LAS Attributes and Classification

This page describes the LAS attributes stored by alsDB and the classification scheme used throughout the processing pipeline.

LAS point attributes

alsDB stores the following standard LAS attributes for each ingested point.

Attribute	Type	Description
Z	float64	Ellipsoidal height above WGS84 or national vertical datum (m)
Intensity	uint16	Returned pulse intensity (instrument-dependent scale)
ReturnNumber	uint8	Which return this point represents (1 = first return)
NumberOfReturns	uint8	Total number of discrete returns for this pulse
ScanDirectionFlag	uint8	0 = negative scan direction, 1 = positive scan direction
EdgeOfFlightLine	uint8	1 if this point is at the edge of the flight line
Classification	uint8	ASPRS LAS classification code (see table below)
ScanAngleRank	float32	Scan angle in degrees relative to nadir (positive = right of flight)
UserData	uint8	User-defined byte
PointSourceId	uint16	Source file ID (useful for multi-file datasets)
R, G, B	uint16	RGB colour (0–65535), if present in the source file

ASPRS LAS classification codes

The Classification attribute uses the ASPRS standard codes:

Code	Meaning
0	Never classified
1	Unclassified
2	Ground
3	Low vegetation (height < 2 m)
4	Medium vegetation (2–5 m)
5	High vegetation (> 5 m)
6	Building
7	Low point (noise)
8	Model key-point
9	Water
10	Rail
11	Road surface
12–14	Reserved
15–17	Wire, transmission tower, connector wire
18	High noise
19+	Overlap / user-defined

Filtering during ingestion

Pass a list of class codes to ingest() or ingest_many() to store only specific point classes:

# Store only ground (2) and vegetation (3, 4, 5)
db.ingest("tile.laz", classes=[2, 3, 4, 5])

# Store everything (default is no filter)
db.ingest("tile.laz")

Reducing the stored classes saves array space and speeds up both ingestion and later queries. For most forest structure applications, classes 2–5 are sufficient.

Role of classification in processing

The processing pipeline uses Classification and ReturnNumber directly:

HAG computation (all processing functions)

filters.hag_delaunay uses ground points (Classification == 2) to build a Delaunay TIN, from which height-above-ground is interpolated for every other point.

Gap fraction (alsdb.processing.gap)

The MacArthur–Wilson estimator uses first returns only (ReturnNumber == 1):

\[P_\text{gap} = \frac{N_\text{ground, first}}{N_\text{ground, first} + N_\text{vegetation, first}}\]

where vegetation = Classification in {3, 4, 5}.

Structural metrics / biomass (alsdb.processing.biomass)

Vegetation points (Classification ≥ 3, or HeightAboveGround above the minimum threshold) are used for height percentiles (h50, h75, h95, hmean) and canopy cover.

CHM / DTM / DSM (alsdb.processing.chm)

DTM interpolates ground points (Classification == 2) using a Delaunay TIN (dtm_method="tin", default), IDW ("idw"), or minimum-Z binning ("min"). DSM rasterises all first returns to the maximum Z. CHM is computed from height-above-ground of vegetation first returns after HAG normalisation against the Delaunay TIN.

Ground point outlier rejection (all processing functions)

Before the Delaunay TIN is built, ground points are screened for outliers using filters.elm (Extended Local Minimum — detects below-ground multipath returns) followed by filters.outlier (statistical — detects above-ground misclassified objects). Outlier ground points are reclassified to Class 1 (unclassified) so the TIN is never corrupted by bad ground returns. Vegetation classification is untouched by this step.

Note

If a dataset uses non-standard class codes (e.g. some national campaigns use code 8 for ground), the HAG computation will be wrong. Verify the classification scheme of your input data and re-classify if necessary before ingestion.

Note

ScanAngleRank (degrees, stored as float32) is available in the schema for future scan-angle normalisation of canopy-cover and density metrics. Current metric computation does not apply this correction; apply a scan-angle filter (e.g. |ScanAngleRank| ≤ 10°) during ingestion if your dataset combines flight lines with very different scan angles.