Toward featureless visual navigation: Simultaneous localization and planar surface extraction using motion vectors in video streams

摘要: Unlike the traditional feature-based methods, we propose using motion vectors (MVs) from video streams as inputs for visual navigation. Although MVs are very noisy and with low spatial resolution, do possess high temporal resolution which means it is possible to merge different frames improve signal quality. Homography filtering MV thresholding proposed further quality so that can establish plane observations MVs. We an extended Kalman filter (EKF) based approach simultaneously track robot planes. formally model error propagation of derive variance merged MVs, provide necessary observation EKF. have implemented method tested in physical experiments. Results show system capable performing localization mapping a relative trajectory less than 5.1%.