Agisoft PhotoScan 1.1.0 pre-release

[Alexey Pasumansky]

Hello Kjellis85,

Do you mean coded targets that were printed from older versions? If you do not wish to re-print new targets from latest 1.1.0 version, then you should check on "Disable parity" option in the Detect Markers dialog.

[andrekje]

Would you please fix the missing file extensions when saving in Linux ?
to reproduce:
save project , enter filename "test"  - and observe that there is no extension on the saved file.
This must be hell for new users that don't know what extension the file is supposed to have.
same problem for many, but not all, exports/other saved files.

[andyroo]

I think I noticed this before (earlier version) and forgot to mention, but when registering GCPs and moving from image to image with [PageUp]/[PageDn] keys, I can't turn my gray wisp into a green flag without moving the mouse/trackball. In other words, let's say I want to speed through a bunch of targets by hitting [PgDn]/<click> and I don't want to move the auto-positioned target at all. I have to carefully move my mouse off and back on the location of the gray wisp each time instead of just lifting my mouse off of the table and pressing [PgDn]/<click> [PgDn]/<click> [PgDn]/<click> [PgDn]/<click> [PgDn]/<click>.

Since for my current project I typically position around 2000 makers per flight, this painfully increases the time it takes.

But I think the same thing happens in 1.0, so I don't know if this is really a comment that belongs in the beta thread. Sorry if it doesn't.

[andyroo]

Batch Process/Save Job/Path defaults to <C:\Program Files\Agisoft\PhotoScan Pro> on my install each time. It would be great if the location of the previous Save Job path was remembered between sessions.

[mirijovsky]

Hello,

do you prepare any new functionalities for orthophoto exporting? Seam lines editor would be very, very helpful. Something similar as in Pix4D.

Best,
Jakub

[Alexey Pasumansky]

Build 2001 has been just published.

Main changes according to the latest requests:
- coordinate system selection in Batch Process -> Export Mode/Points dialogs,
- added world file option for Batch Process export options,
- added Classify Ground Points option for Batch Process,
- added class selection option for Batch Process -> Build Model,
- automatic alignment reset for cameras that could not be optimized (when Optimize feature is used).

Also tiled model export  is now optimized for network (distributed) processing option.

[ppkong]

Build 2001 has been just published.
^

I had download and found the version is 2002

[Alexey Pasumansky]

Yes, actually there were two updates yesterday

[ppkong]

Can I export model tiles use Python API?

[Alexey Pasumansky]

Hello ppkong,

Currently this export option is not yet implemented in Python API.

[andyroo]

Top progress bar on color correction doesn't show progress during individual steps.

I've attached a screenshot showing what my progress showed for I think the whole several hours of RBA. I don't know how many hours, but >3. I was largely through the RBA phase of color correction when I took the screenshot, but it showed 100% all through this phase, at least the several hours that were ongoing when I came in this AM.

This isn't a big issue but would help me get a better idea of the value vs timesink of color correction. I've been debating whether to include it or not, and I think I'll generate another ortho without when this is done to compare processing times. I've included the logfile excerpt, showing where I was as of when the pic was posted. Note that the top progress bar did reset to show progress during the rendering of individual tiles (ongoing now):

Code: [Select]

initializing renderer... tessellating mesh... done (486452588 -> 486452588 faces)
done in 1187.85 sec
Raster size: 7987x26779
Estimating mosaic occupancy...
1 blocks used from 1x1 (100%)
Rendering mosaic...
selected 486452588 faces
initializing renderer... tessellating mesh... done (486452588 -> 486452588 faces)
done in 1157.08 sec
Raster size: 49373x164680
Estimating mosaic occupancy...
1763 blocks used from 49x161 (22%)
Calculating color correction...
collecting control points... collected 16396632 points in 2409.05 sec
analyzing 2414 images... ************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************** done in 513.594 sec
estimating corrections...
disabled 0 out of 14148570 points (rms error 0.142296)
bundle_adjust: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 0.126901 -> 0.0231729
bundle_adjust: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 0.14267 -> 0.0237614
bundle_adjust: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 0.143956 -> 0.0212
disabled 102613 out of 14148570 points (rms error 0.0227518)
bundle_adjust: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 0.0216419 -> 0.0215288
bundle_adjust: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 0.0220428 -> 0.0218924
bundle_adjust: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 0.019221 -> 0.0190423
disabled 129508 out of 14148570 points (rms error 0.0214636)
bundle_adjust: xxxxxxxxxxxxxxxxxxxxxxxxxxxx 0.0213096 -> 0.0212986
bundle_adjust: xxxxxxxxxxxxxxxxxxxxxxxxxx 0.0216685 -> 0.0216544
bundle_adjust: xxxxxxxxxxx

This flight was something like 2557 12 and 16 megapixel images (2 cameras), and has taken roughly 5.5 days to process so far on a dual core Xeon 3.4 GHz with 192 GB RAM, an AMD 290x and NVidia GTX 670Ti. Dense pointcloud extracted 921 981 977 points (about 4 days), and I built a 486 452 588 face mesh from that. Which is probably overkill. Mesh generation and DEM extraction were pretty fast, but it looks like color correction took close to 24 hours.

[tforward2]

How did you fix this?

Hi Alexey,

Thanks for the quick response.
I have finished diagnosing this issue this weekend and it is not related to photoscan version 1.1.0.

This behavior is due to remote access of the computer via MS Terminal Services.

Sorry for reporting this before finishing the troubleshooting.


Orpheas

[Cavi_Ing]

Hi there,

today i have used the new version (1.1.0) for the first time. Have now performed the "Align Photos" for a aerial photogrammetry project with about 700 pictures.

My settings:

Key point limit: 75'000
Tie point limit: 1'000


I am surprised about the result. My point cloud counts only 70'000 points. For a such project i got with the previous version (Point limit: 75'000)  more than 10 Million points after the align process.

Have i done everything right?       

Can i expect the same accuracy as before with significatly fewer points after alignment?

Thx

[Alexey Pasumansky]

Hello Cavi_Ing,

Using Tie-point limit 1000 means that on every image there will be left not more than 1000 matching points. If you wish to reproduce the results of the version 1.0.4, you need to set up tie-point limit to zero, as in this case all found matching points will remain.

[Cavi_Ing]

Hello Alexey,

many thanks for the quick response.

If I understand you, i get the same results with:

v1.0.4:
point limit = 75'000

v1.1.0:
key point = 75'000
tie point   = 0

Are the TIE POINTS a specific selected quantity (best quality points) of te detected key points?

Or what is the different influence of Key and Tie - Points on the calculation?

THX for response