Image-based floor segmentation in visual inertial navigation

This paper presents a floor segmentation algorithm for indoor sequences that works with single grey-scale images. The portion of the floor closest to the camera is segmented by judiciously joining a set of horizontal and vertical lines, previously detected. Since the proposed method is not based on computing the vanishing point, the system can deal with any kind of indoor scenes and adapts quickly to camera movements. A second contribution is the detection of moving features for points within the segmented floor area. Based on the estimated camera ego-motion, the ground plane homography is derived. Then, the expected optical flow for the ground points is calculated and used for rejecting features that belong to moving obstacles. A key point of the designed method is that no restrictions on the camera motion are imposed for the homography derivation.