Skip to main content

Intrinsic calibration

Estimate each camera's focal length and distortion, one camera at a time.

Work in progress

Scaffold page — to be expanded with live-overlay screenshots and stopping criteria.

The flow​

Intrinsics are computed on demand, not in a continuous live loop. Each camera goes through four phases:

  1. Capture — record a sweep while moving the board through the field of view. A live coverage overlay shows which regions and orientations you've hit.
  2. Prepare — replay the recording and tune what the solver will use: trim the clip, set a sampling stride, and cap the number of frames.
  3. Computing — the service selects diverse keyframes and solves.
  4. Results — inspect the estimated parameters, the overall reprojection error and coverage before accepting.

Keyframe selection maximises coverage and pose diversity, not raw frame count: a cheap stride first, then a greedy pick by board tilt and image position (default cap: 25 keyframes).

Under the hood​

realtime-calib follows Caliscope's intrinsic pipeline, adapted to current OpenCV:

  • ChArUco corners are interpolated, then solved with cv2.calibrateCameraExtended — the ChArUco-specific calibrateCameraCharucoExtended was removed in OpenCV ≥ 4.7.
  • Flag: CALIB_USE_INTRINSIC_GUESS only — no distortion-model flags, exactly like Caliscope's plain cv2.calibrateCamera call. That means the classic 5-coefficient distortion model [k1, k2, p1, p2, k3], seeded with a guess, with a free aspect ratio.

→ Explanation & sources: Methodology

→ See also: Calibration best practices — how to capture a good intrinsic sweep.