Forum

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - Alexey Pankov

Pages: [1]
1
Feature Requests / Re: refining markers
« on: November 09, 2012, 01:13:10 PM »
I hate that they move when I just wanted to validate their current position.
+1

Magnifier tool however seems too complicated. Scroll wheel is enough for zooming.
PhotoScasn only needs to be able to distinguish between single click and 'hold and drag' actions.
And single click of course should go only for marker validation without changing its' position.

2
General / Re: Extremely large projects
« on: October 29, 2012, 03:07:26 PM »
Mark, I think procedure you described is similar to mine.
It's just not necessary to build models in separate chunks. Of course you can build and export all partial models one by one in the same chunk. I clone chunks in order to have in the end all partial models saved in PhotoScan project.
When cloning chunks, instead of Save-Append approach I simply go for Duplicate option in chunk context menu.
In my projects I've always been able to align all photos in single chunk, I've never dealt with 6,500 photos in single project. So I'm not sure that separate alignment and merging chunks will not introduce any undesirable artifacts. However your procedure (steps 1-4) still should produce single chunk with all the cameras concordant.
And I also don't deal with OBJ-models. For now I'm using PhotoScan for orthophoto and DEM production. I noticed that if bounding boxes are adjacent orthophotos and DEM match precisely. If boxes were slightly overlap one may see small differences in overlapping areas on orthophoto and DEM. I think these differences are caused by tiny variance in mesh filtration. And I'm able to notice them only in comparison of two pictures.
I don't promise that you will get seamless model using this approach. But the mismatches should be minimal.

3
General / Re: Extremely large projects
« on: October 26, 2012, 01:59:28 PM »
In few occasions I used this procedure I didn't bother about disabling cameras. Though you're right, it could reduce amount of computation.
The main point is that all parameters affecting model (calibration, camera positions and orientations) should be the same in every chunk. That means that all the photos should be aligned and optimized first all together, in a single chunk, and when you later clone this chunk you have not to change camera/markers coordinates, neither perform optimization.
Orthophoto and DEM of adjacent chunks I got in this way matched each other perfectly.

4
General / Re: Extremely large projects
« on: October 25, 2012, 06:58:52 PM »
It is quite easy to get two seamless chunks.
Make copy of the one chunk. And then drag bounding box corner keeping the common edge intact.
When you get more chunks it's safer to have slightly overlapping boxes in order not to get a hole between them. Show Aligned Chunks tool lets to see all chunks and all the bounding boxes at once. You can use Disable/Enable Items buttons on Workspace pane to turn separate chunks on/off.

5
General / Re: orthophoto from standard editon
« on: October 25, 2012, 06:49:10 PM »
Hmmm, I just played with changing the bounding box; this does not have any effect on the created orthophoto. Must me something else.
I also tried to change (orientation of the) bounding box.
When you change it after model is build it doesn't have any effect. But if you change bounding box, rebuild model and export texture, the texture plane corresponds to PS base plane very well. So for me it looks like PS projects texture on the base plane that it 'remembered' from the model generation.
However your experiments with changing point coordinates break this theory.
I dont't have other ideas.
Will wait when you turn Standard into Pro :)

6
Feature Requests / Re: Extremely large projects 1000+ photos
« on: October 25, 2012, 06:23:01 PM »
The remaining question is how it will be with seams, or better neighboring models from the chunks in resulting render.....
The partial models don't have to overlap.
What could be fantastic - is to 'glue' resulting separate meshes back into single one. And to implement LOD builder :)

7
General / Re: orthophoto from standard editon
« on: October 25, 2012, 05:54:42 PM »
As far as I understand how PhotoScan works, it projects texture on base plane of bounding box, the one that is shown by red frame.
PS Standard doesn't have any referencing tools, so the only way to reference such "orthophoto" is to perform it in external tools. I think every GIS has a tool for manual referencing of surface images, even Google Earth.
However, I always thought that texture size is very limited by RAM available.

8
General / Re: Extremely large projects
« on: October 25, 2012, 03:23:24 PM »
Are you building geometry in arbitrary or height field mode?
The thing is that in Arbitrary mode the bounding box turns into a cube with the base equal to the largest dimension of "your" bounding box. In case such as pictured in the first post, "real" bounding box in Arbitrary mode is much larger than it seems. This also could be a reason why you run out of memory.

9
General / Re: Best way to optimise a multi-chunk model?
« on: October 25, 2012, 02:53:36 PM »
In such a case I would make alignment and optimization in single chunk. Then cloned this chunk several times, set smaller bounding boxes and performed building geometry procedure in these chunks. It gives several partial models that would nicely match each other, since all of them are referenced identically.
As to optimization, it is a bit tricky. 20 GCPs may be not enough to get evenly optimized model, 10 meters accuracy of camera positions may be not enough to get accurate referencing of the whole model. Sometimes optimization by cameras and then update (affine transformation) by markers gives better results. Try both ways prior to geometry building and check which leads to a more accurate referencing.

10
General / Re: orthophoto from standard editon
« on: October 25, 2012, 02:34:43 PM »
In the PhotoScan standard edition, I have the possibility to create and export orthophotos.
But.. How it is possible?

11
General / Re: how to improve photo alignment in dense vegetation area
« on: August 06, 2012, 12:32:56 PM »
Dense vegetation is always a big trouble. In windy day it's even more changeable than water surface.
All tree leaves looks the same. And the same tree looks so much different when photographed from above or pictured at some angle, as on image periphery. That makes PhotoScan duty nearly impossible.
We too often have gaps on forestry areas.
Sometimes it helps to increase image overlap and exclude oblique parts of photographs. Sometimes not. Fortunately we don't have big forests and may get photo aligned thanks to some roads and rivers crossing the vegetation areas.

Starting from version 0.8.5 there is an option of using markers as user set tie-points. This works if you run "Align selected photos" from context menu. However it needs a lot of handwork, 'cos you need to find and mark many fixed (i.e. not moving like tree heads under the wind) points.
To be honest, I tried this approach on dense forest only once and didn't succeed.

12
General / Re: infite orthophoto
« on: August 03, 2012, 01:15:54 PM »
Now in heigt field, with version 0.9.0 I put 60 millions .
Well, accordig to Memory Requirements there is a good chance that 12 Gb RAM will be enough to decimate 60M mesh. Try to decimate it to few millions and then export orthophoto.
If you succeed with decimation there should be no problems with export.

13
General / Re: infite orthophoto
« on: August 02, 2012, 06:01:43 PM »
Normally such an ultra slow processing takes place when there is not enough RAM. System starts to use scracth file and it makes processing thousands times slower. So first I would suggest you to watch RAM consumption when you working on this project.

But I'm still wondering why it happened in your project.
500 x 600 meters at 3 cm per pixel is just about 300MPx image. 12 Gb RAM should be enough to process it even in single block.
Low quality model also shouldn't consume much memory. But wait a moment...
457 photos in Arbitraty mode even at low quality may produce quite large model, that consists of dozens of millions of polygons. How many polygons are in your model?

If there are too many (when just the visualization of the model consumes more than half of RAM), then you should go for a smaller model (in number of polygons). There are number of options for it:
 - height field mode (normally it produces lower polygons model, than arbitrary mode);
 - lower quality of reconstruction;
 - splitting model in separate chunks, setting smaller processing boxes in each of them, then building individual smaller size models and exporting separate orthophoto blocks from these individual models (if all the models come from the same point cloud then orthophotos should match well).

Btw, how did you come to the merged chunk - have you built the models in separate chunk and then merged the models or you built the model after merging chunks?

Pages: [1]