Pix4D fails to convert RAW JPEG thermal images to greyscale thermal images

Hello,
I have dataset of thermal images captured using the DJI XT2 thermal camera. Pix4D fails to convert raw RGB images to greyscale thermal corresponding ones which leads to processing failures. Is there any solution for this problem ? (PS: also tried FLIR tools and obtained the same problem)

Thank you in advance.!
PS: a sample image is attached

30m_auto_Th (2)|625x500

Hi @y.zefri Does it give any error? Can you give a screenshot? Are you using rjpg image format? We would need at least 5-6 images to test in Pix4Dmapper.

Hello,
Please find below a screenshot of the obtained error + sample images
sample images b2.zip (3.1 MB)

Thank you very much in advance.

Hi, Pix4DMapper is correctly converting the rjprg to tif images here. The error is due to uncalibrated images which happens if the software is not able to extract keypoints and match them. Generally it is due to insufficient overlap or incorrect camera parameters. Can you send me a screenhsot of the mapview in Pix4D? What was the frontal and side overlap? I will need the logfile for troubleshooting. One more thing, the pixel size is incorrect, you will need to edit it and change the value to 17. Process step 1 again, let me know how it goes.

Hello,
The overlaps were 90% for the end-lap and the same for the side-lap.
For the camera parameters, I have an image bloc having been acquired with the Zenmuse XT2 and that has been processed correctly with the same shown parameters. Changing the value to 17 changes nothing unfortunately, the problem persists.Logfile.txt (1.5 MB)
I have attached the logfile. Thank you very much in advance.

It might be an issue where due to some missing exif tags, Pix4D is unable to extract any keypoints from the tif image (as the rjpeg to tif extraction generates images with no variability in such a case). Can I have 5-10 of your images? I could then confirm and you would need to reach out to the hardware manufacturer. You could also check the converted tif images in QGIS.

Hello,
Please find below a sample of 9 images. Raw images aren’t uniform, they have a variabality and keypoints can be extracted from them, but the ones obtained after the Pix4D conversion have unique pixels values which make it impossible for keypoints extraction. Not all acquired images have this conversion problem. We have obtained from the same UAV/Cemera thermal images, in another flight bloc, who have been succesfully converted for keypoints extraction.

Thank you very much for your help.

ThermalImages.zip (3.0 MB)

Hi,

That means due to some reason some EXIF tags are missing. You will need to contact the hardware manufacturer so this doesn’t happen in the future. For teh conversion, Pix4D uses a FLIR SDK and if some tags are missing, the conversion will be incorrect.

Thank you for the clarification but it seems that i don’t quite understand the situation.
What are the missing tags ? How come that we’ve got by the same camera images that have successed the conversion ? And is there any way to fix those images cause we cannot do again this flight, not anymore.
Thank you in advance.

Hi, Can you send me 5 images from a project that works and 5 images from one that doesn’t? I could try to understand which tags are missing.

Hello,

Here are samples from the images that pass the processing.
images.zip (3.9 MB)
Thank you for your help.

Hi, I downloaded the images and all of them were converted correctly. I would need images that do not work as well to compare.

Hi,
I’ve already sent you images that did not pass the processing. They are in the conversation

Here they are

Hi, sorry for the delay. We think this has something to do with the hardware as I mentioned. I checked in FLIR tools and the images in the folder thermalimages do not show up whereas the images that convert correctly do