Vision-based global localization for mobile robots using an object and spatial layout-based hybrid map

This paper proposes a novel vision-based global localization approach that uses an object and spatial layout based hybrid map. For environment modeling, we use the following visual cues with a stereo camera; local invariant features for object recognition and their 3D positions for object position representation. Also, we use the depth information at the horizontal centerline in images where the optical axis passes through. Therefore, we can build a hybrid local node for a topological map that is composed of a metric map and an object location map. Based on such modeling, we suggest a coarse-to-fine strategy for the global localization. The coarse pose is obtained by means of object recognition and point cloud fitting, and then its fine pose is estimated with a probabilistic scan matching algorithm. With real experiments, we show that our proposed method can be an effective vision-based global localization algorithm.