Printable Turntable Laser Scanner (Page 1 of 5)

All of the software was written by me aside for several supporting libraries (libpng, libwiringPi, libraspicam, etc).

The Raspberry Pi communicates with an A4988 stepper driver with 16x microstepping in order to drive a 0.9 degree stepper motor with a very high level of accuracy.  It signals a laser on/off by using a darlington array to more or less act like a relay.  The camera board takes a picture of the laser on and off and the raspberry pi detects the laser pixels from the resulting image.  Since the internal camera parameters are know and the laser plane is known from the initial calibration, the raspberry pi is able to project a ray into the scene to intersect the laser plane and give us a 3D location.  This location is then rotated by the amount that the steppers have rotated the table.  This produces a very nice point cloud.

I just have a dumb simple meshing algorithm implemented right now so I'm using MeshLab to reconstruct the results with Poisson until a I have a chance to implment one that can run on the Raspberry Pi in a reasonable amount of time.

Thanks.

Thanks! The second laser helps with areas that are occluded from the first laser and where the angle between the laser and the object surface is close to parallel.  If there is enough interest I plan on opening the source and hardware design.

+10

I am very interested in this as well. I just ordered up some blue line lasers and a HD webcam to play around with the David scanner system. Trying to wrap my head around how this form of 3D scanning works so I can improve upon it.

In my research, the blue lasers were recommended for the narrower beam being able to pick up finer details. I will be experimenting with macro lenses and filters on the webcam as well. Limiting the depth of field and filtering out the wavelengths outside of the laser seems like it would make a huge difference. Am I way off base there?

Nope, you're right on.  However, although a blue laser has a shorter wavelength than a red one but that doesn't necessarily mean it has a narrower beam.  You should check the beam divergence value listed in the specs for your laser.  However, you will get more energy in a green or blue beam than a red laser.  This makes it great for scanning but also makes it a lot more dangerous.  Make sure you take the necessary safety precautions if you're using a high powered blue laser!  A macro lens should be great for picking up detail but you'll need a way to correct the lens distortion.  David scanner may have support for this already but I'm not sure.  Lens distortion correction is pretty high up on my "Scanner TODO" list. :-)

It's not too "high powered", only 5mW (I am not planning to stare into it or at the reflected beams!) it has an adjustable (sorta) lens. I bought it from Aziz lasers. I will have to ask about the divergence since I don't see that listed.

As far as the macro lens, I will be looking for 1:1 macro (really just an extension tube) and bigger diameter than the camera lens to fight some of the distortion.

Thanks.  I'll keep you posted.

Thanks for posting the project.  Also thank you for using a Raspberry PI with the Raspberry PI Camera.  A full up linux system with video is VERY usefull.  I happen to have about 10 Pi's and 4 cameras laying around so IF you put the project up I will come.  You might also want to post this project on the Rpi's forum and see what kind of results there are.

Bob

This is very cool! good job, If you decide to do put kits together that would be awesome, and count me in.
could always use more rpi projects.

I'm using the "official" Raspberry Pi camera which is a 5MP camera but the skull scan above was taken at just 1.2MP (for faster scanning).  Also, the above scan was taken with 800 steps but the scanner can do 6400 steps (1/16 microstepping @ 0.9 degrees/step).  However, more steps doesn't necessarily equate to more detail if the bottleneck becomes the laser quality.  But since it will be an open system, you have the flexibility of upgrading the lasers if your project requires it.  I would imagine that as-is it would work well for the vehicle models you're describing but I haven't tested anything like that yet.  If you post a picture/link of a typical vehicle I'll see what I can find and do a test scan.