Messages

466

General / Re: twin cameras processing for aerial photogrametry

« on: July 01, 2013, 05:11:52 AM »

I guess the aerospace's question is not being well understood. If I'm right, the advantage he would want to exploit comes from the wider field of view as a result of the sum of that of a pair of divergent cameras that might overlap even by just a small but constant percent.
I would not expect this setup to work well in general because  Photoscan would try to orient each photo in the pair as an independent camera, not only in terms of intrinsic parameters but also exterior ones. The closeness of the two photo-centers in a pair would lead to weak resections with bad results.
The plan could work if you managed to split the projects in chunks that had no pairs inside. If the successive strips are flown in opposite directions then you could form chunks with every two successive pass halves.
>LLLLLLLLLLLLLLLLLLLL
>RRRRRRRRRRRRRRRR.     TURN
==============}chuk one with R photos
RRRRRRRRRRRRRRRR<
LLLLLLLLLLLLLLLLLLLL<
==============}chunk two with L photos
>LLLLLLLLLLLLLLLLLLLL
>RRRRRRRRRRRRRRRR.     TURN
...etc. This planning would require ground control for every two rows... And that sounds bad.
just thinking, a better solution could be to alternate trigger of left an right cameras so they did not fire simultaneously, this would work well if you let left an right camera overlap over 30% a variant that could also be possible with one single camera capable of tilt from left to right in cycles... Mmmm sounds feasible, but still much more complicated than shortening the focal length of your camera.
By the way, what cameras are you flying with?
Regards

467

General / Re: Simple workflow question - build geometry?

« on: April 09, 2013, 05:20:24 PM »

this is not correct. The actual bundle adjustment (which computes camera calibration and orientation parameters and creates the sparse point cloud) happens during camera alignment. The sparse point cloud contains true 3D points. Nothing 2D here.

Let me respectfully insist in that it is correct from my point of view.
Feature points are detected in every single image (2D) before orientation, and are paired by comparison of clusters of pixels (2D to 2D).
lots of statistically-approved couples (x,y) & (x',y') enter in equations that yield solutions to unknowns:  lens parameters and also center and attitude of cameras and at the same time x,y,z for the feature points are calculated (so, no role is played by the third dimension until now). So the result is 3D but features are two dimensional as they are just singularities in the appearance of a certain groups of pixels not necessarily linked to actual shape-features but to photo-features.

468

General / Re: Simple workflow question - build geometry?

« on: April 09, 2013, 02:33:01 AM »

Does "build geometry" improve the accuracy of the original point cloud points or does it just add more points/facets?
Thanks.

Yes. As far as the sparse cloud is built on matched "feature points", it does not have nothing to do with the actual terrain topography but just with some points identified as features because of their appearance in the photos (vicinity contrast or whatever that made them distinctly identifiable for some detection algorithm). You should understand this is just a 2D step
This nature makes them well-suited to provide optimal correlation values and thus are great for inner and outer camera orientation parameters estimation.
If you build geometry you should get a better terrain description even after a strong decimation (now the third dimension matters) to a number of points below the population of the initial sparse cloud.

I wouldn't expect to infer good breaklines from the sparse point cloud but on the contrary you would see them neatly rendered after dense model decimation.

Regards!

469

General / Re: UAV photos look reversed after align (canon 1400sd)

« on: February 19, 2013, 04:34:08 AM »

1 yes it is normal... or let's better say it is the way it happens. I find it weird too
2 Not normal at all. Help Pscan with three or more pass-points (shared markers between chunks).

470

General / Re: 5 or 6 tries, no results! Basic questions here

« on: February 19, 2013, 04:27:39 AM »

I would suggest one more:
If your subject did not fill the image frame and most part of photo is out of interest you could add good extra-actors (simple stones may help) around your object. Filling the photo frame as much as possible you will get better results and you can mask unneeded parts after alignment is done.

471

General / Re: Modeling a bomb/mines field

« on: February 08, 2013, 03:53:53 AM »

Differential GPS is not possible on cheap UAV 

If by cheap you mean trimble X-100 I'd say you are wrong in two senses. 1. xwing is far from cheap, imho. and 2. dgps for sub-dm accuracy is actually possible on board of small uas... just wait a couple of months. It's gonna take longer for kinematic, but is very on the way for post-processed solution.

On the other side, have you considered manned solutions like parafan or ul (ultralight aircraft)? this could fit perfectly and you could carry with you surveying-grade gps stuff.

472

Feature Requests / bounding box

« on: February 06, 2013, 04:20:42 AM »

• bounding box bottom is too subtle
• once the model is georeferenced, bottom should be aligned to xy plane automatically
• once the model is georeferenced, box rotations about x & y should be locked unless user wanted them free. A contextual menu with lock-toggles for each axis could do while allow panning is not a must here
• render the bottom face differently when is facing up
• ... last but not least. Dear Agisoft team, please, be my heroes!:  Could you please consider using different cursor styles suited to each tool/context?

