Nomad point cloud import queries

[Andrew Lees]

I am new to Vectorworks and this is my first attempt to use point clouds captured with Nomad. Now, my scan of this floor plan seems to be a bit of mash-up, but that is not my main issue right now. I have seen a VW video on this (https://www.youtube.com/watch?v=_cQv1da242Q) and the point cloud model shown in that seems to behave somewhat differently than mine. In the video the presenter is able to zoom in and trace the point cloud - which looks great. When I zoom in however I get two problems:

 

1. The density of the point cloud seems to decrease - to the point of it being too vague to find hard edges to trace once I get close.
2. I have a grid pattern superimposed over my point cloud. The size of this grid (which is manifest as gaps in the point cloud) gets denser and more obscuring the more I zoom in.

 

The attached images show progressive zoom-ins as described above. This is an orthogonal top view (not plan view) with section box cutting the horizontal slice. For info, the point cloud was imported at 100% points resolution.

 

Suggestions?

 

Thanks

[michaelk]

@Andrew Lees

 

This is what I've found to make point clouds from Nomad work successfully.

 

1.  Use a recent, pro version of the iPhone to get better LIDAR.

 

2.  Smaller scans.  Only do one room at a time.  Maybe two if they are small.  The issue is drift.  The larger the area of scan the more likely that it will "drift" a little.   That's why you are seeing some walls duplicated on a different axis.  

 

3.  I like to scan around the other side of connecting wall openings to get wall thickness and to help line up adjacent scans.  It's important to always have objects or architectural features that will appear in at least one other scan to help line them up later.

 

4.  Watch the screen as you are scanning.  There is a slider on the bottom of the screen.  It only controls pass through from the camera to the screen while scanning.  It has nothing to do with what gets recorded (AFAIK).  Try turning it way down so you are seeing primarily or only the points that are being recorded.  Sometimes you will see a "jump" happen while recording.  I can't prove it but I think that's the scan losing its place.  Discard that scan and start over.

 

5.  In the VW drawing, lay out all the rooms from individual scans.  It takes a little while to rotate them all to the same direction.  I've found that if you use the _VW.pts version of the file, the floors will be level, but not necessarily at Z=0.  But they appear to be close to the same - usually negative - Z value.  If you've got one scan of a hallway, then it's easier to align the scans of the rooms off the hallway using the door openings.  If you keep each scan on a separate layer it makes it easier to work with and you are less likely to select the wrong scan by accident.

 

7.  When you have all your room scans laid out, use the clip cube to cut the top and bottom so you only have a sliver showing about 3 or 4 feet above the floor.  Try looking at it in top/plan view as well as top view in shaded rendering to see what shows up better for you.  That will give you a floor plan that looks like all the wall lines are about 5/8" thick.  You sort of have to draw it not too zoomed in.  The closer you get the more you see the spaces between the points.  You will quickly find the sweet spot.

 

8.  Once you draw your walls, guess at window and door sizes and locations.  Select one wall, turn on the clip cube and set the visibility to wireframe.  Now it's really fast to use the new drag handles on doors and windows align them with then scans.

 

9.  It also works outside.  I've had pretty good success using it to get eave overhangs, roof pitches, hardscapes, external columns, existing tree sizes and locations.  About half the time on a building the size you are working on it will hold together long enough to scan the exterior of the entire residence well enough to get an accurate footprint.

 

I haven't tried this yet, but I've heard that using a tripod for your iPhone that rotates around 360 and scanning several times in the same room produces better results.

 

 

[Tom W.]

7 hours ago, Andrew Lees said:

I have a grid pattern superimposed over my point cloud. The size of this grid (which is manifest as gaps in the point cloud) gets denser and more obscuring the more I zoom in.

 

I have not seen this before + have no idea what's caused it...

 

7 hours ago, michaelk said:

I've found that if you use the _VW.pts version of the file, the floors will be level, but not necessarily at Z=0.

