A new way to use Marker tracking to scan

TL;DR at the end
So I’ve mostly been testing the Pop2 since I got it a couple months ago, and I’m working on something I can post to the showcase thread. My main interest in the scanner was using it to scan figurine sized objects, so generally not over 15in/38cm. I’ve been able to extract some very good details out of the one I plan on posting eventually, but it is a difficult scan. Its a mostly PVC figurine that was painted using dark shades of red, blue and black, and has some highlights of canvas tan. Needless to say IR does not play nice with dark colors when scanning, so it has posed some difficulty with tracking or even scanning in general.

I’ve just started playing with an idea today that has shown some promise. Using marker tracking mode works great for the hard to scan objects, but the problem is getting enough markers in the frame around certain parts of a figurine without putting stickers directly onto the figure, which for some features just doesn’t work well. Like trying to scan under an overhanging feature where there is only enough room to fit maybe 3/4 of one sticker, and the angle of the scanner is such that it will miss most other markers in the area. That got me to thinking about how the markers are tracked. Now without looking into the program logic RevoPoint uses I cannot confirm how it is programmed, but for the fastest and most reliable processing it needs to look for a static shape and the ease of detecting that shape is determined by contrast. At least in the machine vision systems I’ve worked with. The stickers provide a white dot inside of a black circle, a high contrast target, but the camera is only looking for the white circle, not the black ring. That train of thought lead me down a rabbit hole and I emerged from the other side with an almost perfect solution.

Sewing needles.

To be specific, ball head sewing needles, the ones with a colored ball on the end.
I got my sewing supplies and grabbed a tin of these needles with white balls. I stuck some of the Blu Tack that came with the scanner onto various spots of the figurine I’ve got set up, poked a bunch of needles through it so the white balls are all around the target area.

Sure enough the scanner tracks the needles like the marker stickers. The ones I’ve got on hand only have a 3-3.5mm ball on them, the tracking has the same issue you can get when at an angle to the stickers and they don’t reflect well (even though the balls reflect very well), I’m hoping it can be made better with larger balls. I’ve ordered some pins with a 6.5mm ball to test if they work better and should be able to update you before long.

One of the great things about this is that the scanner does not pick up reflective metal hardly at all, so the needle part isn’t visible to the scanner. Keep in mind that while the needle portion does not get picked up by the scanner, it will still block the projected light used to scan your object, so placement will effect how well it scans. Lots of needles in front of the object mean lots obstruction in the scan path. You can see the shadows they cause in the black and white image, so be mindful of placement.

I did also test changing the background and noticed that when putting white paper behind my target it had more issues with detecting the pin heads.

Two more great points in favor of this, ball head needles can be picked up for relatively cheap (100 pack of 6.5mm ball with 65mm shaft for 5 USD) and can be reused indefinitely.

TL;DR

White, and possibly pearlized ball head needles can be used under the marker tracking mode instead of the stickers.
Use some tack, putty, or whatever you have on hand to stick the needles around the target so there are plenty visible at every angle/side you are going to scan.
The metal of the needle is not picked up by the POP2, but still blocks light used to scan, be mindful of placement.
May not work well against white backgrounds.
Highly reusable.

If anyone has the supplies on hand, please go ahead and try it and let us know how it works for you.

Edit 1:
P.S. Anytime you’re pushing needles through something, please be careful not to stab yourself.

6 Likes

You see , you can buy a very small markets and use it too , it don’t have to be the same size of 6mm … if your figure has dark color mostly you can use white dots label stickers .
I used it all already , you can use the pins with ball head also in feature mode to add some extra feature , there is lost of possibilities.

And one thing, any visual scanner have issues with dark colors, as the scanners need to capture reflections from the surface and black or very dark paint sadly can’t do that , no matter IR, Laser or LED .

Hi Dalhimar

I have a few tips to possibly get better results.

  1. Add a black background instead of white
  2. Scan where there is not bright light but where its dim then add a directional light pointing the same direction as the scanner on the object again try to defuse the light a bit.

Hope to hear your feedback soon.

2 Likes

Well I received the new ball head pins, which have a 5.8-6.1 mm pearlized white ball. I’m still doing some testing with various scans to confirm my results, but I can definitely say they track better than the smaller pins I was playing with.

@Revopoint
I don’t have the materials to do what I want right now, but sometime in the future I want to construct a scanning box about 2ft on a side and at least as tall, then line it with IR flocking.
Until then, or until I set up a better scanning area, I have a sheet metal cabinet door that is painted black to use as a backdrop. I would have to go into a frame by frame analysis to tell you exactly how much it effects the tracking, but it does appear to track the pin balls better when they are passing in front of it. It is possible that it is a placebo effect, hard to tell for certain.
I have tried changing the lighting, this being my work room I’ve got multiple sources of light that I can adjust. As long as I don’t have anything too bright it doesn’t seem to effect the scan much. I’ve tried using different color temps as well. I’ve got lights around 3100K, 3600K, 4300K, 5500K and 6100K. It doesn’t appear that the color temp alters the scanning with POP2, at least in the normal temp ranges. However these are all LED lights so their spectrum is pretty tight. I’ll have to revisit the color temps when my Mini comes in, since I know that typical LED’s have a spike in the blue spectrum and might interfere. In fact I think I recall someone with the demo/prototypes saying their LED lights seemed to cause an issue.

To reduce how much it influences the scanning, I have kneaded carbon black/lamp black pigment powder into the Blu-Tack I have been using, and now it isn’t picked up unless I set the exposure towards the upper limit. Word of warning doing this, the tack will hold onto the pigment powder pretty well, but there will always be a risk of it leaving some behind.

So far this method of marker tracking seems to work best when you’re able to position the needles so that the shaft does not cross in front of any of the balls. The shadow the shaft cast can interrupt the tracking.

Here is a scan I put together quickly using this method. Done at depth camera exposure setting 6, meshed in CloudCompare at octree level 10 with no cleanup. From the left post to the middle post is about 48mm, and middle to right is 51mm.


4 Likes

saw this on the internet. not sure if it works better or worse, if it works at all…

1 Like

Would a nylon net with white “beads” also work? The method definitely sounds interesting!

@new23d It depends what scanner was used , according to the marker distance looks like Range or MIRACO .

how higher the scanner accuracy how closer the markers need to be to each other , Range can use up to 8 cm apart due to bigger FOV .

@SphaeroX yes if you use general mode

like this

1 Like

I tried this the other day without much success. Revoscan kept saying that there is not enough markers

How smaller the FOV how more markets you need . The markers works best when facing the scan while rotation . I think you need to have at least 3 markers visible all time to continue scanning .

I would 3d print something like this and replace every single marker with it. much more markers per surface area.