Spray & Targets?

[DDRDan]

I’m scanning fiberglass boat decks to develop cad files for CNC machine cutting EVA carpet. They are very reflective surfaces and most have a texture. When using an anti-reflective spray, adding stick on targets is next to impossible. They don’t stick or they slide during application. And with the textured surface the dots are not uniform to the plane. They become a non-planar surface when pressed in place.

1. Are the dot targets the only method of targeting for Revoscan software??

2. Can someone explain the methodology in exactly what the software is looking for when it wants location references? I’d prefer using large spheres with a base.

3. Will it recognize any uniform shape as a location reference and can I identify & instruct it to use those targets?

4. I’m not scanning for features. Only for edge locations to facilitate dimensioning and trace drawings in my CAD software.

PS: How hard could it be for Revopoint to include a .DXF export??? The limited export options are a royal pain in converting because the files are large in my scanning application! You built a large format scanner (Range) and you give file extension choices for printing 3D toys?

[PUTV]

@DDRDan

PS: How hard could it be for Revopoint to include a .DXF export??? The limited export options are a royal pain in converting because the files are large in my scanning application! You built a large format scanner (Range) and you give file extension choices for printing 3D toys?

The software produces all files ( pointcloud and meshed ) originally as rich format of .ply
It is almost the same as OBJ format .
There is not a complex converter build in more of an extension changer .
So you need to use a proper converter to get your specific files .

@DDRDan

Are the dot targets the only method of targeting for Revoscan software??

Marker mode replace the needs of features tracking and uses the markers as base tracking only, ignoring the features of the object .

@DDRDan

Can someone explain the methodology in exactly what the software is looking for when it wants location references? I’d prefer using large spheres with a base.

The scanner is looking for unique features on the object , no simetrical features, so no spheres , as a sphere is seeing by the scanner sensors just as a 2D circle without volume , same goes for a cube .
You can use any other objects as a feature placing it around the surface as supporting points , it can be simple as a bits with more than 4 faces or as simple as a crumbled paper into small fragments , usually painter tape will do that .

Depending if the FOV the features need to be closer to each other or far away , how smaller the FOV how more features need to be closer to each other , usually at max distance of 8 cm for Range series .

Try with the painter tape , stick many crumpled fragments then use 3D spray after and you get yourself best tracking experience.

[DDRDan]

Thanks for the info!

[DDRDan]

OK. Balls of painters (blue) tape is not a good target for scans. Crumpled blue tape is an ‘unstable’ shape. The folds and crimps of the tape are prone to relaxing their shape after application.

Scanned a 82" x 188" deck area. Tape balls on 6cm. I scanned up one side and then down the other. Where I finalized to the origin scan it showed the tape balls had completely change in shape in multiple directions. And after reviewing all the tape targets located near the scan splices … they all showed progressive expansion from the first scan to the last. The longer they sat there, the more they expanded.

This method may be good for objects that can be scanned in a few minutes?? For large format … this was a total waste of (3) man hours, 2 rolls of tape, and 2 cans of spray.

After adding up what I’ve already spent $$$ trying to make this “toy scanner” work, I could have rented a laser scanner. The Range is not an economical large format scanner!! … period. You should stop advertising it as such.

I’m an engineer with years of scanning experience. I should have seen that flexible crumpled tape could never be a valid scan target. Crumpled tape is a terrible idea.

[PUTV]

Hey @DDRDan
Sorry to hear that it did not worked for you ,
I am moderator of the forum , I don’t advertise or sell anything , your frustrations towards me blaming me of something I have nothing to do with , your issues are related to lack of features of your scanned object and nothing else , people scanning whole cars inside and out without issues using Range . You know we’ll the drill and the technology you using and it’s requirements.

I just given you a tips from others that successfully scanned large areas using the tape and scanned bigger areas than yours .

You should invest in some printed targets to use outdoors as features or buy online , very cheap .

Print out yourself some target , it gonna be the cheapest way and works outdoors as well , a simple white cubes or pyramids will make the trick and don’t have to be big . There is plenty of white 3D scanning targets online if you don’t have 3D printer , even magnetic if needed .

[DDRDan]

You addressed my postscript on DXF before my target questions? It was a negative point in the software’s evaluation and you chose to addressed it immediately. That screams you have an investment in providing defense? Stop leading people in that direction and you won’t be seen as having an association.

