Topics

1

General / 2.2.1 issue? - Sparse Point Cloud not showing in Model view port

« on: May 09, 2025, 04:30:29 PM »

Hi,

I have replicated this issue twice now using 2.2.1. I have not seen it in previous versions.

I have create a sparse cloud from 6500 images. The sparse cloud has about 8.4 million tie points. I go to do a gradual selection clean (now called Clean Tie Point in 2.2.1). After selecting 200,000 points using the Project accuracy slider and deleting them, then running optimization, the sparse cloud disappears from the viewer. The tie point counter, in the bottom left of the screen, is showing me the correct number of points (8.1 million) but there is nothing on the screen. I have tried resetting the view, resetting the region but nothing shows the tie points. And obviously the tie point button is clicked to show tie points on.

Luckily i had duplicated the chunk before I started the gradual selection clean. The sparse cloud shows up perfectly in the first chunk

Would any one have any suggestions?

Regards
Ben

2

Feature Requests / Smart Marker detections by Surveyed GCP Location

« on: May 09, 2025, 07:23:25 AM »

Hi,

Could you please consider adding this option to marker detection -

1. a csv or txt file (import reference command) is used to import the GCP locations.
2. Detect Marker command is run but with the option of only searching images within a nominated radius of the imported markers. The user would nominate the radius e,g 20 metres

Using this suggestion Metashape will only search images in the areas of the imported GCPs, which should speed up the process. Instead of every image or selected images

I have a project of 350 hectare area - 6500 images and 32 non-coded markers. I think if my suggestion  was implemented the process of detecting markers would be very quick.

Thanks for reading

Kind Regards
Ben

3

General / Error - Too small max compute workgroup

« on: May 08, 2025, 02:01:42 PM »

Hi,

When starting the "Build Model" command I get an error which stops the command. The problem is listed in the log file -

2025-05-08 20:45:23 Vulkan device Intel(R) UHD Graphics 770 skipped: too small max compute workgroup count 65536x65536x65536 (at least 524288x16384x16384 required)
2025-05-08 20:45:23 Found 2 GPUs in 0.016 sec (CUDA: 0 sec, OpenCL: 0 sec, Vulkan: 0.016 sec)

This job has only around 2000 photos. I have processed jobs many times bigger than that on this computer.

Could someone please tell me why this would happen -  "too small max compute workgroup count 65536x65536x65536 (at least 524288x16384x16384 required) "

I am using professional version - 2.2.0 build 19890

Thanks
Ben

4

General / Shifting Data from Assumed Coordinates to somewhere else

« on: November 15, 2024, 05:29:01 AM »

Hi All,

I have spent alot of time trying to work this specific requirement out, so thought I would share in case it may help others. I'm sure alot of MS users reading this will think this is obvious or have a better way. I would be very happy to hear from those users.

Problem - how do you shift a MS job from assumed coordinates (XYZ) to a either somewhere else in space or to a published coordinate system (E,N,ALT or lat,long,heigh). This solution is for when you have markers in your job that you can calculate the shift (XYZ and rotation) required to move them to required new coordinates.

By Assumed Coordinates, I mean where you have no exif location data and have just nominated a point in your MS job at any coordinate e.g. Point 1 has XYZ of 1000.000,2000.000,50.000. This example of "Assumed" coordinates could also be called "Local Coordinates" in metashape.

I have a project where I have captured the underside of a bridge with a handheld camera.  These photos have no exif location data. This is a visual inspection job so actual true location and true dimensions are not critical but it would be handy if the job was roughly in a known CRS so I need to move it.

I did have a survey staff (3metres long) in some of the photos to act as a scale bar. That scaled the project to roughly actually size.

After aligning the photos, the sparse cloud is created. I rotated the sparse cloud using "rotate object" to put the z axis roughly "up" and also roughly rotated the sparse cloud alignment to match the real world bridge direction. I only did this so it visually looked close to the real world bridge look, making it easier to work on the data. During the photo alignment I had needed to create some markers to help it align. In the reference pane I gave one marker an assumed "source" coordinate ( x=40 y =25, z=10) by writing in the Reference / "View Source" / marker XYZ . After ticking one marker running the "optimize cameras" that meant all my other markers now had "estimated" x,y,z values - These are shown after clicking the Reference / "View Estimated" button. I copied (via click,drag,hightlight,copy) all the marker's "estimated" XYZ values and pasted those values into a Excel spreadsheet and saved as a csv file. I then imported them into MS using the "Import Reference" function. So now a have a project on "Local Co-ordinates" with all markers having "Source" values. I created two additional markers at each end of my bridge. I exported all the marker values to a csv.

