A Descriptive Language for Flexible and Robust Object Recognition Nathan Lovell and Vladimir Estivill-Castro School of CIT, Griffith University, Nathan 4111 QLD, Australia Abstract. Object recognition systems contain a large amount of highly specific knowledge tailored to the objects in the domain of interest. Not only does the system require information for each object in the recog- nition process, it may require entirely different vision processing tech- niques. Generic programming for vision processing tasks is hard since systems on-board a mobile robots have strong performance requirements. Such issues as keeping up with incoming frames from a camera limit the layers of abstraction that can be applied. This results in software that is customized to the domain at hand, that is difficult to port to other applications and that is not particularly robust to changes in the visual environment. In this paper we describe a high level object definition language that removes the domain specific knowledge from the implementation of the object recognition system. The language has features of object- orientation and logic, being more declarative and less imperative. We present an implementation of the language efficient enough to be used on a Sony AIBO in the Robocup Four-Legged league competition and several illustrations of its use to rapidly adjust to new environments through quickly crafted object definitions. 1 Introduction Most object recognition systems use hard-coded, domain specific knowledge. For example, all leagues in Robocup rely on the ball being orange and spherical. If it were changed to be a non-uniform colour or a non-uniform shape then most object recognition systems would have to be largely re-coded 1 . We saw how devastating this was in last years challenge in the four-legged league which required the robots to locate a black and white ball where only eight of the twenty-four teams managed to even identify the ball and no team passed the challenge 2 . This paper presents a higher level descriptive language of the objects we are likely to recognise in a particular domain and thus leave the underlying vision 1 This is certainly true for the team Griffith 2003 code. There are many other examples of systems that are programmed in this way because they are based on the vision algorithms developed by Carnegie Mellon University[1] for the year 2000 competition. 2 http://www.openr.org/robocup/challenge2003/Challenge2003 result.html D. Nardi et al. (Eds.): RoboCup 2004, LNAI 3276, pp. 540–547, 2005. c  Springer-Verlag Berlin Heidelberg 2005