Skip to content

Satellite Imagery Change Analysis

Drishti

An analyst workstation that compares Sentinel-2 satellite imagery of one area across two dates, flags candidate land and water changes, and records every review decision.

Drishti is a geospatial analysis workstation for comparing openly licensed Sentinel-2 satellite imagery of the same area on two dates, finding candidate changes to land and water, and documenting each analyst's review.

An analyst draws an area of interest, chooses two date windows and checks cloud cover. A background worker masks clouds and shadows, aligns the scenes, computes spectral differences and turns changed pixels into polygons with measured areas. The analyst then compares before and after on the map and accepts, rejects or flags each candidate, and every decision is kept in an append-only history.

Detection is classical and deterministic — spectral indices, change-vector analysis and robust thresholds — so the same inputs always give the same result. Every run records its scenes, parameters, software versions and a checksum of each output.

Drishti is evaluation-grade and not certified for operational use. Detected changes are candidates for an analyst to interpret, not confirmed findings, and no accuracy figures are claimed until it has been validated against approved reference data.

Core Capabilities

  • 01Area of interest by drawing, GeoJSON upload or coordinates, with Sentinel-2 L2A scene search across two date windows
  • 02Cloud, shadow, cirrus and snow masking, with each run's valid-pixel share and co-registration shift reported
  • 03Change detection from spectral indices (NDVI, MNDWI, BSI, NBR) and multi-band change-vector analysis, with robust or Otsu thresholds and a minimum mapping unit
  • 04Before/after swipe and side-by-side comparison, true- and false-colour composites, and a 10 m pixel inspector
  • 05Candidate review — accept, reject or follow up — with reasons, append-only notes and a full audit log
  • 06Versioned offline report bundles, GeoPackage candidate exports and GeoTIFF change masks

Engineering & Architecture

  • React and MapLibre analyst workspace, FastAPI service and a background processing worker
  • PostgreSQL with PostGIS under least-privilege database roles; local Analyst and Administrator accounts
  • Imagery read on demand as cloud-optimised GeoTIFFs from the public Earth Search catalogue over an allow-list of hosts