Thermal Cameras for Drivers Who Cannot Rely on Headlights

Reading a thermal specification

• Resolution determines how much detail you can resolve. 256 × 192 is enough to see that a warm shape on the roadside is a person at 85 metres with a 3.6 mm lens. 640 × 512 resolves much finer detail at the same distance, which is what you need for identification rather than mere detection.
• NETD is thermal sensitivity. Lower is better; ≤ 50 mK is a good commercial target. A high NETD camera produces a mushy image in mild weather even though the resolution number looks fine.
• Lens focal length trades field of view against distance. Wide (3.6 mm) suits low-speed manoeuvring and near-field detection; narrow (10 mm) suits highway and long-range work.
• Pixel size tells you about the detector generation. 12 μm is current mainstream; larger pixels generally mean an older, bulkier detector.
• Defrost is not a luxury in a cold climate. A thermal lens ices like any other, and without a heater the camera is blind exactly when it is most needed.

Where thermal genuinely earns its place

• Unlit rural routes where pedestrians, cyclists and livestock appear beyond the headlight beam.
• Mining and quarry haul roads, where dust defeats both headlights and IR cameras.
• Rail and depot shunting, where detecting a person on the track at long range is a safety requirement.
• Agricultural machinery working at night in fields where animals and people may be present.
• Marine and port operations for finding people and vessels in darkness and haze.

What thermal cameras are not good at

Thermal images are low-resolution compared with visible-light video, they show heat rather than detail, and they cannot read a number plate or identify a face. They work best as a second, complementary channel alongside a normal camera — one channel for detail, one for detection. In warm weather the thermal contrast between a person and the background drops, which is why AI classification matters more than raw image appeal.