PL-VIO: Tightly-Coupled Monocular Visual-Inertial Odometry Using Point and Line Features.

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Estimating camera trajectory and building 3D maps are crucial for autonomous systems.
• Existing visual-inertial odometry (VIO) systems primarily rely on point features, which may not capture sufficient environmental structure.
• Integrating inertial measurements with visual data is essential for robust state estimation.

Purpose of the Study:

• To propose a tightly-coupled monocular visual-inertial odometry (VIO) system that leverages both point and line features.
• To enhance the accuracy and robustness of camera trajectory estimation and 3D mapping.
• To exploit the richer geometric information provided by line features compared to point features alone.

Main Methods:

• Developed a Point-Line Visual-Inertial Odometry (PL-VIO) system.
• Employed Plücker coordinates and orthonormal representation for compact 3D line representation.
• Implemented a sliding window optimization framework to fuse pre-integrated IMU errors with point and line re-projection errors.

Main Results:

• The PL-VIO system effectively integrates inertial measurements with both point and line visual features.
• Experiments on public datasets show that PL-VIO outperforms state-of-the-art VIO systems that use only point features.
• The inclusion of line features significantly improves the estimation of camera trajectory and 3D map reconstruction.

Conclusions:

• PL-VIO offers a more robust and accurate solution for visual-inertial odometry by incorporating line features.
• The proposed method enhances the ability to build detailed 3D structural maps.
• This approach represents a significant advancement in sensor fusion for autonomous navigation and mapping.