find_surface_pose — 3D match_pose op

• Data kinds: points × points → pose

• Call: import fullseye as fs; fs.ledger.find_surface_pose(model_points, scene_points, model_normals=None, scene_normals=None, dist_step: 'float' = None, angle_bins: 'int' = 30, k_normals: 'int' = 16, **kw) -> 'dict' (to call the implementation directly, import ppf; ppf.find_surface_pose(model_points, scene_points, model_normals=None, scene_normals=None, dist_step: 'float' = None, angle_bins: 'int' = 30, k_normals: 'int' = 16, **kw) -> 'dict'; from the registry, ops3d.get("find_surface_pose"))

• Return value through the ledger: fullseye.ledger.find_surface_pose(...) returns only the declared out type pose (the underlying function also returns auxiliary values). When you need what was dropped, use fullseye.ledger.find_surface_pose.raw(...) or call ppf.find_surface_pose directly.

• Underlying return: {R, t, rmse, votes, ...}

• GPU: none (CPU reference implementation only)

Usage

Find where a known object (model point cloud) lies in a scene point cloud by point-pair-feature (PPF) voting and ICP (pose).

> The detailed description below is the original text — the summary and the headings are translated.

HALCON find_surface_model にあたる初期値の要らない 6-DoF 照合(Drost et al., CVPR 2010):

ppf_model でモデルの点対特徴を表にし、surface_match で場面の参照点ごとに投票して上位の仮説を ICP で詰める。同じモデルを多くの場面へ当てるときは 2 段の形を使う。素の関数(fullseye.find_surface_pose)は {R, t, rmse, votes, inlier_fraction} の dict を返し、台帳 fullseye.ledger.find_surface_pose は pose の正典 (R, t) を返す(残りは .raw)。(R, t) はモデルを場面へ写す: scene ≈ model @ R.T + t(点は(N, 3)、単位はそのまま)。法線を渡さなければ k_normals 近傍で向きつき法線を推定する。dist_step の既定はモデルの直径の 1/20、angle_bins は角度の分割数。

★回転対称な物体(球・円柱・平板)は姿勢が一意に決まらない —— 対称の分だけ別の答えが同じ票を取る。照合が外れたとき(仮説が 1 つも立たない)は rmse が inf、votes が 0。

空の入力(model_points が 0 要素)は ValueError。拒否(ValueError): 点群が (N, 3) でない、モデル点が 5 個未満、場面の点群が空。

References (sample data, literature)

• 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.

Runnable examples (verified samples that actually call this op)

• find_surface_pose_bracket — py -3.11 examples_3d/find_surface_pose_bracket.py

Ops the type connects to (they accept pose as input)

fuse_to_voxel · pose_error · bundle_adjust · mean_reprojection_error · optimize_pose_graph · relative_pose · mean_edge_error · rotation_translation_error

Same category (match_pose)

match_phase_3d · match_pca · moment_axes · match_logpolar_z


*Provenance: ppf.py — 3D 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.