Drones and robotics

Thermal inspection drone

It flies your site on a set grid and reads a real temperature in every pixel. Each hot spot gets pinned to the exact panel. You get back a graded defect map, not a memory card of pictures to sort.

Talk to engineering See all solutions

The problem

Heat is the first symptom.

A fault shows up as heat long before you can see it. A cracked solar cell, a busbar working loose, a wet patch of insulation under a flat roof: each one runs hotter than its neighbours for weeks or months first. From the ground that heat is invisible. A person with a handheld camera covers a small slice of a site in a day, reads every panel at a slightly different angle and hour, and hands back a folder of loose images nobody can hold to a standard. By the time the fault turns into lost output or a shutdown, the cheap moment to fix it has passed. What you actually need is the whole site measured the same way on the same morning, every hot spot placed on the right panel, and a report that says what is wrong and where. That is a scheduled flight, not a clipboard.

The measurement, at a glance

640 x 512 Radiometric pixels per frame

< 40 mK Thermal sensitivity, NETD

+/- 2 C Measured temperature accuracy

1 cm Geotag precision, RTK

The system

One flight, three jobs.

Imagery in production

It measures, it does not just look

A calibrated temperature in every one of 327,680 pixels, saved into the frame.

The payload is a radiometric long-wave infrared camera working the 7.5 to 13.5 um band, not a colour-mapped thermal viewer. Every one of its 327,680 pixels holds a calibrated temperature you can threshold and defend, saved right into the frame. So an anomaly is a number in Celsius, not a warm-looking blob. The uncooled vanadium oxide sensor reads to under 40 mK, enough to tell a cell running 3 C above its string apart from a reflection, and it holds +/- 2 C against a traceable blackbody reference. A 20 MP visual camera sits alongside it, bore-sighted, so every thermal hit comes with a photo of the actual part. The same head flies on every airframe in the range.

Imagery in production

The whole site, flown the same way every time

Draw the boundary once. The aircraft flies a set grid, at a set altitude, on every visit.

You draw the boundary once. The flight stack plans a serpentine grid at a fixed altitude and overlap, follows the terrain so the ground sample distance stays constant, and flies it the same way on every visit. Month on month, the scans line up, so a growing fault is easy to spot. RTK positioning tags every frame to within a centimetre. That is what lets the software drop an anomaly on the right panel out of tens of thousands, rather than somewhere in a row. Flights are booked for the sun angle and time of day that give a clean signal, usually clear sky with a load on the asset. No two operators, no two walking routes, no drift between inspections.

Imagery in production

A defect map, not a photo dump

Frames stitched into one thermal orthomosaic, every fault located, graded and exported.

Back on the ground the frames are stitched into a single thermal orthomosaic of the site. Analysis runs against the known fault signatures: one hot cell, a hot diode or substring, a hot junction box, an offline string, a wet roof bay, a hot electrical joint. Each hit is placed on the map by its RTK position, then graded by how far it rises above its neighbours and by the pattern it makes. The report names the fault, the temperature rise, the GPS location and the matching photo. You get a PDF for the file, a CSV for the work-order system and a GIS layer for the asset map. What lands on your desk is a repair list.

The kit

Configured for inspection.

Every part here is chosen for scheduled radiometric survey work.

Imagery in production

Airframe

Weather-rated inspection multirotor

35 min class

Sealed carbon body rated IP4x, steady in wind up to 12 m/s and light rain. Hot-swap smart battery, and low-noise props so it can fly close to occupied buildings and live sites.

Request datasheet

Imagery in production

Payload

Radiometric thermal and visual head

640 x 512 LWIR

An uncooled long-wave infrared array bore-sighted to a 20 MP visual camera. Radiometric to +/- 2 C, recording R-JPEG on board with a real temperature saved in every pixel.

Request datasheet

Imagery in production

Payload

RTK positioning module

1 cm geotag

Real-time kinematic GNSS tags every frame to the centimetre, so anomalies land on the correct panel and repeat scans line up between visits.

Request datasheet

Imagery in production

Ground

Ruggedised planning controller

Field ready

A bright-screen controller running the mission planner, terrain-follow grid, geofencing and a live radiometric preview, so a soft or reflective frame gets caught before you land.

Request datasheet

Imagery in production

Software

CalyOS Embedded inspection stack

Onboard + desk

Flight planning, thermal orthomosaic stitching, anomaly detection against IEC 62446-3 and EN 13187 signatures, and one-click export of the defect map and report.

