Environment modeling by a mobile robot with a laser range finder and a monocular camera

This paper presents a framework of building a 3-D environment model that consists of 3-D object models and their locational information. This kind of environment model is useful for the robot to achieve high level tasks including object manipulation and carrying. In the framework, we first build a 2-D map from laser scans, and detect objects by matching a scan template with the 2-D map. Then, we verify the detected object candidates using vision-based object recognition. We also created 3-D object models from a sequence of monocular images. We have conducted experiments in a room having desks and cabinets to show the effectiveness of the method.