Welcome, Autel Pilots!
Join our free Autel drone community today!
Join Us

Comparing Evo II Enterprise to Mavic 3 Enterprise for Mapping

In the Reference Settings what you have is fine, but I always output to the correct Projected system whether UTM or State Plane. If UTM I use EGM96 for my geoid and if State Plane Geoid 12 or 18.
But I see your RS2 setup on a known point. What coordinate system is it setup on? NAD83(2011) or WGS84? I am assuming NAD83(2011) that you obtained from your RTK/NTRIP Service?

Next, I never knew that the Evo 2 RTK had Camera Calibration XMP data. But it does. I had previously used it and it worked fine(Unknowingly), but with this dataset it is the cause of the vertical shift. On this note, what is your camera angle on your Altitude Optimization oblique images?

Next, start from the beginning. I have no luck when I go back and try new settings. Make a new project.

Go into Preferences/Advanced and uncheck the Load camera calibration from XMP data. For this dataset it seems to be the cause of the shift. You would think that using Auto would have it ignored. But I only had luck when it was not imported to begin with.

1673117655965.png


After you align the first time, Import your GCPs. For testing I only mark 10 to 15 images. Remember to mark, keep the control points checked as GCPs. When done marking uncheck the ones you want to be check points. In this case all since we want to see how low of a RMSE we can get with no GCPs.

Press the Update Transform button next to the Optimize button.

You should have pretty tight data at this point.
You can Optimize now. I used the default params checked and Fit Additional Corrections.

I didn't really find a large benefit when trying to use Gradual Selection to remove the most inaccurate points. I would assume because the area was so small and uncomplicated (Pure Guess). I tend to see drastic improvement in RMSE with the Gradual Selection setting with my large beach maps and the nightmare of the water. (Too many images to mask out).

Agisoft has always stated that using Auto will have better results than using the import Camera Calibration params. I recently did a test where I used the Calibration Params from my Phantom 4 RTK and fixed/locked it so it would only use that for alignment. It was not too good, around 10cm total RMSE.

The thing that is weird though is I always have this setting on (Load camera calibration from XMP metadata) and the Evo 2 image sets I ran came out great. Maybe there was a firmware update on the Evo 2 that now places a generic Camera Calibration in the XMP/metadata? I was just really surprised to see it in the Precalibrated section.
 
But I see your RS2 setup on a known point. What coordinate system is it setup on? NAD83(2011) or WGS84? I am assuming NAD83(2011) that you obtained from your RTK/NTRIP Service
I got it from my NTRIP service, so I also would assume NAD83
 
In the Reference Settings what you have is fine, but I always output to the correct Projected system whether UTM or State Plane. If UTM I use EGM96 for my geoid and if State Plane Geoid 12 or 18.
But I see your RS2 setup on a known point. What coordinate system is it setup on? NAD83(2011) or WGS84? I am assuming NAD83(2011) that you obtained from your RTK/NTRIP Service?

Next, I never knew that the Evo 2 RTK had Camera Calibration XMP data. But it does. I had previously used it and it worked fine(Unknowingly), but with this dataset it is the cause of the vertical shift. On this note, what is your camera angle on your Altitude Optimization oblique images?

Next, start from the beginning. I have no luck when I go back and try new settings. Make a new project.

Go into Preferences/Advanced and uncheck the Load camera calibration from XMP data. For this dataset it seems to be the cause of the shift. You would think that using Auto would have it ignored. But I only had luck when it was not imported to begin with.

View attachment 15471


After you align the first time, Import your GCPs. For testing I only mark 10 to 15 images. Remember to mark, keep the control points checked as GCPs. When done marking uncheck the ones you want to be check points. In this case all since we want to see how low of a RMSE we can get with no GCPs.

Press the Update Transform button next to the Optimize button.

You should have pretty tight data at this point.
You can Optimize now. I used the default params checked and Fit Additional Corrections.