Knowing people are scanning cars without issues is half of the story. Were they successful on the first try? Did they lose money learning? At what cost are they scanning cars in & out? If I buy a scanner and then realize I’ll invest an equal amount in expensive accessories and high man hours in targeting to complete the scan it becomes an uneconomical choice in scanners. It’s promoted as shoot and go.

My ‘main’ question in the original post was to see if 3d solid shape targets would work in lieu of stickers. A laser targeting method I am familiar with and an opportunity to eliminate learning other methods recommended that are failing. It was the entire purpose of the post. Everything about the question leads the reader to “Target recognition - solid targets and what shapes”.

You said:
"no simetrical features, so no spheres , as a sphere is seeing by the scanner sensors just as a 2D circle without volume , same goes for a “cube” .

Now you’re saying:
“a simple white “cubes” or pyramids will make the trick and don’t have to be big”

If you are trying to confuse people, you’ve succeeded with contradictory comments.

I didn’t come into an ‘outside’ forum. I came to the “Revopoint” forum. I expected their tech staff to have a presence in here to resolve customer issues. I now realize their tech support is “What I did” tech support. People perpetuating other peoples costly mistakes.

[PUTV]

@DDRDan

You said:
"no simetrical features, so no spheres , as a sphere is seeing by the scanner sensors just as a 2D circle without volume , same goes for a “cube” .

Now you’re saying:
“a simple white “cubes” or pyramids will make the trick and don’t have to be big”

If you are trying to confuse people, you’ve succeeded with contradictory comments

A very small cubes across big scanned areas will show different faces by movement and angle of the scanner so it will be not symmetrical , it is different if you scan just a single cube vs many of them at a different angle .

Good luck with your project !

[PUTV]

@DDRDan

I didn’t come into an ‘outside’ forum. I came to the “Revopoint” forum. I expected their tech staff to have a presence in here to resolve customer issues. I now realize their tech support is “What I did” tech support. People perpetuating other peoples costly mistakes

Technical support is related to the hardware issues not your personal scanning issues , please visit the Forum Tutorial section and learn more about how to work with the device . Tutorials - Revopoint 3D

Your scanning issues are related to lack of experience in 3D scanning using this technology trying to scan objects that are not scannable on its own , I invite you to watch some videos to better understand how it works , practice makes perfect .

Good luck!

[fxtentacle]

First off, I’m not affiliated with anyone, I’m just a curious user and I have my own issues with the scanner and the software.

@DDRDan

PS: How hard could it be for Revopoint to include a .DXF export??? The limited export options are a royal pain in converting because the files are large in my scanning application! You built a large format scanner (Range) and you give file extension choices for printing 3D toys?

I’m not sure if that is useful. DXF is a 2D vector graphics format. The 3D scans will be, well, 3D. Of course, the RevoScan could flatten your result onto a 2D plane, but that would be quite lossy. There’s a reason why for professional 3D scanners you usually spend another $10k for a professional reverse engineering software (e.g. SolidWorks + Xtract) which will do nothing but convert the PLY/STL from the scanner into CAD data like STEP or DXF. But in a pinch, Fusion 360 can be rented by the day and it works surprisingly well for extracting 2D sketches as DXF out of a 3D-scanned mesh.

@DDRDan

Can someone explain the methodology in exactly what the software is looking for when it wants location references? I’d prefer using large spheres with a base.

In my impression, the feature tracking (“SLAM”) part of the RevoScan software is pretty bad at handling large smooth areas. But in general, evenly lit spheres work as a marker, you just need to get the scale right. And you will need 10+ spheres visible in any frame, so that it can resolve the visual ambiguities using geometrical constraints. But as @PUTV correctly pointed out, stuff with a lot of wrinkles works really well, because it produces unique shapes in the depth image. I myself have used crumpled painters tape, too, with much success.

I have 0 inside information here, but to me it looks like the feature tracking uses something like SIFT on the depth preview image. That means the part here where SIFT fails: https://youtu.be/IBcsS8_gPzE?feature=shared&t=548 is likely what causes your tracking and alignment errors.

And that means any kind of marker that works well for SIFT will probably also work well for the RevoScan software.

@DDRDan

Will it recognize any uniform shape as a location reference and can I identify & instruct it to use those targets?

Sadly not. There is currently NO WAY in the software to manually select points to be used as an alignment reference. I agree that this is standard procedure in Metashape or Reality Capture, but RevoScan doesn’t support it yet.

@DDRDan

After adding up what I’ve already spent $$$ trying to make this “toy scanner” work, I could have rented a laser scanner. The Range is not an economical large format scanner!! … period. You should stop advertising it as such.

