A Geometric Framework for Absolute Pose and Velocity Estimation With Event Cameras.

Area of Science:

• Computer Vision
• Robotics
• Sensor Fusion

Background:

• Current geometric methods for event-based motion estimation primarily focus on velocity, neglecting absolute pose estimation.
• Simultaneous recovery of absolute pose and velocity from event streams is a challenging, underexplored problem critical for applications like robotic navigation and augmented reality.

Purpose of the Study:

• To propose a novel geometric framework for simultaneous absolute pose and velocity estimation from event streams.
• To address the limitations of existing methods by leveraging 3D scene lines and triggered events.

Main Methods:

• A geometric framework utilizing two core constraints: orthogonality between 3D lines and event planes, and collinearity of events with projected lines.
• Development of linear and polynomial solvers for absolute pose estimation (6-DoF).
• Development of linear and optimization-based solvers for angular and linear velocity estimation.

Main Results:

• Methods require a minimum of three event-line correspondences for independent pose or velocity determination.
• Achieved state-of-the-art performance in both simulation and real-world datasets.
• Demonstrated significant improvements in accuracy and computational efficiency over existing methods.

Conclusions:

• The proposed geometric framework effectively enables simultaneous absolute pose and velocity estimation from event streams.
• The methods offer a robust and efficient solution for critical robotics and augmented reality applications.
• Publicly available code facilitates further research and development in event-based vision.