HiRo-SLAM: A High-Accuracy and Robust Visual-Inertial SLAM System with Precise Camera Projection Modeling and...

Area of Science:

• Robotics and Computer Vision
• Simultaneous Localization and Mapping (SLAM)

Background:

• Conventional visual-inertial SLAM methods suffer from biases due to imperfect camera models, uneven feature distribution, and outliers.
• These limitations hinder accuracy and robustness in real-world applications.

Purpose of the Study:

• To introduce HiRo-SLAM, a unified optimization framework for high-accuracy and robust visual-inertial SLAM.
• To address limitations of existing SLAM systems through novel innovations.

Main Methods:

• Precise Camera Projection Modeling (PCPM) for accurate camera intrinsic and distortion handling.
• Visibility Pyramid-based Adaptive Non-Maximum Suppression (P-ANMS) for uniform feature constraints.
• Robust Optimization Using Graduated Non-Convexity (GNC) to mitigate outlier impact.
• Point-Line Feature Fusion Frontend combining point and line features for enhanced perception.

Main Results:

• HiRo-SLAM outperforms state-of-the-art visual-inertial SLAM methods on EuRoC MAV, TUM-VI, and OIVIO benchmarks.
• Achieved a 30.0% reduction in absolute trajectory error on the EuRoC MAV dataset.
• Demonstrated millimeter-level accuracy in controlled conditions.

Conclusions:

• HiRo-SLAM offers significant improvements in accuracy and robustness for visual-inertial SLAM.
• The system excels in environments with moderate texture and minimal motion blur.
• Performance may be limited in highly dynamic scenarios with severe motion blur or extreme lighting.