calibrate op• Data kinds: keypoints → table
• Call: import fullseye as fs; fs.ledger.camera_calibration_lm(object_points, image_points_list, dist_model='k1k2p1p2k3', image_shape=None) (to call the implementation directly, import calib_lm; calib_lm.camera_calibration_lm(object_points, image_points_list, dist_model='k1k2p1p2k3', image_shape=None); from the registry, opscalib.get("camera_calibration_lm"))
• GPU: none (CPU reference implementation only)
Jointly refines intrinsics, distortion and all poses by LM from Zhang's initial estimate and returns them with their covariance (table).
> The detailed description below is the original text — the summary and the headings are translated.
Monocular calibration: Zhang's closed form (calib.camera_calibration) gives the start, then a Levenberg-Marquardt refinement minimises the reprojection error over fx, fy, cx, cy, the distortion coefficients of dist_model and every board pose jointly (Zhang 2000, §3.3; the Brown-Conrady model with OpenCV's coefficient order). The Jacobian is analytic; poses are updated on the left by a small rotation vector. Skew is fixed at 0.
object_points: 板の点 (N, 2) の (x, y)(checkerboard_points など)。image_points_list: 視点ごとの (N, 2) の (row, col)。dist_model: "none"・"k1"・"k1k2"・ "k1k2p1p2"・"k1k2p1p2k3"(既定 = OpenCV の既定と同じ 5 係数)。image_shape: Zhang の閉形式が退化したとき(全部が正面平行など)の初期値に使う。
返す dict: K (3x3)、fx・fy・cx・cy、dist [k1, k2, p1, p2, k3]、poses(板 → カメラの 4x4 の list)、rvecs・tvecs、reproj_rms(全体)、per_view_rms、residuals(視点ごとの (N,2) (row, col))、sigma_px(残差から推定した角の雑音)、covariance(fx, fy, cx, cy, k1, k2, p1, p2, k3 の 9x9 の周辺共分散、固定の係数は 0)、param_std(名前 → 標準偏差)、param_names、init(Zhang の値)、iterations・converged、dist_model、object_points、image_shape。★再投影誤差が小さいことは配置が良いことを意味しない(camera_calibration のノートの実測)。配置の良し悪しは param_std と calibration_uncertainty_map で見る。
Raises ValueError: 視点が 3 未満、点の数が合わない・4 未満・非有限、未知の dist_model、初期値が作れない、LM の途中で板が背後へ回る。
• camera_calibration family guide
• Sample-data catalog (download URLs / licences) — 2-D uses skimage.data (BSD/public domain) plus synthetic images; 3-D lists download URLs for real data sources (Stanford, PDS, …).
• Operator provenance and references — the sources of the research/methods this op family came from.
• The canonical algorithm (author, year) and its uses are named in the family usage guide above.
• poc_calibration_truth_map — py -3.11 examples/poc_calibration_truth_map.py
table as input)image_points_to_world_plane · contour_to_world_plane_xld · sim_caltab · disp_caltab · calibration_error_map · calibration_uncertainty_map · calibration_next_pose
calibrate)camera_calibration · find_marks_and_pose · hand_eye_calibration · stereo_calibration_lm
*Provenance: calib_lm.py — CALIB operator registry. This per-op note is generated by tools/opdocs.py md (do not hand-edit).*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.