1. Build error-budget awareness

Total IR temperature error roughly decomposes into: calibration error (blackbody/algorithm), emissivity error (target property), ambient reflection (low-ε surface), atmospheric/distance attenuation (path), optical transmission (lens/window), temperature drift/noise (instrument state), and target/geometry factors (pixel filling/non-uniformity). We take them one by one.

2. Wrong emissivity — the number-one killer

The camera defaults ε=0.95 or 1.0, but real materials vary widely. Typical emissivity (8–14 μm, room temp): human skin / painted metal 0.95–0.98; heavily oxidized metal 0.8–0.9; polished aluminum / stainless 0.05–0.2; thick glass ~0.9 (but reflection+transmission complicate it); plastic film 0.3–0.6 (depends on thickness/color). If true ε=0.15 but you set ε=0.95, the result is far too low — often reading hot metal as room temperature.

Countermeasures: look up tables or infer ε with a contact thermocouple + camera; apply known high-ε tape to low-ε metal then measure; never use the default ε on polished metal.

3. Uncorrected ambient reflection — the invisible trap of low-ε surfaces

IR received by the target = self-emission + ambient reflection × (1?ε). Low ε → large (1?ε) → large ambient contribution. A polished aluminum plate may mainly reflect your body (~37 °C IR), fluorescent lights, hot pipes or the cold sky. If you don't enter the reflected temperature (T_reflect) or ambient parameter, the algorithm defaults ambient to body temperature or ignores it, and the result drifts.

Countermeasures: correctly enter the reflected temperature (approximate with a blackbody or known ambient); shield strong hot/cold reflectors (baffle); prefer covering low-ε targets with high-ε material.

4. Atmospheric absorption and distance — not negligible outdoors / long-range

IR traveling through air is absorbed by H?O and CO? (8–14 μm has a window but still attenuates); longer path, higher humidity, more dust → weaker radiation at the detector → lower temperature. Countermeasures: close industrial measurements (<3–5 m) can ignore it, set transmission τ≈1; outdoors/long-range — enter distance, relative humidity and atmospheric temperature in software for auto correction; with heavy smoke/steam, get closer or wait for stable conditions.

5. Lens / protective-window transmission change

If calibration excluded an added Ge/Si/sapphire window, adding it later → effective transmission drops → lower reading. Countermeasure: recalibrate radiometrically after adding a window, or correctly set the window material/transmission in the menu (high-end models support this).

6. Target not filling the pixel (insufficient spatial resolution)

A thermal camera has an IFOV (instantaneous field of view). If the target size < 2–3×IFOV, the pixel value is pulled down by the surrounding background (partial-fill effect) → low reading. Countermeasures: get closer or use a telephoto lens so the target fills at least a few × few pixels; use a spot region's average rather than a single pixel; watch the minimum resolvable target-size spec.

7. Instrument drift and insufficient warm-up

Just powered on, the detector/lens are still changing temperature → DN drifts → the reading slowly "climbs" or "drops" before stabilizing. Countermeasures: warm up per the manual (typically 15–30 min for high-accuracy); ensure shutter NUC runs normally; watch core heating after continuous hot-target observation.

8. Other common pitfalls

  • Uneven surface temperature / oxide-layer differences → different ε at different spots, a single-point reading is unrepresentative;
  • Detector saturation (target over-range) → temperature clipping, need a higher range or attenuator;
  • Strong stray light / direct sunlight on the lens → off-band energy interference;
  • Trusting false-color but not the number → the palette misleads the eye; always read the digital ROI mean.

9. A reliable thermometry checklist

  • □ Emissivity set correctly per material (or high-ε tape applied);
  • □ Reflected/ambient temperature parameters entered;
  • □ Distance and atmospheric parameters (outdoors) set;
  • □ Target fills the FOV (or area averaging used);
  • □ No uncalibrated added window;
  • □ Instrument warmed up and NUC normal;
  • □ Readings match a blackbody check point (portable blackbody spot check);
  • □ Reading is the ROI mean, not a single pixel.

A blackbody calibration gives the thermal camera a standard ruler, but you must tell the algorithm the target material, environment and geometry honestly. Accurate ruler + correct usage = trustworthy temperature.