Skip to main content

​
Drone Pipeline

agronome.ai processes drone imagery through an automated pipeline that takes raw images from your drone and produces georeferenced vegetation index maps, elevation models, and orthomosaics.

​
Pipeline stages

Upload → EXIF Analysis → ODM Processing → Post-Processing → Tile Generation → View in Analyze Tab

​
1. Upload

Upload drone images through the web interface. Images are uploaded directly to Google Cloud Storage using signed URLs — large uploads never pass through the API server. Supported cameras:
  • DJI Mavic 3 Multispectral (multispectral + RGB)
  • DJI Phantom 4 Multispectral
  • MicaSense RedEdge-MX / Altum
  • Any RGB camera (orthomosaic only)

​
2. EXIF analysis

Once upload completes, agronome.ai automatically extracts EXIF metadata from every image:
  • Camera identification — model, sensor type (RGB vs. multispectral), band configuration
  • GPS coordinates — flight area, altitude, ground sampling distance (GSD)
  • Band detection — identifies available spectral bands (Red, Green, Blue, Red Edge, NIR)
  • Irradiance data — DJI irradiance values for radiometric calibration
This determines which vegetation indices can be generated. A multispectral flight with NIR and Red Edge bands enables NDVI, NDRE, GNDVI, SAVI, and EVI. An RGB-only flight produces orthomosaics and elevation models.

​
3. ODM processing

Raw images are processed by OpenDroneMap (ODM) running on GCP Batch with Spot VMs:
  • Orthomosaic — georeferenced, stitched aerial image
  • Digital Surface Model (DSM) — elevation data for canopy height modeling
  • Point cloud — 3D reconstruction of the flight area
Processing runs on cost-efficient Spot VMs with automatic retry on preemption.

​
4. Post-processing

After ODM completes, the post-processor generates:
  • Vegetation indices — NDVI, NDRE, GNDVI, SAVI, EVI calculated from multispectral bands
  • Canopy Height Model (CHM) — derived from DSM
  • COG conversion — all rasters converted to Cloud-Optimized GeoTIFF
  • Color relief — gdaldem color-relief applies scientifically accurate colormaps

​
5. Tile generation

Pre-rendered PNG tiles are generated using gdal2tiles.py at standard web map zoom levels. Each layer gets its own tile set: 
{org_id}/{job_id}/tiles/{layer_type}/{colormap}/{z}/{x}/{y}.png
Tiles are served as static files from GCS through an authenticated API proxy. No dynamic tile server — tiles are generated once and cached permanently.

​
6. Viewing

Processed layers appear in the Analyze tab on the field detail page. You can:
  • Switch between layer types (NDVI, NDRE, orthomosaic, elevation, etc.)
  • Toggle colormaps for vegetation indices
  • Compare layers side-by-side
  • View a timeline of all drone flights and satellite passes for the field

​
Job statuses

StatusMeaning
pending_uploadSigned URLs generated, waiting for image upload
uploadingImages being uploaded to GCS
analyzingEXIF metadata extraction in progress
readyAnalysis complete, queued for ODM processing
processingODM stitching running on GCP Batch
post_processingVegetation indices and tiles being generated
completedAll layers available for viewing
failedProcessing error (check logs for details)