A two-view based multilayer feature graph for robot navigation

Abstract

Abstract-To facilitate scene understanding and robot navigation in a modern urban area, we design a multilayer feature graph (MFG) based on two views from an on-board camera. The nodes of an MFG are features such as scale invariant feature transformation (SIFT) feature points, line segments, lines, and planes while edges of the MFG represent different geometric relationships such as adjacency, parallelism, collinearity, and coplanarity. MFG also connects the features in two views and the corresponding 3D coordinate system. Building on SIFT feature points and line segments, MFG is constructed using feature fusion which incrementally, iteratively, and extensively verifies the aforementioned geometric relationships using random sample consensus (RANSAC) framework. Physical experiments show that MFG can be successfully constructed in urban area and the construction method is demonstrated to be very robust in identifying feature correspondence.I. INTRODUCTION When a mobile robot travels in a modern urban environment, the robot often needs visual signals from its onboard camera to assist navigation. A typical modern urban environment is usually rectilinear and consists of many structured objects and distinctive features such as vertical walls, parallel edges, orthogonal planes, etc. Extracting such features from video frames to form a quick scene understanding can directly benefit navigation tasks such as localization, mapping, obstacle avoidance, and motion planning.Here we design a multilayer feature graph (MFG) to facilitate the scene understanding in urban area. Nodes of an MFG are features such as scale invariant feature transformation (SIFT) feature points, line segments, lines, and planes while edges of the MFG represent different geometric relationships such as adjacency, parallelism, collinearity, and coplanarity. MFG also connects the features in two views and the corresponding 3D world coordinate system. Fig. 1 illustrates the MFG in the 3D world coordinate system. We design an MFG construction method using a feature fusion process which incrementally, iteratively, and extensively verifies the aforementioned geometric relationships using random sample consensus (RANSAC) framework.We have implemented MFG construction algorithm and tested it in physical experiments. Results show that MFG can be successfully constructed from raw image data. Since the process utilizes multiple types of geometric relationships,

Featureless Motion Vector-Based Simultaneous Localization, Planar Surface Extraction, and Moving Obstacle Tracking

“…Low level features like edgelets [6] and lines [13] are also studied, and combined for better performance. Recently, high-level features like 3D lines and planes [9,10,14,16,17,20,25] are introduced to vSLAM works to construct hierarchical environment representations, and semantic features such as vertical and horizontal lines [8] also attract attentions. All of these works require feature transform, which is often computationally expensive.…”

Planar building facade segmentation and mapping using appearance and geometric constraints

“…Previously, a building facade mapping method was proposed in [13] where they use a multi-layered feature graph [14], [15] for robust estimation of building facades. For better reconstruction results, the reconstruction process enforces geometric constraints such as parallelism and coplanarity.…”

High level landmark-based visual navigation using unsupervised geometric constraints in local bundle adjustment

“…However, this method limits itself to a single image for now. At almost the same time, Li et al propose the initial MFG concept based on two views [15], which is then applied to building exterior mapping under an EKF framework [16]. Inspired by [4], we present an LBAbased approach to constructing MFG from a video stream.…”