Introducing a novel vision based obstacle avoidance technique for navigation of autonomous mobile robots

This paper introduces a novel vision based obstacle avoidance technique for indoor navigation of autonomous mobile robots. The indoor environment is considered as office environment with homogenous surfaces. In this technique, a color image taken by a monocular vision camera is clustered by mean-shift algorithm, then the clustered image is classified by a novel classification technique based on graph partitioning theory. The classified image includes meaningful information such as floor, walls and obstacles for robot to navigate around office environment. The simulation results show the effectiveness of proposed technique for further real-time implementation and experiments.