Vision based localization system for a mobile robot in RoboCup environment

Localization of mobile robots is one of the most important research topics in autonomous robots. In this paper we present a system which uses only the information from a vision sensor (Video camera) of a middle size robot in RoboCup[1] environment to localize the robot in the RoboCup soccer field. The main idea is based on comparison of an image frame with a set of images in a database. The images in the database are obtained from the soccer field only once during a training stage. For each such image a feature vector based on Wavelet transform and some other local data are calculated and saved in the database. During the game, whenever the robot needs to find its position, an image frame is frozen from robot camera and its feature vector is calculated By comparing the feature vector of this image and those in database, the most similar (if exists at any) is selected The position from which this most similar image in database was taken is considered to be the location of robot. This algorithm was implemented on Sharif CE middle size RoboCup team[2] and relatively satisfactory results were obtained.