 

Yes. Nomad generates a 'xxxxxxxxxxxxxxxxx.pts' version of the point cloud which is designed to be viewed on Nomad (because it is oriented correctly) + a  'xxxxxxxxxxxxxxxxx_VW.pts' version which is the one you are meant to import into VW (which is upended when viewed on Nomad). Make sure you name your scans + save them to a dedicated folder as you create them as it's a nightmare trying to find the files you just created otherwise! (the names Nomad gives them are completely random + they can't be arranged in date-created order... And for the life of me I can't see how you rename them or move them into new folders after the fact!! That could be me though...)

[Andrew Lees]

Thanks @michaelk some good tips there. I think shaded view makes it a little more punchy. I definitely will do it room by room next time. I can also imagine that a tripod would make it much better - kind of more like a proper fixed point cloud device. The Nomad instructions just said start it up and walk around - so that is what I did.

@Tom W. I was hoping that the weird tartan pattern was something easily fixed - like a grid setting I don't know about that can just be toggled off. As I said, I'm new to VW so what is obvious to seasoned users might be completely unknown to me. 

 

It seems a little more useable now though, so I will persevere and see if I can translate this into a serviceable plan. 

[zoomer]

11 hours ago, Tom W. said:

I have not seen this before + have no idea what's caused it...

 

I assume a GPU or GPU driver issue ?

 

 

19 hours ago, Andrew Lees said:

my scan of this "floor plan" ....

 

I only tried to scan more than one room at a time with Room Plan Mode

and was disappointed. I do not understand why it lost the angle.

All Walls should have been perpendicular but Room Plan was about 15° off.

I think Nomad needs more artificial or better human intelligence.

 

 

18 hours ago, michaelk said:

This is what I've found to make point clouds from Nomad work successfully.

 

This is all great @michaelk

 

 

18 hours ago, michaelk said:

There is a slider on the bottom of the screen.  It only controls pass through from the camera to the screen while scanning.

 

That was new to me.

 

18 hours ago, michaelk said:

It takes a little while to rotate them all to the same direction.

 

It does.

 

But I do not get where Nomad takes the angle from.

It is neither true north angle nor did I find any possibility to control this

in any way, like starting perpendicular to a Wall.

 

18 hours ago, michaelk said:

but I've heard that using a tripod for your iPhone that rotates around 360 and scanning several times in the same room produces better results.

 

I already tried to mimic that manually but did not find any benefit.

 

 

18 hours ago, michaelk said:

the floors will be level, but not necessarily at Z=0.  But they appear to be close to the same - usually negative - Z value.

 

I assume negative floor height level comes from iPhone to Floor distance (?)

 

 

18 hours ago, michaelk said:

5.  In the VW drawing, .....

 

7.  When you have all your room scans laid out ....

 

Looks like you made a secret from the most important 6. point ......

[michaelk]

36 minutes ago, zoomer said:

Looks like you made a secret from the most important 6. point ......

 

🤣

 

6.  Enjoy not spending an entire day measuring a building knowing that you will miss getting at least one (but probably more) critical measurements. 

 

Nomad point clouds aren't perfect, but they are way better than trying to read hand written measurements off of a crowded, incorrect old drawing two weeks after you measured the building 🙂 

[michaelk]

41 minutes ago, zoomer said:

I only tried to scan more than one room at a time with Room Plan Mode

and was disappointed. I do not understand why it lost the angle.

All Walls should have been perpendicular but Room Plan was about 15° off.

I think Nomad needs more artificial or better human intelligence.

 

I agree about room plan.  It can get confused by soffits, vaulted ceilings, timber frame roof trusses, etc.  

 

With point clouds you can usually get a reasonable slope to a vaulted ceiling, number and approximate size of beams, placement of HVAC supplies, light switches, wall thickness, difference in elevation between floors, etc.  Not perfect.  But so so so much better than spending a day with two people and a tape measure.