Request datasheet

How a survey runs

From boundary to report.

Most sites are one flight in the morning and a report the same day. No memory cards to sort, no specialist to hire.

Draw the boundary once

You mark the site on the map and set the altitude and overlap. We save that plan, so every future visit flies the exact same grid.
Fly the grid on a clear morning

The aircraft follows the terrain and holds a steady path while it captures the whole site. A live radiometric preview means a bad frame is caught in the air, not back at the desk.
Get a graded defect map

The frames are stitched, every anomaly is located and ranked, and you receive a PDF, a CSV and a GIS layer. What you hand the repair crew is a list, not a folder.

What you get

A morning's flight, a day's repairs.

The deliverable

Report

You don't get a drone and a memory card. You get a scheduled survey turned into a graded list of defects. Every one is located to the panel or the metre, ranked by how serious it is, and comes with the thermal image, the matching photo and the temperature rise. Hand it straight to whoever holds the spanner.

Imagery in production

Every fault graded

Anomalies are ranked by temperature rise and pattern, so you deal with the string outage before the single warm cell.

Located to the panel

The RTK geotag drops each hit on the right module, or the right metre of roof, not just the right field.

Baselined over time

The same grid every visit means a fault that's growing shows up as a trend instead of a surprise.

Radiometric data kept

Every pixel keeps a real temperature in the exported file, so any finding can be checked again and defended later.

Questions

Asked and answered.

Can it fly in wind or light rain?

The airframe is sealed to IP4x and rated for wind up to 12 m/s and light rain, so a normal working morning rarely grounds it. A clean thermal signal still wants clear sky and a load on the asset, which is why flights are booked for the right conditions rather than just any slot.

How do you make sure two inspections match?

You draw the boundary once and every visit flies the same grid, at the same altitude, following the terrain. RTK positioning tags each frame to a centimetre, so this month's scan lines up with last month's and a growing fault is easy to see.

Which standards does the analysis follow?

Anomaly detection runs against the IEC 62446-3 and EN 13187 fault signatures, covering a hot cell, an offline string, a hot junction box, a wet roof bay and more. Each detection is graded by temperature rise and by pattern.

What do we actually receive?

A thermal orthomosaic of the whole site, plus a report that names each fault, its temperature rise, its GPS location and a matching photo. It comes as a PDF for the file, a CSV for the work-order system and a GIS layer for the asset map.

“A finding you can defend beats a picture that looks alarming. That is why every pixel we hand back holds a real temperature, and every defect comes with the spot it sits on.”

Our inspection principle

Ready when you are

Book the survey.

Tell us the site and what you need to prove, and we'll plan the flight and the report that comes with it.

Talk to engineering See all solutions

The payload

How it measures.

Imagery in production

Radiometric core

A temperature in every pixel

The sensor is a 640 by 512 uncooled vanadium oxide microbolometer sensitive from 7.5 to 13.5 um, the long-wave band where objects near ambient emit most strongly. Sensitivity is under 40 mK, so a cell a few degrees above its string stands clear of noise, and absolute accuracy holds to +/- 2 C or +/- 2 percent against a factory blackbody calibration. Each frame is saved as radiometric R-JPEG, so the temperature of any point is recoverable on the desk, not baked into a colour map.

Imagery in production

Visual pairing

The photo that proves the hot spot

A 20 MP visual camera sits on the same gimbal, bore-sighted to the thermal frame, so every anomaly comes with an ordinary photo of the exact part. That is what turns a warm blob into a work order: the roofer sees the wet bay, the electrician sees the scorched junction box, and nobody argues about which panel. Both channels ride a three-axis gimbal, so ground sample distance stays honest at survey altitude.

Imagery in production

Emissivity and calibration

Set for the surface, not the guess

A reading is only as good as the emissivity you tell it, so the software carries per-surface presets for glass, painted metal, bare metal, felt and membrane, with reflected-temperature compensation for a clear sky. The camera is calibrated across its range against a traceable blackbody source and re-verified on a schedule, so the numbers in the report survive scrutiny from an insurer or a warranty claim.

The measurement, at a glance

< 40 mK Thermal sensitivity, NETD

+/- 2 C Radiometric accuracy

16-bit Radiometric data per frame

30 Hz Thermal frame rate

Read across

One camera, many faults.