Best regards
Jos? Mart?nez (Geobit)

473

Feature Requests / Re: Cut and trim by masks

« on: November 25, 2012, 09:51:49 PM »

I tend to agree this could be a very good idea but with exceptions (so mine is +0.99). I've noticed that oversizing masks helps to get better geometry. I mean that if you mask object zones nearby the very silhouette or intentionally mask image regions corresponding to object's faces that are very oblique to point of view, this improves geometry and help to get cleaner model. In this case if you trim the model by-mask, you'll actually destroy some good geometry parts ...
So, a good idea that needs to be fine tuned and carefully engineered

474

Feature Requests / Re: "base height field"

« on: November 25, 2012, 04:33:54 AM »

I guess that you might be wrong. height-field does not model the top envelope, if it was that simple you could just rotate your cloud 180 about x or y and model it as height-field and then rotate again and invert face normals... but my bet is that you would have the very same surface.
Height-field is agisoft's name for one implementation of some of the classic 2,5D triangulation algorithms,  probably Delaunay.

I'm not saying that your proposal is a bad idea, but just wanted to suggest a workaround.

475

Feature Requests / lasso selection

« on: November 24, 2012, 05:10:37 AM »

I voted for an irregular form selection tool, so THANKS.
But I just can't manage to do fine with the lasso tool unless you have a very steady hand or a digitizing board instead of a mouse. So I'd be thankful if polygonal selection by entering vertices of a fence was also possible as it works for image-masking

476

Feature Requests / Re: "base height field"

« on: November 24, 2012, 04:50:37 AM »

Have you tried exporting points in *.LAS format and processing them with whatever you used to generate terrains from lidar data?
Some time ago I asked one of my collaborators for a routine that sub-samples a point cloud in this way. It superimposes a regular grid on top of the cloud with the step being in size N times the typical point spacing in the cloud.
Each prismatic cell encloses a number of points usually between 0 and NxN so  if it keeps the one with the lowest z you are likely taking the ground of the sample (you can also choose keep ceiling optionally) ?does it make sense for you?. I could try with this very same dataset for you, so if you liked we could speak about how to get a copy of our program.
Anyway I suppose that much better approaches must be built in Lidar programs such as Terrasolid.
Regards

JMR

477

Feature Requests / refining markers

« on: November 08, 2012, 10:45:27 AM »

IMHO there is a need for improvements in the gray marker's refining workflow

I hate that they move when I just wanted to validate their current position.

In my opinion, when you click on the marker, you could open a magnifier window with accept and skip icons where the user could reposition the marker by dragging, validate the gray position as is calculated and turn it blue, or skip to close the magnifier letting marker remain in gray status. (Options for magnifier might be: magnification factor, show upsampled and border/contrast enhanced crop, crosshair cursor on button down, automatic target detection by key press... )

Thanks

JMR (was LFA)

478

General / GCP faster (Edited)

« on: November 06, 2012, 10:16:54 AM »

I have not seen this posted before... well may be not that unseen 
My trick to save time when using GPC
1. Align your project without GCP
2. Create blank markers using the right mouse button on the corresponding branch of the object tree. You must create as many as you have in your project. (To save time, I recommend using the same naming scheme for field collection work. (Since pscan 0.9.x you can just go to step 6 as it creates automatically ALL empty markers during import, and then back to 3... sorry, I didn't notice)
3. Choose three of your project GCP that frame your work area.
4. Locate at least two pictures containing each of the three GCP
5. Place markers on each photo selected by using the right mouse button on the image. This makes six unassisted marker placement.
6. (skip if done before) Go to the Control Points dialog, define its coordinate system using the settings button and then import an ASCII file containing: name, x, y, z. for each GCP
7. Click the Update button and observe gray markers for all GCP on their tentative locations.
8. select one marker Filter images by marker and refine the position of each gray marker dragging it to its exact position. Occasionally you can click Update to improve the absolute orientation of the model and thus the remaining gray markers will approximate better to their precise positions.
9. repeat step 8 with all of your GCP.
I hope this hint help you to save some time!

I take this point to suggest AGISOFT team to improve the gray marker's refining workflow. I hate that they move when I just wanted to accept their current position.
In my opinion, when you click on the marker, you could open a magnifier window with accept and skip icons where the user could reposition the marker by dragging, validate the gray position as is calculated and turn it blue, or skip to close the magnifier letting marker remain in gray status. (Options for magnifier might be: magnification factor, show upsampled and border/contrast enhanced crop, crosshair cursor on button down, automatic target detection by key press... )

Regards,

JMR (Geobit.es)

479

Feature Requests / Advanced selection. Was "Cull/Cut by Low-Density tool."

« on: September 07, 2012, 11:33:03 AM »

Hi all:
Thanks Alexey, that was a very good feature addition
I have a more ambitious suggestion. I'd call it Advanced Selection Tool dilalog
On top three Radio buttons for Selection mode: "select" "deselect" "toggle"
Main select box for: Islands, Triangles
Options for "Islands" aka "shells":

• With poly count >= than X
• with poly count <= than Y
• with poly count between X and Y

Options for Triangles: by area, by aspect(Lmax/Lmin), by diedral angle, by longest edge lenght

• value >=X
• value <=Y
• value between X and Y

This would help to clean meshes by deleting "spikes", noisy areas, and the so-called low density areas

Apply button should run selection over existing selection set if existed according to selection mode: "Select" will add new elements to current selection set, "deselect" will do the opposite and "toggle" would alternate current status for selected and non selected elements. A similar selection mode behaviour should also be available for all already existing selection tools in addition to the hopefully comming very requested irregular fence/laso selection.
Very best Regards!

JMR (Geobit)

480

General / Re: arc of point clouds

« on: August 31, 2012, 01:46:16 PM »

Bad calibration.
Autofocus enabled?
Sensor-shift or lens shift IS enabled?