Topics

1

General / Technical description of method used for 3D-reconstruction

« on: April 24, 2024, 12:49:01 PM »

Dear Community

I am looking for a paper or technical description of the method applied by Photoscan and/or Metashape or to reconstruct 3D-models out of the orientated imagery (standard cameras).

Reason is to get a better understanding of the imagery used and possible improvements while acquiring images in difficult environments (caves).
A better understanding could help to improve the images taken in the caves to have better results of the modelling process later (e.g. contrast, …).

Looking forward to any hints.

Cheers,
Yvo

2

General / Question about camera orientation using measured location and rotation

« on: May 16, 2019, 12:54:23 PM »

Dear List

I have the following setup:

• Several cameras are located in a room and are taking simultaneously images.

With a total-station I am able to measure all the 6 orientation parameters of all the cameras:

• For each camera, I know precisely the Easting, Northing and Altitude of the sensor.
• For each camera, I know precisely the direction North (Azimuth) the camera is looking.
• For each camera, I know precisely the angle of the camera looking down- or upward (horizontal or elevation angle).
• For each camera, I know precisely the angle of the camera tilting to the left or right (horizontal or elevation angle).

And the restriction is:

• I cannot place any markers on the walls.

The attached sketch tries to summarize the setup and the known values.

But I am unable to get any correct model. Currently, I am even struggling, if the three rotation values I have are Omega-Phi-Kappa or Roll-Pitch-Yaw.

Question:
What is the correct approach to get the correct rotation angles as input from the Azimuth, Dipping and Tilting of the camera?

Looking forward to any hints.

Best regards,
Yvo

3

General / Value for Pitch in case of upward looking camera

« on: May 16, 2019, 08:14:16 AM »

Dear List

I try to create a model of a room including the ceiling. The camera locations and orientations are accurately surveyed using a total-station.

But I am failing to create a correctly orientated model of the ceiling.

According to the definition of pitch (see page 65 in https://www.agisoft.com/pdf/metashape-pro_1_5_en.pdf) I set the values for pitch close to 180 degrees.

But the model of the ceiling does not appear correctly orientated. Instead, it looks like a wall.

Question:
How do I set the value for pitch correctly? Yaw looks ok and roll is set to 0.0.

Looking forward to any hints.

Best regards,
Yvo

4

General / Memory usage of large projects

« on: March 01, 2016, 08:36:05 PM »

Dear List

I am processing around 850 images (6000 x 4000 pixels). I would like to get a mesh and later an orthoimage over all the images.

Currently I am running constantly out of memory. Are there some tricks to overcome this problems, or is there a rule of thumb of the dimensions to be processed?
Is processing with a cluster helping (3 nodes and 1 master)?

Looking forward to any ideas and hints.

Best regards,
Yvo