Automatic floor segmentation for indoor robot navigation

Indoor autonomous robots can perform desired task in indoor environments without continuous human guidance, so navigation is indispensable for them. In the state of the art of vision-based navigation, one or more cameras are usually installed on a robot. This has led to a larger workload to deal with the data collected by cameras and arises the problem of delay. In the paper, we propose a novel vision-based navigation framework for indoor autonomous robots in which the camera is fixed on the ceiling. In the navigation scheme, a robot takes floor as their moving regions. So floor segmentation algorithm has to be designed to get floor regions in navigation images automatically. We adopted clustering analysis to implement automatic floor segmentation, and we also proposed a PCA based improved version of the algorithm to remove negative effect of shadow for segmented results.