Single-Shot Intrinsic Calibration for Autonomous Driving Applications

Area of Science:

• Computer Vision
• Robotics
• 3D Reconstruction

Background:

• Accurate camera intrinsic calibration is crucial for 3D computer vision tasks.
• Existing methods often require multiple images or complex setups.
• Single-shot calibration with multiple checkerboards offers a streamlined approach.

Purpose of the Study:

• To develop repeatable guidelines for single-shot camera intrinsic calibration.
• To optimize the placement and orientation of multiple checkerboards for accurate calibration.
• To validate the proposed guidelines in simulation and real-world scenarios.

Main Methods:

• Utilized a simulator to determine optimal position and rotation intervals for checkerboard placement.
• Generated and evaluated thousands of multiple checkerboard poses against ground truth values.
• Tested guidelines on diverse cameras and validated projections using 3D LiDAR (Light Detection and Ranging).

Main Results:

• Identified specific intervals for optimal corner detector performance.
• Defined robust guidelines for creating effective multiple checkerboard setups.
• Achieved accurate intrinsic parameters for 3D applications with at least seven checkerboards in a single image.

Conclusions:

• The developed guidelines enable accurate single-shot camera intrinsic calibration.
• The method is robust across various camera types and distortion profiles.
• This approach facilitates precise 3D perception in applications like robotics and augmented reality.