I then used a external website that has a map viewer and shows location values for the cursor when over the map. The web site allows the user to extract location values in many different Coordinate Systems e.g. WGS 84, MGA2020 etc. From this I could get two very rough MGA2020 coordinates for either end of my bridge (at similar locations to the additional points I created in Metashape). I used some survey software to rotate and shift all my markers from assumed coordinate to MGA2020 based on these two common points. I exported out the MGA2020 coordinates of all markers from the survey software to a csv.

I then -
10. Using the "Import Reference" command - Re-imported markers (using the new csv - When the "Import csv" dialogue box opens, if needed, change the "Coordinate System" from Local Coordinates to the defined CRS you need. In the example I have attached I am going from Local Coordinates to MGA 2020 (EPSG 7856), click OK and import the data. This updated all my source marker values to MGA2020 values
11. Under the Reference Settings the "Coordinate System" should now be showing the CRS you want (not Local Coordinates)
12. Tick all your markers on and click " Update Transform"
13. In the Reference Tab / "View Source" list - your markers should have the correct coords and labels (e.g. E,N,Altitude or Lat,Long  not XYZ)
14. The data should have moved. Also click View Menu/ Reset View to see your data.
15. If you get the error message" Error- vertical datum outside tolerance" you have a marker or markers that did not update and still on their original coords. Either delete or update these.

Notes
20. For Step 10 - if you are just moving your data from a "Assumed" CS to another "Assumed" CS to can just leave the "Coordinate System" as "Local Coordinates"
21. Obviously this method could be adopted if you have "accurate" marker values in both local and CRS. You would not have to use external software and resources.

I hope this may help someone with the same need.

Kind Regards,
Ben

5

Feature Requests / Be able to use FIND in camera list in Reference Tab using Wildcards

« on: November 02, 2024, 02:47:38 AM »

Hi,

I would like to suggest MS adds the functionality of FIND within the the Reference tab / Cameras list. This would work like any FIND function in programs like Notepad, Word, excel etc

I have projects with thousands of images (cameras). Sometimes i want to find an image or a several images within the list of thousands. I know the image name but instead of scrolling down the list, I just want to be able to  chose "FIND" and enter the partial name of the image using wildcards e.g. *0170* for image name  DJI_20241031120609_0170.JPG.  MS would then automatically scroll down this list to the first occurrence of *0170*. if that wasn't the image I could then click "NEXT" and I would be taken to the next image satisfying the criteria in the list of images.

This would be a much faster way to find images that currently is possible.

I do know that in the Workspace Tab when right clicking on the Images Folder there is the option to "Select Items" and that allows wildcards. It selects the photos but it does scroll to them. But this functionality is not available for the Reference Tab.

Thanks for reading. I hope this suggest can be implemented in a future version

Kind regards,
Ben

6

General / Detect Marker Issue or Tie Point Discrepancy issue?

« on: October 12, 2024, 11:10:06 AM »

Hi All,

I'm  using MS Pro v2.1.3.

I have carried out a UAV flight with a DJI M300 and P1 camera of a pedestrian foot bridge. I have flown nadir and oblique flights and also captured images with a Canon hand held camera. Prior to the flight we placed many black and white targets on the bridge.

After aligning the images using highest accuracy setting and then running the detect targets command, Metashape has picked the centre of many targets correctly. Metashape allocated Marker numbers for those targets.  I have noticed that Marker No 6 has been correctly identified (center of target correctly picked) in 36 images. I also notice that Marker No 14 has been correctly identified in 13 images. The problem is the target is the same target! Marker 7 and Marker 11 are about  15cm apart.

The images Metashape has picked the centre for Marker 6 are different to the images Metashape has picked the centre of Marker 14.  So for Marker 6, the marker is centered on the target in its images. And the  Marker 14, the marker is centered on the target in its images. How can this be?

In the Reference Tab  Markers list the difference is shown "View Estimated" Table  -
point 14   355681.7410382535   6308320.882072767   3.6456287776534353         13   
point 6   355681.8124383928   6308320.808663793   3.7618934840380938         36   

Does this mean the photos have not aligned correctly and Metashape thinks there a two targets in that area even though there is only one target there?  The detect marker settings were "Tolerance"  = 1 and " Maximum Residual (pix)" = 5.

Does this also mean there is Tie Point Discrepancy in the sparse cloud in that area?  If that is the case, I would have never have found this if I didn't have a target there. How could I have identify this issue with no target? It would also suggest there could be issues in other areas but I would not know until I have created a model or point cloud and could see the issue.

