Separate names with a comma.
Some features disabled for guests. Register Today.
Discussion in '3D Scanners' started by Ticky-Tack, Jun 21, 2019.
A mid-sized version of a Photogrammetry 3D Scanner.
Ticky-Tack published a new build:
Read more about this build...
Awesome. Really want to see where this is headed.
I've went through a number of ideas, before arriving to this one. My objective was to make a high quality open sourced 3D scanner. And it started with investigating with LIDAR methods. I still intend to do LIDAR in the future, I've always wanted to play with highly sensitive Time-of-Flight circuits. But the costs are just too much for what I'm trying to achieve atm.
For the LIDAR idea, I was going to build a circuit based off Texas Instruments TDC7201 TOF chip http://www.ti.com/lit/ds/symlink/tdc7201.pdf
Using differential time between the two timer circuits, TI claims you can achieve sub-mm accuracy with their built-in 128 sample averaging.
The circuitry made sense, and I could manufacture my own dev board for $30, or buy theirs for $80. Where the problem came in was with the laser and optics. The optics would require a beam splitter, and some nice quality lenses to achieve a small diameter laser dot. If I can't get the laser dot focused to a mm/sub-mm level, the TOF sampling would be useless. I'd also have to use a higher powered laser with the beam splitter in place, and for the receiving sensor to detect the defracted light reflecting off the object. Which puts the laser/optics unit itself into the $200-$300 range. Still cheaper than the majority of the LIDAR systems out there, but not really within my budget.
My next venture was to use a DSLR, with a similar design as to the one I'm building now, just had more robust support from the top of the linear rail. But then the costs were really high with that idea, since DSLR's are heavy and bulky.
I sacrificed the DSLR, thinking the higher resolution will not provide any added benifits for the objects I will be scanning. If I was doing full body scans of people, I can see the benifit of a DSLR, or multiple camera setups. But for the distances I'm looking at, I don't think the lower resolution will make much of a difference. It also slims up the design a lot.
I have parts on order to make a "500mm" version of the design. The CAD designs were made for the 1000mm design. What is nice about this design, is you don't have to change much to scale. Just swap out the linear rails and V-Slots for smaller dimensions, and that's it. There maybe an upper limit to how tall you can make the scanner, and would require additional support, but that's simple, just add more V-Slots.
Too bad we haven't invented portals yet, I want my parts to arrive now lol.
OpenBuilds Portal! LOL.
Some updates to the designs.
Turntable wheel support now sketched up
Slight adjustments to plastic part dimensions, mostly adding more clearance for screw holes
Turntable mounting spool designed
New files have been uploaded to build page.
Well, the build process has been a lot of trial and error. There were a number of problems. But I finally have a working prototype.
What went wrong:
The linear rail for the camera would not fold down with it's own weight. Had to add a linear spring to keep tension on the timing belt
Turntable wheel holder sucked, they broke easily. I made a new bracket, 1 stencil assembly for all 4 wheels, and more robust.
The timebelt bracket at the top of the linear actuator was flimsy, it broke one day causing the camera rail to swing down, slam into the turn table, and broke the hinge on the carriage. The hinge and timebelt bracket has been beefed up
The turntable stepper motor mount was stupid, it's hard to get reliably flat prints from plastic, just too much room for temp flexing, over extrusions, etc. Replaced with a standard metal collar for stepper motors, the turntable is now level
Reduced the size of the camera carriage to only have 2 linear bearings instead of 4. 4 was over kill, and again with 3d printers not being the most accurate, having 2 bearings offset from eachother by a small amount, adds a lot of friction, meaning more strain on the motors.
My motor drivers fried themselves, don't trust china
The movement of the linear rail was not what I was expecting. My imagination doesn't play by the laws of physics, but the movement still serves the purpose I need it for. Just had to write some more trig into my program
What has been done since the last update:
Arduino program, and serial protocol has been written
Node js API server to interface with serial protocol, and provide a standard http protocol for control
Camera coordinate and angle translation
Scanning movement program done
What's to come:
Actually start taking pictures
Make a UI to show and control the scanner
Test and debug, cause we always know that happens
Polish CAD designs, all of my new parts are in the CAD design locally, just they aren't placed correctly, and don't have screws, washers, etc placed
Looks great! I want a photogrammetry setup, but unfortunately it would likely be primarily for people, which for all practical intents and purposes means the hundreds-of-DSLRs-on-a-cage method. Even at ~$200 a camera, that racks up real quick!
A really nice hand-held laser unit at affordable (ie. not $100k) prices would be pretty handy too though. A couple times a month I wish I had one!
It's fun to do, anyway. This was a park bench I shot a couple years ago:
Some fairly standard photogrammetry precautions and PhotoScan made very short work of it, it was really easy. It's sitting on a quick Substance Bitmap2Material floor underneath the vignette of ground.
Not sure how valuable this method is in terms of actual calibrate-able measurements though, unfortunately.