Industries / Industrial equipment and services / Industrial services / Thermography
Drones, aircraft and robots
This work carries the heat camera on aircraft, helicopters, drones and robots, so that whole towns, power lines, buried pipe networks and solar farms can be surveyed faster than a person can walk them.
5
Milestones achieved, from 1977 to 2026.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1977
A town sees its roofs
In the mid-1970s Canadian government agencies flew airborne thermal scanners over towns to find heat escaping from buildings. The Canada Centre for Remote Sensing (CCRS) reported that three years of flights had found damaged flat roofs, faulty heating lines, and heated ramps and sidewalks left running after the snow had melted. The Ontario Centre for Remote Sensing (OCRS), flying since early 1975, settled on 480 m above the ground for its scans. For May 1977 it and the Ontario Ministry of Energy planned to show the thermal images of Lindsay, Ontario (population 12,872), at an interpretation centre where each householder could see the image of their own house and have it explained, which the paper calls the first large-scale community thermography presentation in Canada.
Credit
J. Cihlar and R. J. Brown (CCRS), G. Lawrence (OCRS), J. N. Barry (Philip A. Lapp Limited) and R. B. James (Canada's Office of Energy Conservation)
1981
Wind spoils the pictures
NASA's Lewis Research Center surveyed all of NASA's field centers from the air and estimated first-year net savings of $386,000. City demonstrations of 3 to 5 square miles followed, mostly tied to the US Department of Housing and Urban Development's winterization program. Many of the images, NASA's and private contractors' alike, could not be read: buildings blended into the background, known roof losses did not show, and users began to question the method. NASA engineers John R. Jack and Robert L. Bowman modelled how wind and air temperature change roof temperatures under clear skies. They found that above a surface wind of about 15 miles per hour (24 km/h) a small heat loss could appear as a large one and the reverse, and drew a map of the conditions under which flights give readable images.
LessonA heat survey is only as good as the weather it was flown in, so the conditions have to be chosen in advance and stated in the report.
Credit
Jack and Bowman, NASA Lewis Research Center, for the US Department of Energy
2013
Software reads pipe flyovers
District heating networks carry hot water from a central plant to buildings through pipes under the streets, and a leak or failing insulation warms the ground above it. Ola Friman and Peter Follo of the Swedish Defence Research Agency, Jörgen Ahlberg of Linköping University and Thermal Systems Technology, and Stefan Sjökvist of Thermal Systems Technology, a Swedish company, published methods that analyse aerial thermal images automatically to find leaking pipes, with results from large-scale leak surveys in several cities in Sweden and Norway, in IEEE Transactions on Geoscience and Remote Sensing (online November 2013).
LessonIn 2024 Elena Vollmer and colleagues at the Karlsruhe Institute of Technology (KIT) noted that the published leak-finding methods all report high scores, each on a very different set of images, which made a true comparison impossible.
Credit
Friman, Follo, Ahlberg and Sjökvist
2017
Solar farms get rules
A solar module with a cracked cell, bad soldering or a disconnected string shows a heat pattern unlike its neighbours', so a thermal camera can find faults across a whole plant while it is producing power. On 15 June 2017 the International Electrotechnical Commission (IEC) published Technical Specification 62446-3, which sets requirements for the camera, the ambient conditions, the inspection procedure, the report and the qualification of the people doing it, with a table of thermal abnormalities to guide the calls. A March 2018 review by Task 13 of the International Energy Agency's photovoltaic programme (IEA PVPS) estimated that a 4 MWp plant (megawatts peak, its rated output) takes about 15 to 20 hours to survey with a hand-held camera and 5 to 10 hours by drone, and that drones can image 2 to 3 MWp, typically up to about 12,000 modules, an hour.
Credit
IEC Technical Committee 82; IEA PVPS Task 13, lead authors Ulrike Jahn and Magnus Herz of TÜV Rheinland
2026
Heat patrols in wildfire plans
Infrared inspection shows abnormal heating in transformers, splices and connections that the eye cannot see. A February 2026 review of US utilities' wildfire mitigation plans, written by Pacific Northwest National Laboratory (PNNL) for the US Department of Energy, lists it among the practices in use: Redding Electric Utility patrols its overhead lines with infrared on a five-year cycle and flies drones to hard-to-reach spots, San Diego Gas & Electric gathers infrared data from helicopters, and Bear Valley Electric Service flies drones whose infrared images engineers review before ordering repairs. The review adds that utilities are still weighing the reliability, costs and false alarm rates of these tools, and that infrared surveys can need trained operators or consultants.
Credit
the utilities named; compiled by André Coleman and colleagues at PNNL (report PNNL-38528)
Watch for
U-values from the air
A U-value is the rate at which heat passes through one square metre of wall or roof for each degree of difference between inside and outside, in watts per square metre per kelvin (W/m²K); it is usually measured with heat flux sensors fixed to the inside of a wall for several days. Kestrix, a UK company, says it flies off-the-shelf drones with thermal cameras 30 to 60 m above homes, stitches the images into a 3D model and estimates U-values without going inside. Its co-founder Lucy Lyons told Ingenia that the company is working with academic partners to validate the accuracy, first for certain housing types. An Innovate UK project with Kestrix (£240,000) captured data from 104 properties.
