Hi everybody,
I Have some questions. I’m working since 2018 in the creation of 3D models and 2D of electrical substations of my company (I read a lot of articles and proved with diferentes techincs in take the photograps and processing options in a great amount of projects).
In order to obtain the orthomosaic I usually take nadir photograps with a simple grid and to process the template “3D Maps”.
Furthermore in order to obtain the dense point cloud I take just photograps at 45 using a double grid and to process the template “3D model”.
For my surpsise in the last processings (in diferents electrical power stations projects) I am having better details in the the “dense point cloud” using the simple grid and nadir photographs than the projects that I use double grid and 45° photograps.
Obviously using nadir photograps there are some details in that the camera “can´t see”.
I attach 6 photograps of the same place of 3 projects, that the photographs were taken in the same conditions light (cloudy day):
-simplegrid and 90, using 3d maps template (great sharpness, great color and recognizes the wire that is very important)
-Double grid and 45 using 3d model template (more noise than 2D and didn´t tecognize the wire, bad sharpness)
-Project merggin two projects (a lot of great of noise and just put very few wires from the first project, horrible sharpness)
I have this questions:
-If I made a double grid at 45 to obtain the the densified point cloud, what parameters are better? the 3d model or the 3d maps??. Because in the step 1 in the tab “matching” the 3maps has selected “Aerial Grid or corridor” (in this casi it is like this), and for 3d model it has selected “Free flight or terrestrial”. My case is a double grid, but I need the 3D model.
In the step 2 (according to the recomendations) I disabled the multiscale option. And what is better in the matching windows size?, I believe that in this case is better 9x9.
-Looking the images that I attach. Why the merge project have that much of noise and the quality is so worse than the others?
-Why if in the 2D project recognize a great amount of wires, and when I create a merge project (using the 2D project as input) the program just put a few amount of wire and on the other hand put a lot of noise.
-Have someone some tip to this kind of project?
Thanks for the help,
Hi, I will made the question more simple.
If I take photograps obliques in a double grid flights in order to obtain a dense point cloud of 3D model.
Is better use the 3D maps template or 3D models template?
What is the “philosophy” behind the boxes “Aerial Grid or Corridor” and “Free Flight or terrestrial”. Because 3D maps have activated “Aerial Grid or Corridor” and 3D Models “Free flight or terrestrial”. In my case I have photos from an Aerial Grid but I need a 3D Models.
Thanks.
1 Like
Hi there. These names are based on general use cases where the setting applied would be appropriate. For situations like this where you are modeling such complex geometry, it’s better to drill more into the processing options set by the templates. In general, you should not change matching settings for most projects until you have ensured that all other settings are optimal. For a scene like yours, you would want the “Aerial Grid or Corridor” to start. For a complex scene like this I would recommend starting with the standard 3D maps template and fine-tuning your settings from there. The following are some that will likely help.
STEP 1 -
- Calibration method: Standard
- Keypoint Image scale: 1 - 0.5
- Geometrically verified Matching: Enabled (Reduce noise)
STEP 2 -
-
Minimum Number of Matches: 5-6 (This will help reduce noise)
-
Multiscale: Disabled
If these settings don’t provide you a good result then you can consider diving in and tweaking the matching settings in step 1 to improve a bit more or consider flying with more overlap.
1 Like
Holdene, thanks for your complete answer. I processed using your recomendation. The problem is that when I increse the minium numbre of matches to 5 or 6, reduce the noise but also reduce the details, for example I can not see the wires.
In this case I do not believe that the problem is the overlap. Because for my surprise as I said in the first post, I had better result in the details of the wire, and the less noise using a simple grid at 90°, that double grid at 45° (obviously using 90° I do not have some details on some side).
I attach a photograph using your recomendation. Above there are another using simple grid
Thanks for the answer.
Wires are very difficult to recreate in most cases especially with complex geometry like the towers here. It may be that you should consider even higher overlap otherwise the scene may not be able to be recreated to the degree you are looking for. Can you share a screenshot of the image positions graphic in the quality report? Maybe it will tell if you have room for more images.
Thanks for the answer.
I attach the images of the two cases, and the complete quality report
:
Double Grid - 45° Parameters Holden:
NORcustomForo_report.pdf (1.7 MB)
Simple Grid 45°
NOR2D_report.pdf (1.5 MB)
Thanks. A doublegrid is definitly going to give you better results in this case. I would also expect that a higher reoslution camera may provide a bit of an improvement. generally 20MP is a good size rather than 12MP that you currnetly have.
Other than that I think you may still be able to increase the vertical and horizontal overlap a bit more and that should allow you to increase the min number of matches without loosing too much detail around the model
Thanks for the anwer.
Yes, I use double grid with 45 in order to create the 3d model and simple grid 90 in order to obtain the 2D orthomosaic. My surprisse was when I saw that using the simple grid 90 in a great amount projects the result were better in some aspects than using double grid and 45.
When you said increase the vertical and horizontal overlap. Would it be making double grids at different heights? Or what other strategy? When I use this strategy I have problems whit the heights, duplicate stuff or inclined planes.
Thanks!
Apologies for the confusion. When I said vertical it would be more appropriate to say forward. Meaning the space between subsequent images is smaller. That should help aswell as decreasing the distance between flight lines.
1 Like
