Forum

Author Topic: Transform for Projecting Point Cloud onto Orthomosaic  (Read 2368 times)

shwetankumar

  • Newbie
  • *
  • Posts: 20
    • View Profile
Transform for Projecting Point Cloud onto Orthomosaic
« on: March 29, 2017, 03:38:29 AM »
Hi,

I recently started generating orthomosaic images using Agisoft and the quality of the results have been extremely good. I am trying to do a small experiment for demonstration purpose. I am tusing geotagged images for performing scaled 3D point cloud generation, and following up with orthomosaic generation.
Since the point cloud quality of the model at High density resembles the orthomosaic view very closely, the idea for the demo will be to show how I can project the points of the dense cloud onto the orthomosaic in a way that they line up correctly.
I am guessing there is a transform that can take me from the scaled point cloud (in metres) onto the 2D orthomosaic photo. The transform seems to be an orthographic projection but I am not able to compute the exact projection vector along which I should get consistently good projections. I have tried some random vectors based on hit and trial values, but obtaining the actual transform will be very helpful.
The goal is to show how well the projected point cloud lines up with the orthomosaic. Furthermore, I am trying to show dimensions in each and show that the dimensions in the 2D orthomosaic is the same as the scaled point cloud.

Thanks,

Shwetank
« Last Edit: March 29, 2017, 05:09:28 PM by shwetankumar »

shwetankumar

  • Newbie
  • *
  • Posts: 20
    • View Profile
Re: Transform for Projecting Point Cloud onto Orthomosaic
« Reply #1 on: March 29, 2017, 05:12:29 PM »
Any solutions to achieve this ?

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 13297
    • View Profile
Re: Transform for Projecting Point Cloud onto Orthomosaic
« Reply #2 on: March 29, 2017, 06:41:35 PM »
Hello shwetankumar,

I think that the easiest way would be to Capture the Orhto view and than Model view, that is switched into orthographic  - TopXY projection for the dense cloud visualization. If you have transparency option turned on, you can overlay the captured points on top of the orthomosaic.
For the scale you can create a couple of scale bars in the dense cloud view mode and they will be displayed in the Ortho view automatically. They can help, when you will be combining the capture view results.

Alternatively you can try to use the exported product in the external applications and overlay there (for example, in Global Mapper).
Best regards,
Alexey Pasumansky,
Agisoft LLC

shwetankumar

  • Newbie
  • *
  • Posts: 20
    • View Profile
Re: Transform for Projecting Point Cloud onto Orthomosaic
« Reply #3 on: March 29, 2017, 06:49:11 PM »
As usual, thanks for the prompt response Alexey !

It would be great if you can elaborate a little bit on the TopXY projection ...
Is it simple a vector (x,y,z) with x and y values set to zero and z = 1 or -1 ?
Also, how can I compute the distance factor to transform along the vector ? (assuming unit normal vector to orthomosaic)
Is it safe to assume that the orthomosaic is indeed along the z = 0 plane ?

Thanks,

Shwetank

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 13297
    • View Profile
Re: Transform for Projecting Point Cloud onto Orthomosaic
« Reply #4 on: March 29, 2017, 06:53:32 PM »
Hello Shwetank,

Actually I meant that you can switch to TopXY projection in the Model view and then switch to the orthographic projection mode. Then you can use a Capture view option to get a kind of screenshot of the Model view contents.
I think it may be easier, if you need it only for the demonstration purposes in a presentation, for example.
Best regards,
Alexey Pasumansky,
Agisoft LLC

shwetankumar

  • Newbie
  • *
  • Posts: 20
    • View Profile
Re: Transform for Projecting Point Cloud onto Orthomosaic
« Reply #5 on: March 29, 2017, 08:05:47 PM »
Oh, got it.
That is definitely an easy solution to view it.

But if I was to do a mathematical projection operation, what would be the best way to obtain the plane along which I need to project the point cloud in order to get the desired alignment ?

I am trying to do the exported style of operations where I have the orthomosaic and the point cloud as separate files.

It would be great to know a transform (like a gravity vector), which can tell me the direction of projection. If I understand it right, an orthographic projection along that vector along any plane should line up with the orthomosaic.
« Last Edit: March 29, 2017, 08:20:11 PM by shwetankumar »

shwetankumar

  • Newbie
  • *
  • Posts: 20
    • View Profile
Re: Transform for Projecting Point Cloud onto Orthomosaic
« Reply #6 on: March 30, 2017, 04:52:29 PM »
Hi Alexey,

Can you please help me in getting this working as desired ?
Could it be as simple as the vector from centroid of camera coordinates to the centre of the dense point cloud ? (but of course this is just a guess)

Thanks !
« Last Edit: March 30, 2017, 05:33:57 PM by shwetankumar »

JMR

  • Sr. Member
  • ****
  • Posts: 442
    • View Profile
Re: Transform for Projecting Point Cloud onto Orthomosaic
« Reply #7 on: March 30, 2017, 09:34:53 PM »
if I understand well your desire it is to flatten dense point cloud, so why dont you just set all points elevation to zero?

shwetankumar

  • Newbie
  • *
  • Posts: 20
    • View Profile
Re: Transform for Projecting Point Cloud onto Orthomosaic
« Reply #8 on: March 30, 2017, 10:59:47 PM »
Hi JMR,

Practically, what you say is the exact operation I want to perform.   :)
However, the point cloud has (x,y,z) coordinates where z does not necessarily correspond to the axis of top-down flattening (makes sense ?).      :-\
I am asking for the direction along which I do the flattening, unless the point cloud is always axis aligned and either one of the axes correspond to the cartesian coordinate space for the depth (or elevation) with respect to the camera flight.

Since it is an orthographic projection, only the direction of projection (the true vertically down direction) should be sufficient to get the desired result, which remains an unknown.

Thanks
« Last Edit: March 30, 2017, 11:07:36 PM by shwetankumar »