locan.process.filter.exclude_sparse_points

locan.process.filter.exclude_sparse_points(locdata, other_locdata=None, radius=50, min_samples=5)

Exclude localizations by thresholding a local density.

A subset of localizations, that exhibit a small local density of localizations from locdata or alternatively from other_locdata, is identified as noise and excluded. Noise is identified by using a nearest-neighbor search (sklearn.neighbors.NearestNeighbors) to find all localizations within a circle (sphere) of the given radius. If the number of localizations is below the threshold value min_samples, the localization is considered to be noise.

The method identifies the same noise points as done by the clustering algorithm DBSCAN [1].

Parameters:

• locdata (LocData) – Specifying the localization data from which to exclude localization data.

• other_locdata (LocData | None) – Specifying the localization data on which to compute local density.

• radius (float) – Radius of a circle or sphere in which neighbors are identified (equivalent to epsilon in DBSCAN).

• min_samples (int) – The minimum number of samples in the neighborhood that need to be found for each localization to not be identified as noise (equivalent to minPoints in DBSCAN).

Returns:

All localizations except those identified as sparse (noise) points.

Return type:

LocData

References

[1]

Martin Ester, Hans-Peter Kriegel, Jörg Sander, Xiaowei Xu, A density-based algorithm for discovering clusters in large spatial databases with noise. In: Evangelos Simoudis, Jiawei Han, Usama M. Fayyad (Hrsg.): Proceedings of the Second International Conference on Knowledge Discovery and Data Mining (KDD-96). AAAI Press, 1996, S. 226-231, ISBN 1-57735-004-9.