Monocular SLAM using vertical straight lines with inverse-depth representation

For vision SLAM, most of the current work focuses on employing point features as landmarks, few take advantage of lines. In this paper, an effective EKF-based monocular SLAM approach using vertical straight lines is proposed. The vertical straight line is represented with the inverse-depth form which allows an efficient and accurate representation of its depth uncertainty and can also increase the computation speed of the SLAM algorithm, compared with the arbitrary direction line. Detailed extraction, representation, initialization and matching method of vertical line features are described. The estimation process is depicted, and experiment results in a static indoor environment are presented and discussed.