Robust Real-time Landmark Recognition for Humanoid Robot Navigation

Landmark recognition is identified as one important research area in robot navigation systems. It is a key feature for building robots capable of navigating and performing tasks in human environments. However, current object recognition research largely ignores the problems that the mobile robot context introduces. We developed a landmark recognition system which is used by a humanoid robot to identify landmarks during its navigation. The humanoid landmark recognition system is based on a two-step classification stage which is robust and invariant towards scaling and translations. Also, it provides a good balance between fast processing time and high detection accuracy. An appearance-based classification method is initially used to provide the rough initial estimate of the landmark. It is followed by a refinement step using a model-based method to estimate an accurate classification of the object. The goal of our work is to develop a rapid, robust object recognition system with a high detection rate that can actually be used by a humanoid robot to recognize landmarks during its navigation.