Robust object Recognition Using SURF feature model applied in NAO Robot

Object recognition is one of the most important processes for robot soccer in the standard platform robot league. Main task of the vision system of robot soccer while playing is recognizing and tracking the objects like ball, goalposts, robots of the same team and rival robots in the game. The basic idea of several object recognition methods, especially in robot soccer and RoboCup environment, is to use the algorithms based on the color feature of pixels. One of the significant challenges in color feature-based algorithms is the illumination changes of an environment. Since colors are different in different illumination conditions and are influenced by environmental factors like noise, the vision system of a robot will be disturbed if the environment’s illumination changes. Therefore, in this paper a robust object recognition approach against illumination changes of an environment is proposed by using the matching algorithm based on SURF feature. Experimental results on 5 different datasets obtained from the top 5 teams of the world shows that the proposed approach has a good result which includes a recognition rate of more than 96% and an error rate of 4%.