Vectorize ingest processor

The vectorize ingest processor turns a categorical raster into indexable geometry — with no model, no GPU, and no inference endpoint.

It is the deterministic sibling of image_segment. Where that processor asks an inference provider to find regions, vectorize reads the regions straight out of the pixel values, because for a whole class of imagery they are already there:

• A USDA CDL pixel of 75 is "Almonds".
• A Sentinel-2 SCL pixel of 9 is "cloud high probability".
• Flood masks, burn scars, NLCD and CORINE land-cover tiles, and most change-detection products are all the same shape of data.

Asking a neural network to rediscover a label the file already states is wasteful anywhere, and disqualifying in an air-gapped deployment, where it would make a GPU mandatory in order to read an integer.

It emits the same contract as image_segment

Each detected region carries label, score, area_fraction, bbox, mask_xy_shape and centroid_xy_shape — exactly what image_segment emits. So reproject carries the result to a geo_shape, and every downstream query, aggregation and tile layer keeps working untouched.

A pipeline can swap one processor for the other and nothing else changes. That interchangeability is the point of the design, not a coincidence of it.

Because the classes are known rather than predicted, score is always 1.0 — this is a measurement, not an estimate.

Request fields

Field	Type	Description	Required
source_uri_field	String	Document field holding the raster URI. Defaults to source_uri.	No
target_field	String	Field that receives the emitted regions. Defaults to segments.	No
band	Integer	Raster band to read. Defaults to 0.	No
values	List	Pixel values to vectorize. Omit to vectorize every distinct value found.	No
exclude_values	List	Pixel values to skip — nodata, background, water.	No
class_labels	Object	Maps pixel value to label, as {"75": "Almonds"}. Values without an entry fall back to the raw pixel value.	No
thresholds	List	Bins continuous data into classes before vectorizing.	No
min_area_fraction	Double	Discards regions smaller than this fraction of the raster. Defaults to 0.0005.	No
simplify_tolerance	Double	Douglas–Peucker tolerance for the emitted polygons. 0.0 (default) keeps full detail.	No
max_regions	Integer	Ceiling on emitted regions. Defaults to 256.	No
max_pixels	Long	Ceiling on raster size, as a guard against reading an enormous scene.	No
holes	Boolean	Emit interior rings for regions containing holes. Defaults to true.	No
streaming	Boolean	Stream the raster rather than reading it whole. Defaults to true.	No
window	Object	Read only a pixel window, as {x, y, width, height}.	No
window_field	String	Document field holding that window, instead of a fixed one.	No
reference_config	Object	Settings for resolving the raster reference (for example, credentials or region).	No
on_failure_action	String	fail (default) or skip. Any other value is rejected.	No

Example

Vectorize a cropland layer, keeping only orchard classes, then hand the result to reproject:

PUT /_ingest/pipeline/cropland
{
  "description": "Vectorize CDL orchard classes and project them to WGS84",
  "processors": [
    {
      "vectorize": {
        "source_uri_field": "source_uri",
        "target_field": "segments",
        "band": 0,
        "values": [75, 76, 204, 69],
        "class_labels": {
          "75": "Almonds",
          "76": "Walnuts",
          "204": "Pistachios",
          "69": "Grapes"
        },
        "min_area_fraction": 0.001
      }
    },
    {
      "reproject": {
        "field": "segments",
        "shape_field": "mask_xy_shape",
        "target_field": "mask"
      }
    }
  ]
}

Each entry in segments then looks like:

{
  "label": "Almonds",
  "score": 1.0,
  "area_fraction": 0.0421,
  "bbox": [612340.0, 4183220.0, 613880.0, 4184960.0],
  "crs": "EPSG:5070",
  "mask_xy_shape": { "type": "polygon", "coordinates": [ [ [612340.0, 4183220.0], "..." ] ] },
  "centroid_xy_shape": { "type": "point", "coordinates": [613110.0, 4184090.0] }
}

Choosing between vectorize and image_segment

Use	When
vectorize	The pixel values already carry the classes — land cover, cloud masks, flood extent, change detection, any thematic or categorical raster.
image_segment	The regions must be inferred — buildings in an aerial scene, a nodule in a CT slice, vessels in port imagery.

info

Because both emit the same fields, you can start with vectorize on the data that has labels and add image_segment for the data that does not, without changing your mappings, queries, or tile layers.

Related

• Image segment processor — the inference-based sibling
• Image tiling processor — grid tiling with spatial indexing
• Reprojection processor — carries emitted geometry into a geo_shape
• Geospatial