SDPS 2011,

Improving Vision Based Robot Navigation Through Specific Location Markers

Cengiz Toğay and Utku Bayram

ABSTRACT:Navigation is one of the most studied problems in the robotics area. To navigate a target, a robot has to know its own position and target positions and its orientation in surrounding environment. Robots are guided to a target based on some inputs. The input can be some sensor readings such as ultrasonic, GPS, infrared, etc. or cameras. In this study, we utilized vision based inputs for navigation. Landmarks defined as identifiable objects with known real position in an environment are utilized to estimate
the position of a robot. One of the pioneering studies introduced by Sugihara utilizes a single camera to estimate the position through beacons (landmarks). Our developed
robot named Troy has a webcam to capture images from the environment. We utilized an open source image processing API named OpenCV to process images and identify landmarks.
When the landmarks are identified, we can estimate distance of the landmarks to the robot, and then these distances are utilized for estimating the robot’s position and
orientation. In this paper, we present a monocular vision based range measurement method for indoor applications for navigation of a robot with two identified landmarks and
also identified experiences are presented.