RTK EVO TEST 2 with attached data

[jmason702]

RTK TEST 2.zip


Here is the flight with 6 recorded points ALL in WGS84 Ellipsoid. 80 AGL- 5MPH- 280 photos, I am processing now with DD using points 3 and 6 as checkpoints and the other 4 as GCPs

Feel free to process in Meta and Pix, smooth flight no issues... ALL fixed solutions using NTRIP

 

[Dave Pitman]

Google Earth reports the center point of image MAX_0001.jpg as 738.8 meters (2424 feet)
If you add 80 feet to Google Earth's surface level, the camera should actually be close to 763.2 meters (2504 feet)

MAX_0001.jpg exif data reports elevation of the camera as 736.6 meters (2416.7 feet)
An approximate difference of minus 26.6 meters (87.3 feet)

Google Earth is not using ellipsoid height I guess.

 

[jmason702]

Am I doing something wrong here? first time using Metashape?

 

[jmason702]

Google Earth reports the center point of image MAX_0001.jpg as 738.8 meters (2424 feet)
If you add 80 feet to Google Earth's surface level, the camera should actually be close to 763.2 meters (2504 feet)

MAX_0001.jpg exif data reports elevation of the camera as 736.6 meters (2416.7 feet)
An approximate difference of minus 26.6 meters (87.3 feet)

If you capture an image with the drone on the ground, shouldn't the exif elevation be very close to the ground elevation at that point as referenced to WGS84? Why is it so much lower?

Not sure, unless the drone is trying to project surface elevation?

 

[jmason702]

MUCH Better... need to hit the update tab

 

[Dave Pitman]

Not sure, unless the drone is trying to project surface elevation?

I'm guessing that Google Earth uses the EGM96 model for height and that is the difference to the ellipsoid used by the drone. But I was hoping someone would know definitively.

 

[jmason702]

I'm guessing that Google Earth uses the EGM96 model for height and that is the difference to the ellipsoid used by the drone. But I was hoping someone would know definitively.

Thats a very good possibility. I have noticed that in my area I am getting between 60 to 80 feet difference when comparing ellipsoid to the geoid

 

[jmason702]

DD Report....................... Thoughts??

 

[Dave Pitman]

It still looks like DD isn't giving a sufficiently heavier weight to the GCPs.

I'm also curious about the GSD calculation of .97 inches/px. That seems like too low of resolution for 80 feet capture height. But I'm referencing a P4P's camera.

What were your overlap settings?

 

[jmason702]

80 front, 70 side overlap

 

[jmason702]

The EVOs PRO camera in my opinion is better than P4P, I own both, Then again the P4P has done great also

 

[jaja6009]

I'm guessing that Google Earth uses the EGM96 model for height and that is the difference to the ellipsoid used by the drone. But I was hoping someone would know d

It still looks like DD isn't giving a sufficiently heavier weight to the GCPs.

I'm also curious about the GSD calculation of .97 inches/px. That seems like too low of resolution for 80 feet capture height. But I'm referencing a P4P's camera.

What were your overlap settings?

I don't know about Drone Deploy, but it does (or did) use Pix4D's Engine, and Pix4D on Mapper Desktop has a default Accuracy setting for GCPs as .02. His images are rated .01.

 

[jaja6009]

How about you take an image on the ground of the GCP you measured with the Emlids. Nadir, perfectly centered over the target. We then should know what elevation its giving.

 

[jaja6009]

Pix4D

I took images as WGS84 and elevations as ellipsoid.
I took Control Points as WGS84 and ellipsoid
Out was WGS84 with EGM96 applied.

Orthometric Height is around 737 to 739m







Bottom right corner shows the elevation of the selected point on the cloud.

 

[jmason702]

THANK YOU!

 

[Dave Pitman]

I'm also curious about the GSD calculation of .97 inches/px.

The EVOs PRO camera in my opinion is better than P4P, I own both, Then again the P4P has done great also

I'm referring to the Ground Sample Distance, not really the quality of the camera. With a flight height of 80', I would expect a GSD of somewhere around .3 inches/px, not .97 inches/px.

EDIT: I see I was looking at the DEM GSD. Elsewhere in the report, it states .24 inches/px. which makes sense.

 

[Dave Pitman]

I don't know about Drone Deploy, but it does (or did) use Pix4D's Engine, and Pix4D on Mapper Desktop has a default Accuracy setting for GCPs as .02. His images are rated .01.

If I'm reading the report correctly, is says that the GCP mean RMS error is .69 inches (.018 m) and the camera position mean RMS error is .42 feet. I think the GCP should be set to at max .005 meters which is what you have in P4D, correct?

 

[Dave Pitman]

Pix4D

I took images as WGS84 and elevations as ellipsoid.
I took Control Points as WGS84 and ellipsoid
Out was WGS84 with EGM96 applied.

If I'm reading the report correctly, you used 1 point as a GCP and 5 points as check points. And for the checkpoints, the largest mean RMSE was less than 4cm. Is that correct?

 

[jaja6009]

If I'm reading the report correctly, is says that the GCP mean RMS error is .69 inches and the camera position mean RMS error is .42 feet. For contrast, I usually set a surveyed GCP to .001 meters (.04 inches) in MS.

If I'm reading the report correctly, is says that the GCP mean RMS error is .69 inches and the camera position mean RMS error is .42 feet. For contrast, I usually set a surveyed GCP to .001 meters (.04 inches) in MS.

Yes

If I'm reading the report correctly, you used 1 point as a GCP and 5 points as check points. And for the checkpoints, the largest mean RMSE was less than 4cm. Is that correct?

Yes.

I left accuracy for the cameras with what's in the metadata. I used .005 for the GCP's.

With my PPK kit i set the cameras to .04 and the Control Points at .2 (Default Pix4D).

I don't know if me only marking 5 images on each point is the problem, as I was in a hurry. I can go back and mark them all if needed.

 

[Dave Pitman]

I don't know if me only marking 5 images on each point is the problem, as I was in a hurry. I can go back and mark them all if needed.

I think error less than 3cm horizontal and 5cm vertical is considered good.