Car bumper without target markers

I’ve been working on scanning a front bumper for a race car I am building. I’m conducting two scans, one of the outside face and one of the inner face. The outside has to be done with markers but I thought I would see if the inner could be done with multiple scans and merging then given the surface is unpainted and fairly disordered bare fibreglass - here is the WIP results so far, the object is 1.8 metres long for scale.

I used common features (light wells, mounting plates) to create about 12x scans so far and join them together using ‘align’ and ‘merge’ in Studio, this is simply using features to align, not markers. Using this method I can identify missing detail and scan small sections, filling in the work without the risk of failing a scan or losing alignment during the scan. If I mess up, I can easily go back without losing my work or risk of the odd software crash.

This is a really tricky part to scan, mainly because you would normally scan from the centre and work outwards, but there are very few features to lock onto in the centre, meaning I have had to start from one side. This means on such a large object you get scanner tolerance accumulation, but so far the POP is giving me about 1mm / metre which is exceptional, especially since this is a real stress test for this little unit.

Pop is taking a well deserved rest right now, from being on and working all day :grin:


Alignment evaluation.
I’ve split this up into a few sections - see image.

Points [A] are the light fixing holes, they all align very well horizontally and are in 1mm of their position
Points [B] locate onto the car and are aligned, horizontal under 1mm accuracy
Intakes [C] all align in the horizontal, suggesting the scan is straight across the front of the part and to a high degree of accuracy.
[Arrow] shows where the scan has ‘dropped’. This was the last part of the scan and I scanned it at a funny angle which I think confused the scanner’s position as it was trying to deal with a compound change of direction.
[Yellow} this is show by the skew in alignment to the mounting plates - it can be seen the right is lower than the left.

I’ll fix this by going back to a previous ‘save’ as mentioned, re-scanning and merging the right hand side again by aligning the scanner ‘normal’ to the surface this time.

Hope this helps give people some ideas with tracking and work flow. The results speak for themselves, the device is highly accurate and even using a stacked method can get great results.


This is really promising. I’ll tell you that I just spent the last 3 hours or so trying to scan the engine bay of my DeLorean without much luck at all, and this gives me some hope. That’s a great tip to use multiple scans of small areas and then merge them. Maybe I’ll give that a shot.

Are you using the ‘feature’ mode in the scanner? What platform are you using to do the scan? (Android, iOS, PC?)

Feature only - Windows 10
You have sister car, DeLorean to my Esprit :grin:
There are some limitations to using feature scan - imagine it like you’re putting a jigsaw puzzle together and then doing the outer loop (with the straight edges) what are the chances you’ll get the last piece to fit both ends of the loop perfectly (ie you wont) this is similar to feature scanning which is why you have to back it up with physical dimensional checks.

Keep in mind thou this bumper is a really tough thing to scan without a CMM and dots due to the complex shape, dots are king in this field. Share DeLorean scans when ready :grinning:

Sweet, I LOVE the Esprit!!

I’ll be adding dots shortly, and I’ll see how it goes :slight_smile: I also made myself some matting spray with rubbing alcohol and baby powder - it actually made a massive amount of difference and was even easier than I expected.

I will keep in mind needing to confirm dimensional accuracy. There are a few straight flat pieces that I can easily measure and then do crossways measurements between the most important points. The scan will be made freely available. :slight_smile:

1 Like

Yeah the more matt white the surface is, the better returns on the scanner. I use a custom tripod to position the scanner and constantly adjust the gain and brightness, sometimes lower values than you expect offer more detail / exposure.

Cant wait to see results.

Do you just mix alcohol and powder? How viscous is it? Can it be used in a pumpspray?

Yeah, I mixed 91% isopropyl alcohol and generic baby powder. Roughly a 50/50 mix, maybe? I mixed it in a spray bottle and shook it up and it sprays into a mist just fine. Initially it is mostly clear, but then when dry it produces a very hazzy matte finish that the camera picks up much better.

That said… My results from doing the scan have been absolutely terrible, even using markers. :frowning:

Here’s a short-ish video: Revopoint Pop 3d Scanner - Not really working - YouTube

I will say that the matting spray doesn’t show up very well on camera and looks dark, but in person it is much whiter - and you can also see that the IR camera picks it up very well. I’d love to get some tips before going out and trying it again.

