Fog can reduce the effective detection range of perimeter thermal surveillance and make intrusion alarms less reliable. Thermal cameras may retain an advantage over visible-light cameras in some fog conditions, but sufficiently dense fog can leave too little target contrast for dependable detection. A clear-weather range specification therefore cannot establish how far a camera will protect a perimeter during fog.

The practical limit depends on the fog’s droplet distribution and liquid water content, the infrared wavelength band, the distance through the fog, and the target’s contrast against its surroundings. For an evaluator, the useful question is whether the complete system can detect and track the required target early enough for a response under the site’s adverse-weather conditions.

Fog reduces contrast along the entire viewing path. A thermal camera forms an image from infrared radiation reaching its detector. Suspended water droplets scatter and absorb some of that radiation, while the atmosphere also contributes radiation of its own. As transmission falls, the difference between the radiation received from a target and its background becomes smaller. Edges soften, features disappear, and a person can merge into the surrounding scene.

Distance compounds this effect. A target near the camera may remain distinct while an identical target farther along the fence becomes difficult to separate from the background. Fog can also vary across a site: a low-lying section, drainage channel, or coastal approach may have a substantially different optical path from the area beside the camera. A weather reading at the control building may therefore describe conditions poorly at the most vulnerable boundary.

Visible meteorological range is useful for recording weather severity, but it does not uniquely determine infrared transmission. Fog with the same visible visibility can have different droplet populations and different effects at thermal wavelengths. Relative humidity alone is an even weaker basis for a range prediction because it does not describe the amount or distribution of suspended liquid water. Claims that a camera “sees through fog” need a stated weather condition, viewing distance, and performance criterion.

The infrared band matters, but it does not provide a universal answer. Long-wave infrared cameras, commonly operating around 8–14 micrometres, and mid-wave infrared cameras, commonly around 3–5 micrometres, interact differently with the atmosphere. Long-wave imaging can offer a transmission advantage over visible imaging and, under some fog conditions, over mid-wave imaging. That advantage depends on droplet size, water content, and the system’s spectral response; dense fog can seriously impair both thermal bands.

Band comparisons also need to account for the complete camera. Lens aperture, detector sensitivity, resolution, image processing, and target temperature all influence the result. A cooled detector may improve sensitivity, but it cannot recover information when atmospheric transmission and scene contrast have become insufficient. Comparing two products by wavelength band alone can obscure larger differences in their usable performance.

Target-to-background contrast creates another limit. A person’s exposed skin may provide a strong thermal signature, while outer clothing can appear much closer to the temperature of the surroundings. Wet clothing and wet backgrounds can alter that contrast further. During periods when target and background radiance become similar, often called thermal crossover, even a relatively clear image may contain little useful separation. Fog can then push an already marginal target below the detection threshold.

Detection, recognition, identification, and alarm generation require different evidence. Detection establishes that an object is present. Recognition supports a broader classification, such as distinguishing a person from an animal or vehicle. Identification requires more specific distinguishing information, which the evaluator should define explicitly. A thermal “identification range” should never be assumed to establish a person’s identity.

A camera may show a moving warm region while providing too little detail for reliable human classification. Automated intrusion detection adds further requirements: the software may need a minimum target size, stable contrast across successive frames, and sufficient track duration before issuing an alarm. Fog can cause intermittent detections or broken tracks even when an operator can occasionally see the target. Increasing sensitivity may recover some events while also increasing nuisance alarms.

For perimeter protection, range should therefore be tied to an operational result. “A person is visible” is a weaker acceptance condition than “a person crossing the protected zone triggers an alarm with sufficient time for intervention.” Pixel-based range calculations can help screen camera and lens combinations, but they do not establish adverse-weather alarm performance.

Testing should reproduce the protected task and record the conditions. Ask for evidence that connects weather, target movement, camera configuration, and alarm outcomes. A useful evaluation should cover:

• Weather along the viewing path: record visibility, temperature, humidity, and available fog measurements, including changes during each trial. Identify whether conditions are reasonably uniform across the protected zone.
• Representative targets and routes: test people in relevant clothing, at different distances, crossing and approaching the camera. Include slow movement and locations where backgrounds provide weak contrast.
• The installed configuration: use the proposed lens, mounting position, analytics settings, and alarm rules. Preserve the configuration record so results can be compared meaningfully.
• Operational outcomes: record successful and missed events, alarm delay, track continuity, and nuisance alarms over an appropriate observation period. Review recorded footage alongside event logs.

Acceptance criteria should state the required probability of detection, acceptable nuisance-alarm rate, and maximum alarm delay for defined conditions. A single successful demonstration does not establish reliability. Where representative fog cannot be tested before purchase, treat the claimed adverse-weather coverage as an assumption requiring field validation.

Also separate atmospheric degradation from moisture on the camera window. Condensation, droplets, or contamination on the external surface can reduce image quality independently of fog along the viewing path. Window heating and suitable environmental protection can help manage local moisture, but they do not remove atmospheric attenuation. This distinction matters when diagnosing whether an installation needs maintenance, a different enclosure arrangement, or shorter viewing distances.

Where fog compromises the required detection distance, coverage design becomes more important than a larger nominal range figure. Shorter camera-to-target paths, revised mounting positions, and overlapping views can improve resilience. A longer focal length increases target sampling but does not restore contrast lost through the atmosphere. Radar or fence-mounted intrusion sensors can provide complementary detection, although their own coverage limits and nuisance sources require evaluation. Thermal video may still support assessment when it can no longer sustain dependable independent alarms.

The decisive specification is the distance at which the installed system meets its detection and response requirements under defined fog conditions. If that distance leaves an approach route exposed, the design needs additional sensing or closer coverage before a clear-weather range claim becomes operationally useful.