An Accurate 3d Vision System Using a Projected Grid for Surface Geometrical Study

mcmlogical use. The optical appamtuscs which compost ARSTRACT our 313 active sensor are one (or more) camera and one projector, the camera calibration has been widely malysed The 3D active vision system ilivstrated in this paper notably in stereo-vision systems [FAURT] but with regard has been developed far a geometrical study of object to acrive vision system using a projector. few works have surfaces. Indecd, al l our work has k e n directed towards been done in this area. the geometrical exploitation of the calculated 3D data. This For the projector calibration the 3D rcfmncc data is the reason for which the projected grid is considered as are the 3D projected grid ndes, in order to avoid dl a network of curves rather than a graph. manual measuremtnrs of these data source of mis&xs In order to calculate the 3D coordinates we must we propost , in section 4 an optical and autonomous first calibrate the camera and the projecror. An originat method for the projector calibration. calibration system is proposed in this paper which The major problem of 3D active vision system calibrate the projector front the camera and permits to projecting an entire grid of lines is the correspdtnce avoid the manual measurement of the 3D projected grid. problem solving between the original and imaged The features of our calibration system are the efficiency, @&.We propose in section 5 a reliable method for rhc the autonomy, the precision of the reference data calculus, correspondence: each imaged curve i s labeled and the sped in calculating the system parameters. independently from the other curves by using global and Another imponmt conwibutian of our w o ~ k is the geomemcal cons~aints. Thus, all the 3D ~d points can correspondence probl'ern solving. Different from orher k calculated and not only the grid ndes, published methods. the correspondence is done curve per Our 3D sensor provides two families of 3D curves curve without any ambiguity by using geometrical and and permits to tackle easily a geometrical study of object global constrainrs. Thus, with the help of our 213 and 3D surfaces. mis third level of our system will be djscussed methods. all the 3D curves on the object surfaces can be jn section 6. calculated and not only the Prid nodes. Rnally, the two'famices of 3 0 curves provided by our system allow us to calculate easily the shape parametem by using the concepts of differential geomety.