A neural network based localization from distorted image in omnidirectional catadioptric vision system

Abstract

Omni-directional catadioptric vision system has been used increasingly in RoboCup(Robot World Cup) medium-sized league for localization. Omni-directional vision system can greatly increase the view range of robot. The main disadvantage of this vision system is the image distortion. This paper proposed an efficient method, using an Artificial Neural Network, to localize the robot in the field based on the distorted video image captured by mobile robot.