I didn't really find a large benefit when trying to use Gradual Selection to remove the most inaccurate points. I would assume because the area was so small and uncomplicated (Pure Guess). I tend to see drastic improvement in RMSE with the Gradual Selection setting with my large beach maps and the nightmare of the water. (Too many images to mask out).

Agisoft has always stated that using Auto will have better results than using the import Camera Calibration params. I recently did a test where I used the Calibration Params from my Phantom 4 RTK and fixed/locked it so it would only use that for alignment. It was not too good, around 10cm total RMSE.

The thing that is weird though is I always have this setting on (Load camera calibration from XMP metadata) and the Evo 2 image sets I ran came out great. Maybe there was a firmware update on the Evo 2 that now places a generic Camera Calibration in the XMP/metadata? I was just really surprised to see it in the Precalibrated section.
I always remember seeing it, This is a V1 PRO RTK, Mabey the V2 did not have it not sure


Thanks for ALL your time helping with this, you are AWESOME!
 
I get similar results with the Evo set as Jeff. The results on the M3E are great. I'm surprised at the Evo's results because Jeff's Evo sets have been really tight in the past. The pixel error on the CPs in the Evo set are really high in my processing for some reason.

On the CRS issue, in this case, the drone and the CPs are all referencing the same system so there really shouldn't be any difference if you tell MS that its WGS84 or NAD83 so long as they are the same. I just leave it on WGS84 in this case.

evo.png



m3e.png
 
Last edited:
  • Like
Reactions: jaja6009
I get similar results with the Evo set as Jeff. The results on the M3E are great. I'm surprised at the Evo's results because Jeff's Evo sets have been really tight in the past. The pixel error on the CPs in the Evo set are really high in my processing for some reason.

On the CRS issue, in this case, the drone and the CPs are all referencing the same system so there really shouldn't be any difference if you tell MS that its WGS84 or NAD83 so long as they are the same. I just leave it on WGS84 in this case.

View attachment 15473



View attachment 15472

I had pretty much the same outcome as you and Jeff.
My vertical shift only went away after I unchecked the Load Camera Calibration from XMP.
And this is confusing since I thought that even if it imported it, if you choose Auto in the Camera Calibration it would ignore it.
The only thing I did was to start a whole new project (I never have any luck when I try new things on a aligned project), uncheck the Load Camera Calibration from XMP, confirm that the Precalibrated Camera Calibration data was not imported and then Align Images.
My data then was a tad tighter than my Mavic 3E data.

I am multi-tasking. You are in my Emlid thread with my baffling Mount Point not going online fiasco. I can take problems, but intermitent problems are the worse!!!!
 
I am multi-tasking. You are in my Emlid thread with my baffling Mount Point not going online fiasco. I can take problems, but intermitent problems are the worse!!!!
Absolutely.

I just set up an RS2 (on 30.b2) and it hooks up with E.Caster. Likely Emlid will have to look at your system report and see what's up. I know that doesn't help much.
 
Last edited:
  • Like
Reactions: jaja6009
Hey, I made a mistake.

I mixed up the images with the Evo 2 and the Mavic 3 E after I renamed them.
But I still get the Evo 2 Precalibrated Camera Calibrations.
I am sorry. I was occupied trying to get my Emlid to connect to the caster Mount Point (Long Story).

I still have the vertical shift.

I do apoligize for the circle of connufusion. I only found it after I saw that your results were not matching.

I am still trying to get sub 6cm.
 
Hey, I made a mistake.

I mixed up the images with the Evo 2 and the Mavic 3 E after I renamed them.
But I still get the Evo 2 Precalibrated Camera Calibrations.
I am sorry. I was occupied trying to get my Emlid to connect to the caster Mount Point (Long Story).

I still have the vertical shift.

I do apoligize for the circle of connufusion. I only found it after I saw that your results were not matching.

I am still trying to get sub 6cm.

That makes sense.
 

Latest threads

Members online

No members online now.

Forum statistics

Threads
11,226
Messages
102,644
Members
9,818
Latest member
redwingaerials