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.

Topics - jan

Pages: [1] 2
General / too large marker coordinate errors
« on: January 10, 2014, 01:47:51 PM »

I'm working on a large project (5000 photos). I want to make 5 chunks of 1000 photos each, and merge these chunks later on. There are no camera coordinates, but I have 9 GCPs.

Thus far I aligned, made a mesh and optimized a first chunk. 4 GCPs are visible on this chunk, and the GCP error coordinates are very good (errors are less than 0.005 m, while the chunk itself is bigger than 100 m in both dimesions).

Now, for the second chunk, only 2 GCPs are visible, but these 2 GCPs are both also visible on the first chunk. The strange thing is that for the second chunk, after giving in the marker coordinates in the ground control pane, the GCP erros are huge (> 100 m for 1 of the 2 markers). How come this is possible as the marker errors are tiny in the first chunk (for the same two markers)?


General / checking all cameras at once in ground control pane
« on: January 10, 2014, 11:33:53 AM »

I'm using ps (64 bit) for the moment in win7.

I want to check all cameras in the ground control pane, but it seems not possible.
I tried to first select all cameras (ctrl A), then right click mouse, and press 'check'
but this doesn't work.
Is there another way as checking each camera manually can be a lot of work with many photo's...

kind regards,


I can run a previous stable Photoscan version ( in Linux (ubuntu 10.04 LTS), but not the last stable version (i.e. on the same OS.

Normally I double click on the file, and then press the button "run in terminal" to open photoscan (this normally works), but with the latest version, nothing happens.

Anybody knowing a solution?


General / alignment mistakes
« on: November 13, 2013, 07:31:42 PM »
Hi all,

I often encounter alignment mistakes by PhotoScan. Often Photoscan aligns subsets of photo's right, but these subsets are at angles with each other (resulting in intersecting planes). I should have enough overlap between the pictures. I mostly use a lot of pictures (2000 and more), but I have a quiet powerful PC (24 GB RAM, i7 hexacore, GeForce GTX 580) so here is no problem. Mostly I align at medium quality, and this will take about 30 to 40 hours or so to align. Do you think that aligning at high quality will reduce this problem? Is there anything else I could do? I'm using aerial photos from a desert, but there are a lot of bushes present, so there are in my opinion enough features to be able to align correctly. In matter of fact, as explained above, PhotoScan does align correctly, but only within subsets. These subsets are however at angles with each other...

General / too many overlapping pictures (how to thin them automatically?)
« on: November 11, 2013, 08:39:55 PM »
Hi, I've not been using Photoscan for a while, but I'm going to use it again now.

I've a new collection of photos (about 5000 of them) of a landscape acquired by interval-shooting from a drone. However there is too much overlap-area between the photos. Although this is of course better than too few overlap, aligning 5000 photos will take too much time. Do I have to thin this photo collection manually (which also will take a lot of time with 5000 pictures!), or can this photo thinning be performed automatically in Photoscan.

In case that such function is not (yet?) implemented in Photoscan this may be considered as a feature request... It would be so nice if Photoscan could automatically take a suitable (or even optimal in terms of processing time) subset of photos from a photo-set with too many pictures (i.e. too much overlapping area between all pictures in the set). In this way computing time could be (sometimes highly) reduced by omitting superfluous data.

best wishes,

General / "Can't build model" when trying to Build Geometry
« on: April 19, 2013, 08:13:01 PM »

I want to make a low quality model from 175 pictures, as I want to have it fast to have a preview.

the pictures were aligned well in low quality mode. However when executing the "Build Geometry..." step, PhotoScan immediately comes with a tiny pop-up window that says: "Can't build model".
The camera positions are displayed above the point cloud, and they seem to be correct. However in the cameras panel, all camera coordinates are all 0.

I'm using PhotoScan 0.9.0 build 1586

What could be wrong?

General / "ground control" pane - "view estimated" button
« on: April 08, 2013, 05:21:43 PM »
What is the purpose of the "view estimated" button in the "ground control" pane.

