A 3-D scene interpreter for indoor navigation

A 3D scene interpreter for visual navigation of a mobile robot is presented. The interpreter provides a partial understanding of scene contents. With the help of a priori knowledge and world constraints, it can be used for route planning and obstacle avoidance. The interpreter consists of data-driven low and intermediate-level stages and a model-driven high-level analysis stage. The high-level analysis interprets the scene contents by matching the description of the scene provided by the lower-level stages to semantic models stored in the database. Experimental results are presented to verify the performance of the interpreter.<>