Please see attachment showing the issue

Any help appreciated

Kind Regards
Ben

7

General / Issue with "automatically placed" blue flag markers

« on: September 29, 2024, 09:15:50 AM »

Hi All,

I'm using the latest version of Metashape Pro (v2.1.3) for mainly processing UAV M300 / Zenmuse P1 or Phantom 4 pro data. This issue has arisen when flight subject was a Building.

I aligned the photos with no problem. From 490 images the average cammera error results were -   X Err (m) 0.004,   Y Err (m) 0.004,  Z Err (m) 0.013

After the initial photo alignment, I had to add some more photos and needed to manually pin some targets (these are green flagged)  in the photos to get these to align. As i pinned the same marker, in different photos, Metashape started to create Blue flag markers (as expected). The issue is -  these blue flag marks are nowhere near the actual marker in the photo. Metashape should be predicting very close to the correct position but it is not. They are metres out of location,

What is strange is if I delete the projections of a blue flag marker, a new grey marker is automatically created, but it is automatically created exactly over the target in the photo. As soon as I pin that grey marker, turning it green, Metashape goes and creates more incorrect blue markers in other photos . I then have to delete those blue markers (I use python code for this) and keep pinning the gray markers.

Also, you can see the marker error (pix) get worse when there are blue flags for that marker. As soon as I delete the blue flags the marker error (pix) is good again.

I have now seen this in two projects.

Would anyone have a suggestion on what is going wrong. Could it a bug in the new version?

I am using the following python code delete all the blue flags in the project. It works very well.

It was found here - https://www.agisoft.com/forum/index.php?topic=10969.0

import Metashape
chunk = Metashape.app.document.chunk
for marker in chunk.markers:
    for camera in list(marker.projections.keys()):
        if not marker.projections[camera].pinned:
            marker.projections[camera] = None


Any help appreciated

Regards
Ben

8

General / Shift project from GDA2020 to a local cordinate system

« on: September 15, 2024, 02:33:34 PM »

Hi All,

I have imported DJI M300 images into a new project. The images are on WGS84 lats and longs. I then run the "Convert" command and transform from WGS84 to MGA2020 (Projected CS - Geocentric Datum of Australia).

The images have co-ordinates roughly in the area of  Easting 376400m Northing 6353500m

What I would like to do is shift the images to a Local Coord system of about Easting 400m Northing 500m.  By doing a Easting shift of -376000m and a Northing shift of -6353000m

I then want to do all the processing (sparse cloud, dense cloud, model, Tiled Model etc) on the Local Coords. But when I'm finished shift the whole project back to MGA2020.

Unfortunately I do not know python. I have seen other posts that do shift via the console but that is mainly on imported objects e.g.

Metashape.app.document.chunk.importModel(path, shift=Metashape.Vector([-450000, -4480000, 0])) #values specific for the provided model

Could anyone help on how to shift a project from a Coordinate System (MGA 2020) to local coordinates and then back again?

Thanks again for any help

Ben

9

General / Filter by selection for sparse cloud?

« on: September 04, 2024, 03:28:20 AM »

Hi All,

I have a sparse cloud that has some misalignment and am trying to isolate with photos are in error. I can see there is a filter photos by tie points command, but I really need to be able to select a small area of sparse cloud and filter that to make it easier. Most other objects (point cloud, mesh) can be filtered by selection. Can the Sparse Cloud (or thin cloud as it is sometimes called) be filtered by selection??

It would be also be great if Metashape had limit boxes that could define the extent of the area to be shown.

Thanks for any help

Ben

 

10

General / Recovery possible on Tiled model after computer crash?

« on: August 08, 2024, 10:17:47 AM »

Hi All,

I was creating a tiled model. It had been processing for 40 hours and there was an electricity blackout!!. Worst was the UPS didnt kick in and computer shut down!
I have not started Metashape yet. Is there anyway that Metashape may have done a incremental save? Can I do anything to recover the 40 hours of tiled model processing?

Any help very much appreciated (especially some good news  )

Ben

11

Feature Requests / UAV Mission Planning

« on: May 05, 2024, 08:57:06 AM »

Hi,

I would like to suggest the UAV mission planning tool functionally is extended to include nadir, oblique and linear route mission creation.

Currently Metashape allows for mission planning and creation by exporting waypoints.  And this works, but the missions can take a very long time compared with being able to create a constantly moving UAV following a route.

It would be great if Metashape has similar mission planning to DJI Pilot 2 app.

Kind Regards,

Ben

12

General / Close Holes in Mesh - by selection

« on: March 26, 2024, 02:06:34 AM »

HI All,

I'm using Metashape Pro v2.1.0 build 17532.

I have  UAV collected images of a civil site with stockpiles, vehicles and machinery. I have created a mesh / point cloud / DEM / Ortho. I was testing the best / fastest way to remove the vehicles and machinery from the mesh. 

