Infrastructure starts moving long before anyone can see it.

Finium turns open satellite data into continuous measurement. We track railway embankments, bridges and the ground around them at millimetre scale — without a site visit, and without waiting for a clear sky.

Book a demo See the use cases

Sentinel-2 image of open terrain, October 2021

Sentinel-2, October–December 2021: military infrastructure appears and expands near Yelnya, Russia. Four passes out of an archive that updates every few days.

Passes over a single point in one month

Sample point in Poland, 31 days

Sentinel-2
Landsat 8/9

Sentinel-1
NISAR

Combined

potential observation no observation

≈20+

observations per month over any point on the map, using open Copernicus and USGS constellations alone. The data is already public — the hard part starts when you turn it into a decision.

A satellite image is a matrix of measurements, not a picture

Stacking observations across time, spectral bands and sensors gives us a data cube in which every pixel carries its own history. Instead of comparing two images, we analyse a trend.

Data cube: latitude and longitude on the base, time along the vertical axis — Latitude and longitude on the base, time along the vertical axis.

Four constellations feed the same stack

optical

Sentinel-2

13 bands at 10 / 20 / 60 m. Vegetation condition, moisture, built-up areas.

optical + thermal

Landsat 8/9

11 bands, 30 m multispectral, 15 m panchromatic, 100 m thermal infrared.

radar

Sentinel-1

C-band SAR, 5 × 20 m in IW mode, VV/VH polarisations. The backbone of InSAR measurement.

radar

NISAR

L- and S-band SAR, 7 m azimuth, 2–8 m range, up to quad-pol channels.

Radar keeps measuring when optics cannot

An optical sensor is passive: it needs daylight and a clear sky, and over this part of Europe most passes come back as cloud. Radar illuminates the ground with its own signal, so it works at night and straight through weather. That is the difference between seasonal sampling and continuous measurement.

Optical image of the coast, largely obscured by cloud — Optical — the same coastline, most of it under cloud.

SAR image of the same coast, fully visible — SAR — the same pass, the whole scene usable.

What two radar passes give us

Displacement

Vertical movement of a few millimetres, derived from the phase difference between passes.

Coherence

Pixel stability between two observations — the signal that something on the ground has been disturbed.

Elevation model

A digital elevation model reconstructed from radar pairs and kept current with the rest of the series.

From measurement to a maintenance decision

Three areas where the radar archive answers the question faster and cheaper than a field survey.

Railway line on an embankment with slopes and toe zone marked — Track bed, left and right slopes, embankment toe — each is measured separately along the corridor.

Railway embankment stability

Repeated Sentinel-1 observations reveal settlement trends along an entire corridor before instability becomes an operational risk. The result tells you where to measure on the ground — it does not replace that measurement.

deformation monitoring along the line, kilometre by kilometre
classification of stable and unstable persistent scatterers
settlement that matched the in-situ RILA measurements

Chang L., Sakpal N. P., Oude Elberink S., Wang H., “Railway Infrastructure Classification and Instability Identification Using Sentinel-1 SAR and Laser Scanning Data”, Sensors 20(24), art. 7108, 2020.

Aerial view of a motorway bridge and its approaches — Deck and superstructure, piers and abutments, the ground under the foundations, approaches and embankments.

Bridge assessment at network scale

Instead of reviewing structures one at a time, we screen the whole network for abnormal displacement patterns and rank what deserves an engineer's attention first.

displacement trends in mm/year for every structure
shortlist of structures for detailed inspection or SHM instrumentation
prioritised ground validation

Malinowska D., Milillo P., Reale C. et al., “Global geo-hazard risk assessment of long-span bridges enhanced with InSAR availability”, Nature Communications 16, 9048, 2025.

A reference satellite tile next to two tiles found by similarity search — A reference tile on the left, the closest matches found across the archive on the right, with their similarity scores.

Change detection from satellite embeddings

Vector representations of image tiles let you search by similarity instead of browsing scenes by hand. You point at a reference area, and the system finds its counterparts — and the moment they stopped matching.

search for locations similar to a given reference tile
alerts when new development or infrastructure appears
works across a global archive, not just the latest scene

Czerkawski M., Kluczek M., Bojanowski J. S., “Time, space, and nightlights: Global evaluation of Major TOM Earth embeddings”, EGU General Assembly 2026, EGU26-13425.

Define one area of interest, the rest runs itself

Today an analyst loses hours to scene search, pansharpening, writing intermediate results to disk and merging them back into a single layer. Finium closes that loop into one workflow built around a defined AOI.

Step 1

Define the AOI

Mark the asset you want to watch — a bridge, a section of track, a sensitive area.

Step 2

Choose the measurement

Select the parameters, the data layers and the models to run on incoming data.

Step 3

Subscribe to new data

Every new pass over your AOI triggers processing and a notification.

Step 4

Work in the report

The interactive report stays in the project and is shared with the whole team.

Open data first, commercial tasking when it matters

The open constellations give continuous baseline coverage. Where a moment needs higher resolution, the same pipeline ingests commercial sources and drone imagery, and delivers them through one searchable catalogue with an API.

Copernicus
USGS
ICEYE
Airbus
drone imagery
searchable catalogue with an API

Sovereign, open and cloud-native architecture

We build on open standards and cloud-native formats, with no dependency on a single vendor. Data stays where it has to stay, including deployment inside the customer's own infrastructure.

Python
GDAL
STAC
Zarr
PostGIS
GeoServer
QGIS
Kubernetes
Linux
OGC standards

Roadmap

We start with one measurable use case and widen the scope only once operational value is proven.

Phase 1 — in progress

A working demo

Railway embankment stability monitoring on a real section of track, with validated results.

Phase 2

Validation with partners

Present the product to PKP PLK, PKP Cargo, ORLEN and the Polish defence sector, and secure the commercial path forward.

Phase 3

Scale across the rail domain

Extend to bridges, stations, depots and risk monitoring along entire corridors.

Tell us about your corridor

Send us the extent of your area and we will come back with what the available radar archive covers, and what it can already tell you about your assets.

Get in touch

Marcin Niemyjski

GIS and remote sensing

mniemyjski@finium.dev

Rajmund Matusik

Software engineering

rmatusik@finium.dev

Grzegorz Pustulka

Software engineering

gpustulka@finium.dev