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 - Faulty_geo

Pages: [1]
General / Re: reset selected cameras removes tie points
« on: April 18, 2021, 03:12:15 AM »
Hi Alexey,

I'm having problems now aligning any project with "film camera with fiducials" checked. I've started a new project to try and remove some of the variables. This problem only occurs when I have the "film camera with fiducials" checked in camera calibration, but works fine when I just go straight to alignment (see attached image) without any camera calibration information and without checking the box.

These are my steps:

1) load photos into project (these are high-res scanned air photos)
2) import masks (I saved them from the old project)
3) in the Camera calibration window:
     select all camera groups, select "film camera with fiducials"
This is where the problems start (second chunk in attachment image). I did a third trial by also including the following steps (third chunk in attached image):

    merge all calibration groups
     add fiducial locations and focal length
     fix "f" parameter
4) add fiducials to photos (these were exported from the original agisoft project that was malfunctioning. A colleague wrote a python script, but they all seem to have imported fine)
5) align photos
     accuracy high
     generic selection
     60,000 key points, 10,000 tie points
     ignore stationary points

The processing takes a reasonable amount of time given the size and number of photos. All things appear to work until the end- when it says 0 of the cameras have aligned. What weird is it says I have points in the reference frame when I click on the project, but none of them display since none of the cameras are aligned. (see attached image)

This is all very strange because I've run a gazillion of these types of models before, with the same photos even, and only now and hitting this problem.

The attachment shows three chunks, labeled with the differences. The only one that works doesn't have any camera calibration information (first chunk in attached image).

Let me know if you have any ideas. I'm using v1.7.2

General / Re: reset selected cameras removes tie points
« on: April 11, 2021, 11:34:29 PM »
Hi all,

Sorry I took so long to get back, I was trying to figure out why this problem only seemed to happen sometimes.

I think the missing tie point problem occurs when I have the "film camera with fiducials" box checked in camera calibration. After the initial alignment, I had set some of the camera parameters (f) but didn't have any of the fiducials placed. I'm guessing that caused things to glitch when I tried to realign the cameras without fiducial information.

I originally went down this path because I had problems with cameras not aligning in my original project (one where I had placed fiducials). l need to spend a little bit of time trying to figure that out first before I ask more questions.

I think this problem (missing tie points) is due to missing fiducial information.

General / reset selected cameras removes tie points
« on: April 08, 2021, 07:08:33 AM »

This might sound like a stupid comment, but after performing a camera alignment, if I select the cameras, right-click, and choose "reset select "reset camera alignment" all of my tie points are removed. It used to be that when I chose this option, I would keep the tie points and could align the selected cameras again. This cut down a ton of reprocessing time because I already had the tie points.

I'm not sure if this is a version issue, but I recently updated to 1.7.2.

Is there a workaround, where I can make adjustments to the model parameters/gcps/whatever, reset the alignment (but keep tie points), and then realign the cameras? I like to make several chunks and try out different things, and if I have to start from scratch every time, it takes a really long time.

Feature Requests / Re: batch editing for camera calibration groups
« on: July 22, 2020, 05:21:42 AM »
Thanks! I could have sworn I tried to merge the calibration groups and it wouldn't let me because the images are different pixel sizes, but maybe I tried to do something else. This solution worked.

Feature Requests / batch editing for camera calibration groups
« on: July 01, 2020, 04:26:35 AM »
Hi, I've been using high-res scans of historical aerial photos and would love to see a couple of changes in a future version of Agisoft. It's possible I just haven't discovered the solution for some of these and they already exist, so sorry if that's the case.

Camera calibration dialogue box:

     - Batch selection of calibration groups for editing: photos get split into many groups with scanned images even if they’re the same survey set because scans may have a slightly different number of pixels. It would be helpful to be able to select all the groups to edit variables that are the same, like focal length, fiducial names/labels and coordinates, etc. If I try to mass-edit the information by selecting multiple groups at once and adding information it causes the program to freeze for a while.
     - The problem mentioned above is fine for checking the “film camera with fiducial marks” box, I can select all the groups and check that box and then run the auto-detect for fiducial.
     - Separate the auto-detect of image border masking from the fiducial recognition in case fiducial types aren't recognized, but the mask extent is.
     - Some fiducial types aren’t supported, such as "dot" fiducials in the center of the image borders (e.g., top, bottom, left, right). For example, from for Zeiss RMK cameras.

