ARASAT
Centre-pivot fields seen from a satellite

Products · Agriculture

Field Boundary Extraction

The parcel layer every other agricultural analysis is built on.

Processing service

Processing service delivered per area of interest

Overview

Where exactly is each field, as it is actually farmed today?

Deep segmentation models extract the exact boundary of every cultivated parcel directly from satellite imagery, independent of cadastral records, so the result reflects the land as it is farmed this season. Vectorisation and geometry clean-up produce closed, clean polygons that become the base layer for crop classification, irrigation mapping and water accounting.

Deep-learning segmentation
Method
No cadastral input required
Independence
10 m standard, sub-metre with high-resolution input
Resolution
Clean vector parcels
Output

Example outputs

  • Centre-pivot fields in imagery and the same fields as detected boundary polygons

    Imagery (left) and extracted parcel polygons (right).

  • Parcel polygons outlined in pink over mountainous terrain

    Delineated parcels in mountainous terrain with small, irregular fields.

How it works

Pipeline and maturity of each stage

  1. 01

    Imagery selection

    Operational

    Cloud-free optical scenes across the season, or high-resolution imagery where field sizes demand it, optionally enhanced by super-resolution.

  2. 02

    Segmentation

    Operational

    A deep-learning model predicts parcel interiors and edges simultaneously, trained on parcel polygons from a range of climates, field sizes and farming patterns.

  3. 03

    Vectorisation and clean-up

    Operational

    Edges are converted to closed polygons, geometry is regularised and small artefacts removed, producing a topologically clean parcel layer.

  4. 04

    Delivery

    Operational

    Parcel layer with area statistics, updatable each season, delivered as shapefile or GeoJSON with a report.

Data used

  • Sentinel-2
  • High-resolution optical imagery on request
  • Super-resolved imagery for small parcels

Methods and models

  • Deep-learning segmentation
  • Edge and region prediction
  • Vectorisation and geometric regularisation

Outputs

  • Parcel boundaries (shapefile / GeoJSON)
  • Parcel area statistics
  • Seasonal update of the layer

Who it is for

  • Cadastre and land administration
  • Agricultural statistics and subsidy control
  • Irrigation scheme management
  • Input layer for crop and water products

More in Agriculture

All agriculture products

From space. For Earth.

Start with ARASAT

Tell us what you need to monitor, understand or protect. Our team works with you from the first conversation to a running program.