Thermal images and drones
Drone professionals are increasingly using thermal imagery to provide high-value deliverables to clients. Applications for thermal imaging include components for electrical utilities, solar panel inspection, roof inspections, live flare stacks for oil and gas, and water stress on farms. A number of drone professionals have approached Raptor Maps about generating thermal maps and analyzing radiometric data. In this blog post, we are addressing the most common thermal question we receive:
What are the best practices for creating a thermal map?
478 images were collected by an Inspire-1 with a radiometric 640x512 camera with 80% overlap and sidelap at 200 ft. AGL
Must be radiometric. 9 mm or 13 mm lens. Higher resolution (640×512) yields better results.
Fly at least 200 ft. above your area of interest with 80% overlap and sidelap.
Ensure your files are radiometric JPEG (R-JPEG).
Make deliverables that contain thermal analysis with actual temperature values.
Obtain a radiometric camera
Radiometric thermal cameras outputs a value for every single pixel that can be converted to a temperature. These cameras are often designated with a “radiometry” or “R” designation. These are accurate to within ±5 °C under ideal conditions. Radiometric images can be fused into a radiometric orthomosaic, which is a larger map in which every pixel contains temperature data. The file formats are radiometric JPEG (R-JPEG) or radiometric TIFF. Raptor Maps recommends R-JPEG, as it is compatible with most photo viewers and easier to work with. The most common radiometric thermal cameras for drone professionals are the Zenmuse XT Advanced Radiometry, the FLIR Vue Pro R, and the senseFly ThermoMAP.
Non-radiometric cameras output an RGB image that highlights relative temperature differences. This data is typically useful in applications such as search and rescue (SAR) and wildfires, where detecting hot spots is sufficient. Some of these cameras can record spot temperatures of the center pixels (similar to a temperature gun), and these are accurate to within ±20 °C. The file formats are typically normal JPEGs, and can be processed like any RGB photo. The most common non-radiometric thermal cameras for drone professionals are the Zenmuse XT, the FLIR Vue, and various analog thermal cameras with FLIR cores.
Examples of the different file formats for thermal cameras can be found here.
Optimize your image set
Creating a radiometric orthomosaic (aka thermal map) involves fusing the temperature data from multiple radiometric images. In our experience, these are the flight plans that lead to successful radiometric orthomosaics:
- Fly at least 200 ft. above your surface of interest
- At least 80% overlap and sidelap
- Radiometric camera with 9 mm or 13 mm lens
Typical resolutions for radiometric cameras are 640 x 512 and 336 x 256. Radiometric orthomosaics can be made from both cameras, but drone professionals with 640 x 512 cameras tend to have more success. The 336 x 256 cameras have 74% fewer pixels, making it more difficult to distinguish between features.
“Short” lenses (fewer mm) are known for the “fisheye” effect (with significant distortion at the edges). This works great for GoPro action photos, but not well for creating radiometric orthomosaics. Conversely, “long” lenses have less distortion, but are more zoomed in. This means that more passes will be needed to achieve adequate overlap of your area of interest. You can see some examples of these phenomena here.
We have seen several drone service providers experience these pitfalls and have to re-fly. Don’t let these happen to you!
Set camera parameters in-field
With radiometric cameras, you can set parameters such as emissivity and background temperature. This will ensure that the temperature measurements of your areas or objects of interests are accurate. FLIR has written a helpful technical note about these parameters for drone users, which can be found here.
Take RGB imagery
Thermal orthomosaics and analyzed thermal data are a high-value deliverable for drone professionals. Often, great care is spent capturing quality radiometric images, at the expense of RGB imagery. If you spot a thermal anomaly, RGB imagery can be critical in assessing whether you are recording a “real” phenomenon (as opposed to glare or a person standing in the wrong place). You can also overlay the RGB pins on the thermal orthomosaic for a more complete picture.
Be aware of changing temperature and lighting
If your area of interest is long and skinny, it may be tempting to fly out-and-back a few times. But in the time it has taken your drone to fly to the furthest point and return, some features (especially metal surfaces) may have changed temperature. Consider flying a pattern parallel to the “short” side instead.
A common misconception is that clouds do not affect thermal images, since the temperature of the surface hasn’t changed dramatically. However, the camera is not measuring temperature, it’s measuring a signal that is converted to temperature. Objects with a low emissivity (e.g., mirror-like objects) will look very different between sunlight and shade, making it more difficult to generate a radiometric orthomosaic.
Verify your data
Before you leave the field, make sure your SD card contains radiometric thermal images, either in radiometric JPEG (R-JPEG) or radiometric TIFF format. If you are unsure, load it into the Raptor App™ or FLIR Tools and verify that there is temperature data. Flight planning software (e.g., DJI GS Pro), may override the R-JPEG setting in the DJI Go app, so make sure to double-check after your pre-flight.