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.


Topics - jazzyj

Pages: [1] 2 3 4
1
I've been able to use the ground classification tool to remove about 80% of above ground objects. I have some areas of shrubs that I want to remove all points within the edges of the shrubs.  I've of course tried about 6 or more iterations of the ground classification tool using difference settings but it would just be easier to manually delete the points in the area of the shrubs.  Howver I can't get the point cloud and orthomosaic view to sync together so when I move the view in the X/Y plane it moves the view for both the layers.  Nor can I figure out how to select points while being able to view the orthomosaic as a base layer in the X/Y plane when editing the point cloud.  Is there anyway I can accomplish this?

2
Bug Reports / ignore [wrong thread]
« on: April 12, 2023, 08:35:25 PM »
Wrong thread sorry.

3
General / Comparison to Correlator 3D?
« on: October 01, 2021, 07:57:41 PM »
Wondering if anyone has also used Correlator 3D and how they would compare it to Metashape - mainly for Drone Photogrammetry for large area mapping and surveying type applications.  One of the frustrating things I've found in all the software I've used is the automated extraction of a DTM (Digital Terrain Model) from the Surface Model by filtering the trees, structures, and other objects that are not bare ground.  I've had to spend hours trial and air with the Metashape settings to get it to work well. Yes, you can manually edit the point cloud but that can be really time consuming when you are talking hundreds of acres of area.  And the best settings for one job aren't necessarily the best for another.  The other solutions that do it automatically that I've used that don't give you access to the settings are hit and miss (mostly miss) as far as doing a good filtering job.  I've heard Correlator 3D does a good job although a bit heavy handed, but I'd prefer a littler over filtering (lops off a few small mounds) than under filtering (leaving elevation rises where there were bushes/trees/cars on flat ground.)

But also interested in overall how it compares, not just the DTM extraction.  Not sure if it can handle multispectral image processing?  Processing time/efficiency? And also how many export formats it has compared to Metashape?

4
General / Metashape Not Responding Texture Generation
« on: June 25, 2020, 08:01:01 PM »
I've been trying different parameters for creating a model of a large area with a group of buildings (3,250 20MP photos).  I made several models fine, but they all were 3D instead of 2.5D Height Field.  I decided to try the Height Field as I think that's more appropriate for planar surface.  The model generated fine, I limited to 15 million faces (same as before).  Arbitrary Texture, Mosaic blending, 4096 x 16 (before I used 8192 x 10 so memory requirements would actually be a little lower for 4096 x 16).  I have 120GB of RAM and Tesla M60 GPU. 

I ran a batch process overnight and woke up to find the Metashape in the "Not Responding" state and was stuck blending textures.  This was all on version 1.6.2.  So I upgraded to 1.6.3 and restarted the texture process.  It took about an hour to finish the parametizing step.  It got 1% into blending and now not responding again. However, in Windows performance monitor I can see it is using CPU (around 13.5%) which typically indicates it is doing something, where as before I think after overnight the CPU had come down to 0%.  However the memory usage is not changing at all, so that indicates to me it is indeed locked up, maybe in some sort of a loop?

Still I have very very rarely ever seen Photoscan/Metashape go into the not responding state. 

5
General / Decimating Mesh - What About Texture?
« on: June 24, 2020, 07:14:03 PM »
I decimated a mesh from 25 million to 15 million faces (it was of a group of buildings).  The geometry of the buildings looks good but the texture looks terrible.  (Texture was arbitrary mappying 10 X 8192px mesh pages)  Do I need to regenerate the texture after decimating the model?

6
General / One Ground Control Point to Shift Map?
« on: June 24, 2020, 04:36:06 AM »
Typically, my drone's GPS is off by about 1-2 meters and up to 15 meters vertical.  If all I want to do is shift the coordinates of the resulting map to be within 0.5 meters  horizontal and 1 meter vertical, can I optimize the cameras with one ground control point that has 0.1 meter accuracy?

7
General / Generating Different Textures
« on: June 23, 2020, 09:58:27 PM »
I noticed that 'texture' isn't shown in the workspace.  If I want to test different textures, do I need to duplicate the Chunk for each different texture I want to generate, or just duplicate the model?

8
General / Mesh + Texture vs Tiled Model for FBX Export
« on: June 23, 2020, 09:25:07 PM »
I know the tiled model is apparently quicker to navigate in Photoscan, but does it make any difference as far as exporting the model to FBX if I generate a tiled model or not versus just the mesh and texture?  I have a huge model so the processing takes really long to do trial and error on all this.

9
General / Textured Model Settings - Large Area of Buildings
« on: June 20, 2020, 12:05:49 AM »
I'm tasked with building a 3D Textures Model of several buildings over a 20 acre area.  I have about 3000 20 megapixel source photos.  I want to generate the best quality that an average well spec'd workstation can handle (let's assume like an i7 processor with 16GB ram and dedicated GPU). Would the default Metshape settings be adequate or for such a large data set would I need to make some adjustments and if so, what?

