Perception in the Dark-Development of a ToF Visual Inertial Odometry System

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Visual-inertial odometry (VIO) is crucial for visual simultaneous localization and mapping (vSLAM).
• Time-of-Flight (ToF) cameras offer accurate depth sensing and robustness to lighting variations.
• Existing VIO systems face challenges in low-light or GPS-denied environments.

Purpose of the Study:

• To develop and evaluate a real-time VIO system utilizing a low-cost ToF camera.
• To enhance VIO performance in challenging environments such as darkness and areas without Global Navigation Satellite System (GNSS) coverage.
• To demonstrate the system's suitability for Unmanned Aerial Vehicle (UAV) applications.

Main Methods:

• Implemented a real-time VIO system integrating a ToF camera and an Inertial Measurement Unit (IMU).
• Employed the Iterative Closest Point (ICP) algorithm with salient point selection and robustness weighting.
• Fused ToF-VIO data with IMU measurements using an error-state Kalman filter.

Main Results:

• The ToF-VIO system demonstrated accurate trajectory estimation when compared to motion capture ground truth.
• Successful operation and performance validation were achieved in variable and dark indoor lighting conditions.
• The system showed high accuracy and efficiency in real flight experiments on a UAV platform.

Conclusions:

• The developed ToF-VIO system is accurate and efficient for UAV navigation in GNSS-denied and low-light environments.
• The system's robustness to ambient light variations makes it suitable for diverse operational conditions.
• This research contributes a viable solution for autonomous navigation in challenging scenarios.