AI-generated editorial illustration; not a documentary photograph.
Finnish space-technology company Kuva Space has completed a pilot with an unnamed European security stakeholder to assess whether hyperspectral satellite imagery and AI can improve regional monitoring of illicit crops. The Afghanistan study combined imagery from Kuva Space's Hyperfield-1 satellites with Copernicus Sentinel-2 data to classify fields that were likely to contain opium poppies.
Mapping Agricultural Parcels
The pilot covered 648,450 square kilometres using 363 hyperspectral images collected between January and June 2026. Because precise agricultural parcel maps were unavailable, Kuva Space fine-tuned a separate field-boundary model with 10-metre-resolution Sentinel-2 data. It identified 248,889 fields in the analysed area of Helmand Province, creating the parcel map used for subsequent classification.
Fewer False Positives
Kuva Space's in-house foundation model evaluated parcels for spectral patterns associated with poppies, using observations from different stages of the growing season to help distinguish them from legal crops. The company reported 75% accuracy when combining Hyperfield-1 and Sentinel-2 data, compared with 71% for Sentinel-2 alone, with the improvement attributed primarily to fewer false positives.
The combined system flagged 8,835 parcels as likely poppy fields. Those results are intended to narrow the areas requiring expensive high-resolution imagery or analyst review, rather than establish that poppies were present in every flagged field.
Verification Remains Limited
Kuva Space said the regional cultivation patterns broadly matched recent reporting from the United Nations Office on Drugs and Crime, but the individual classifications have not been verified field by field. More in-situ ground-truth data would be required to confirm specific detections, making the pilot a prioritisation tool rather than definitive evidence of cultivation.
Expansion to Myanmar
The company is extending its poppy-detection work to Myanmar and plans to incorporate factors such as location and altitude. It is targeting accuracy of at least 90% through additional ground-truth data, second-generation Hyperfield-2 satellite imagery and other inputs, including radar, weather and elevation data. That figure is a company target, not a demonstrated result.
Source: Kuva Space takes on illicit crop monitoring with hyperspectral satellites and AI
SIGNAL TITANS
Leave a Reply