Skip to content
ZINDORA

Technology

Where Earth Observation meets Artificial Intelligence.

Zindora builds the analysis layer: the processing, models and systems that turn openly available Earth Observation into something a decision-maker can use.

Earth Observation

  • Satellite imagery processing
  • SAR analytics
  • Multispectral analysis
  • Temporal change detection

Machine learning

  • Geospatial AI
  • Computer vision
  • Anomaly detection
  • Multimodal data fusion

Platform

  • Spatial analytics
  • Cloud-based processing
  • APIs and data services

Two sensors, two different questions.

Radar and optical imagery are not interchangeable and are not ranked. They fail in different conditions, which is precisely why using both is worth the complexity.

Comparison of synthetic aperture radar and multispectral optical Earth Observation.
Property Sentinel-1 C-band SAR Sentinel-2 Multispectral optical
What it measures Backscattered radar energy — surface roughness, geometry, dielectric properties Reflected sunlight across 13 spectral bands
Through cloud ✓ Yes. C-band passes through cloud and rain × No. Cloud blocks the scene entirely
At night ✓ Yes. The sensor supplies its own illumination × No. Requires daylight
Sees a vessel because A metal hull returns strongly against a rougher sea surface It is visually distinguishable from the water
Main limitation Sea state and wind change the clutter floor; small wooden craft return weakly Cloud cover, which on this coast is seasonal and persistent

The Kenyan coast has a long monsoon season. A monitoring approach that depends on a cloud-free sky is unavailable for much of the year, which is why radar carries the detection task and optical corroborates it when conditions allow.

The processing chain.

What happens between a satellite product and a statement someone can act on.

  1. 01

    Ingest

    Sentinel-1 GRD product retrieved for the area of interest

  2. 02

    Calibrate

    Digital numbers converted to radar backscatter (sigma nought)

  3. 03

    Speckle

    Multiplicative speckle noise reduced without erasing small targets

  4. 04

    Terrain

    Geometric correction and geocoding to a map projection

  5. 05

    Detect

    Targets separated from sea clutter against an adaptive threshold

  6. 06

    Associate

    Detections matched against AIS reports in space and time

  7. 07

    Contextualise

    Environmental, historical and geographic context attached

  8. 08

    Publish

    Evidence, confidence and provenance emitted together

Provenance is carried through every stage, not reconstructed at the end.

Earth Observation sources.

We exploit Copernicus Sentinel data to build downstream intelligence applications. The missions are operated by ESA and the European Commission; Zindora operates no satellites.

Sentinel-1

C-band SAR

SAR-based observation for all-weather, day/night monitoring.

Radar is unaffected by cloud and does not need daylight, which matters on a coast with a long monsoon season. Metal hulls return strongly against a rougher sea surface, which is what makes vessel detection possible at all.

  • Active radar
  • Cloud-independent
  • Day and night

Sentinel-2

Multispectral optical

Multispectral Earth Observation for land, coastal and environmental applications.

Thirteen spectral bands at 10, 20 and 60 metre resolution support water quality, coastal habitat and land-cover analysis, and give radar detections a visual corroboration when the sky is clear.

  • 13 spectral bands
  • 10-60 m resolution
  • Optical, needs daylight

Complementary sources

Open and commercial

Additional EO integrated where the application justifies it.

Higher-resolution commercial imagery, thermal and altimetry data can be brought in where a specific question needs them. The architecture treats sources as interchangeable inputs rather than building around any single one.

  • Source-agnostic design
  • Open data first

Built for where it runs.

Earth Observation is a heavy data problem. A single radar scene is gigabytes, and a useful time series is many of them. Architecture choices that are invisible on a fast connection decide whether a system is usable at all on a constrained one.

Move the compute
Process next to the archive and ship results, not scenes.
Open formats
Standards-based outputs, so results are usable outside our own tools.
Bandwidth budget
Interfaces designed for the connections users actually have.
APIs first
Intelligence that connects to existing workflows rather than replacing them.

Technical collaboration.

If you work on Earth Observation, geospatial AI or maritime data, we would like to hear from you.

Get in touch → See it applied →