Radiometric consistency in UAV thermal photogrammetry for Earth-science surveys: preview artefacts and cross-domain photogrammetric benchmarks
This study presents an open-source GUI tool for converting DJI UAV thermal R-JPEG previews into float32 GeoTIFFs, demonstrating that this preprocessing workflow significantly improves photogrammetric accuracy by increasing tie-point counts and reducing reprojection errors across diverse Earth-science survey sites.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine a drone flying over a landscape, its camera capturing the invisible heat radiating from the earth. This technology, known as thermal imaging, allows scientists to see temperature differences that are invisible to the naked eye. They use these images to find cracks in bridges, detect stress in solar panels, or map the flow of heat from a volcano. However, there is a hidden trap in how these cameras save their pictures. When a drone takes a photo, it stores two versions of the data in a single file. One version is a colorful picture designed for human eyes, where the camera software stretches the colors to make the image look dramatic and easy to interpret. The other version is a raw, precise record of the actual temperature at every single point, stored as a long list of numbers. The problem arises when scientists try to build 3D maps from these photos. The software that stitches images together is designed to read the colorful, human-friendly version. If it uses that version, it is working with a distorted picture where the colors have been artificially adjusted, losing the precise temperature data needed for accurate scientific measurement.
A team of researchers from Istanbul set out to solve this problem and to prove that using the raw, precise data makes a measurable difference in the quality of the final map. They developed a free, easy-to-use computer program that acts as a translator. This tool takes the raw temperature numbers hidden inside the drone's files and converts them into a new format that preserves every decimal of the temperature reading, while also keeping the location and time information intact. They then tested this method on three very different locations: a bridge, a large solar power plant, and a geothermal field near a volcano. In each case, they built 3D models twice. First, they used the standard, colorful preview images that most people would use. Then, they used the new, precise temperature files created by their program. The results were clear and consistent across all three sites. The models built from the precise temperature files were significantly more accurate. The software was able to match the images together more tightly, finding many more connection points between photos and reducing the errors in how the images aligned. In some cases, the error in the alignment dropped by nearly thirty percent, creating a much sharper and more reliable 3D map.
The researchers discovered that the issue was not simply about the number of colors in the image, but about how the camera changed the contrast from one photo to the next. The standard preview images are automatically adjusted by the camera to look good, which means the brightness and color scale shift depending on what is in the picture. This constant shifting confuses the computer trying to stitch the photos together. The raw temperature files, however, kept a steady scale throughout the entire flight, allowing the software to see the true physical relationships between the objects. While the study did not measure the absolute temperature against a physical thermometer on the ground, it proved that the method preserves the relative differences in heat perfectly. By making this conversion process simple and automated, the researchers have provided a way for field teams to turn their drone footage into scientifically rigorous data without needing to write complex code. This ensures that when scientists map the heat of the earth, they are working with the actual temperature, not just a pretty picture.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.