ThermalVariations
Capture guide

Drone thermal inspection guide for solar PV

Most false “faults” on a thermal inspection aren’t faults — they’re capture artefacts. Sun glint reads as a hotspot, an oblique gimbal warps apparent temperatures, a washed-out range hides real ΔT. Our pipeline screens every uploaded frame for exactly these problems before analysis; this guide is how to fly so that screen has nothing to catch.

Each section below corresponds to a check the analysis actually runs. Warnings don’t block your report — flagged frames are listed in its capture-issues section — but every warning is a frame whose findings deserve less trust, so the goal is zero.

01Irradiance — fly at 600 W/m² or better

Fault-driven temperature differences scale with how hard the modules are working. Below the 600 W/m² floor required by IEC TS 62446-3, a faulty cell and a healthy cell look nearly identical, so a clean scan proves nothing and a flagged one can't be graded.

  • Fly within a few hours of solar noon on a clear day; avoid broken cloud that pulses irradiance mid-flight.
  • Measure irradiance on site if you can, and declare it when you create the job — an operator-declared value is the authoritative input to our irradiance check.
  • If you can't measure it, keep the GPS and timestamp metadata intact: we estimate a clear-sky ceiling from each frame's position and time as a fallback.

02Camera angle — shoot straight down

Thermal cameras read apparent temperature, which drifts with viewing angle: oblique shots lower the glass's effective emissivity, distort module geometry and invite sky reflections. Our gate reads the gimbal metadata and warns when a frame is captured well off nadir.

  • Lock the gimbal to −90° (nadir) for the survey passes and keep roll level.
  • Stay within a few degrees of straight down; frames more than about 25° off nadir get flagged.
  • On steeply tilted arrays, plan lines so the camera stays close to perpendicular to the module plane.

03Sun glint — don't mirror the sun into the lens

Module glass is a decent mirror. At the wrong geometry the sun reflects straight into the camera as a blown-out white blob that looks exactly like a catastrophic hotspot. The gate detects saturated, achromatic regions and warns, but a re-shoot beats a warning.

  • Plan flight lines so the sun-module-camera reflection angle never closes; a small heading change usually fixes it.
  • Glint moves when you move — if a 'hotspot' slides across modules between frames, it's a reflection.
  • Watch the live feed on the first pass; if panels flash white, adjust heading or time of day.

04Focus and motion — keep frames sharp

A soft frame smears a one-cell hotspot into the background. Blur comes from focus set wrong, flying too fast for the integration time, or gusty conditions. The gate scores every frame's sharpness and warns on soft ones.

  • Set focus at altitude on real modules, not on the ground before take-off.
  • Slow down: thermal sensors integrate longer than RGB — cruise speeds that are fine for mapping can smear thermal frames.
  • Avoid flying in winds that make the gimbal work hard; besides blur, wind also cools the glass and shrinks ΔT.

05Thermal contrast — use a sensible range/gain

If the camera's thermal range is set too wide, the whole array renders as one flat grey slab and real ΔT is quantised away before we ever see it. The gate measures scene contrast and warns on washed-out frames.

  • Use auto-gain or a manual range spanning just the scene's temperatures (roughly ambient to ambient + 40 °C).
  • Check the histogram on the first frames: module detail should be visible, not a uniform mid-grey.
  • Ironbow or white-hot palettes both work — the analysis detects the palette automatically.

06Upload the right files — thermal frames, original bytes

Dual-sensor drones save a thermal frame and a visible-light twin per trigger. Only the thermal frame is analysable — an RGB photo has no temperature information, so it is the one check that hard-blocks a frame rather than warning.

  • Upload the thermal ('_T') files; visible-light twins ('_W') are detected and skipped, not analysed.
  • Upload original files straight from the card. Re-exporting or emailing strips the EXIF, GPS and gimbal metadata used to locate findings and verify capture angle.
  • Keep the RGB twins on hand anyway — they're invaluable for confirming soiling versus real hotspots during verification.

Before you take off: the 60-second checklist

  • Clear sky, within a few hours of solar noon, irradiance at or above 600 W/m² (measure it if possible).
  • Gimbal locked to nadir; flight lines planned off the sun-reflection angle.
  • Focus set at altitude; speed low enough for sharp thermal frames.
  • Thermal range/gain producing visible module detail on the live histogram.
  • Enough overlap that modules aren't clipped at frame edges, and altitude low enough to resolve individual cells.
  • After landing: upload original thermal files with metadata intact, and declare the measured irradiance on the job.

Fly it well and the report comes back clean: every finding graded against the anomaly taxonomy, classified per the typical interpretation of IEC TS 62446-3, with nothing in the capture-issues section.

Flying a site soon?

Flat-rate analysis: you fly the site, we grade every thermal frame and deliver a severity-ranked report. Estimate the cost for your site size in seconds — no account, no sales call.

ΔT figures and abnormality classes on this page describe the typical industry interpretation aligned with IEC TS 62446-3 — indicative engineering guidance, not normative text from the specification. Temperature differentials are only meaningful when captured at ≥600 W/m² irradiance (per the TS) and verified on site before repair work.

← Back to the anomaly library