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 - Bruno Andrieu

Pages: 1 2 [3] 4
General / Re: Big Holes after align pictures
« on: April 13, 2014, 01:20:17 PM »

From my experience with plants, wind is a serious issue.
You could get missing parts for branches moving with wind;

Also, when building the mesh, try the "mild filtering" .


General / Re: calculating area of a plant
« on: April 06, 2014, 04:40:52 PM »
Thanks Foodman

I am ok with one side if this is always the case.  I would like to be sure of what is done, so that my calculations are right.

For instance the two sides of a leaf may differ more or less in texture or colour. And generally they are illuminated differentely. I would like to be sure that this cannot result in representing the two sides by different polygons.


General / calculating area of a plant
« on: April 06, 2014, 12:17:50 AM »
I build 3d models of plants, from which I would like to calculate, for instance, the cumulated area of leaves.

Is leaf area  precisely given by the cumulated area of the triangles that represent a leaf ?
For instance, a leaf is very thin, but I am not sure wether both sides of a leaf are represented by a same triangle or could it be that they are sometime different triangles for each side of a leaf ?

thanks ....

Bug Reports / interpolation results in an incomplete mesh
« on: April 05, 2014, 11:40:26 PM »

I repeteadly obtain incomplete meshes when using interpolation (defaut) compared to no interpolation.
This is whatever the number of cameras or the visual quality of the point cloud.
I am working with plants, meaning thin objects (leaves)

I attached two captures showing meshes built from the same point cloud; only difference is enabling or not interpolation.

Any hint ?


Thx for this hint.
In my case the focus distance is always between 1m and 2 m, so (hopefully)  there should not be to much of this.

Since  Agisoft propose the lens software, I would expect that they are significant cases where a fixed calibration would be the best choice but I am usure if I am in this case.

Hello Alex

I understand.  Does it mean also that photoscan will not use the imported camera positions , and will recalculate the camera positions ? (I use fixed calibration option for the lens). I would like to be able to impose imported camera positions as well.

Hello Bruno,

After you import cameras you need to use Build Points dialog. If you cancel that no tie points will be found and PhotoScan will not have any imformation regariding overlapping images.


After having build a point cloud, I exported the camera (positions) and calibration.
Now I build another chunk with the same images (a subset of them).  I  import cameras position and calibration. I do not re-align cameras because I want to use the imported positions.

But photoscan cannot build the dense point cloud in these conditions. What is wrong ?

My question is about using or not a fixed camera calibration.
I repeatedly use a same camera  (Nikon D5200) with a same single focal lens (50mm)
There is typically 0,5% variatiability in Photoscan estimated focal lenght between independent calibrations on similar scenes. I expect this to come from the calibrating process plus some flexibility in the camera+lens systems.
If for instance the true camera+lens characteristics would be stable within 0,1%, I should probably use camera fixed calibration.

I would like to know
- Is this 0,5% variability in the range of what is expected  (or low or high ?)
- Does someone knows what is the part of variability expected from  the camera+lens characteristics in my case (camera is attached to a monopod; variability over a period of several weeks, with usually mount/dismount of the lens, some change in air temperature, etc ) ?
Thx for any hint


Bug Reports / Mild filtering filter more than moderate filtering
« on: March 26, 2014, 11:07:53 PM »
I generated a model of a pot containing wheat plants. With the same images I generated point clouds with several accuracy (ultra high, high, medium) and compared between mild and moderate filtering.

I observe consistently that dense points clouds generated with moderate filtering have more points than dense point clouds generated with mild filtering. Moreover these extra points are often artefacts. I expected  the opposite behavior . Is there an explanation?

Attached are capture showing the pont clouds generated with "ultra high" accuracy and mild or moderate filtering.
The yellow margins around the leaves (specially in moderate filtering) are artefact coming from the soil background.  They are more of them in "moderate" than in "mild". You can also see that the sides of the pot, which lacks good texture, have been eliminated with "mild" but not with "moderate" filtering. So I am confused by this.
BTW : what would be the best method to avoid the margins coming from background ?


Feature Requests / Comparing results obtained with different options
« on: March 22, 2014, 05:40:06 PM »

I am using photoscan for reconstuction of plants and comparing how various options change the results (which are very nice actually !).
At the moment for storing results , I  create duplicated  chunks containing the same cameras. Also  I cannot reorganize the order of the chunks in the list

It would be great to have facilities for organizing the work space in directories
(directories of chunk, and within a chunk, directories for storing several dense clouds or models, etc.


Feature Requests / camera tools
« on: March 21, 2014, 06:18:35 PM »
I frequently use a large number of cameras, for which I have not full control of the position. Then I must delete or disable redundant cameras based on image quality.

It would be very usefull to be able to select a camera based on their position on the model window (ie right click on the blue rectangle), get the information about camera (eg image quality, etc) and diseable or delete that camera.

it would be nice that user can define (in preference) the size of the rectangle showing a camera.
it would be nice that the photos pane contains columns showing camera position.

Feature Requests / Re: Request: Dense Cloud Filtering
« on: March 21, 2014, 06:00:47 PM »
+2 for me !!

Bug Reports / Re: no pointcloud is produced when using openCL
« on: February 18, 2014, 05:02:27 PM »


It seems as we have hardware issues with the graphic card.
Sometimes a hard life to be a newbye...

Thanks for feedbacks,


General / Re: Generating 3D point clouds in forested areas
« on: February 13, 2014, 12:40:01 AM »

What is the pixel size, compared to a leaf of a ramet diameter ? I would think you need pixels to be very small compared to these elements.

can you post some images ?


PS I am trying 3D reconstruction of wheat canopies, with pixel size of about 0,2 mm

Bug Reports / Re: no pointcloud is produced when using openCL
« on: February 12, 2014, 11:54:35 PM »

I have ask our informaticien to run these tests and hope to report soon.


Pages: 1 2 [3] 4