estimate_normals — 3D curvature op

Data kinds: pointsnormals

Call: import fullseye as fs; fs.ledger.estimate_normals(points, k=25) (to call the implementation directly, import curvature3d; curvature3d.estimate_normals(points, k=25); from the registry, ops3d.get("estimate_normals"))

Usage

Point-cloud normals made consistently outward (pointing away from the neighbourhood centroid). → (N,3).

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

手順(各点、Python ループ): `cKDTree で自身を含む k+1 近傍(k は N-1 に切り詰め)を取り、クエリ点を原点にした近傍座標の散布行列 local.T @ local` の最小固有ベクトルを法線にする(単位長)。向きは「近傍重心との内積が正なら反転」= 近傍重心から離れる側に揃える。

• 近傍が 5 点未満の点は固定値 `(0, 0, 1)` を返す(推定していない)。

• 向き付けは局所ヒューリスティクスで、閉じた凸形状なら外向きだが、開いた面・薄板・凹部では隣接点どうしで向きが食い違い得る(大域一貫性は保証しない)。大域的に揃えるには `orient_normals に通すか、最初から estimate_oriented_normals を使う。organized 深度画像なら normals_from_depth` が視点向きで速い。

• 返り値は float64 (N,3)。`k` 既定 25。決定論的。

• 内部は `principal_curvatures` と同じ計算を通る(法線推定にも二次曲面フィットまで走る)ので、点数が多いと遅い。

Background guides (the physics and conventions behind this op)

blas_threads_and_memory — 行列分解が遅い理由の知識 — BLAS スレッド・キャッシュ・メモリ配置

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)

cylinder_axis_metrologypy -3.11 examples_3d/cylinder_axis_metrology.py

feature_registerpy -3.11 examples_3d/feature_register.py

oriented_normalspy -3.11 examples_3d/oriented_normals.py

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

icp_point2plane · compute_fpfh · shot_descriptor · fuse_to_voxel · reflect · refract · normal_consistency · ransac_cylinder

Same category (curvature)

principal_curvatures · mean_curvature · gaussian_curvature · shape_index


*Provenance: curvature3d.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.