Turn imagery, survey records, and operational data into maps with traceable evidence.
The operating problem
An attractive map is insufficient when an engineer needs to know when an asset was surveyed and how reliable its position is. We connect capture, quality checks, registries, and field workflows so each object has an accountable source.
See the mechanism
Compare two captures. Flag, then verify.
Change detection compares observations of the same ground over time. The output is a list of candidates with their evidence, and ground hidden by cloud is marked as not assessed.
Change detection · synthetic scene
C1
New structures on parcel P-102
Bare soil and six new roofs where a crop field was. Queued for a permit check.
Field check
C2
Canopy loss, north edge of P-104
Tree cover replaced by bare ground. Needs a field visit before anyone calls it clearing.
Slide to view
C3
New unpaved track across P-105
A linear change joining the main road. Could be a track or a drainage cut; unconfirmed.
Field check
M1
Cloud over P-105 south
Masked in the later capture. No change is claimed until a clear acquisition exists.
Not assessed
A drawn scene, not real imagery. Slide to compare the earlier and later captures. A detected change is a candidate for review, and cloud-covered ground is masked rather than guessed.
What we build with you.
01
Capture & constraints
Existing-record review, imagery selection, survey requirements, permissions, and an explicit accuracy budget.
Web maps, change detection, asset registries, routing, and GIS/BIM integration appropriate to the workflow.
04
Handover
Documented exports, catalogues, maintenance responsibilities, and the evidence needed to investigate a disputed object.
Standards we build on
Open catalogues, stated accuracy.
We catalogue imagery and derived layers with the SpatioTemporal Asset Catalog (STAC) specification, which gives a common structure for describing and cataloguing spatiotemporal assets.1 A STAC API search filters by collection, bounding box or geometry, and time range, so every derived map can be traced back to the scenes that produced it.2
Copernicus Sentinel-2 samples 13 spectral bands at 10, 20, and 60 metre resolution and is designed to revisit the same point every 5 days at the equator.3 That suits land-cover and vegetation change across a district. Parcel or asset work usually needs higher-resolution tasking, drone capture, or field survey.
Before comparing captures we mask cloud, cloud shadow, and cirrus with the scene classification map produced for each Level-2A product.4 A scene-level cloud percentage cannot tell you which pixels over your site are clear.
Pixel size is not positional accuracy. The ASPRS standard sets accuracy thresholds independent of ground sample distance and requires at least 30 checkpoints for a fully compliant assessment.5 Rasters are delivered as Cloud Optimized GeoTIFF, an OGC standard that uses HTTP range requests so a client downloads only the parts of an image it needs.6
For technical reviewers
Every analysis starts from a reproducible catalogue query. This one asks the public Earth Search STAC API for Sentinel-2 Level-2A scenes over central Dhaka with under 10 percent estimated cloud.2,3 The cloud filter uses the STAC Query extension, which this endpoint supports; clear pixels are then chosen with the scene classification layer.4
Choose one area and reconcile the available records before commissioning additional capture. Specialist field delivery is scoped with the relevant partners.
Pilot area or asset class
Available surveys, as-builts, and data licences
Required accuracy and decisions the map must support
Sources
Evidence and further reading
Primary sources for the standards and practices referenced on this page. They describe the field, not Verne's own results.
STAC API: Item Search (opens in a new tab)radiantearth/stac-api-spec on GitHub · v1.0.0, April 2023The /search endpoint and its core parameters: collections, bbox or intersects, datetime, and limit.
03
S2 Mission (opens in a new tab)Copernicus SentiWiki (ESA) · undated13 spectral bands at 10, 20, and 60 m, a 5-day design revisit at the equator, and Sentinel-2C replacing 2A in January 2025.
04
S2 Processing (opens in a new tab)Copernicus SentiWiki (ESA) · undatedThe Level-2A scene classification map, with classes for cloud, cloud shadow, and thin cirrus.
In 7–10 working days, Verne maps your workflows, data sources, repetitive decisions, automation opportunities, and AI risk areas. You receive a prioritized roadmap showing what to automate, integrate, avoid, and build first.