[zoomer]

49 minutes ago, michaelk said:

6.  Enjoy not spending an entire day measuring a building knowing that you will miss getting at least one (but probably more) critical measurements. 

 

 

That was my manual measuring experience so far too.

I always hoped Lidar Scans would prevent this.

[Andrew Lees]

Picked up messing around with this Lidar scan again and I wonder if anyone can enlighten me about this 200mm / 8" thick "fuzz" that makes up the scanned walls? What is this?

Is it generated to create a gestural wall thickness to make the scan more legible?

Am I right to assume that the outermost (if considered from the wall) or innermost (if considered from the room space) surface of dots is the actual scanned surface?

Hopefully my little annotated screenshot clarifies what I am talking about - the dashed line being the fuzz depth (as I have called it). The issue is particularly clear I think in this scan as it is a massive masonry wall in an old building which scales at over 1200mm thick outer dots to outer dots.

I think I will experiment with a new scan in my immediate environment to see if I can figure this out. But any suggestions welcomed.

[Andrew Lees]

Okay, I've run my little experiment on a simple space. The attached is a small stairwell in which I also took a key dimension using a Leica Disto. Confirms it is the inner face of wall points that denote the surface, with - in this case - 62mm of poche wall thickness (compared to the 200mm in my earlier marked up screenshot).

 

What is also interesting is the amount of drift even in this small space - I have two alignments of the walls. It would be interesting to see if this could be improved by using a tripod and making sweeps from a fixed point.

[Jeff Prince]

@Andrew Lees there are a lot of things that can affect the quality of a scan and your software’s interpretation of the data collected.  62mm is pretty accurate for the technology currently.  Moving smoothly and without rapid changes in direction/twisting seems to produce the best results.  Incidentally, I have the best results outdoors by moving up and down slopes rather than along them.  Indoors I pretend I’m spray texture with the iPad and have good results.

[Andrew Lees]

Thanks @Jeff Prince. I've not tried outdoors yet, but will adopt that up / down technique. I also do a kind of spray technique indoors in the smooth circular motion as per the instructions on Nomad. In that stairwell scan I think I did accidentally make a more sudden shift and that might explain the superimposed plans - there is quite a technique to this it seems...

 

What is your experience with this fuzz / poche effect that was my original query though? It definitely seems to be the case that this is applied mass / material behind the scanned surface. I have just been playing around with this some more and there is also a technique to interpreting scans in VW it seems - little tilts into 3D every now and then so you can check up that you are correctly tracing the wall surface. This does indicate that the room facing points show the surface and then this is given mass by density of points behind them. I'm intrigued by this as it suggests that these "mass points" are not therefore scanned points, but added points (behind the scanned plane) to aid the legibility of the outputs. But that's just a guess. This technology is very mysterious and I must admit that this scattering dust and seeing it taking shape into the room does seem a bit like magic!

 

I also just discovered the automated points reduction edit in VW - makes the rather ropey scans I am working on much clearer.

Edited May 8 by Andrew Lees
typo

[Peter Neufeld.]

On 4/25/2024 at 8:15 AM, michaelk said:

2.  Smaller scans.  Only do one room at a time.  Maybe two if they are small.  The issue is drift.  The larger the area of scan the more likely that it will "drift" a little.   That's why you are seeing some walls duplicated on a different axis.  

 

3.  I like to scan around the other side of connecting wall openings to get wall thickness and to help line up adjacent scans.  It's important to always have objects or architectural features that will appear in at least one other scan to help line them up later.

 

 

Hi @michaelk

 

From this Planet Vectorworks article about Update 4 :

 

"The new capability to scan multiple rooms and seamlessly combine them into a single model, plus the added support for round and slanted walls, provides significant time savings when surveying existing building or site conditions."

 

So it now takes into account different rooms and keeps the rooms in their position when importing into Vectorworks. No doubling up of walls etc.

 

Cheers,

Peter