I viewed the mesh in wireframe and deleted triangles around a vehicle. Leaving a hole in the mesh.

If I select the triangles just around the hole and run the close holes command, Metashape seems to be able work out that there has been a area selected because the "Apply to selection" option is automatically ticked. I leave the "Level" slider set at 100% but when I click "OK" nothing happens.

Does anyone know what I'm doing wrong?

Interestingly, if I select nothing and run the "close holes" command Metashape analyses the whole mesh and finds the hole! It takes a while though!

Any help appreciated

Ben

13

Feature Requests / Cameras State Manager

« on: March 16, 2024, 08:04:05 AM »

Hi,

I think it could be very helpful to have a "Cameras State Manager" - maybe needs a better name. This would work like AutoCad's "Layer State Manager".

You would open the Cameras State Manager from an icon in the Reference Panel. Maybe next to the "Update Transform" icon. It would be also listed under the Tools Menu.

The Cameras State Manager would enable the user to save the current "state" of the Cameras in the Reference panel to a name. This would save the "state" of all variables for every camera in the Cameras Reference Panel. If all cameras are ticked - that state could be saved as the  name "All Cameras Ticked". If all cameras are unticked - that state could be saved as the name "All Cameras Un-Ticked". The power of the layer state manager would come when the user has some cameras ticked and some unticked, because this state could be saved to a name e.g. "Only Accurate cameras ticked". 

The Cameras State Manager would be a dialogue box (see attached example). This box would list all the saved "States" . There would be a "restore button". The user would highlight the name they would like to restore and click the "restore button" and the cameras would return to exactly the state they were saved  (ticked or unticked).

The Cameras State Manager would not only be saving the ticked / unticked state of a camera but also the accuracy of the camera. This would then allow the user to import camera accuracies or if DJI  RTK have them import directly when the photos are imported. Then save that to a cameras state name e,g, "All Cameras with original accuracies" . But then the user could change the camera accuracies and untick some cameras, say if if they were having alignment issues, and then save that state e.g. Only Cameras with Changed Accuracies". Because the Cameras State Manager is saving all variables to a named state the changed accuracies could be restored at anytime with one click of a button.

After importing photos, I often will turn off cameras if there are issues and run optimizations. But later will want to test with all cameras on. It is easy to tick (check) all camera or untick all cameras but it is painful if you have to un- tick some but not all.  A Camera State Manager would make this very quick.


Thanks for reading, I hope you make find this suggestion worth incorporating into Metashape.

Regards
Ben

14

General / How to split model imported into Metashape

« on: March 02, 2024, 02:40:53 AM »

Hi All,

I have imported a 2.5d mesh into Metashape. The mesh was in autocad dxf format of 3d face triangles and of a ground surface about 3km x 3km. It is a big file and because of this I need to split the mesh into tiles of about 500m x 500m to make the data more manageable to our clients.

The data imports correctly into Metashape.  And is the only object in the Metashape file. But I cant do any thing with it. I found a script from Alexey that is suppose to split a mesh into tiles but it would not work for me. I know nothing about scripts but when i run it nothing happens. That script is attached.

I have tried creating a Point Cloud from the mesh (at 0.05m interval) but it does not work. The point cloud object is created but it only contains one point. 

Would anyone know how i can split my dxf mesh into tiles of a nominated size or nominated number of tiles. Or even better - how can I split my mesh into tiles but also create a scheme of the tiles generated. Just like happens when exporting a kml of an ortho.

Any help appreciated

Ben

15

General / DEM - can it be shifted down in height?

« on: February 23, 2024, 04:49:49 AM »

Hi All,

I created a project from UAV images (Zenmuse P1). Several GCP's were used as control. The data processed very well with very low RMS, Total Errors and good Check point results. Every check mark under 0.03m for Easting, Northing and Height.

I then created a 3d model, tiled model, DEM and ortho.

It was after everything was created when i found out the GCP values have an absolute error of 0.04m in height from a bad benchmark value. The values I used need to be lowered by 0.04m. I went through the GCP values in the Reference pane and lowered the source values by 0.04m in height and ran the "update transform" button, not knowing what would happen to my 3dmodel, tiled model and DEM.

I was very pleasantly surprised to find that my 3d model and tiled model updated perfectly! That is, there was a total shift down in height of 0.04m across the whole 3d model and Tiled model. This is exactly what i needed.

But when I look at the DEM, it did not shift at all. So I then went to Tools/ DEM/ Transform DEM and ran that. But no shift happened.

My question is - does anyone know how I can shift the DEM 0.04m down in height. Or do I have to delete the DEM and run "Build DEM" and create a new one from scratch?

Any help appreciated
Ben