The same radiometric head reads a different fault signature on each asset class.

Solar

Cells, diodes, strings

Hot cells, hot bypass diodes, PID and whole offline strings show as distinct thermal patterns, scored against IEC 62446-3 delta-T bands.

Electrical

Joints and connections

Loose or corroded bolted joints, overloaded conductors and failing connections on lines and in substations run hot long before they fail.

Envelope

Moisture and heat loss

Trapped moisture under a flat roof and missing insulation in a facade read as thermal anomalies, assessed against EN 13187 building thermography.

Engineering questions

About the sensor.

Is it radiometric or just a thermal picture?

Radiometric. Every pixel stores a calibrated temperature in the saved file, so you can re-measure any point, change the emissivity and re-threshold on the desk. A colour thermal picture cannot do any of that.

How accurate are the temperatures?

Plus or minus 2 C or 2 percent against a traceable blackbody calibration, provided emissivity and reflected temperature are set correctly, which the software presets handle per surface. For fault-finding the delta between a part and its neighbours matters more than the absolute, and that resolves well inside a tenth of a degree.

Does weather stop the survey?

Thermography needs a thermal load and a clean sky, so solar is flown under good irradiance and roofs at the right hour. The airframe is rated to 12 m/s wind and light rain, but the physics, not the aircraft, sets the window.

“We stopped arguing with the maintenance contractor about where a fault was. The report has the panel ID, the photo and the temperature. They just go and fix it.”

Reliability engineer, a renewables operator

Flight & autonomy

It flies the grid for you.

You set the boundary and the schedule. The aircraft plans the coverage, holds the altitude and overlap, and flies the same path every time, so the data lines up between visits.

A single survey

From boundary to map.

Draw the site

Mark the asset boundary once on the planner and set altitude, overlap and the fault types you care about. The stack generates a serpentine grid and a geofence.
Fly the grid

The aircraft flies the route on RTK guidance, following terrain to keep ground sample distance constant, tagging every frame with a centimetre-accurate position.
Capture both channels

Thermal and visual frames are recorded together at each waypoint, with a live radiometric preview on the controller so a soft or reflective frame is caught before landing.
Stitch and detect

On the ground the frames become a thermal orthomosaic and anomaly detection runs against the fault signatures for the asset class.
Export the report

Graded defects drop onto the map with location, delta-T and photo, and export as PDF, CSV and a GIS layer for the asset system.

Autonomy

Repeatable beats heroic.

The value of an aerial inspection is not a spectacular one-off flight; it is the same flight, flown identically, every quarter for years, so that a fault which grows two degrees a month is caught while it is cheap. That only works if the aircraft removes the operator's variability. Grid altitude, overlap, speed and the sun window are fixed in the plan, not left to the person on the sticks. Terrain follow holds ground sample distance constant over a sloped array or a stepped roof. Geofences and no-fly volumes live in the flight stack and are enforced, not advisory. Link loss triggers a return-to-home, and a low battery ends the current lane cleanly and resumes after a hot-swap. The pilot supervises and keeps the site safe; the aircraft flies the survey the same way it did last time.

Per flight day

50 MW Solar capacity scanned per day

35 min Endurance per battery

12 m/s Rated wind tolerance

1 cm RTK frame position

Specify it with us

Map your assets.

Send us the site boundary and the asset class and we will scope the survey, the flight schedule and the report format. You keep the aircraft, the data and the licence outright.

Talk to engineering OEM & ODM

Reporting & data

The point is the report.

The drone is the cheap part. What you buy is a graded, located, defensible defect list that goes straight to the crew who fix it and the file that proves you inspected.

The pipeline

From frame to fix.

Imagery in production

Stitch

One map of the whole site

Thousands of geotagged frames are mosaicked into a single thermal orthomosaic, so instead of scrolling images you see the whole solar field, roof or line run as one map with real coordinates. Overlap and RTK position let the software place every pixel correctly, which is what makes the next step trustworthy.

Imagery in production

Detect and grade

Faults found and ranked

Detection runs against the known signatures for the asset: a single hot cell, a hot diode, a hot junction box, an offline string, a wet roof bay, a hot electrical joint. Each is graded by its temperature rise above the local baseline and by pattern, so a whole dead string outranks one warm cell and the crew works the list in the order that protects output.

