Agisoft Metashape
Agisoft Metashape => General => Topic started by: Dragline on December 08, 2018, 08:49:16 PM
-
I recently acquired a Phantom 4 RTK, and have been working with it for the past week. However I seem to be encountering systematic errors in the Z axis. The X & Y values seem to be spot on though.
For instance, I just created a map of an area with significant vertical relief (100m within 30ha). Within this area I have 3 points that I have measured with our survey equipment, I know these points to be good within ~2cm vertically. When I compare these points to a DEM produced by photoscan I find results like this:
Surveyed point, Elevation (ft) DEM, Elevation (ft) Difference (ft)
1557.6 1557.6 0
1428.1 1430.4 -2.3
1382.6 1386.1 -3.5
Some ideas that I have to fix this problem:
*Enable Dewarp in flight planning software before flight. I believe that in the drone the camera calibration will be applied
*Import the dewarp camera parameters into Agisoft, somewhat covered at the end of this thread (http://www.agisoft.com/forum/index.php?topic=9910 (http://www.agisoft.com/forum/index.php?topic=9910))
*Very heavy use of the gradual selection tool (http://www.agisoft.com/forum/index.php?topic=8016.0 (http://www.agisoft.com/forum/index.php?topic=8016.0))
If there is any interest in taking a look at this data set, I can send it out along with control points.
-
Dragline,
it would be very interesting to have some data set to test it... Using RTK to just fix the camera perspective centers can often bring a problem in elevation as focal is directly correlated to flight altitude... So the higher you are from ground, the Z error will increase relative to some error in focal…..
-
Paulo,
Thanks for the consideration. I was planning to collect a test data set today with lots of control points. However it has begun to snow quite a bit. However, when I do collect this data, I will be sure to send it on over.
-
Great,
to decorrelate the calibration parameters as well as to eliminate bowl effect, I would suggest capturing additional obliques photos to the traditional grid pattern coverage, maybe a few going up to the block and a few after terminating the grid collection....
-
Hello Paulo,
I have had a busy few days! However I got a chance to collect another dataset yesterday. It includes obliques. This seems to have fixed most of the problem, however there is still some error there. The error seems to be just a vertical shift though. I'm not quite sure where this is coming from yet. I will PM you a link to the data for you to take a look at.
The control points are in EPSG: 26734. I just use the Agisoft coordinate calculator to convert the photo locations to this system.
-
Hello Dragline,
Can you provide the PDF report for this project and specify, which accuracy settings have been used for the cameras in the project?
-
Hello Paulo,
I have had a busy few days! However I got a chance to collect another dataset yesterday. It includes obliques. This seems to have fixed most of the problem, however there is still some error there. The error seems to be just a vertical shift though. I'm not quite sure where this is coming from yet. I will PM you a link to the data for you to take a look at.
The control points are in EPSG: 26734. I just use the Agisoft coordinate calculator to convert the photo locations to this system.
Dragline,
I was wondering if you sent me a PM regarding your last data set... Would like to have a look if possible,,,,
-
Dragline,
it would be very interesting to have some data set to test it... Using RTK to just fix the camera perspective centers can often bring a problem in elevation as focal is directly correlated to flight altitude... So the higher you are from ground, the Z error will increase relative to some error in focal…..
Paulo,
Can you elaborate a bit on this for me? What do you mean "RTK to just fix the camera perspective centers?
Thanks,
S
-
Hello Paulo,
I have had a busy few days! However I got a chance to collect another dataset yesterday. It includes obliques. This seems to have fixed most of the problem, however there is still some error there. The error seems to be just a vertical shift though. I'm not quite sure where this is coming from yet. I will PM you a link to the data for you to take a look at.
The control points are in EPSG: 26734. I just use the Agisoft coordinate calculator to convert the photo locations to this system.
Dragline,
I was wondering if you sent me a PM regarding your last data set... Would like to have a look if possible,,,,
Hello Paulo
I just bought a Phantom 4 RTK + D-RTK 2 unit.
If you want I can send you a set of data measured on February 20, are taken with the latest activation of DJI, released that same day.
I can provide you with the photographs and all the files generated by the Phantom 4 RTK.
I can also send you the data measured by the base D-RTK 2, "RTCM data" and "User Log"
and the Rinex data (1 second) from a nearby refence station (10 km)
Greetings.
-
Hello Domingo,
yes please send the dat, and would certainly have a look at it… If you measures a few GCPs from the base station, then we can have an independant check on the quality of airborne RTK positions….
-
I was wondering if there has been any progress made on this issue. I believe I am also encountering this Z error in proportion to height. If anyone has worked out a solution I would be very interested to learn more.
-
The key here is to take oblique photos of the area of interest. Use these for the initial alignment but disable them before dense cloud generation.
-
is there a reason why you want to leave them out during dense cloud processing ?
Wont the model be better if you have oblique photos in the dense cloud ? (better walls of buildings etc) ?
(im a noob)
-
I've found that there is significantly more noise on flat surfaces when I leave the oblique images on during dense cloud generation. This noise makes the rendered DEM's much uglier.
-
Does the workflow include optimising the cameras after alignment? Would it make much difference using the precise camera positions (no additional GCP) to optimise?
-
Hello Dragline,
Can you provide the PDF report for this project and specify, which accuracy settings have been used for the cameras in the project?
Alexey, do we have to change the accuracy settings for the cameras (e.g. average reported precision from EXIF) or are the accuracies read directly from each camera?