Real-time SLAM Using New Line Descriptor (LEHF)(<Special issue>Real World Imaging)

Abstract

Simultaneous localization and mapping (SLAM) is a technique to simultaneously performs mapping of environments and localization of a camera in real-time. Most existing monocular vision based SLAM techniques use point features as landmarks. However, the use of line segments as landmarks has some advantages. We propose a novel method for a real-time SLAM system that uses line segments as landmarks and computes a Line-based Eight-directional Histogram Feature (LEHF), which is our new line descriptor, to achieve correct matching. LEHF is a fast and efficient way of describing features of line segments, which are detected by the line segment detector (LSD) method. The line-based orthogonal iteration (LBOI) method takes the confidence of each 3D line segment into consideration in order to estimate a camera pose from 2D-3D correspondences made by line descriptor matching. The RANSAC algorithm is applied for 2D-3D correspondences to estimate the correct camera pose. We conducted an experiment in order to test our SLAM system in a desktop environment and to perform augmented reality (AR). Mapped 3D line segments were also evaluated in a planar scene.