The short definition

Thermal cameras use a microbolometer sensor sensitive to long-wave infrared (LWIR), typically the 8 to 14 micron band. Every object above absolute zero emits LWIR proportional to its temperature. The sensor reads that radiation and produces an image where intensity correlates with heat: humans, vehicle engines, and electrical hotspots appear brighter than the cooler background.

Because thermal works on emitted radiation, it needs no illumination. Pitch dark, fog, light smoke, and thin foliage don't degrade the image the way they degrade visible-light cameras. That's the advantage, and the only reason thermal earns its 3 to 10x price premium over a visible-light bullet.

Thermal vs IR night vision

Both work at night, but they aren't the same thing.

  • IR night vision. A visible-light camera with a built-in 850nm or 940nm LED illuminator that floods the scene with IR; the sensor captures the reflection. Range limited by LED output, typically 30 to 150 meters. The image is a grainy black-and-white photograph.
  • Thermal. No illuminator; the sensor detects emitted LWIR. Range limited by lens focal length, typically 300m to 2km for human detection. The image shows heat as intensity, not a photograph.

Most enterprise installs that go thermal also keep visible-light cameras for face and license-plate ID. Thermal triggers, visible confirms.

Detection range and lens choice

Thermal lens focal length, paired with sensor resolution (typically 320x240, 640x480, or 1280x1024), drives detection distance. NATO STANAG 4347 defines four target categories: detection (something is there), recognition (it's a human), identification (a specific human), and forensic (court-quality). For security:

  • 9mm lens, 320 sensor. Human detection at 250m. Recognition at 80m. Common at building-perimeter installs.
  • 19mm lens, 640 sensor. Human detection at 800m. Recognition at 250m. Common at industrial perimeter and distribution centers.
  • 50mm lens, 640 sensor. Human detection at 2km. Recognition at 600m. Common at substations, ports, and large perimeter walks.

Where thermal earns its premium

  • Long perimeter fence lines. A 1km perimeter would need 6 to 10 visible-light bullets with IR illuminators to cover. Two thermal cameras with 19mm lenses cover the same perimeter, with better fog and rain performance.
  • Substations and critical infrastructure. NERC-CIP-driven installs at electrical substations. Thermal detects humans approaching from any angle and routes to a verified-monitoring center. See the NERC CIP context in the main glossary.
  • Distribution-center perimeter. Yards typically run 5 to 20 acres with limited lighting. Thermal plus AI human/vehicle analytics keeps false alarms low and detection range high.
  • Ports, channels, and waterfront. Visible-light cameras lose detail across water due to humidity and reflection. Thermal sees heat signatures through both.
  • Wooded and rural sites. Foliage hides intruders from visible cameras. Thermal sees through light brush; humans appear as bright spots against cooler vegetation.

Pairing thermal with AI analytics

Raw thermal triggers on any heat anomaly: a deer, a raccoon, a fence hot from sun. AI analytics on the thermal stream classify the anomaly as human, vehicle, animal, or other before alarming. That filter cuts false positives dramatically and is the difference between a thermal install operators trust and one they ignore.

Common analytics platforms that handle thermal: FLIR's own United VMS analytics, Genetec restricted-area analytics, and camera-agnostic platforms like Dragonfruit AI and Briefcam. Some thermal cameras (Axis Q1942, Hanwha TNO) ship with built-in classification analytics that don't require external server hardware.

When to ask Tec-Tel about thermal

Thermal is the right answer when the perimeter exceeds 200m, the lighting is bad, or false positives are killing your operators. We'll do a perimeter walk, calculate camera-count and lens-focal-length, and pair it with verified monitoring.