3D cognitive map construction by active stereo vision in a virtual world

Abstract

In this study, a multi-scale phase based disparity algorithm is developed. This algorithm is then applied in a simulated world. In this world there is a virtual robot which has a stereo camera system simulated with the properties similar to human eyes and there are 3D virtual objects having predefined simple shapes. The virtual robot explores its environment intelligently based on some heuristics. Only stereo images rendered from the virtual world are supplied to the robot. The robot extracts depth information from the stereo images and when an object is seen, it investigates the object in detail and classifies the object from the estimated shapes of the object parts. © Springer-Verlag 2004.