Motion-error-free calibration of event camera systems using a flashing target

Area of Science:

• Computer Vision
• Robotics
• Biologically Inspired Computing

Background:

• Event cameras offer high dynamic range and temporal resolution, crucial for advanced vision tasks.
• Accurate calibration of event cameras is vital for 3D vision but challenged by the cessation of gray frames in popular models.
• Traditional motion-based calibration methods suffer from movement-related errors.

Purpose of the Study:

• To introduce a motion-error-free calibration method for event cameras.
• To enhance the accuracy and reliability of event camera calibration for 3D vision.
• To develop a robust calibration technique adaptable to hybrid event-based systems.

Main Methods:

• Utilized a flashing target generated by an electronic display to elicit high-fidelity events, avoiding motion artifacts.
• Developed an improved event accumulator to reconstruct detailed gray images for calibration.
• Implemented an optimization method for simultaneous adjustment of camera parameters and control point positions.

Main Results:

• Achieved significantly higher calibration accuracy with a reprojection error of 0.03 pixels compared to 0.96 pixels for motion-based methods.
• Reduced 3D reconstruction error to approximately 0.15 mm, a substantial improvement over the 8.00 mm error of motion-based methods.
• Demonstrated the method's adaptability for hybrid calibration in event-based stereovision systems.

Conclusions:

• The proposed flashing target method provides a highly accurate and robust solution for event camera calibration.
• This technique overcomes limitations of traditional methods, enabling more reliable 3D vision applications.
• The developed approach is suitable for diverse event-based vision systems, including hybrid configurations.