conference

Tightly-coupled Line Feature-aided Visual Inertial Localization within Lightweight 3D Prior Map for Intelligent Vehicles

Xi Zheng, Weisong Wen, Li-Ta Hsu

2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC) · Sep 2023

Abstract

Visual-inertial navigation system (VINS) is widely used for autonomous platforms but suffers from drifting over a long time. To remedy this situation, a lightweight 3D prior map-aided visual-inertial navigation system is presented in this paper, which tightly couples the visual-inertial data stream with a lightweight prior map involving 3D line information. To fill the gaps between 3D maps and 2D images, the mutual geometric feature of line segments is utilized to connect these two types of information in different dimensions. By detecting and matching line features in two data sources, the line pairs are utilized as constraints in the nonlinear optimization model and added to the existing factor graph framework in a tightly coupled form. Meanwhile, a fast line feature tracking strategy is employed to monitor and remove extreme outliers, which will further improve the reliability of this structural characteristic during the cross-modality localization. The effectiveness of the proposed method is evaluated by public indoor unmanned aerial vehicles (UAV) datasets, and outdoor unmanned ground vehicles (UGV) datasets generated by the CARLA simulator.

Problem Addressed

Visual-inertial navigation drifts over long periods, and exploiting lightweight 3D prior maps for drift suppression requires bridging the gap between 3D maps and 2D images.

Key Contributions

Line-feature constraint linking 3D prior map lines with 2D image lines via mutual geometric features
Optimization model converting minimum line-to-line distance into a point-to-point reprojection problem
Fast line tracking strategy based on direct optical flow to monitor and remove extreme outliers
Tightly coupled factor graph fusing prior-map line features with visual-inertial data to suppress drift

Conclusions

Evaluations on public indoor UAV datasets and outdoor UGV datasets generated with the CARLA simulator demonstrate the method’s effectiveness, especially in autonomous driving scenarios. Future work includes analyzing degeneration of line feature constraints and GNSS-aided initialization between the VINS local frame and the global map frame.