That will only save you money if the laser scanner can capture the entire area in one go. Otherwise, you’ll have exactly the same issues once you need to merge multiple laser scans into one result.

What I would probably try in your situation would be to stick matte painter’s tape into the corners that you want to measure and then take some high-res pictures with a cheap DSLR and throw the images into the Reality Capture “pay as you go” app. You’ll get a free preview and then you can decide if you want to pay $2 to get the actual data.

@DDRDan

I didn’t come into an ‘outside’ forum. I came to the “Revopoint” forum. I expected their tech staff to have a presence in here to resolve customer issues.

I believe they see this more as a platform for users to talk to each other.

@DDRDan

The Range is not an economical large format scanner!! … period. You should stop advertising it as such.

As you can see here: Feature Request: Turntable + Single frame and here: Feature Request: Pre-Fusion + Keyframe ICP + Final Fusion I bought the MINI 2 which is advertised as a jewelry scanner and while the hardware is fantastic, the RevoScan software is still severely limiting the result quality. But there are workarounds and with a bit of training, I could get amazing results.

In my opinion, Revopoint is a great example for a Kickstarter. They did deliver the hardware, it works, it arrived on time, so I would cut them some slack while they iron out the software issues. And some of us, like me, actually enjoy the tinkering process.

But if you need industrial-grade results now, then Revopoint is probably not the right vendor for you. (my personal opinion)

[DDRDan]

First off: Revopoint contacted me to address my issues. I failed to educate myself completely in SL Scanning. And, as you stated, Range has problems with certain large flat surfaces. I was expecting too much from it. I should have contacted Revopoint directly before coming to this forum. I found that Revopoint has excellent customer service.

@fxtentacle

I’m not sure if that is useful. DXF is a 2D vector graphics format. The 3D scans will be, well, 3D. Of course, the RevoScan could flatten your result onto a 2D plane, but that would be quite lossy.

My application (flat decks) only requires a 2D vector file. Flattening the x/y plane is exactly what I want. It would eliminate my haveing to produce a trace. If RevoScan can give me an averaged flat plane (Z@0.0) I need to read up.

I’ve used 3D to DXF files and the conversion imports all have 3 axis every time? It never removed the Z axis? It’s a lousy mesh but I don’t need that.

@fxtentacle

In my impression, the feature tracking (“SLAM”) part of the RevoScan software is pretty bad at handling large smooth areas. But in general, evenly lit spheres work as a marker, you just need to get the scale right. And you will need 10+ spheres visible in any frame, so that it can resolve the visual ambiguities using geometrical constraints.

CNC’d flat top pyramids (20 mm) with 6mm circular indents last week from HDPE stock. Milled HDPE is dull and gives me an acceptable surface tolerance. Doing a test this week.

@fxtentacle

That means the part here where SIFT fails: https://youtu.be/IBcsS8_gPzE?feature=shared&t=548 is likely what causes your tracking and alignment errors.

And that means any kind of marker that works well for SIFT will probably also work well for the RevoScan software.

Thanks for that link. I reduced my scan distance, and I’m trying to stay consistent with matching the angle of the scan to scan lap areas when moving the scanner. Both helped.

@fxtentacle

That will only save you money if the laser scanner can capture the entire area in one go. Otherwise, you’ll have exactly the same issues once you need to merge multiple laser scans into one result.

I may be misreading your point? But, I’ve scanned large helical arc shapes with 50+ scans at multiple z axis levels and never had any major problems with Leica or Faro software scan merges?

@fxtentacle

But if you need industrial-grade results now, then Revopoint is probably not the right vendor for you. (my personal opinion)

I would agree. I wasn’t looking for commercial grade but I was looking for the ease and quality that is being “user” mentioned on the web. Two things went wrong when I researched “which” scanner to purchase?? I believed what I was seeing in videos and odd posts on SLS and I neglected researching the SLS technology before purchasing. My fault.

Thanks for the info!

[lajoosh]

@DDRDan

My application (flat decks) only requires a 2D vector file. Flattening the x/y plane is exactly what I want. It would eliminate my haveing to produce a trace. If RevoScan can give me an averaged flat plane (Z@0.0) I need to read up.

