A technique for estimating submersible motion parameters based on stereo pair acquired images, using low performance computational equipment and a simplified video camera calibration procedure employing neural networks and genetic algorithms, is suggested.