Study and optimization of a system for 3-D vision

It is shown that the problem of three-dimensional vision can be tackled using stereoscopic techniques. the stereo-visual model adopted is based upon two parallel optical axis TV cameras with a short stereo baseline (12 cm). It is noted that three essential steps are required to measure stereo disparity; select significant points in one of the two images: identify the same points in the other image and determine the exact correspondence between the two sets of points selected; and measure the position in space of such sets of points. The authors put particular emphasis on the third step. The approach they propose consists of: exact optoelectronic tuning and calibration of the system to evaluate how measurement errors and camera resolution affect the precision of the stereo transformation; detection of the position of a large number of points in the three-dimensional and two-dimensional domains; and calculation of the stereo transformation parameters by application of nonlinear minimization techniques