Dendromatics

Changed

- Take advantage of multiple CPU cores during `DBSCAN` clustering.

Fixed

- Take into account all iterations in `individualize.compute_axes` to fix inconsistencies with progress bar.

- Update `CSF-3DFin` to version 1.3.0. Thanks to Daniel Girardeau-Montaut,
and the CloudCompare project, it contains many bug fixes and speed
improvements. Most notably, it fixes a race condition that resulted
in non deterministic executions.

- Fixed a bug that prevent using voxelate with custom xyz component indices

Changed

- Update `CSF` to point to the CSF-3DFin fork. This add a fix that improve numpy array - CSF interop.
Height normalization processing is faster (depdending the size of the cloud it could be of several orders of magnitude)

Added

- `ground.check_normalization_discrepancy` returns the indicator and the percentage of discrepancy
between the original area and the slice area.

Fixed

- Fixed a malformed call to `voxelate` inside `ground.check_normalization`.

- Fixed the documentation build int RTD.

Added

- Added an optional group of dependencies for documentation (`docs`) in the project file.

Changed

- Changed `scipy.spatial.cKDTree` for `scipy.spatial.KDTree` and use parallel queries to speed up computations.

Fixed

- Removed few harmless unused variables.

- Modified boolean filtering in `clean_cloth()` so it accounts for cases where DTM is completely flat. This happens, for instance, when a DTM is fitted to an already normalized point cloud.

Added

- Added clearer error messages in certain situations where point density of input point cloud was low and no clusters were found in step "1.-Extracting the stripe and peeling the stems". Previous error messages were Python's defaults "zero-size array to reduction operation minimum which has no identity" and "min() arg is an empty sequence".

- Added `ground.check_normalization`. This function slices a normalized point cloud and compares its area vs. a scalar. It's intended use it's to compare the area of a point cloud to the area of a slice of points around ground level from the height-normalized version of the same point cloud. This is useful to check for inconsistencies in the height-normalization.

- Added support of Python 3.12

Fixed

- Fixed a bug in clean_ground(). It caused that, during denoising step before dtm computation, the removed points (noise) were independent of the voxel resolution used.