Finnish space technology company Kuva Space has completed a pilot with a major European security stakeholder to show that hyperspectral satellite data can enhance monitoring of illicit crops across broad areas. The study concentrated on suspected opium poppy cultivation in Afghanistan's Helmand Province. It combined the company's Hyperfield imagery with Copernicus Sentinel-2 data and artificial intelligence to distinguish poppy fields from legal crops.
Pilot Scope and Satellite Coverage
Kuva Space's Hyperfield-1 satellites collected 363 hyperspectral images between January and June 2026, covering 648,450 square kilometers. The company integrated this imagery with Sentinel-2 data and used AI-based classification to identify likely poppy fields. Because Afghanistan lacks precise agricultural boundaries, Kuva Space built a separate parcel delineation model tuned to the country's small and fragmented fields.
Building a Parcel Level Detection System
The detection system first identified field boundaries before classifying each parcel according to spectral signatures associated with opium poppy. Kuva Space trained its in-house foundation model on 1.2 million image patches, then applied it across a large operational area. The model used a relatively small set of expert annotations and produced parcel level predictions rather than individual pixel flags, while geolocation accuracy averaged 21.75 meters.
Accuracy Gains and False Positives
The combined hyperspectral and Sentinel-2 model achieved 75 percent overall accuracy, compared with 71 percent for Sentinel-2 alone. The practical gain came mainly from reducing false positives, which matters when screening nearly 249,000 fields. Across Helmand, the system identified 248,889 agricultural fields and flagged 8,835, about 3.6 percent, as likely poppy fields, helping direct costly high-resolution imagery and analyst time toward the most promising locations.
Why Spectral Detail Matters
Hyperspectral sensors measure light in many narrow bands, allowing the AI models to detect subtle differences in chlorophyll, water content, and pigmentation. These spectral cues help separate poppy from visually similar legal crops even in small or mixed fields. The company also uses observations from different times to recognise growth cycles and harvest timing associated with poppy cultivation.
A Shifting Global Drug Map
The pilot arrives as the geography of opium production is changing. The United Nations 2026 World Drug Report states that Afghanistan's output remains far below pre-ban levels, while Myanmar recorded a 17 percent increase in poppy cultivation in 2025. Kuva Space views scalable monitoring as essential for authorities that cannot rely on ground access in volatile or remote regions.
From Pilot to Operational Use
Kuva Space is already extending its poppy detection work to Myanmar, where geographic and altitude data can sharpen accuracy. The company also plans to adapt the method to other illicit crops such as coca and cannabis based on customer demand. Chief Executive Jarkko Antila said the approach closes a long-standing gap between wide area coverage and crop level detail.
Limitations and Verification
Kuva Space was clear that the flagged fields are model predictions rather than confirmed detections. The results broadly align with recent United Nations Office on Drugs and Crime reporting at an aggregate level, but individual field verification remains necessary. More in-situ ground truth will be needed to confirm individual fields and to move from likely detections to verified intelligence.
The pilot offers a concrete example of how hyperspectral Earth observation and purpose-built AI can support security and law enforcement priorities across vast territories. While the current 75 percent accuracy is described as a first generation result, Kuva Space is targeting at least 90 percent through improved Hyperfield-2 satellites and additional data such as radar, weather, and elevation models. The flagged fields remain predictions for prioritisation, but the method could make illicit crop monitoring more focused and cost effective.