1 Like

Watching yer vid, move the POP back about twice, even 3x the distance. You started too close and it lost contact, I can see that by the lack of red marker triggers on the screen, then when you pulled back it saw more but you already ‘lost’ the scan - once it’s gone it cant recover. This is why I wish they had a ‘scan dots only’ option. once you have a large scan of dots you can move in for better detail for your car I wouldn’t worry - your point cloud will be astronomical for an object such as… on a flat plate.

Forget what it says about ‘too far’ etc that only affects mesh density, and for flat panels you dont need it unless your PC can handle 20 million triangles on a flat panel. When you move it back you’ll get a much larger field of view and it’ll track better.

I suspect the AI is getting WAY confused trying to sew the dots and the scanned features together.

Next, ditch the hand held - get an awesome kent-faith tripod like mine and use that. What happens is your arm gets tired, you’ll start to shake and the AI cant keep up. I’ve lost so many scans now when I get about 3 mins in and my arm is going wobble wobble and the PC is going “you wot mate” :grin:

I’ve also noticed POP wont lock back onto the scan with my giggy giggy arm movements on restart - if its on the tripod it take a few secs and then locks on, nice :ok_hand:
I’ve not used the hand held since I moved over.

I’ll try and do a time lapse when I get my dots as I might scan my chassis which is stripped, but my vid skills are dreadful :joy:

1 Like

I can definitely give that a shot - I’ve got plenty of tripods available! My main for camera work is a Manfrotto with the fluid video head. I had someone else suggest more matting spray and more dots, so I’ll do it all. :slight_smile:

One limitation I hit is that there seems to be some kind of clip plane set back more than about 18 inches or so. Anytime I go farther out than that, it’s all black. I’ll try to move it as far as I can without running into that.

You need the tripod or a lazy arm, sudden small movements throw the POP off quite easily.
Yes you cannot go infinitely back or the correlation between the two cameras and the IR wont work, that’s why there is a threshold where the scanner works best in for detail (the good, too far scale).
Just look at your field of view - as the other poster says you want 4-6 dots always in shot. If you’re close, you need more dots in a denser arrangement.

I’ve got another thread on here with the bumper dotted up - and it scans nicely.
You don’t need any more matting spray as its scanning perfectly.

I’ve trained people with $100,000 machines that find scanning tricky, you’re constantly learning - it is an art form that requires practice, unless you have a CMM arm. What the POP can achieve is very impressive for its cost but it requires more input from the user due to the cost reduction.
once you nail your method down, you’ll be really pleased, but it does take an investment in time.

Always happy to help.

1 Like

Well versed. What you save in the initial cost of the unit, is definitely offset with the time investment.

For sure you pick up technique with every scan you do.

1 Like

Did another attempt tonight. I put about double the number of markers, maybe more, sprayed a bit more matting spray in some areas, removed some wires and things that were sometimes in the view, moved the camera back as far as it would go without clipping during the scan, and put it on a nice video tripod mount so that I could move it verrry slowly.

The scan definitely came out BETTER, but still not great. It’s definitely getting closer to being useful, though.

I think the biggest deficiency is with the software’s handling of markers. I strongly believe that as soon as the markers are no longer in view, the software completely forgets about them. In other words it doesn’t save them into the saved mesh in memory, and since they are totally flat when the camera comes back it can’t re-align to them. I have tried to do passes where I start in one area, pan to the left, move the camera up a bit, and then pan to the right. When I do that the mesh NEVER lines back up.

I’m wondering if maybe I should 3d print some star shapes or something, or maybe use nuts and blue tack, in order to make some better reference points at the start and end of the scan?

Gizmos work, if I have an organic shape I want to align, I sometimes scan in an engineers square or something to help and then edit it out later.
I’ve not scanned with dots for a long while but the ‘scan dots only feature’ is what I have been banging on about for a bit here, creaform use it and it certainly works. From what I can tell though creaform puts them into a ‘layer’ so a separate entity to the point could, I don’t think this is how the handyscan software works.

I still think the key to using POP is slow movements and smoothness. Every time I have had a failure it’s been where I have made a sudden movement. You can see it when resuming a scan, it sometimes takes 2-3 seconds to lock on where it is working out the geometry.

Oh try not to ‘pan’, you want to make linear movements which is why I use a boom tripod - ideally the scanned face needs to be normal with the scanner head - which is why you use a rotary turn table. Imagine you’re spraying the object with a rattle can - you wouldn’t use rotating motions (unless you need to get the into recesses). I really need to do a video on this but I’m useless at that side of things.

BTW I have never had any joy resuming a dot scan so I would wager you are correct in that it doesn’t remember the positions.

1 Like