Messages

31

General / Re: Volumes on tilted grounds where it is not possible to position ground vertices.

« on: October 10, 2019, 04:45:11 PM »

Hi Isaac,

You could use Virtual Surveyor to estimate the volume. See here:
https://youtu.be/e7fizU7aQk0

I agree with you, it would be great if these more complex volume measurement tools were also available in Metashape.

Best regards,
SAV

32

Feature Requests / Oriented scale bars

« on: October 08, 2019, 01:34:16 AM »

Allow to add/import orientation values (YAW & PITCH / AZIMUTH & PLUNGE) for scale bars which are then used to orient the model, similar to camera orientations

Regards,
SAV

33

Feature Requests / Texture editing

« on: October 08, 2019, 01:12:30 AM »

Allow to edit/adjust the texture

- Tone
- Contrast
- Color
- Highlights
- Shadows
- Sharpen


Best regards,
SAV

34

General / Re: Project with 220K images for processing

« on: October 06, 2019, 03:10:03 PM »

Great to hear 

All the best for your massive project.

Regards,
SAV

35

General / Re: More holes in mesh with markers. How can I get rid of small holes in Mesh?

« on: October 05, 2019, 07:39:42 PM »

For the small holes in 3DandVR's 'no marker mesh' (1st image) it should work.
But I agree, it won't help to fix the isolated/singular meshes that are shown in the 2nd and 3rd image.

Regards,
SAV

36

General / Re: Project with 220K images for processing

« on: October 05, 2019, 05:15:27 AM »

Hi again jivago,

Looks like a powerful system. But I would not go for a Xeon processor, based on a statement by Pugetsystems:

Xeon Scalable processors are not worthwhile for PhotoScan

Here are the details: https://www.pugetsystems.com/labs/articles/Agisoft-PhotoScan-1-4-1---Xeon-Scalable-CPU-Performance-and-GPU-Scaling-1164/

I would only go for a Xeon based system if I wanted/needed more than 128GB RAM.
Intel's Core X an Core i9 series are much better value for money. I personally would go for the i9-9900K. It's 'only' a 8 core processor, but it outperforms a 18 Core 9980XE, as shown here: https://www.pugetsystems.com/labs/articles/Metashape-1-5-1-Intel-Core-i7-Core-i9-vs-AMD-Threadripper-Performance-1465/

So don't get hooked by the number of cores only. That's not the only thing that controls your processing times.

Regarding your GPU choice. I suggest to use the money that you save by not using a Xeon setup and invest it in the latest RTX 2080 Ti 11GB or even a dual RTX 2080 Ti 11GB setup. You might even have money left to upgrade to 128GB RAM as well.


All the best.

Regards,
SAV

37

General / Re: Using a LIDAR dataset from an ofitial source to georeference a Metashape project

« on: October 05, 2019, 04:54:04 AM »

Hi Isaac,

Nope, I am not working for Agisoft but I am a long-term user of Agisoft Photoscan/Metashape. I am also an official Agisoft reseller and provide training courses on best-practice workflows for digital photogrammetry (using Metashape).

Regarding your question. After importing the dense cloud from CloudCompare into Metashape you are still able to compute a mesh from this shifted dense point cloud in Metashape. You can also generate a DEM from it, but you won't be able to compute a texture or orthophotograph from it because the camera locations haven't shifted and are still at their original locations. However, you could use CloudCompare to generate the orthophotgraph (at the resolution of your dense point cloud) by making use of RASTERIZE. http://www.cloudcompare.org/doc/wiki/index.php?title=Rasterize

There would be another solution to your problem, but I fear that would include some programming/python scripting.
A) You could use the Open3D package in a custom python script that you run in Metashape to run ICP inside Metashape instead of externally in CloudCompare. Not a quick/simple solution though.

B) Write your own custom Python script that uses the transformation matrix that you get out of CloudCompare when you perform ICP and then apply these values to everything in your Metashape project.


Last but not least (a low tech approach), you could try and generate a mesh from your LAS file which might help you to pick your ground control points better and then use the technique that I explained in my earlier post. Use CloudCompare to generate a mesh from a point cloud. PLUGINS > PoissonRecon.

All the best.

Best regards,
SAV

38

General / Re: Difference between Ground Control Points vs Check Points ?

« on: October 05, 2019, 04:29:30 AM »

Hi Probert1968,

There are quite a few terms used, such as ground control points (GCPs), ground markers, control points, check points, markers .... and more. It can get quite confusing. Let's try to clarify a few things.

In Metashape, so called markers are used to georeference and scale photogrammetric models and for camera calibration purposes. Sometimes markers are also used to help the Structure-from-Motion algorithm to align cameras, but that's a special case.

Ground control points are points that you have placed/marked in the field and surveyed before acquiring the (aerial) imagery. The coordinates of these ground control points are later imported into Metashape in order to link them with the pictures. As soon as you import ground control points into Metashape they become markers.
Metashape lists all markers in the reference pane. By default they are all checked/enabled (there is a checkbox on the left next to each marker), which means they are used as reference information to control the model and camera calibration. That's when a marker becomes a check point.
If you uncheck a marker (no tick in the check box next to each marker), then this marker/control point becomes a check point. This means that it is not directly used in the photogrammetric modelling process anymore. But it's still available to check the accuracy of the model that you've generated. Basically you compare X/Y/Z values from your model to the X/Y/Z values of your measured ground control points (which are now a check points). You check the error between the model and the real world measurement, hence the name check point.

