Tagged: 3D scanner
October 13, 2018 at 10:43 pm #71951
New to all this. I’ve buried myself in videos and searched the net far and wide. Know what I didn’t find?
“MPCNC 3D Scanner”, that’s what.
You guys have machines that can do a 360 around an object with the X and Y controllers and I mean perfect circles: I’ve seen you do it. The Z lets you move up and down. The controller for the “spindle” would let you rotate the camera to keep pointing into the center of the circle.
Storing of the images would be the standard Octopi time-lapse deal.
Jumping catfish, you could just print a phone bracket on a rod and chuck it into the mill head. Do a 360 while you take a time lapse. A couple of laps at 2- or 3-different heights and then shoot the Octopi recording at one of these free photogrammetry apps. If you can get a working scan with a hand-cranked turntable and a cell phone, one of your rocking monsters would get you some high-quality meshes.
Given all that, I’m curious why no one is posting something like that. It’s very possible that, being a noob and an idiot to boot, there’s something I’m not getting here. Am I stupid or would that be doable thing?October 14, 2018 at 5:07 am #71956
There CNC would need to have 4 axis. The spindle control is good in the thousands of rpms. Other than that, what you described is doable. Instead of using octoprint timelapses, there are some features in octoprint to add gcode that gets intercepted and calls system commands. It could take a snapshot from a webcam. A phone would be hard to interface with to tell it to take a picture.
The really problem is that it’s fishing with a rocket launcher. A dc motor on a turntable can automate the same thing and has the advantage of having just one direction for the camera. So only one direction to worry about lighting and a background. If you can fit it inside the MPCNC, then you can fit it on a turntable. Even adding a second axis to a turntable would be pretty easy.
IMHO, it’s doable but not a good fit.October 14, 2018 at 9:25 am #71969
@jeffe3b: yeah. I can see that. With so many guys using an Octopus setup with a webcam, that webcam would be hugely better than a phone, but you wouldn’t need the print server software. It MIGHT work. But better ways to do it.
Thanks! I’m a little smarter now!October 16, 2018 at 1:20 pm #72271
My 3D scanner uses an old Xbox Kinect for the scanning, plugged into my PC.October 16, 2018 at 5:31 pm #72306October 17, 2018 at 11:43 am #72406
@jeffeb3 you think the 5 axis I’m working on would be a good fit? Im having issues getting enough torque to cut anything with a spindle attached, but it should turn a camera just fine. That will give you camera angles and hieght, and the rotary table will spin the piece. X and y will mostly move to adjust to maintain consistent distance from the piece.October 17, 2018 at 12:45 pm #72428
Every type of 3D scanner is different and needs to be used how it was designed. For example using the old Xbox 360 Kinect as a scanner had similar results and quality as this scanner featured on a video by Ben Heck. I have a version 1 of these and I am not really impressed. No usable detail on anything smaller than a cupcake. These types of scanners are designed to go around the object getting scanned. A modified MPCNC might be able to work with one of these but you would still have to buy the 3D scanner and those are either poor quality or really expensive.
The original idea posted in this thread mentioned using a camera to take hundreds of pictures that get spliced together using software to build a 3D model. I have looked into a few offerings like this a few years ago. Again the detail of the results is usually not that great. Sure you could scan a human and print out a doll with blurry or smoothed over features. Not good enough for what i wanted to do at the time. And again in that case why would you build an MPCNC big enough to put a person inside of. … … Actually that might be fun.
Other types of 3D scanners hold the camera or scanner in one location and spin the object getting scanned. If you are trying to scan small objects then I suggest looking here.
These have two types of scanner tech. One uses a camera that takes lots of pictures. The other uses a laser that draws a line. The laser line is more accurate for small fine detail. (The laser line usually has a hard time scanning shinny or reflective surfaces.) Both options you need to use software that builds the models.
In conclusion. Depending on what you are trying to scan and how big it is you will want different types of 3D scanners. For small objects I suggest using the ones that spin the object. For objects up to the size of a briefcase I suggest building a platform with a long arm like Ben Heck did in the first video I linked. For really big objects like people or cars I suggest you just use a hand held scanner.
I don’t see a great use case to use an MPCNC as a scanner framework. It could be done but I think it would be easier to do using different methods.
That’s just my opinion.October 17, 2018 at 12:50 pm #72429
I remember using my cell phone one time to take a bunch of pictures of a hot wheels car that I then blew up and 3d printed double size.
Not sure how that fits into this post, but it was the only 3d scanning I’ve ever tried 🙂October 17, 2018 at 2:48 pm #72440
@jeffeb3 you think the 5 axis I’m working on would be a good fit? Im having issues getting enough torque to cut anything with a spindle attached, but it should turn a camera just fine. That will give you camera angles and hieght, and the rotary table will spin the piece. X and y will mostly move to adjust to maintain consistent distance from the piece.
It’s possible, but I still think it would be easier to do without moving the camera and just moving the thing you’re scanning.
And as David points out, there are several good algorithms for localizing the camera, so just walking around with the camera works pretty well.October 17, 2018 at 4:21 pm #72454
Naomi Wu did a full body scan at one of those scan and print yourself places. They use a lightfield scanner that has several cameras. Even that scan wasn’t the most detailed, but turned out alright. Just shows that even a professional 3d scanner doesn’t always produce the best files. She uploaded the obj to thingiverse. I wouldn’t suggest looking for either of these at work, her standard outdoors attire would piss off most church goers.October 17, 2018 at 6:56 pm #72476
Dui, ni shuo de duiParticipant
It shall be possible to do on the MPCNC but I don’t see the point to do it this way.
Basically, you would need to add two more axis, one for the camera to stay oriented to the center of the object you want to scan, and the other to change the Z angle, in order to scan the object from the top or the bottom, so you can get almost all details. The trajectory of possible positions of your cameras would look something like a sphere.
But I think it is a huge waste of time compared to just a simple rotating table.
What could make this project interesting would be to get compiled data of the images + the actual position of each stepper motor at the exact moment the image would be taken. Then, you’d need a small dark room, get yourself a nerd mathematician and throw it in there to crunch numbers and trigonometry and you’d probably come up with a 3D scanned object file and a lawsuit for kidnapping a mathematician.August 10, 2019 at 7:01 am #109176
Could you please help a newbie?
I need a scanner to foam casts – depth precision is important. Would a Kinect 360 do it? If so, what software do you recommend?
You must be logged in to reply to this topic.