 

10
I'm testing a PPK processing software with Phantom 4 RTK drone files which updates the geocoordinates inside the EXIF file.  I compared EXIF data before and after and it appears the software I'm using only modifies the 6 decimal GPS coords in the EXIF section and not the 8 decimal coords in the XMP section.  I was assuming Metashape reads the 8 decimal coords from the XMP data however I loaded both images into Photoscan and instead of the coords being the same they are different, indicating Metashape is only using the 6 decimal coords which is 0.1 meter accuracy level! 

Metadata Before

XMP
Absolute Altitude   +367.33
Relative Altitude   +50.20
Gps Latitude   46.46966340
Gps Longtitude   -120.67038065

EXIF

GPS Latitude Ref   North
GPS Latitude   46.469663 degrees
GPS Longitude Ref   West
GPS Longitude   120.670381 degrees
GPS Altitude Ref   Above Sea Level
GPS Altitude   367.328 m

After PPK Updates Image Metadata

XMP

Absolute Altitude   +367.33
Gps Latitude   46.46966340
Gps Longtitude   -120.67038065
Relative Altitude   +50.20

EXIF

PS Latitude Ref   North
GPS Latitude   46.469672 degrees
GPS Longitude Ref   West
GPS Longitude   120.670374 degrees
GPS Altitude Ref   Above Sea Level
GPS Altitude   366.4672 m



11
General / Camera Position Accuracy Horizontal vs Vertical
« on: June 05, 2020, 04:58:16 AM »
I have PPK corrected camera positions where accuracy is 0.05 meters for LAT/LONG and 0.2 meters for altitude what value should be entered for accuracy? I am mainly concerned that the resulting outputs have the most accurate horizontal coordinates and not as concerned with the vertical accuracy of the point cloud but would that make any difference to what setting I should use?

What if I'm only interested in relative accuracy of the coordinates and not absolute position.  So lets say the camera GPS receiver has an absolute position error of 3 meters for LAT / LON and 20 meters for altitude but the change in the error stays relatively small varying only by 1 meter LAT / LON and 2 meters altitude between each camera position?  (so absolute error is 2-4 meters and absolute altitude varies between 18-24 meters) What accuracy value should be used?






12
General / Python Scripts for Altitude Adjustment Won't Run
« on: May 21, 2020, 04:19:13 AM »
Is it just me or do the scripts Agisoft reference here not run in 1.6.2?

https://agisoft.freshdesk.com/support/solutions/articles/31000152491-working-with-dji-photos-altitude-problem-

13
I've encountered a strange issue.  I have GCP markers that the ground control markers appear in upwards of 20 photos per marker.  I loaded the marker coordinates, filtered the photos by each marker and manually centered the marker on the visible control point in the photo then unchecked the cameras, left only the markers checked, then optimized the cameras.  However, I later come to find that some of the photos are not showing the green flag indicating that I manually aligned the marker.  So I went through all the markers again.  Did some processing. Saved the project. Then when I loaded the project again, I once again found markers without the green flag!  Either there is some sort of bug, or, the only explanation would be I didn't scroll down far enough in the photo window to discover all the images.

With that being said, when I later move the markers after doing the optimization step, do I just need to run the optimization step again, or do I need to reset the camera alignment, re-align the photos, THEN run the optimization step again?

The workflow in the documentation always shows you align photos, then place markers, then optimize.  Is the advantage to doing the marker placement after alignement so that you can generate the mesh based on the sparse cloud so you can locate the photos with the markers, or if the photos are already geotagged can you place the markers before aligning the photos?  I would guess if the coords in the photos can have an error of say up to 5 meters, it may miss some of the photos with the markers in it when you use the filter by photos function?

14
General / Camera Position Accuracy Setting For PPK + GCP
« on: April 28, 2020, 09:37:29 AM »
My photos are geotagged using PPK however I did not have accurate coordinates for the position of the base station.  So relative camera position accuracy is high but absolute world coordinate position accuracy is low - typical GPS (3m horizontal, 10-20m vertical)

I also have several GCPs (I was unable to set the base station on one of them) which I have accurate coordinates for.

Theoretically I would think the PPK generated camera positions would improve relative accuracy of the ground surface between GCPs, where the GCPs would correct the absolute positional accuracy. 

I see there is a setting for camera position and marker position accuracy. My question is, in this circumstance do I leave the camera position accuracy at the default 10m because that is for absolute position accuracy, or do I change it to something like 0.1m?  I'm assuming I leave it assuming that is absolute position accuracy.  But the model should still be more accurate using PPK as the error in the GPS coordinates between photos should not fluctuate that much as compared to non PPK?

15
General / Are These DTM Instructions Incorrect Now?
« on: March 09, 2020, 02:34:40 AM »
https://www.agisoft.com/pdf/PS_1.1%20-Tutorial%20(IL)%20-%20Classification%20and%20DTM.pdf

This says to build a mesh then using the Export DEM function.  It looks to me like after classify the ground points, you need to use the Build DEM function not the build Mesh function in the current version of Metashape.  Is that right?

This files pops up all the time when searching for Metashape DTM generation.


Pages: [1] 2 3 4