Posted on Sun 03 December 2017

A few results on 3D reconstruction

Disclaimer : This post contains lots of Highdef images. Please be patient and wait for them to load.


Secondly, this post just displays my outputs in 3D reconstruction. I had developed a 3D reconstruction arrangement this last year.


Cool stuff: I’ll be making a complete blog series on making a 3D reconstruction program. Everything from using the camera, … all the way to working with 3D files.


I’ll probably host it as a GitBook. :)


Setup:

I used the single camera and projector setup. Most stereo setups use a 2 camera setup for stereo vision. However, I was implementing Brown University’s paper on using a single-camera-projector arrangement for 3D geometry acquisition.

setup1
Initial Setup using projector and camera

setup2 setup3
Camera Calibration using SL patterns

setup4

Outputs


crop12
Registration of David Bust, and Bottle

david1
Surface View 1 of David Scan (from dataset)

david3
Surface View 2 of David Scan (from dataset)




Although, for live condition, there is a lot of noise.
bottle_set
Structured Light in runtime with Bottle example

bottle1
Tupperware Bottle Scan - Stage 1 (Improves later on)

cap
Closeup of Bottle Scan (next stage). If you look closely, I have made a dimension as well. The cap length shows to be 26.23 mm.

images_crop4

im_crop5
Structured Light Sequences projected on Bottle 2

bottle2
Random Bottle Scan (later stage). I cannot find a picture set for the above tupperware bottle.

bottle_closeup2
Cleaner scan of above object (End Stage)


woodlandmontage
Structured Light Sequences on Woodland Shoes

woodland woodland
And here’s a 3D scan of my shoes :P


the current outputs are just as a temporary showcase. I will be hosting the complete steps of how to make a structured light 3D reconstruction engine, with only 1 camera and 1 projector soon. That’s for sure to be expected by the coming year.


Results and Summary :

results

As seen, there is a error of about 3mm. Which can be resolved by further robust calibration, and less interference/disturbance of surrounding light - SO that an accurate BRDF value can be computed. Also, the material/object under scanning can be coated with a MATT-FINISH POLISH, to avoid surface reflection (as seen in the bottle).

Anyways.. I’ll get the complete series online soon. Stay tuned ! ;)


comments powered by Disqus