Disease Control
Satellite screens for standing water, drones confirm breeding sites, precision spray response.
Our most mature, furthest-along implementation. Nepal's Terai districts carry the highest malaria and dengue burden in the country, and manual surveillance takes 2–4 weeks to find a breeding site — by which point an outbreak is already underway. Drishti closes that gap: satellite screening flags candidate water bodies weekly at zero marginal cost, autonomous drones verify only the flagged zones, and confirmed sites get a precision larvicide response — all with a full georeferenced audit trail from pixel to intervention.
Sentinel-2 acquisition
ESA Copernicus satellite passes over target districts weekly. Cloud-masked Sentinel-2 L2A tiles are ingested automatically via Google Earth Engine at zero cost.
NDWI water detection
Normalized Difference Water Index isolates standing water. Week-over-week change detection flags new or growing water bodies. Permanent rivers and reservoirs are excluded via historical mask.
Mission planning
Flagged candidate zones are ranked by area, proximity to settlements, and historical case burden. Top N zones are queued for drone validation. Operator approves the mission in ~15 minutes.
Survey drone flight
Autonomous drone flies only to flagged zones at 30m altitude, capturing 5–10cm GSD imagery. YOLOv8 detects standing water, containers, blocked drains, and tire piles across the full zone.
Nano-shot confirmation
For each high-confidence water surface, the drone descends to 2–5m. EfficientNet-B0 classifier analyzes macro close-ups for larval signatures — turbidity, organic film, container type, visible larvae.
Intervention dispatch
larvae_confirmed detections trigger an intervention mission. Drone returns to base — camera payload swapped for larvicide tank in <5 minutes. Precision spray applied at exact confirmed coordinates.
Risk prediction
XGBoost model fuses satellite trends, confirmed habitat density, IoT sensor readings, and historical case data to generate ward-level outbreak risk scores 4–6 weeks ahead.
Closed-loop audit
Every intervention traces back through detection → flight → mission → satellite pixel. Full georeferenced audit trail with timestamp. FCHVs receive alerts in Nepali. Risk scores updated.