Achieved whena university or test laboratory publishes Kestrix's estimates beside heat flux sensor measurements on the same walls, with the error.
Claimed by
Kestrix, September 2025
Robot dog on night rounds
Since mid-2024 Cargill's Amsterdam Multiseed oilseed plant has run a pilot of Boston Dynamics' four-legged Spot robot carrying a thermal camera, an acoustic sensor and a pan-tilt-zoom camera. A technician teaches it a route by driving it once; it then walks the route alone, each mission taking about two hours, on the second and third shifts on weekdays and on all shifts at weekends, and staff review its findings in the morning. The plant says the robot's repeated readings caught a ventilator that overheated at one point in the day, which it would not have found otherwise; manual inspections continue alongside.
Achieved whenCargill or an outside party publishes how many faults the robot rounds found and missed against the manual rounds over the same months.
Claimed by
Boston Dynamics, in a case study with Cargill, June 2025
Open
01
Flyover calls against repair records
For one district heating network that has been flown with a thermal camera, a public table of every hot spot the flight analysis flagged, each matched to what the network operator later found there from its own excavation, pressure test or repair records, coded as leak, failing insulation, another heat source (a manhole, a car, a building vent) or nothing found. Add the leaks the operator repaired in the same period that the flight did not flag, and a scoring script that grades any detection method on the same images.
The briefCredit · unclaimed02
Drone U-values against heat meters
A public set of homes in which wall U-values estimated from drone thermal flights sit beside U-values measured on the same walls with heat flux sensors over at least 72 hours, as ISO 9869-1 sets out, with the weather during each flight logged, and a table of the error for each common way of turning the images into a U-value.
The briefCredit · unclaimed
Open 01 · the brief
Flyover calls against repair records
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
For one district heating network that has been flown with a thermal camera, a public table of every hot spot the flight analysis flagged, each matched to what the network operator later found there from its own excavation, pressure test or repair records, coded as leak, failing insulation, another heat source (a manhole, a car, a building vent) or nothing found. Add the leaks the operator repaired in the same period that the flight did not flag, and a scoring script that grades any detection method on the same images.
Why it is still open
Elena Vollmer's 2026 KIT dissertation, which published several drone image sets for leak detection, says that despite diverse data "there are few confirmed ground truth leaks", so the work measured how well methods find thermal anomalies rather than leaks. A December 2025 study of a 64 km network in Nuremberg, Germany, by Thomas Risch, Alexander Weidauer and Ahmad Hadavand checked its results by comparing two operators and an expert's review (the expert judged 384 of 908 conventional findings false), not against digs.
Built with what exists
An operator's existing flight images and its work-order and repair records; KIT's open code and image sets on Zenodo as a baseline; QGIS to match flagged spots to pipe maps and work orders; Python for scoring. The reader handles records only; no digging and no work on the network.
First step
Ask one district heating operator that has flown its network in the last three winters for its list of flagged sites and the repair records for the same streets.
Done looks like
At least 200 flagged sites on one network traced to an outcome, the share confirmed as leaks given for each severity class, the number of repaired leaks the flight missed, and the table and images released under an open license.
Open 02 · the brief
Drone U-values against heat meters
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of homes in which wall U-values estimated from drone thermal flights sit beside U-values measured on the same walls with heat flux sensors over at least 72 hours, as ISO 9869-1 sets out, with the weather during each flight logged, and a table of the error for each common way of turning the images into a U-value.
Why it is still open
Xiaojia Zhang, Elena Lucchi and Andrea Garzulino (Buildings, June 2026) compared four formulas from one drone case study with a heat flux reference and found "significant discrepancies" that depended on wind, air temperature and surface temperature; in their case simple formulas gave physically impossible negative U-values. Kestrix says its own validation with academic partners is under way. In EDF's Project Sojourner, a UK government-funded trial with Kestrix, 79 homeowners had drone thermal surveys and a heat loss report, and the published case study (November 2025) gives no accuracy figures. Searches in November 2026 found no public set that pairs drone and heat flux measurements across many homes.
re.public.polimi.itingenia.org.ukassets.publishing.service.gov.uk
Built with what exists
An off-the-shelf drone with a radiometric thermal camera, flown by a licensed pilot under the local rules (the open category in the UK and EU, Part 107 in the US) with each owner's consent; heat flux sensor kits sold for U-value measurement, such as greenTEG's gSKIN kit; a small weather station; Python for the calculations.
First step
Fit heat flux sensors to two walls of one volunteer house for a week and have the house flown on two cold nights in that week.
Done looks like
20 volunteer homes, two walls each, every wall measured with heat flux sensors and flown on at least three nights; the median and 90th-percentile error in W/m²K for each method published; all images and readings released.