I want to use it to retrieve coordinates of some objects by putting flags (markers) on these objects in a georeferenced model in agisoft.

Am I getting precise object coordinates with this method?


Hi all,

I'm sure some people here make their own aerial pictures with a balloon, kite or drone (and use these as input for the PhotoScan software).

Which camera do you use/recommend?

I currently use a Ricoh GX200, but the intervalometer with this camera is limited to 5 seconds. However, I would like to have a camera with an intervalometer range starting from 1 or 2 seconds. The only one I found with this specification is a Pentax Q10, but this camera has no lens of 24 mm (35 mm equivalent) focal length, which I prefer.

Anybody suggestions?
thank you,

General / markers
« on: June 29, 2012, 08:43:12 PM »

after placing 4 markers on 4 targets on a model (from about 1000 pictures), I refined the position on each of them on individual photos. This was last week. Now I reopen this project again, and all the markers are drifted away from the target objects as viewed from the model. However if I look at the marker positions on each individual photo, they are still correctly placed.

So now I don't know whether PS uses the correct marker positions or not during orthophoto generation


General / too much processor consumption?
« on: June 05, 2012, 08:12:27 PM »
Hello again,

I'm wondering whether it is normal that photoscan consumes so much processer power during inspection of a model which is already finished (aligned, geo built and textured).

By inspection I mean: rotating the model, zooming in, moving and placing markers...

I'v a pc with a i7 hexacore processor (12 threads), 24 GB of RAM and nvidia GTX 580 gpu.

when i don't move anything the pc relaxes, but when I do something with it (even the smallest thing) all of my 12 processors go to 100% for a while. It is difficult to work this way. Consumption of RAM memory stays well below the limit (only maximum 25 % is used). I can't remember that I had this problem in the past.

Is it normal that my computer needs to consume almost all processor power for every small move of the finished model?

General / align/merge chunks with markers
« on: June 03, 2012, 10:10:10 PM »

To align or merge chunks, markers could be used in PhotoScan.

But is it necessary to know the coordinates of these markers? The idea was to put 3 markers on each of my two chunks, so that pairs of markers point simultaneously to one of the three locations.

It seems to me that the coordinates are not important at this stage, but aligning chunks without georeferenced markers did not work for me...

General / misaligned piece
« on: May 11, 2012, 03:19:26 PM »

sometimes I encounter a piece of my model being misaligned with respect to the other part of the model, while the piece is well aligned in itself (see attachment for an example).

Does anybody knows a good technique to bring such a misaligned piece back in the right position?

The misaligned piece consists of about 20 pictures, the total model consists of 1100 pictures (overlap should be ok). I've no data on camera positions.


Feature Requests / writing space for text
« on: May 08, 2012, 12:15:16 PM »
It may be nice to just have some space to write text.

for example for: project description; setup and description of markers; collaborators; source of coordinates; what went wrong, or whatever..

just a blank field where one could write text;

Of course a separate document could also be used for this, but I think it would be more convenient if everything is in 1 file.


in the manual is stated the following (with regards to the building geometry step of the model):

"If the height field reconstruction method is applied, the red side of the bounding box will define the reconstruction plane. In this case make sure that the bounding box is correctly oriented."

I guess 'correctly oriented' means for aerial images: the horizontal plane?

But how can one know the horizontal plane at this stage. It is only known after the reference system is set by placing markers with known 3D-coordinates on the 3D model? 

Does "correctly oriented" here merely means "more or less correctly oriented"? What could go wrong if the bounding box is not precisely horizontal?

another question: If I choose top XY when producing an orthophoto in Photoscan, will it project the 3D model on the red side of the bounding box, or on the z=0 plane from the reference system which is calculated from the coordinates of the markers? (probably the latter, but just asking to be sure.)

kind regards,

Python and Java API / starting
« on: March 02, 2012, 12:37:04 AM »

where to find

as I understand with my little knowledge about python, has to be somewhere,

because the tutorial begins with

import PhotoScan

I've found a map in photoscan-pro/python/lib/python3.2 with a lot of *.py files; but no file has the name

Pages: [1] 2