Hi Alexey,

Thank you for the tip. I upgraded to Metashape 1.6 and tried disabling the markers. I picked up where I left off in the workflow I described above, where the model was not optimized again once adding in a few local coordinates to generate X_est, Y_est, Z_est.

I also tested if the marker locations would change if I optimized the model again, with the markers disabled. I figured since I already optimized the model once with scale bars and no markers, the estimated marker coordinates should not change if they were disabled.

The result: the marker estimated coordinates only changed a little bit (generally less than 0.05 mm). My guess is this change could be explained away by updating Metashape versions, if the way the models are optimized was updated between 1.5 and 1.6. I am satisfied enough with this small change, since it's low enough for the project I'm working on to be inconsequential.

I think this is issue is resolved, since I have found a way to get estimated coordinates within the model by only using scale bars to constrain the model geometry.

This appears to be working, I now have estimated XYZ for all markers, even the ones I didn't assign coordinates to.

One more question: Since I don't want the markers to influence geometry, does it make sense to skip optimizing the model after adding the XYZ for 4 markers? For example:
1. place scale bars and optimize
2. place markers, export locations for marker X, Y, Z
3. clear locations/reference for marker check points
4. set marker accuracy to 10 m
5. import local coordinate locations for 4 markers using file from step 2
6. export auto-populated X_est, Y_est, Z_est. Do not optimize.

It seems like if I optimize at step 5 the marker locations will slightly influence model geometry, even if the marker accuracy is set to something as high at 10 m. Or am I thinking about this wrong?

The key problem here is that none of the markers have estimated locations (X est, Y est, Z est) and at no point in the process that do those field populate. When I place a marker on a photo, it generates X, Y, and Z coordinates but the estimated field are blank. The markers do not have outside control- they are added to the model as a check point and have not been surveyed, etc. I want the location in local coordinates.

This was true before clearing the coordinates as well. They do not appear after optimizing the model, or as far as I can tell, making any changes. The only items that have estimated values are the scale bars (est distance), which are "checked" in the reference pane.  I've attached an example of the screen when estimated values are selected from the reference toolbar.

I'm using 1.5.1 because I do not have administrative privileges on this computer to update the project. Also, I started these projects before I know 1.6 was available and did not want to switch part way through to keep everything consistent. Do you think updating will fix the problem?

Thank you for your response! I thought if I didn't check the markers in the reference pane, they would remain independent.

A couple of clarification questions:
1. When you say "clear location for reference point," do you mean right click the marker and clear? This removes XYZ and keeps the markers on the photos, but when I export the markers there are no coordinates, which I still need. I was hoping for estimated X, Y, and Z.
2. I tried your suggestion and added a marker to the project tree, then placed on the photos. However, when I export the markers I have the same problem and have no X, Y, or Z information. I'm using the "export" icon and selecting "markers" in the dialogue, leaving all options checked including "save estimated values."

You are right that these two options seem to have the same end result.

I still want to know what the coordinates are in model space (local coordinates), based on the geometry that Agisoft generates. I realize they are not real-world coordinates since I'm not adding coordinate information, but that's sort of the point.

I am using coded target scale bars to control model scale and geometry. I am running several different models with the same photo set to test changes in accuracy with different scales.  I am using the check point markers to understand model differences.

General workflow: align photos, find coded targets, create scale bars, update and optimize cameras, add markers as check points.

The check point coordinates (local coordinates) show up as soon at I place them, but if I adjust the location in any other projections, the coordinates do not change. If I move the original marker I placed, the coordinates don’t change. If I change the scale bars and re-optimize, the coordinates don’t update. If I keep the check point markers and realign the model from scratch, the coordinates don’t update.

Is there any way to verify that the coordinates are accurate? I seems they’re pulling XYZ from a single photo’s estimated location. I’m using Metashape 1.5.1

Pages: [1]