Agisoft PhotoScan 1.2.0 pre-release

[Alexey Pasumansky]

Hello Giancan,

We were unable to reproduce the problem on the latest build of 1.2 version.
Dense cloud export saves the same points you see in the model view (providing that they are inside the bounding box).

DEM generation based on the dense cloud may interpolate the missing areas, if they are visible from at least one camera. So could you please check whether you see the results of interpolation for DEM based on the cropped dense cloud?

[Diego]

Hi Alexey,

Unfortunately with the update build 2172 now I get an error when you run the script, something changed, you can help.

dist = calib.error(T.mulp(points[point_index].coord), proj.coord).norm() ** 2


Error =  AttributeError: matrix * vector: vector length does not match matrix row size minus 1

Thank you

[giancan]

The issue definitely shows up when exporting to potree. All points are exported from the dense cloud although I surely delete some of them from the dense cloud.
In the dem, this is more difficult to show but I will try to capture some screenshots to demonstrate it.

[Alexey Pasumansky]

Hello Diego,

In the latest build we've slightly changed the representation of the sparse cloud points via Python.
Now point.coord is a 4-dimensional vector. If the fourth element is zero, it means that the point is actually a ray (applicable for points detected for single camera station). Otherwise regular points have 1 as fourth element.
.mulp() function accepts 3-dimensional vector on input, so when working with the sparse cloud points coordinates they should be treated accordingly to the new representation.
I've sent you updated script via e-mail.

[Alexey Pasumansky]

I have some questions dealing with (new) parameters in PhotoScan 1.2.0

In optimize camera alignment dialog there is p3 and p4. Are these parameters additional terms (like p1 and p2) to create an aproximated function that describes the nonsymetric distortion function?

The other thing is fit aspect (before 1.2.0). What happens if its checked or not? Is it the aspect ratio between pixel in x direction an y? My idea was that optimzing this parameter forces the adjustment to create square pixels? I havent found anything in the manual...

Gradual selection:
What does projection accuracy mean?

Hello Paul,

P3 and P4 are tangential distortion coefficients with the higher degree items.

Aspect defines the relation between pixel size in x and y directions. Unchecking the aspect in the optimization dialog will keep the same ration between pixel size (and therefore between Fx and Fy).


Projection accuracy is related to the size of the matching points (that are not actually points, but some region).

[Alexey Pasumansky]

I do not know to use the console pane one by one, I use only .py files.
Find below my script.

When I put the save command (doc.save(os.path.join(dossierProjets, dossier + ".psx"),compression = 6,absolute_paths = False)) just before aligment, all other commands work, but at the end of the process, there is nothing (see attached file "window").

When I put the save command (doc.save(os.path.join(dossierProjets, dossier + ".psx"),compression = 6,absolute_paths = False)) just before buidDem command, all commands executed before are OK but there is no Dem (see attached file "window_2")

Hello Maxime,

I think it is related to the PSX treatment. After saving the project file in PSX format it is being re-loaded to the project space both in GUI and via Python. So you need to relink chunk variable to the document. If you've assigned doc = PhotoScan.app.document, it will work when trying to access the doc variable after saving. But previously assigned variables like chunk = doc.chunk will be now pointing to the isolated chunks outside the project.

[Helimax]

Thanks Alexey, it's running now 

Just a remark :

In PhotoScan Python Reference, for buildOrthomosaic, we can choise surface between OrthoSurfaceDem and OrthoSurfaceMesh.

It seems that neither works. (my syntax : surface=PhotoScan.OrthoSurfaceDem or surface=PhotoScan.OrthoSurfaceMesh)
error message : AttributeError: 'module' object has no attribute 'OrthoSurfaceDem'

But if one puts nothing, it works and it takes DEM.

Maxime

[Alexey Pasumansky]

Hello Maxime,

I think you need to use: PhotoScan.DataSource.ElevationData or PhotoScan.DataSource.ModelData in .buildOrthomosaic() function.

[bisenberger]

Split in chunks script doesn't work with current version.

http://wiki.agisoft.com/wiki/Python

[Airgo]

Hello,

buildding Orthomosaic from mesh gets slow after 3% and freezes after 10%.
Dense point cloud has 38Mio points, mesh 2,4 Mio faces. 8GB RAM, 100GB free Disc space (which is still 80GB while freezing).
In 1.1.6 the orthofoto export works without problems.

Regards,
Mathias

[Alexey Pasumansky]

Hello bisenberger,

you need to comment or remove the following line:
   

Code: [Select]

doc.addChunk(new_chunk)And also change the following arguments in the Build Mesh function:

Code: [Select]

source = PhotoScan.DataSource.DenseCloudDataand

Code: [Select]

source = PhotoScan.DataSource.PointCloudData

[Alexey Pasumansky]

Hello Mathias,

Are you using the same parameters, maybe you can post the screenshots and attach the processing logs?

[Airgo]

Hello Alexey,
I was using the same parameters (standard options).
Hers is  a screenshot and the logfile. During the first attempt my computer crashed.
Mathias

[pastouf]

What is the right flow to create the google map title.

Impossible to have real zoom.

Align
Dense Cloud
Mesh
Texture
Tile
Orthomosaic

Calcul and export Ok, but index.htm doesn't work (only google maps, no orthophoto with multiple zoom???)

Thanks

[giancan]

Dear Alexey,
as I mentioned earlier, in this beta version I experienced some problems when creating dem or orthomosaic from dense cloud. Indeed, although I "clean" the dense cloud, when I create a new DEM or ortho in different coordinate system, the results ignores the editing I did to the dense cloud, and instead it uses the original cloud.

Cheers,
G.