Forum

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 - gEEvEE

Pages: [1] 2 3 ... 5
1
General / Re: Did the SfM or Reprojection Error computation change?
« on: March 03, 2023, 11:58:19 PM »
Dear Alexey, done! Cheers for any insights. Geert

2
General / Did the SfM or Reprojection Error computation change?
« on: March 02, 2023, 06:12:16 PM »
Dear all, I have been processing photos with the old Metashape (v. 1.8.4).
Now, I have processed the same photos on the same PC with two different Metashape 2 releases. All images are identical, all processing parameters are identical, but still, version 1.84 gave me an RMS Reprojection error (RE) of 0.66 (2.8 px) and in the 2.0 releases, it is always 2.1 (6.4 pix).

I tested this with several datasets, and the RE is always much higher in the 2.0 versions. So, did the SfM algorithm change, the mathematical model behind the interior orientations or the way that RE is computed? I tested this with GoPro images (so fisheye model).

Cheers, Geert

3
General / Re: Render to image files rather than video
« on: March 10, 2022, 01:08:18 PM »
Dear Paul, this worked like a charm! Cheers!

4
General / Re: Render to image files rather than video
« on: March 10, 2022, 10:35:34 AM »
Dear Paul, thanks for your input.  I could make it semi-work. My renders are still done from the positions of the real-world cameras, not the positions defined as keyframes in the camera path.  Any idea on how to get this? Cheers, Geert

5
General / Render to image files rather than video
« on: March 10, 2022, 02:57:45 AM »
Dear all,

I need to render artificial images from a textured 3D scene . I can create a camera path and then render that into a video file. However, I want to render the views from the camera positions to image files. Let's say I defined a path with 20 camera positions, so I want to end up with 20 images, each rendered from the defined camera positions. Right now, I make a video of 1 second that has a framerate identical to the number of camera positions, and then I extract individual frames. This is time-consuming and suboptimal, hence my question.

Thanks for any help! Geert

6
Python and Java API / Re: Asking user multiple input in dialogue
« on: August 05, 2019, 09:40:58 PM »
Alexey,

You're a legend! many thanks.
One question: why do you use the <return 0> or <return 1> for?
Could I use <raise Exception("some string")> instead?

Cheers,

Geert


7
Python and Java API / Re: Asking user multiple input in dialogue
« on: August 05, 2019, 04:22:29 PM »
Hi Alexey,

yes, such an example would be very, very helpful. Besides, I would also like to mimic a part of the batching dialogue (the part in which I can select specific chunks).
If you could share some pieces of code without much effort, that would be great!

Cheers,

Geert

8
Python and Java API / Asking user multiple input in dialogue
« on: August 05, 2019, 02:20:46 PM »
Hi All,

I want to make a dialogue in which I ask the user something (with Metashape.app.getInt) and also ask if this should be applied on all the cameras or only the selected ones. So all this should be in one dialogue, not in consecutive ones. How can I program this? Does Metashape support this?

Cheers,

Geert

9
General / Re: Work Flow for GCP
« on: December 18, 2017, 07:30:53 PM »
Hi JMR,

of course, the amount of GCPs and their distribution are as important as the accuracy of their coordinates as well as the accuracy by which you indicate them on your images (and - not to forget - the fact that you have to include them in your bundle adjustment, and not just indicate them after the SfM step and simply continue).

The relationship between all these variables is non-trivial and as such, there are indeed no hard guidelines. However, the following article in the Journal of the American Society for Photogrammetry and Remote Sensing should shed some light on this issue: http://www.asprs.org/a/publications/pers/2017journals/PERS_April2017_Flipping_Public/HTML/files/assets/basic-html/index.html#255

All the best,

Geert

10
General / Re: Agisoft PhotoScan 1.4.0 pre-release
« on: October 23, 2017, 09:43:04 AM »
Hi All,

Am I blind or did the extremely useful "Gradual selection" option just vanish in the latest release?

Cheers,

Geert

11
General / Re: Tetracam multispectral imagery processing
« on: July 04, 2014, 02:57:28 PM »
Hi All,

the only solution to do low-cost multispectral imaging at high spatial resolutions is to use two reflex or mirrorless cameras. Whatever LDP LLC (MaxMax) or other companies claim, the current Bayer patterns that cover the silicon sensor make it impossible to create a decent (i.e. spectrally pure) NIR - Red - Green camera.

