<XML><RECORDS><RECORD><REFERENCE_TYPE>3</REFERENCE_TYPE><REFNUM>7466</REFNUM><AUTHORS><AUTHOR>Boyling,T.A.</AUTHOR><AUTHOR>Siebert,J.P.</AUTHOR></AUTHORS><YEAR>2004</YEAR><TITLE>Foveated Vision for Space-Variant Scene Reconstruction</TITLE><PLACE_PUBLISHED> Proceedings of the 35th International Symposium on Robotics, ISR2004, Nord Villepinte, Paris, France, March 23-36, 2004. </PLACE_PUBLISHED><PUBLISHER>N/A</PUBLISHER><PAGES>1-6</PAGES><LABEL>Boyling:2004:7466</LABEL><ABSTRACT>We report the design and successful implementation of an autonomous active vision system based upon a binocular robot head [9,10] that can quickly create a 3D model of its environment by integrating the 3D models recovered through photogrammetry from multiple observations. We outline the architecture of a complete foveated active vision system and describe its components. These include a multi-resolution foveated vergence algorithm, and gaze control based upon analysis of recovered multi-resolution regularized rangemaps, Finally, a multi-resolution approach to aligning the recovered surface models is presented, which allows for the creation of an extended model of the scene. </ABSTRACT></RECORD></RECORDS></XML>