Support Website Contact Support Blog

Processing RJPEGs and edited TIFs (based on the RPJEGs) : Georeferencing and Scaling questions

Hi,

I was trying to process a subset of the following dataset : https://www.rmus.com/pages/data-set-examples
I used the pictures DJI_0198.jpeg to DJI_0208.jpeg to try the following:

I would like to create a reflectance map from this pictures, which kind of works.

However when I display this in ArcGis pro, I see that the produced reflectance tif is shifted compared to the original picture set. I do not understand how this happens. Furthermore I try to extract the raw temperature values from the RJPEGs by reading them (first CSV extraction with FLIR tools) in python. I do succeed to read the temperatures here and play with them to compensate for the temp drift caused by uncooled drone thermal cameras. From this I create a 16 bit tif file that contains the temperature data directly (or I at least know the formula to extract it later). I copy the exif data from the original jpeg to the tif, this way I keep the GPS tag. 

As seen on the screenshot shared here : https://www.dropbox.com/sh/93n35ltbpouzfim/AAA7sbbOOXZwi-zdl7nkLt4Ka?dl=0 ,
the JPEG and TIF are placed on the same location, thus the geotag is maintained (Points overlap). However the reflectance map is different. The one based on the tif loses scale and is out of position. I cannot understand why this happens. How is the scale managed of a reflectance map? It would be useful to display it on the world map as a reference.

Best regards

Paul

Hello Paul,

I think that what you are facing is the same as discussed in this post. It is in Spanish but the screenshots show something similar.

When working with thermal cameras, it is important to check whether pixel size and focal length are correct.

Please check that and in case it is not correct, type the correct values by following this article

In case you have not yet read it, I would recommend this article too. It gives a lot of information on how to work with thermal images in Pix4D.

Regards.

 

 

 

Hello Daniel,

 

Thanks for your reply. Honestly the other post just indicates there might be something wrong with the pixel size of the sensor. I will check that. However the problem in that post was solved internally thus that does not help me much. 

Still, if I can solve the first part, I wonder how I can solve the second part of my question. What kind of other data is embedded in the RJPEG? I don’t see how the second processing based on the tiff file with same exif data and geotag can give such a different result! It is just a correct camera setting and an accurate geotag right? I don’t understand what I am missing there.

Best regards

 

Paul

Hello Paul,

The RJPG is a compression format which includes the radiometric images (usually RGB) together with the thermal images in TIFF format.

When you import RJPG format into Pix4D, the software automatically extracts the thermal images in TIF format and they can be used for the process and you can generate a reflectance map among other products.

If the internal parameters are not right, the reflectance map can have a wrong scale or be displaced.

I think that is what is happening here.

Please check the internal parameters and let me know the result gets better.

Regards.

Hello Daniel

I checked again and realise that the parameters where correct. So I do not understand… The thing that made me puzzled the most is that two almost identical datasets created a different result. The unmodified vs the modified set.

Can the size of the dataset explain the problem? I am using only 11 out of 200 pictures to test faster. Should a bigger area give better scaled reflectance maps?

Best regards

Paul

Hello,

The size of the project is not the problem.

If I understand properly, you have processed the thermal tiff images with Pix4D.

The JPG images you are comparing with, how were they generated? Are you comparing with an orthomosaic? What kind of images are they?

Sorry for the confusion.

 

Hi

I compared two sets:

  • One original set, a subset of https://www.rmus.com/pages/data-set-examples , thus RJPEG format
  • One modified set: based on the original set here above, I created TIF format files which contain the raw temp data on which I copied the exif data from the first set to have the right Geotag and camera information. 

 

The two give way different results. Therefore I am wondering what else pix4d needs to process them properly. As far as I understand the raw value of the RJPEG varies between 1500-15000 whereas my raw values are spread around 25000-40000.

 

Would this affect the stitching for a reflectance map?

 

Best regards

 

Paul

Hi again,

I downloaded and processed the same images that you are using from your link.

As they are in RJPG format, when I import them into Pix4D, the TIFF files are extracted and saved in a folder called “Converted”.

The software then uses these TIFF images for the process.

The TIFF images have values from 5 to 60 approximately so I guess they have some kind of post-process to get real temperatures and they are not the raw values :

I did the complete process and I generated the reflectance map and load it into google, I do not get any scale or shift problem, it is located approximately at the same place as the flight.

What images did you use in Pix4D? Why do not you import directly the JPG files?

Regards.