Note that if you have surveyed your GCPs with high/cm level accuracy, then you can expect to get a total error of about 1-3 times your ground sampling distance (GSD).


I hope this helps. All the best.

Regards,
SAV

39

General / Re: More holes in mesh with markers. How can I get rid of small holes in Mesh?

« on: October 05, 2019, 04:00:56 AM »

This tool works for both large and small holes. That's what you can control with the level slider (see attached images).

Regards,
SAV

40

Feature Requests / Re: Custom point classes?

« on: October 04, 2019, 05:41:03 PM »

+1

Custom class names would give the user more flexibility.

41

General / Re: Physical targets to create a 3D model of topography

« on: October 04, 2019, 05:38:40 PM »

Hi Kobi,

Looks like you are working in a controlled lab environment, hence you should be able to make use of a 'scanning with projection' workflow. Have a look at this website for more information: http://www.pi3dscan.com/index.php/instructions

All the best.

Regards,
SAV

P.S. I am just curious what kind of research/experiments you are doing...

42

General / Re: More holes in mesh with markers. How can I get rid of small holes in Mesh?

« on: October 04, 2019, 05:31:58 PM »

Hi 3DandVR,

Try TOOLS > MESH > CLOSE HOLES...

All the best.

Regards,
SAV

43

Bug Reports / Error: XML syntax error

« on: October 01, 2019, 04:01:44 AM »

Hi All,

Just a warning about a known bug related to shapes in Agisoft Metashape (as confirmed by Alexey):

Unfortunately, there seem to be a flaw in the way of shape data storage within the Metashape project files. If the number of shapes and their vertices gets too high the corresponding XML file inside the shapes.zip archive gets corrupted and leads to the project open failure. We are aware of the problem, but have not yet implemented an alternative and therefore can only suggest to avoid keeping enormous number of shapes within the single chunk.

--
Best regards,
Alexey Pasumansky
Agisoft LLC

Most of the time you would not stumble across this problem. But recently I was dealing with a large project where I computed 1m contour lines for an area of 18km2. Everything worked without any problems. I could open, view and export these contour lines. I then saved the project and closed Metashape. When I tried to open the same project the next day, I was greeted by Error: XML syntax error and the project wouldn't load.

Workaround/fix:
Manually edit the doc.xml file for the related frame. Simple remove the line pointing to shapes.zip archive. You can even delete the whole shapes folder the doc.xml is referring to. DONE.


Regards,
SAV

44

General / Re: Using a LIDAR dataset from an ofitial source to georeference a Metashape project

« on: September 30, 2019, 06:43:54 AM »

Hi Isaac,

There are several ways to do this. Here are two:

A)
1. Find unique objects/features that are clearly visible in both datasets (LAS and your imagery). 10-15 features/objects if possible.
2. Extract the exact X/Y/Z coordinates of these features from the LAS dataset and store them in a CSV file
3. Import the CSV file into Metashape (in reference pane) and use the points to generate new markers
4. Manually link the markers to your images in Metashape

Make sure to adjust the accuracy setting for your markers (default is 0.005m) to match your LIDAR dataset accuracy.

B)
If you want to align your computed dense point cloud to your LAS dataset then you could also to this in CloudCompare (free, open source).
1. Import/open the Metashape dense cloud and LAS file in CloudCompare
2. Manually move/rotate/scale the Metashape dense point cloud so it 'roughly' matches your LAS file
3. Use ICP in TOOLS > REGISTRATION to get the best possible fit. More info on ICP here: https://www.cloudcompare.org/doc/wiki/index.php?title=ICP

All the best.

Best regards,
SAV

 

45

General / Re: Specifications to reduce processing time

« on: September 28, 2019, 07:50:44 AM »

Hi raphaelfp,

Check out this website. They have tested Metashape performance on various hardware:
https://www.pugetsystems.com/recommended/Recommended-Systems-for-Agisoft-Metashape-145/Hardware-Recommendations

It seems like following components do a good job:
- i9 9900K
- RTX 2080 Ti

I've recently processed a large scale dataset (18km2, 2800 images, 20MP each). In total it took about 3 days (including manual masking of lakes/rivers/etc to generate a clean point cloud). All was processed in a single chunk with image alignment, dense cloud and mesh at HIGH quality settings. The system it was run on had 64GB RAM, an Intel Xeon W-2155 CPU and a Nvidia Quadro RTX5000 graphics card. The i9-9900K and RTX 2080 Ti should give you a very similar performance (I think even better). The whole system should cost you much less than US$ 19k.

I also highly recommend to update to the latest version of Agisoft Metashape instead of using 1.2.0 as mentioned in your post. A lot of performance improvements have been built-into the code and GPU acceleration is much better in the recent version of Metashape.

All the best. If you need some more specific advice, feel free to PM me.

Regards,
SAV