Localization of RGB-D Sensors for Robotic and AR Applications

Robot localization problem has been addressed using variety of methods such as ultrasonic, laser range sensors, beacons, landmark based methods and map-based approaches. This research describes a vision based localization approach using low cost, information rich RGB-D cameras which localizes a moving RGB-D camera relative to another static RGB-D camera in the environment. The features extracted from overlapped images of two cameras are matched and fed through an iterative non-linear pose estimation algorithm to obtain the relative pose of the moving camera. Experiments and analysis show similar or improved accuracy compared to other localization methods and datasets.