A new method to detect 3D Obstacles using a stereo vision system and a 2D laser range finder is presented. Laser range finder measures distance to obstacles, but only on a plane parallel to the floor; and stereo vision is not able to estimate distances to surfaces with little or no texture at all. This paper explores a form to take advantages of both kind of sensors. The main idea is to project 3D points detected by the laser telemeter in the form of initial values for pixel disparities into a trinocular stereo vision system. Experimental tests using a mobile robot in a indoor environment showed promising results.