Some years ago, such an approach was tried by the Eastman Kodak Company with their now discontinued Kodak DCS-200 CIR, DCS-420 CIR and DCS-460 CIR. Although their spectral response approximately matched Kodak’s NIR-sensitive emulsions, both these film emulsions and these digital cameras acquired spectral information in very broad wavebands, hence taking significant portions of the other spectral bands into account: the red sensitive photosites (and to a lesser extent also the green channel) take large portions of the NIR into account, a drawback Kodak counteracted by subtracting the NIR’s Digital Numbers (DNs) from the initially captured green and red DNs. However, this approach could never yield pure spectral information.

Because the spectral response curves do not even coincide on the long wavelength side, it is impossible to remove the precise NIR-contributing part of the green and red channel. Moreover, the NIR and red channel still took a significant portion of the red edge region into account, which made these broad band imagers less suitable for quantitative spectral analysis as they masked unique spectral features to a large extent. As also the analogue media were characterized by a low spectral fidelity, Kodak’s NIR-enabled digital cameras emulated the Kodak CIR film rather well, and were therefore often used in several vegetation studies.

Current digital sensor do, however, suffer from the same spectral overlap (see the figure of a bare Nikon D80 sensor that I measured several years ago). A two-camera system based on a NIR-modified with an unmodified camera can, however, deal with these issues. In this solution, the spectral fidelity of the acquired information is, obviously, superior to the CIR film and any current, full-spectrum modified camera.

A possible solution to capture CIR imagery with only one DSC was presented in the United States patent 20060066738, in which Hershey and Zhang proposed a Colour-Filter Array pattern consisting of four different coloured filters, three passing the blue, green and red visible bands, while a fourth is dedicated to transmit pure NIR or UV radiation. It is, however, highly questionable whether any company will ever market such a device, in which case a multispectral imaging system utilizing a multitude of cameras would still offer a higher resolving power and more possibilities in choosing spectral combinations.

Cheers,
Geert

12
General / Re: Photoscan & Australis camera
« on: June 20, 2014, 11:44:31 AM »
That is absolutely normal as the distortion formulas are different. Using lens, you can convert from Australis to PhotoScan (Tools > Convert).

Cu, Geert

13
Hi BigBen,

I did this already many times to document paintings and mural frescoes.
  • Just print some PS markers and place them around you object to be digitized;
  • Measure the distance between 7-10 markers (the longer the distance, the better);
  • If the object is perfectly flat, place three markers so that they form a perfect right-angled triangle. If the object is not flat, place two rulers alongside the object in a right angle. Make sure that these rulers are flat;
  • Then, photograph the whole object with an overlap of 80 %, using a good macro lens (do not zoom or change the focus);
  • Import all images in PS and let PS look for the markers (verify afterwards as some markers might be place wrongly);
  • Align the images;
  • Indicate the distances between the markers and optimise the image alignment with radial and decentring distortion active (k1-k3, p1-p2). Before performing the optimisation, make sure that you indicate the accuracy of your markers and distances in the ground control panel;
  • Then, provide the coordinates for the markers that are in a perfect triangle (or indicate new ones on the rulers so that they form a perfect right-angled triangle. One of them you give coordinates (0, 0, 0). The other two you vary in X and in Y (e.g. 10, 0, 0 and 0, 10, 0). The exact coordinates can be obtained by measuring the distances between the markers or  by reading the distances you see on the rulers. Also set the coordinate reference system to “Local”. This allows you to make a perfect orthophotograph;
  • Build a dense point cloud (medium suffices) and mesh the surface;
  • Mask the outer parts of the images, as lenses loose sharpness there;
  • Export an orthophotograph;
   
Good luck, Geert

14
Camera Calibration / Re: Strange Calibration Parameters, or okay?
« on: May 27, 2014, 06:03:14 PM »
Seems OK to me.

Radial distortion can have both positive (outward, away from the principal point) and negative (inward) values. Negative radial distortion is denoted as pincushion distortion (since an imaged square will appear to have its sides bow inward), while positive distortion is termed barrel distortion (because straight lines bow outward). Either positive or negative radial distortion may change with image height, and its amount is also affected by the magnification at which the lens is used. It can also occur that one lens system suffers from both negative and positive distortion, like your example.

Cheers, Geert

15
General / Re: New book chapter on orthophoto production with PhotoScan Pro
« on: February 06, 2014, 02:13:16 PM »
Hi Matt, thanks. Yes, I know about the error. There are more of them, as Springer really managed to mess up despite several emails to rectify this. >:(

Pages: [1] 2 3 ... 5