Comparison of Local Descriptors for Humanoid Robots Localization Using a Visual Bag of Words Approach

摘要

In this paper, we address the problem of the appearance-based localization of a humanoid robot, in the context of robot navigation. We only use information obtained by a single sensor, in this case the camera mounted on the robot. We aim at determining the most similar image within a previously acquired set of key images (also referred to as a visual memory) to the current view of the monocular camera carried by the robot.The robot is initially kidnapped and the current image has to be compared with the visual memory. To solve this problem, we rely on a hierarchical visual bag-of-words approach. The contribution of this paper is twofold: (1) we compare binary, floating-point and color descriptors, which feed the representation in bag-of-words using images captured by a humanoid robot; (2) a specific visual vocabulary is proposed to deal with the typical issues generated by the humanoid locomotion.