I have never heard about such thing. But if you wanted that, you would have to use a third party scan data processing software like geomagic wrap, to align the scan data to the coordinate system properly first, then maybe try a non-uniform scaling operation, where you basically set the scale on your required axis to zero. Let say you flatten your vertical axis, you would see surfaces that are perpendicular to the vertical axis as a sparese point cloud area, and the surfaces that were vertical, as a denser point cloud area. I don’t know how that would be useful though. I did a bunch of stuff for boats, including deck anti-slip covers, and I honestly never thought of flattening my 3D scans, I dont see the point personally. Just loaded up the scans in cad, attached some planes to the surfaces I was working on by 3 points and traced the contours manually.
3D scanning will never produce 2D vector files for you. You will have make those on your own.

Regarding scanning large featureless surfaces with markers: Here is what I did: spray the surface enough that it is not reflective anymore… stick markers on transparent plastic sheets that were used for oldschool presentation projectors. attach these transparent marker sheets to the surface by blu-tack, tape, whatever. The scanner can see right through these sheets, no problem, and you can quickly cover large areas with markers, and they are also re-usable. Scan an area, move sheets to the next and repeat!
This will work well enough and produce a good result, however, never expect the same ease of use speed and quality of end result to be on par with multi thousand euro scanners, thats just not fair.
If you want to do this job as a business, you might want to invest a bit more.

[lajoosh]

Just out of curiousity, I did what I suggested on my own, flattened a pointcloud and this is the result:

Exported as DXF, CAD software sees this as millions of sketch points.
Useful? Probably not.

Edit: you can have the same result just by exporting a pointcloud from a specific view angle to a DXF file. No preliminary flattening is necessary.

[DDRDan]

Thanks for the info. Also, this is not a business venture. Personal projects only. If I were doing this commercially I’d get a Faro unit.

I’m getting good results by laying continuous 1"x 1" strips of lumber across the deck on 4" min. spacing and somewhat parallel with each other. Then adding a light spray. I don’t need the horizontal or vertical area surfaces, I just need the termination of the edges across long flat distances. Getting from point A to B that have no recognizable features. Losing the scan in the middle has been the problem no matter how many dots or targets I stick on it. And trying to scan just the perimeters edge had the same problem.

The strips get me from edge to edge. And deck features like hatches, cleats, etc. have their own edge features that are being recognized. And I’m using a custom macro I developed for CAD, before they had a flattening command, to complete my trace. It’s not as finite as the software I’ve used recently but it’s doing the job.

By ‘flattening’ I am referring to having the software generate a common plane among the points and also establishing an edge outline. With a square as an example, it can find the most common plane / edges (within the pointcloud) and generate a 2d outline of all the planes. The same for a trapezoid, curves, etc.

I’d rather not manually trace the scan and waste material on the CNC with possible errors when I go to fit them. AI is smarter than IM. Simply dropping a point cloud into CAD and applying planes and a manual trace has it’s pitfalls with pointclouds if you’re looking for accuracy.

And that feature is available. Most commercial laser scanner software’s have it. (And yes I agree this is not a commercial product) Drop that file into AutoCad and it can flatten most any shape into a fabrication profile. Right up to conical and helical shapes. And, provide a spline line review of those lines.

Had it not been for those features I would have never been able to produce the multifaceted ACM watertight panels to meet the helical framing to a 2mm tolerance on both sides of the stair pictured below. The side not shown had a 3 level soffit return with 2 different slopes. There are no 90 degree angles in any of those panels at the handrails. And only 3 similar angles in the whole project. We factory fabricated ‘one’ time and it fit like a glove. I did the same thing on another project with a 150’ long sloped, tapering, helical arc facade, atop 8 stories, on a hospital.

Architects are being crazy and 3D scanning has been the only economical method in fabrication to meet their wants.

[jyardley]

@DDRDan

Architects are being crazy and 3D scanning has been the only economical method in fabrication to meet their wants.

Lol - Not just us then Great work!
It makes me seriously question how ‘old school’ fabricators / machinists stay in business… I guess they are the ones moaning about ‘no profit in the industry anymore’

[DDRDan]

I had to do 6 presentations on 3D scanning possibilities to the company I worked for before they were convinced that making the product ‘twice’ or cut & fit prefabricated panels wasn’t the way to approach profitability. And if it wasn’t for a US tax incentive to buy equipment I think they’d still be field measuring and fabricating twice. The amount of high dollar materials that went to the trash was outrageous. I turned the scanning into a sub business in scanning projects we lost the bid on and they paid for a $70,000 scanner purchase in one year.

This technology has grown exponentially. Architects are elated when you provide this kind of data on their projects. To a point where they were contracting scans to evaluate billing for completed work reviews every month. Before I retired they were convinced to look at Geospatial Lidar / Laser for multi story scans with drones. Prior to that I had to platform or crane scan tall buildings.