cartagen.regularize_building_fer

regularize_building_fer(polygon, length=6.0, area=6.0)

Regularize a polygon using feature edge reconstruction.

This algorithm was proposed by Yang et al. [1]. FER enforces orthogonality and clean edge alignment on raw segmentation polygons. It first estimates the dominant orientation of a building, then iteratively snaps near-parallel and near-perpendicular edges to that principal axis, using Douglas-Peucker simplification, minimum-rotated-rectangle fitting, and a local reconstruction loop.

This is slightly modified version of the SamGeo Python package available here. It is basically the same but without the osgeo dependencies.

Parameters:

• polygon (Polygon or MultiPolygon) – Input building footprint(s) to regularize.

• length (float) – Minimum edge length used during the reconstruction pass (default 6 m).

• area (float) – Polygons whose area is below this value (m²) are discarded (default 6 m²).

Returns:

Polygon or MultiPolygon or None – The regularised geometry, preserving the input type. Returns None if the geometry is entirely filtered out.

See also

enclosing_rectangle

Construct an enclosing rectangle from a polygon.

regularize_building_rectangle

Transform a polygon into a rectangle.

regularize_building_regression

Regularize a polygon using recursive linear regression.

Notes

FER preserves overall shape well and handles irregular, non-axis-aligned buildings. But it assumes roughly rectilinear architecture and tuning length and area may be needed for dense urban scenes or very small footprints.

References

[1]

Min Yang, Renwei Zou, Tinghua Ai, and Xiongfeng Yan. Regularizing building outlines extracted from remote sensing images by integrating multiple algorithms. Geocarto International, 2024. URL: https://api.semanticscholar.org/CorpusID:270784716.

Examples

>>> polygon = Polygon([(0, 0), (0, 2), (1, 2), (1, 1), (2, 1), (2, 0), (0, 0)])
>>> regularize_building_fer(polygon, 1.0, 1.0)
<POLYGON ((0 0, 0 2, 1.5 2, 1.5 0, 0 0))>

(Source code)