In camera calibration, what is the purpose of using multiple images of a known planar chessboard?

• A. To estimate intrinsic parameters by detecting corners and minimizing reprojection error.
• B. To estimate extrinsic parameters between the camera and the robot.
• C. To measure lens distortion using a single image.
• D. To evaluate image sharpness under motion blur.

Answer: (A)

Using multiple images of a known planar chessboard is about determining the camera’s internal optics, the intrinsic parameters, in a robust way. The chessboard gives a precise, repeatable set of 3D points on a plane (z = 0) and their 2D image locations. Each image provides a mapping from the plane’s coordinates to the image via a homography that involves the camera’s intrinsic matrix and the camera’s pose for that view. By collecting many such views from different angles and positions, you can solve for the intrinsic parameters (focal lengths, principal point, any skew, and lens distortion) in a way that is consistent across all images. This is typically done with a method like Zhang’s, which uses the relationships embedded in the multiple homographies to recover K (the intrinsic matrix) and distortion coefficients.

After the intrinsics are estimated, the extrinsic parameters (the camera pose for each image) can be refined, and the overall fit is improved by minimizing the reprojection error—the difference between the observed corner locations and the points projected using the estimated parameters. Using a planar pattern makes corner detection robust and the math tractable, enabling accurate intrinsic calibration.

The other options describe tasks that aren’t the primary goal of this multi-image planar calibration approach: extrinsics are per-view results after intrinsics are known, a single image isn’t enough to reliably estimate distortion, and image sharpness isn’t the calibration objective.