Imagery in production

Deliver

In the format the work uses

Every anomaly carries a GPS location, a delta-T, a severity, the thermal crop and the matching visual photo. It exports as a PDF report for the record, a CSV that drops into a maintenance or work-order system, and a GIS layer for the asset map. Nothing has to be retyped for a job card.

What lands

Three outputs, one flight.

The same survey serves the crew, the system and the auditor.

For the crew

A located repair list

Ranked anomalies with panel ID or grid reference, photo and temperature, so a technician walks to the fault, not the field.

For the system

A CSV and GIS layer

Structured records that import into the maintenance system and drop onto the asset map without re-keying.

For the record

A defensible PDF

A dated report with radiometric evidence and method, the kind an insurer, an auditor or a warranty claim will accept.

The difference

Walk versus flight.

Scheduled drone surveyHandheld walk-round

Coverage Whole site in a morning, every panel measured A sample, whatever fits the day on foot

Consistency Same grid, altitude and sun angle each visit Different route, angle and operator each time

Location Every fault geotagged to 1 cm A note and a memory of roughly where

Output Graded defect map, CSV and GIS export A folder of loose images to sort by hand

Trend Visits line up, growth is visible Hard to compare one year to the next

Evidence Radiometric data kept per pixel Colour snapshots, the values not recoverable

Data questions

Who owns the data.

Where does the analysis run?

On your own desk or server. The stitching and detection run locally in the CalyOS inspection stack, so footage and asset data do not have to leave your network, and there is no per-report cloud meter.

Can we get the raw radiometric data?

Yes. You keep the radiometric R-JPEG frames, the orthomosaic and the exports. Every pixel's temperature is recoverable, so a third party can re-check a finding independently.

Does it fit our existing systems?

Exports are standard PDF, CSV and GIS layers, so anomalies drop into a maintenance or work-order system and onto the asset map without a proprietary viewer.

See a sample

Read a real report.

We will send a sample defect map and report from an anonymised survey so you can see the output before you commit. Then we scope it to your assets.

Talk to engineering See all solutions

The library

Briefs and cases.

The documents and worked examples behind the solution.

Brief

Imagery in production

Solution brief

The one-page case.

2 pages For decision makers

White paper

Imagery in production

Radiometric inspection white paper

Method and accuracy.

16 pages For technical evaluators

White paper

Imagery in production

Data and integration guide

For your engineers.

18 pages For integrators

Use case

Imagery in production

Solar farm survey

50 MW, one morning.

AssetPV portfolio Outputdefect map

Use case

Imagery in production

Flat-roof moisture survey

Before the re-roof.

Assetwarehouse roofs Outputwet-area map

Use case

Imagery in production

Substation and line survey

Hot joints, live asset.

Assetdistribution network Outputgraded joint list

Request the files

Get the documents.

The datasheet, the integration guide and the white paper, sent on request.

Thermal inspection drone datasheet Datasheet PDF, on request Data and integration guide Integration PDF, on request Radiometric inspection white paper White paper PDF, on request

How the pack is supplied

6 Documents and cases

3 yr Warranty standard

10 yr Spares and images held

0 Per-report cloud fees

Specify it with us

Start a survey.

Tell us the asset class and the site and we will scope the aircraft, the schedule and the report. You own the drone, the data and the software outright.

Talk to engineering OEM & ODM

Thermal inspection drone

Heat is the first symptom.

The measurement, at a glance

It measures, it does not just look

The whole site, flown the same way every time

A defect map, not a photo dump

Weather-rated inspection multirotor

Radiometric thermal and visual head

RTK positioning module

Ruggedised planning controller

CalyOS Embedded inspection stack

Draw the boundary once

Fly the grid on a clear morning

Get a graded defect map

Report

Every fault graded

Located to the panel

Baselined over time

Radiometric data kept

Book the survey.

A temperature in every pixel

The photo that proves the hot spot

Set for the surface, not the guess

The measurement, at a glance

Cells, diodes, strings

Joints and connections

Moisture and heat loss

It flies the grid for you.

Draw the site

Fly the grid

Capture both channels

Stitch and detect

Export the report

Repeatable beats heroic.

Map your assets.

The point is the report.

One map of the whole site

Faults found and ranked

In the format the work uses

A located repair list

A CSV and GIS layer

A defensible PDF

Read a real report.

Solution brief

Radiometric inspection white paper

Data and integration guide

Solar farm survey

Flat-roof moisture survey

Substation and line survey

How the pack is supplied

Start a survey.