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 - toxicmag

Pages: 1 2 [3] 4 5 6
General / Re: Automatic point classification of high noise / water
« on: March 15, 2019, 08:38:31 AM »
 The gradual selection could help you.
 At least you  can automatically select points of low confidence, high reprojection error and other values.

Feature Requests / Re: Live Ground Classification Viewing Window
« on: March 15, 2019, 08:34:50 AM »
Hi Jason,

I totally agree and it would be a great improvement. It could also be combined with the feature request in this post:

General / Re: Classify ground points on a region of the project only
« on: March 15, 2019, 08:27:37 AM »
Sorry for not  having a solution.  :-\
Your project is quite big and i am curious what the answers will be.

I would also like to have a selection option before classification. Either by selecting points or by drawing a polygon and giving this area to the classification tool.

 let's see what is possible.  ;)

Dear users,

i processed a pointcloud, mesh and DSM from a UAV survey some weeks ago and used 14 GCP for the model.
Some weeks later a second survey of another 21 points was randomly measured on the surface of the area.

Is there a way to compare the distances in Z-direction (X and Y are not marked and cannot be judged) directly in Metashape?
What would be your way to see how valid the model and the new points are?

Greetings and thanks in advance


Metashape 1.5.0 (7492)

General / Re: Dense Point Cloud generation // far slower in Metashape?
« on: February 01, 2019, 09:16:54 AM »
Dear Alexey,

The processing took a while and now the model looks awesome.  :D
But one full week with quite a nice workstation is far too long for just the “generating densecloud“ step.

Do you and the team plan to provide GPU support for this step some day? The question comes up because i would like to add another graphic card into the system but it would not help in this case.

Thx a lot.


General / Re: Dense Point Cloud generation // far slower in Metashape?
« on: January 10, 2019, 07:43:32 PM »
Are you using the latest Metashape release version?

Yes - latest. Always.

And do you have the processing logs or at least the timing information from the Chunk Info dialog for 1.4 and 1.5 processing?
Not at the very moment - the machine is busy, but i guess i will cancel that now...

Usually long dense cloud filtering process is related to the excessive overlap. If it is the case, the processing time can be reduced by limiting the number of pairs for each camera to be estimating during filtering. It can be done by creating main/dense_cloud_max_neighbors tweak and setting its value to 60, for example.
This is the case. Right. High overlap in flying AND also halv of the dataset are obligue photos. I need high precision in the pointcloud to match it against terrestrial scanning.

I guess i will reduce it to "medium" and maybe also try your advice with the neighbours



General / Dense Point Cloud generation // far slower in Metashape?
« on: January 10, 2019, 01:10:50 PM »
Dear all,

i finished a job in Photoscan last December and was curious if i could get even better results now in Metashape.

It's ~1300 cameras with 9MB each and i need the dense cloud as result. Nothing else.  ;)
Alignment in high was fast and got all cameras well. Dense cloud generation was set to high as well, but it definetely takes too long.  :(

Was the dense cloud generation set to a lower priority level in developing Metashape? Everyone speaks about depth maps now which might be great for texturing but not for the survey industry.

GPU is on (GTX 980Ti)
128 GB RAM
i7 5820K @3.3GHz

Thx for your feedback.

Dear Alexey,

i converted to WGS84 (EPSG: 4326) and did the process (undistort...) again for writing the GPS params into the EXIF.

I am using Win10 and in the file details it just shows the height param (which is correctly taken from Photoscan). No LAT or LON or even Roll, Pitch, Yaw (which i need for the client to reposition each camera).

When watching the files in XnView it shows completely weird params in the "GPS section" of their EXIF menue.

Thanks a lot


Hello Alex,

In the Undistort Photos dialog (File Menu -> Export section) there's a possibility of GPS data updating by the estimated values for the camera locations.

It's going into the right direction but until now only ALT was written in the EXIF (which was blank in GPS section before).
Also strange is, that est ALT is far lower than the viewable GCP (got from terrestrial survey measurement).

I played with checkboxes in the undistort photos panel but no change.

Thanks a lot, Alexey...

Sounds great!
I will try if it will also export the orientation besides to the camera position.




Dear Photoscanners and Pre-Metashapers  :D

after the alignment there's very precise information available about each cameras position and orientation (used GCPs as well).

Is there a way to export these value back into the photos EXIF for later use?

Thanks a lot in advance


Dear Photoscanners...  :)

Anyone using a Sony a7R3 for aerial photogrammetry in general and processing with Agisoft PS in special?

What is your feedback with regard to measurement, accuracy, sensor size and its optical geometric stability?
I heard about issues of the a7R3 due to its shake-reduction.

Thanks so far.


Dear community and Support-Team,

i have the great chance to re-construct a top level infrastructure which is known around the world... And i do not want to spoil that project.  ;)
A lot of terrestrical laser scans and aerial photogrammetry will end up in a BIM model in Revit.

Now the question is if there is any tool, script, workflow where i could also keep the cameras information from Photoscan (position, direction, angle).
The BIM user then could also view the original photos that covered a specific part of the infrastructure.

I know Autodesk BIM 360 claims to do so, but i do not want to leave Photoscan for this project.

Any input welcome...


Dear PS-community,

i am working with aerial data of multispectral images taken by a professional fixed-wing system of an international company. According to that i assume that the camera and its light sensor are installed perfectly.

The camera is a Micasense RedEdge-M.

When calibrating the reflectance PS automatically finds the right images and marks the panel correctly. I then choose "use reflectance panels" and also "use light sensor". Nothing changes when using the light sensor and also it is unchecked when i enter this menu again later.

Depending on the formulas used neither the HeatMap nor the NDVI palette is normalized correctly. The index at the bottom right is not from 0 to 1, or -1 to +1 but has values up to 9 and even above. So are the values later in external software like QGIS or eCognition.

NDVI values of streets are pretty close to the values of forest and trees. For example 0.8 or higher compared to 0.95 for forest.

Any ideas welcome...



Hey Alexey,
thanks for the quick reply.

It's 1.4.3 build 6529 (64bit) which i am using and the reason might be, that for the first chunk i copied all Micasense files to a single folder before adding them in PS. I was doing so, cause i did not like the Micasense way to create all these subfolders on the SD card.

At the second dataset i simply uses the "add folder" option and gave PS all folders and subfolders at one...  :D  A little lazy, but this one is displayed in RGB.  ::)

Is there a way to correct the first dataset other than re-processing it all?



Pages: 1 2 [3] 4 5 6