Is there a way to load a DEM from outside and correct the orthomosaic image

[ppant]

Hi,
I see the problem in orthomosaic result when my surface is concave in nature. The orthomosaic image is stretch. Because the flight lines are straight but between each camera shot there is a change in altitude. I am seeing the resulting orthomosaic is stretched. So is there a way to correct geo location in such case surface with concave nature.

Thanks,
ppant

[Alexey Pasumansky]

Hello ppant,

I have seen the report file that you have sent to our support. From what I've noticed: the calibration parameters for MicaSense sensors are loaded from the XMP meta-information, I suggest to disable this auto-load in the Advanced preferneces tab, then switch the Type in the Camera Calibration dialog to Auto (uncheck Fix Calibration flag if it is enabled) and let PhotoScan to estimate the calibration parameters automatically (with Adaptive camera model fitting disabled).

If the camera accuracy is also loaded from XMP, I can also suggest to reset the accuracy or at least set it to a few meters and then optimize the alignment (enabling all coefficients except K4, P3 and P4).

Currently the residuals graphs are not good, so the incorrect calibration seems to be a reason of inaccurate georeferencing.

[ppant]

Alexy
Thanks for the information

I suggest to disable this auto-load in the Advanced preferneces tab, then switch the Type in the Camera Calibration dialog to Auto (uncheck Fix Calibration flag if it is enabled) and let PhotoScan to estimate the calibration parameters automatically (with Adaptive camera model fitting disabled).

I am working with a python script, it's unclear from document how to change calibration flag to auto with a script. Could you give some pointers or code snippet?

If the camera accuracy is also loaded from XMP, I can also suggest to reset the accuracy or at least set it to a few meters and then optimize the alignment (enabling all coefficients except K4, P3 and P4).

How to reset the accuracy via a script to metes.

Thanks

[Alexey Pasumansky]

Hello ppant,

These lines will set the accuracy for all cameras in the Reference pane to 5 meters and 500 degrees:

Code: [Select]

for camera in chunk.cameras:
    camera.reference.location_accuracy = PhotoScan.Vector([5,5,5])
    camera.reference.rotation_accuracy = PhotoScan.Vector([500,500,500])

And these lines will reset the precalibrated values for all bands and set them to Auto mode:

Code: [Select]

for sensor in chunk.sensors:
    sensor.user_calib = None


[ppant]

Alexey
Thanks for the information. At the moment I am able to get a reasonably good result after setting calibration parameter to auto.
I will work with reference parameter and see how much improvement I will get.

Thanks

[Alexey Pasumansky]

Hello ppant,

I can also suggest to optimize the camera alignment after the alignCameras stage is completed and reference accuracy information is adjusted (usually the set of parameters recommended for optimization is: F, Cx, Cy, B1, B2, K1, K2, K3, P1 and P2).

[ppant]

Thanks for this information. I will update you the result. How it turns out.

ppant