dense point cloud only?

[RalfH]

While processing a number of somewhat larger data sets (between 150 and 1000 photographs) I noticed that in many cases I can not achieve what I want because PhotoScan tries to do more than I want. Sounds odd? It's not.

What I really need in most cases is a dense point cloud. When giving PhotoScan the task to create a model, it computes such a dense point cloud, no problem, even at high resolution with hundreds of photographs.

The problem is that there is no way to stop the process right there and save/export this dense point cloud. Instead, PhotoScan goes on to try and create a meshed model, and this is the processing step which (for large data sets) requires a lot of memory and a lot of time. And in most cases, I am not able to process at more that low or medium resolution (in arbitrary mode) because PhotoScan runs out of memory. Why can't I save the dense point cloud? It's frustrating to see in the programme console that a dense point cloud with several hundred million points was successfully created, but then the programme quits because it wants another few dozen GB of RAM to create a meshed model which I do not need!

I suspect that many users of PhotoScan want a dense point cloud rather than a meshed model, because this would be much more suitable for post-processing with other software (e.g. to apply vegetation filtering).

It would be very great if there was an option to create only the dense point cloud (at the known resolution steps of lowest, low, medium, high and ultra) but skip creating the mesh. That way, it would be possible to achieve higher resolution results in less time with less RAM. At the very least, the dense point cloud should be saved to a temporary file before the meshing is attempted.

[Alexey Pasumansky]

Hello RalfH,

You can export dense point cloud right after Align Photos stage is finished. Please use Export points command from File menu and select Dense export type.

[RalfH]

Hi Alexey,

thanks for the quick answer. Is the dense option new in version 0.9? I am still using 0.85 because I didn't like the sound of the problems some users reported when moving on to 0.9. Or is a a pro feature?

But I do think that at this stage (directly after photo alignment) it is only the sparse point cloud created for photo alignment, and the dense point clouds created in the process of building the model at medium, high or ultra resolution are much denser. At least, the console reports much higher point counts. Or did I misunderstand something?

[Alexey Pasumansky]

Hello RalfH,

Dense point cloud export is a feature of 0.9 version (available both in Standard and Pro).

When dense point cloud export is selected (with the corresponding quality) PhotoScan starts depth maps calculation (like in Build Geometry stage, but Mesh generation is skipped). So this kind of export is time consuming since it requires additional information that is not stored in the project.

[RalfH]

Great, this is exactly what I was looking for!

But what do you think about the problems some users in the forum have reported after switching from 0.85 to 0.9? Have they been solved in the newest release?

[Alexey Pasumansky]

We consider version 0.9.0 stable and workable.

If you have any problems with it let us know.

[Mohabon]

Hello RalfH,

Dense point cloud export is a feature of 0.9 version (available both in Standard and Pro).

When dense point cloud export is selected (with the corresponding quality) PhotoScan starts depth maps calculation (like in Build Geometry stage, but Mesh generation is skipped). So this kind of export is time consuming since it requires additional information that is not stored in the project.

Hello alexey,

i understand that the operation is time consuming, but certainly less time consuming than building mesh after generating depth maps.

What is not completely clear is: the memory requirement for exporting dense cloud at ULTRA-HIGH resolution is the same memory required for building Geometry at ULTRA-HIGH resolution?

[Alexey Pasumansky]

Hello Mohabon,

Memory requirements for dense point cloud export depend on scene type, but in general it should be lower than for mesh generation.

[RalfH]

Yes, memory consumption for dense point cloud is much less than for meshed model - in one case I tried yesterday (~200 images, medium quality), memory consumption peaked at ~3 GB, but meshing went above 16 GB.

[Mohabon]

Hello Mohabon,

Memory requirements for dense point cloud export depend on scene type, but in general it should be lower than for mesh generation.

Hello Alexey,

sorry if i insist on this side, but i find this Dense Cloud extremely useful for my field of application because i can use laptop instead of 3D workstation if memory consumption is so lconsiderably lower than meshing and exporting DEM.

In http://www.agisoft.ru/wiki/PhotoScan/Tips_and_Tricks, i read that 200 Photos modeled in Height-Field Mode are using about 12.8 Gigs of RAM.

If i align and then export Dense Point Cloud, what ratio factor could i reasonably apply to the above mentioned value in the same conditions?
If ratio could be 25% or 30%, it means that majority of laptops can complete the processing that would be great for certain missions.

Kind regards,

M

[Alexey Pasumansky]

Hello Mohabon,

The difference in memory consumption (exporting dense point cloud vs. mesh reconstruction, both including the same stage of depth maps generation) depend also on the target face count in Height-Field mode.
For example, I've tried to process the project with the similar characteristics like in your previous post, and 1 M face count mesh generation peak was lower than memory peak for depth maps generation, while 10-20 M face count led to 10-20% more memory consumption.

[RalfH]

Hello Alexey,

I did enconuter some small problems when upgrading from 0.85 to 0.9: when loading one .psz file (with aligned photos) created in version 0.85 into PhotoScan 0.9, about 20% of all camera positions appeared to be mirrored UNDER the model, so I had to repeat the alignment step.

The "export points" feature works really well and saves a lot of processing time for some of my projects. In the cases I have tested, it requires only a small fraction of RAM compared to creating a meshed model. This means that I can now even create point clouds at ultra high quality for large sets of photos without having to put huge amounts of RAM into my PC. Great!

Two issues could still be improved though:

1. After exporing a dense point cloud at a given resolution, the corresponding depth maps appear in the workspace. But if I want to export the same dense point cloud to another file format (for example, also to LAS after having exported to OBJ), PhotoScan does not use the existing depth maps but re-computes them. Things could run even faster if the existing depth maps were used.

2. For processing the exported point clouds in external software, it would help if there was an option to split the point cloud into chunks (at the moment, I think the entire workspace is processed in the "export point cloud" feature) or into files of pre-defined size.

[Alexey Pasumansky]

Hello RalfH,

Is it possible to provide a sample project saved in 0.8.5 that has problems with camera positioning upon opening in 0.9.0?

1) Unfortunately, in the latest release version depth maps are rebuilt for dense point cloud export even you have them already reconstructed with desired quality. But this problem will be fixed in the next update.

2) Do you mean an option similar to Split in blocks for Orthophoto/DEM?

[RalfH]

Hello Alexey,

Unfortunately I don't have the .psz file anymore, because I have repeated the photo alignment in PS 0.9 and have overwritten the 0.85 file. It would have been too large anyway to send over the net (>200 MB).

Good to hear that the "export point cloud" feature will be further improved.

As for the file splitting, yes, a "split in blocks" option would be great. Exporting the dense point cloud in chunks would also help.

[pixel_Killer]

Hello,

I have exported a dense pointcloud ply file which is somewhat heavy. Around 4.5gb generated on high settings. The point cloud doesn't import into meshlab without crashing so I thought I give rhino 5 a go.
What is the unit scale that the dense pointcloud is exported to? In rhino it is asking me if my file is in centimeters, mm, ft, etc.

Since I don't have a lot of ram, I rather use a dense pointcloud as a Lidar alternative.