Hi,
I have about 3 thousend photos and i tried to define point cloud without using gcp, i have several geometric discrepancies related to laser scanner data?
is there a better way to optimize asset derived from rtk flight?
Thanks
Hi SMT,
what do you mean by those geometric discrepancies? What were your alignment settings? Are the laser data also georeferenced? What kind of scanner have you used?
i used default alignment settings, i told about geometric discrepance related to difference of buiilding height in general related to Z error problem (xt plane seems work), laser is georeferenced end defined with rtc 360
Can you show the problem here?
You ca try to set higher weights for the RTK coordinates of the images. Also, you can try to use some control points for better merging with laser scan data.
hi, i’m going to try again, so i found this topic and i think is interesting for me,
but i have one doubt where is the dewarp option? i’ve never ever play with this…thanks
It should be next to mechanical shutter, like https://forum.dji.com/forum.php?mod=redirect&goto=findpost&ptid=280115&pid=3026790
Thaks Orthan, i thought it’s related to RC options 
anyway i restarted process defining customs value for pose accurancy
Lat acc: 0,03, Long acc: 0,03, Alt acc: 0,06.
I have to wait to understand if it’s works also for mee, do you have other suggestion?
Thanks again
Something doesn’t work, I have about 1 meter error in 60 meter, I made a test on 140 image for fast test, i’ve never ever have problem like this (without RTk-dji air2s) on small area.
I tryed a very simple chunk to avoid errors derived from union of different flight but problems persist, on x y plane error is about 7/9 cm on 60 meter and is proportional to a reference point, but the error on z axis in quite hard to understand, 40 cm on 4/5 meter.
I don’t have any idea to solve the problem.
coordinate system: 4326 - Wgs 84
photos prior pose : position
absolute coordinates: 4326 - wgs 84
prior pose accurancy lat: 0.03 meter
lang: 0.03 meter
alt:0.05 meter
prior calibration:
prior: aproximate
prior lens distortion:
prior: approximate
camera model: no lens distortion???
in attached an example of settings
thanks to everyone have some good ideas…
Hello,
I would start with the Dewarp that was mentioned. We do not recommend de-warping images (fixing the lens distortion) before using them in RealityCapture. We have our own lens distortion algorithm and therefore de-warping them twice can cause issues.
If you don’t want to use the lens distortion model in RealityCapture, make sure that all Prior lens distortion settings are set like this (CTRL+A to select all inputs):
Are your inputs in the same coordinate systems? I assume you are using a combination of laser scans and images.
You are mentioning a reference point. Are you using a manually created point in RC? Is it a ground control or ground test point?
The 40cm deviation on the Z axis could be related to the elevation system. The WGS84 system is a horizontal system and does not define heights. Are your heights related to ellipsoid or geoid?
Hi Kumate,
i only taken photos in dewap mode so i think the problem is non not related to that.
My inputs are in the same coordinate system.
Laser scannig survey walk parallel to this process i need drone photogrammetry to integrate some parts.
I used laser as comparison not in the same process.
I mentioned reference point simply to put attention about the deviation from a same points in laser susrvey and in photogrammetry software.
I don’t think the elevation problem is related to the elevation system couse seems non linear in different area of survey, and z is deviated also in a singular simple element like a roof height 4 meter form the same goround.
The rtk data is in epsg 4326 and related to the elissoid, do i have define some trasformation?
I don’t understand the problem becouse i’ve never have this experience with RC without using rtk and using GCP in local coordiantes that was my routine until this survey
Hi,
as it was mentioned. you shouldn’t use dewarp mode. You need to keep use RealityCapture to use its own camera calibration and undistortion.
Can you show here the resulting errors?
In this case it could help to use also exact calibration parameters as is described here: Banana Effect - What To Do If My Model is Bent | Tutorial