A drone over a highway incident does three jobs no ground unit can do quickly: it shows the whole scene at once, it finds ejected occupants in darkness or roadside brush, and it reads the temperature of anything that might be about to fail. The payload requirement follows directly from those three.
Key takeaways
- Three jobs — scene documentation, casualty search and hazard assessment — each demand a different sensor emphasis.
- The first fifteen minutes decide the value of the flight; readiness and launch time matter more than payload specification.
- For hazmat, radiometric measurement is mandatory, not optional — a colour image cannot support a containment decision.
- The comparison is against ladder trucks and lane closure time, which is where the staffing and cost argument is actually settled.
On this page
Three jobs, three sensor requirements
Scene geometry. Command needs the whole picture: vehicle positions, debris field extent, blocked lanes, approach sight lines. This is a wide-FOV job and it is usually done in the first 90 seconds of flight.
Occupant search. Ejected occupants end up in ditches, brush and median vegetation where headlights and torches fail. Thermal finds a body-temperature target against cool roadside ground almost instantly — this is the single highest-value drone task at night-time crashes.
Thermal state of the wreck. A hot battery pack, a heating pressure vessel, a tanker shell warming toward relief-valve pressure — all of these are visible in LWIR long before they announce themselves. Radiometric measurement turns “that looks hot” into a number that supports an evacuation decision.
The first fifteen minutes
This sequence is what we see working on well-run programmes. Times are from drone launch, not from dispatch.
| T+ | Action | Sensor mode | Output to command |
|---|---|---|---|
| 0–2 min | Launch, climb to 90–120 m, wide orbit | Thermal wide FOV | Scene overview, lane blockage, vehicle count |
| 2–5 min | Perimeter sweep 50–100 m either side of carriageway | Thermal wide, slow lateral | Ejected occupants, secondary casualties |
| 5–8 min | Close orbit on involved vehicles | EO zoom + thermal PiP | Plate and placard read, occupant status |
| 8–12 min | Radiometric spot on tanks, packs, cargo | Radiometric thermal | Temperature trend, evacuation radius |
| 12–15 min | Approach and egress routes, traffic queue extent | EO zoom + wide thermal | Diversion planning, queue length |
| Ongoing | Static overwatch from downwind standoff | Thermal + zoom on demand | Change detection, crew accountability |
Hazmat specifics
For a placarded load, the drone earns its cost in the standoff it provides. Fly downwind-offset, not directly overhead, and keep the aircraft outside the projected failure radius. Read the placard with EO zoom from 150–300 m rather than closing in.
Radiometric thermal matters here more than anywhere else in traffic work. Absolute surface temperature on a tank shell, tracked over time, is decision-grade information: a rising trend on the vapour space is the classic precursor to a BLEVE. Non-radiometric imagery shows you it is hot; it will not tell you it is 68 °C and climbing 3 °C per minute. The difference is covered in radiometric vs non-radiometric thermal.
Be explicit about emissivity. A polished stainless tanker shell has low emissivity and will under-read badly if the payload is left on the default 0.95 setting. See emissivity in drone thermography.
Recommended UAVThermal payloads
OP-90A — multi-sensor pod with thermal, 30x zoom and laser ranging. The all-round choice: one payload does scene overview, plate and placard read, occupant search and target coordinates.
MV-2P — dual-sensor micro pod for rapid-deploy vehicles. Small enough to live in a supervisor car and be airborne before the second engine arrives.
LX-6U — long-reach zoom pod when the requirement is standoff observation of a hazmat scene rather than close search.
Related reading
Technology: radiometric temperature measurement and starlight and low-light imaging.
Field practice: how far a drone thermal camera detects a person, using PiP and fusion display modes and building a public safety drone programme.
Adjacent missions: firefighting and emergency response and event and crowd safety.
Cost, staffing and the case to a traffic authority
Highway drone programmes are almost always justified on lane-closure duration rather than on search outcomes. Every minute of carriageway closure has a quantified cost in most jurisdictions — congestion, secondary collisions, freight delay — and a drone that shortens scene documentation by 15 minutes produces a number a traffic authority already knows how to value.
The staffing model matters more than the hardware. A payload in a supervisor vehicle that is already dispatched costs nothing extra to deploy; a dedicated drone unit that has to be separately called adds response time and headcount. The programmes that survive budget review are bolted onto an existing role.
Reconstruction photogrammetry is the second line of value. The same aircraft that flew the thermal sweep can produce a scaled site model in eight minutes, replacing an hour of total-station work and letting the carriageway open sooner.
Where this goes wrong
Highway work has failure modes that other missions do not.
- Flying overhead on a hazmat call. Directly above a pressure vessel is the worst place for the aircraft and, if it falls, for the responders. Offset downwind and accept the oblique view.
- Rotor wash on light debris. Descending to 15 m over a scene scatters evidence and paperwork. Keep documentation passes at height and use zoom.
- Assuming the thermal reading is calibrated. A polished tanker shell with default emissivity under-reads badly. If the number feeds an evacuation decision, set emissivity or read comparatively against a known surface.
- No airspace agreement. Motorway incidents attract medical helicopters. A programme without a pre-agreed deconfliction procedure will be grounded at exactly the incidents where it is most useful.
FAQ
Can a drone read a licence plate or hazmat placard at night?
A placard, usually yes — the characters are large and retroreflective, and a 30x zoom EO channel reads them from 150–300 m with NIR illumination. A licence plate at night is much harder; expect to need to close to under 60 m and accept that motion blur and plate reflectivity often defeat it.
How high should the drone sit over an active carriageway?
90–120 m AGL gives full scene coverage while staying clear of ground operations and most rotor-wash concerns. For hazmat, move to 150–300 m and offset downwind rather than overhead.
Is thermal better than a spotlight for finding ejected occupants?
Substantially. A spotlight illuminates a narrow cone and is defeated by vegetation shadow; thermal reads the whole swath at once and picks up a 33 °C body against 12 °C roadside ground with no ambiguity. Thermal is the search sensor, and the spotlight is what you use once you know where to send the crew.