast, and was thoroughly tested in hundreds of real stereo frames.