RTK correction with Autel EVO 2 rtk

Good morning everyone
I recently got an Autel EVO 2 with the RTK module for real-time photo correction.
In theory each photo contains the coordinates with centimeter precision and there is no need to take the GCP on the ground. I wanted to understand if it is possible to use WebODM also in this mode, obtaining a survey with centimeter precision without having to insert the GCP. Among the pre-processing parameters I see that there are some ticks not to consider the GCP if you have photos already georeferenced but the precision that comes out of the model does not take into account the high-precision coordinates. Has anyone tested this procedure, even perhaps with the Phantom 4 pro rtk or even better with the EVO 2 rtk? Can you tell me how to get a high precision model? Thanks in advance

How are you finding that the georeferenced data generated are not taking into account the RTK tags in the images?

I saw in the report that the average error is about 0.54 m, while in rtk the error should be a few cm

If it was RTK Fixed or falling back to lower precision, that’s not unreasonable. Are you positive that you were locked in RTK Float with high precision the entire collection?

Just having RTK-grade hardware does not guarantee centimeter-class accuracy all the time.

I agree that rtk hardware alone is not enough. The thing is, I would like to test if this system also works on ODM. I am at the beginning with the rtk system, and I do not exclude that I am doing something wrong. But the same data set (all recorded in fix and without GCP) with metashape gives me an error of 7 cm, which I consider to be improved but acceptable. I would like to understand how to handle rtk data in ODM, if I get something wrong when I launch the processing.

I’m curious how metashape is calculating that.

So the dataset is RTK Fixed, not Float? Hmm… That means that .5m isn’t unreasonable.

I wonder if Piero has any insight into what we’re doing with our accuracy calculations and if there’d be any reason for ODM to not handle the RTK tags from that sUAS properly.

Can you provide an example image with the EXIF intact?

I did some more testing with other datasets as well but in the end the error discrepancy remains.
If you want to do some checks, at the following link I put an image of the evo 2 rtk, with its exif, the two reports of the results of metashape and ODM, the orthophoto generated by ODM and a shape file with the control points that I took. The latter have been used only to verify the results but in no case have I used them to create the alignments.
Confirm if I am wrong or if there is any problem reading the images. If necessary, I can also provide the entire set of images.
Thank you

2 Likes

Thanks so much for proving this extra data!

Are you willing/able to open an issue on our GitHub to help us track this?

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.