Rotation alignment of a camera-IMU system using a single affine correspondence

Area of Science:

• Robotics and Computer Vision
• Sensor Fusion
• Navigation Systems

Background:

• Accurate rotation alignment between camera and inertial measurement unit (IMU) systems is crucial for reliable sensor fusion.
• Existing methods often require complex procedures, specific motion patterns, or additional equipment.
• Precise calibration is essential for applications like autonomous driving and augmented reality.

Purpose of the Study:

• To develop an accurate and easy-to-implement method for camera-IMU rotation alignment.
• To utilize known initial rotation angles for simplifying the alignment process.
• To achieve precise alignment without demanding specialized hardware or camera movements.

Main Methods:

• Formulating the alignment model as a polynomial equation system based on homography constraints.
• Employing a first-order approximation of the rotation matrix.
• Solving the equation system to recover rotation alignment parameters.
• Joint optimization using multiple stereo image pairs for enhanced accuracy.

Main Results:

• Successful recovery of rotation alignment parameters using a single affine correspondence.
• Demonstrated efficiency and precision on both synthetic and real-world datasets.
• Achieved accurate camera-IMU rotation alignment without auxiliary equipment or specific motion constraints.

Conclusions:

• The proposed method offers an efficient and precise solution for camera-IMU rotation alignment.
• It simplifies the calibration process by leveraging initial rotation angles and homography constraints.
• The technique is robust and applicable to various real-world scenarios, improving system performance.