Hello,
Back in 2021 I flew a Zenmuse XT2 sensor over experimental ice to try and map ice surface temperature. Due to my limited experience, I unfortunately flew the drone a little too low over the ice, and now the thermal imagery is struggling to process. I have played around with the initial processing parameters (calibration method, key points image scale), however there are still a lot of uncalibrated cameras in each project, leading to holes in the final orthomosaic.
The RGB imagery from the XT2 has processed properly, so I have correct RGB orthomosaics for each dataset. Since the RGB and thermal sensors were triggered at the same time, I am wondering if there is a way to use the RGB point cloud/camera positions as a base for my thermal image processing. Other recommendations are welcome on how to optimize thermal imagery processing over a ‘smooth’ surface.
Heres some images of the data I am working with:
Thank you!
Here is the quality report for one of the thermal projects.
Feb19_1400_xt2_report.pdf (650.7 KB)
Hi @mharasyn
Welcome to Pix4D Community!
I see in the quality report that you didn’t choose any template for processing. It is recommended to use the “Thermal Camera” template for this case. Please could you try and let us know if it works.
Have a very nice day!
Best regards
Hi Heydi,
Thank you for your help. I initially processed my data using the Thermal Camera template (I have attached the report), but as you can see this leads to fewer matches between cameras. Through trial and error, I have found that processing my thermal data using the ‘3D Maps’ template and adjusting the initial processing parameters until optimized has been my best bet so far, hence why my quality report showed that I was using ‘No Template’.
Let me know if you have any other ideas!
Maddie
Feb19_1400_xt2_report_ThermalCamera.pdf (119.9 KB)
Hello @mharasyn, Thank you for attaching the quality report. Looking at the quality report, the reason for the low fewer key points matching is the image acquisition. The overlap and uniformity of the images captured are not good. For thermal, we recommend capturing images with 90% front and 90% side overlap for better results. For more information on processing thermal images, visit
