Hi, Has anyone got/had any experience of Agisoft photoscan software (which takes a bunch of digital photos and makes point/dense cloud 3D image), if so can the data points be exported into AutoCAD?
Partly want to use it as a record of walling and old buildings, but wonder how it could work with CAD.
Thanks in advance, C

Not sure if autocad can import pointcloud data but I’m sure auto desk Revit does

Yes. Photoscan will output to stl which inventor should be able to import. However, you can’t do anything with that once it’s imported other than look at at. Can’t measure it, can’t intersect it, nothing. You could draw round it by eye.

What you need is likely some sort of reverse engineering software- we have geomagic design x, revit should do it, there are others. That should let you turn dumb mesh or point cloud into ‘proper’ nurbs/cad geometry.

Most reverse engineering software tends to be expensive though. Not seen much open source stuff. You could work with the mesh in something like blender, depends what you want to do with it.

Thing to note with photoscan is that the basic version doesn’t let you scale the mesh so you’d need to find another way of doing that

Thanks for the info, have access to Revit. I use CAD and so was asking as more familiar, don;t think we are at BIM yet especially for old listed stone buildings (or piles of stone in some cases), but worth looking at as a recorded, even if quality photos maybe as good.
So may have a go anyway with the trial and see where it goes. May just have to use a scale pole when I take the photos.
Thanks.

You can scale photoscan, you just need GCPs recorded accurately (rtk GPS etc).

Or you can introduce something with scale and scale off that, or manually measure GCPs. Different results depending on what you’re capturing and how the accuracy matters for you.

You can load into mesh lab say and you have a scaled point cloud then go from there.

TBH depending what you’re doing you’re better off hiring a Faro or similar, or doing something with RGB-D.

If you’re not familiar with PS, key thing is a) RAM b) RAM c) MORE RAM. And also a decent-ish GPU.

Can’t really say more than the above, as i’m dangerously close to giving away professional advice for free 😉 IF you’re at all interested in robotic or autonomous survey though then we should talk…

EDIT: sorry when i’m talking about scaling it’s the pro version.

Aye, it’s the basic version doesn’t do the scaling- my interest is in ‘things’ rather than landscapes. We have access to a faro but find often the limitations are in the time you need and (in our case, it’s old) laser safety area control. you can throw several people at something with multiple cameras and get enough photos for some passable geometry almost before there’s a tripod set up. Very dependent on what output you want though.

My advice for photoscan using a handheld or tripod camera (rather than uav/rpv) would be practice, practice, practice. Surprising how many people don’t take the time to learn to take the photos properly and then spend ages swearing at the software.

For a truck, for example, stand 2-3m away, face truck, take picture, sidestep, take picture, sidestep, repeat all the way round. You should get good overlap. Do different heights too. Bog standard compact and you can generate mesh almost ready to print if the object isn’t too complex.

Also, the quality of the results generally varies massively depending on camera, camera settings, proper photo coverage, light conditions in the day, so it really pays to spend a bit of time trying lots of different things- office furniture, Street objects, well lit etc to find what works. Play around with the sodtware settings too.