Back to selected work
Case 07 · Verdant · Crop Health Survey · Shipped 2024
// Case 07 · 2024 · Precision Agriculture · Autonomy

Verdantcrop survey.

Autonomous multispectral crop-health surveys for large-scale farms. PX4 mission planner, 4-band NDVI capture, field-level health maps — flagged stressed zones across 320ha before they were visible from the ground.

EngagementFixed-price
StackPX4 · ROS 2
Timeline12 weeks
StatusIn production
verdant.internal/field
Verdant — NDVI field health map
Chapter 01 · The Brief

Stressed crops, invisible
until it's too late.

Large-scale farms were scouting on foot or by tractor — a process that covered maybe 20ha a day and spotted stress only after it was visible to the naked eye. By then, the window for intervention had often already closed.

The brief: fly autonomous multispectral surveys over entire fields, generate NDVI health maps before stress is visible, and surface the exact zones that need agronomist attention — so inputs go where they matter, not everywhere.

Brief at a glance
Client
Large-scale grain farm, Midwest USA
Site
320ha · mixed wheat and corn
Problem
Foot scouting: 20ha/day · stress visible too late
Goal
Weekly full-field NDVI survey in a single morning
Payload
4-band multispectral camera · radiometric calibration
Chapter 02 · The Approach

Weekly maps,
automated end-to-end.

Multispectral Capture Pipeline

A PX4-based fixed-wing airframe flies pre-mapped grid patterns at 120m AGL, capturing 4-band imagery (Red, Green, Red-Edge, NIR) with a radiometrically calibrated multispectral camera. A reflectance panel calibration shot is taken before each flight and applied in post-processing to ensure comparable results across seasons.

MAVSDK handles mission upload and telemetry; a custom ROS 2 node monitors the payload sensor and re-triggers if any frame falls below confidence thresholds.

NDVI Map Generation

Raw imagery is processed on an edge server running OpenDroneMap + custom Python pipelines that compute NDVI, NDRE, and GNDVI indices per pixel. The resulting rasters are tiled into a web map and thresholded into zone classifications — healthy, watch, stressed, critical.

Agronomists receive a zone report with GPS coordinates of every stressed area, ready to load into their prescription mapping software for variable-rate input application.

Chapter 04 · By the numbers
320ha
Surveyed
per flight.
vs 20ha/day on foot
4-band
Multispectral
capture.
NDVI · NDRE · GNDVI
7×
Faster
than foot scouting.
full field in one morning
Early
Stress detected
before visible.
2–3 weeks earlier
Chapter 05 · Inside the mission

Fly the field,
read every pixel.

Mission grid, live telemetry, NDVI heatmap, zone classification, and prescription export — the full precision-ag pipeline.

// 01 · Plan
Verdant — mission grid over field boundary

Field boundary → auto-generated grid at optimal overlap and altitude.

// 02 · Fly
Verdant — live flight telemetry

Live position, battery, and sensor status across the flight.

// 03 · NDVI
Verdant — NDVI heatmap

Full-field NDVI map generated on-site after landing.

// 04 · Zones
Verdant — stress zone classification

Healthy, watch, stressed, critical — every zone GPS-located.

// 05 · Export
Verdant — prescription export

Zone coordinates exported for variable-rate input application.

click to expand · drag to explore
Closing

The
credits.

  • Engagement
    Fixed-price · 12-week program
  • Airframe
    Fixed-wing · 2.5m wingspan · 90-min endurance
  • Flight Stack
    PX4 · MAVSDK · grid mission planner
  • Payload
    4-band multispectral · radiometric calibration panel
  • Processing
    ROS 2 · OpenDroneMap · Python NDVI pipeline
  • Dashboard
    Next.js · Mapbox GL · raster tiling server
  • Output
    GeoTIFF orthomosaics · zone shapefiles · CSV prescription maps
  • Status
    In production · weekly survey schedule · 2 farm clients