Wednesday, 16 November 2016

Looking into RTI (Reflectance Transformation Imaging)

Hi all, Today I saw a post on the 3d scanning users group on Facebook about a technique called Reflectance Transformation Imaging or RTI for short.

Also thanks to Jeffrey Ian Wilson for posting this.

The basic idea of it is to capture a series of images of a surface with different points of light, as to calculate the normals of the surface below. I won't go into any more detail but here is a list of really facinating websites which go into detail about all of the processes required to build an incredible set of images.

http://makezine.com/projects/shoot-super-detailed-macro-photographs-with-an-rti-camera-rig/

http://culturalheritageimaging.org/Technologies/RTI/

http://culturalheritageimaging.org/What_We_Offer/Downloads/index.html

This is a great post going into how to extract all of the maps for textures
Part One: http://practiceround.blogspot.co.nz/2015/10/faux-photometric-stereo-part-i-intro.html

Part Two: http://practiceround.blogspot.co.nz/2015/10/faux-photometric-stereo-part-ii.html


Some cool videos of the process

https://www.youtube.com/watch?v=Xlc8hN1hvm4&feature=share

https://www.youtube.com/watch?v=TjucGJiUkdY&feature=youtu.be

I hope to get something working myself over the next week, I'll probably do a build log of the process along with some example images.

the end goal is to get a substance designer graph with all of the inputs for the images and a bunch of write nodes for the different textures, as I have no programming or scripting experience to automate the process any other way.

Monday, 14 November 2016

Scanning a pub

Hi all, This week has been a bit shaky, with a large earthquake hitting the south island of New Zealand.

On the Sunday I scanned an old building in a place called Newtown which had beautiful bricks and some great arches. Sadly I didn't have my boom pole for higher angled images but I got what I needed.

Stats for this scan are as follows

Triangles: 182m
Images: 615/621
Processing time: 3h:48m:46s

Also today I received a shipment of polarizing film! So I can now go ahead and finish building my mobile scanner for cars and other things. Its all just a test until I  have access to a larger workshop to build the whole thing out of aluminum.


Monday, 7 November 2016

Everything that goes up Must come down.

Well that is what happened to the Dome. But not to worry, it is getting an upgrade.

The basic idea is that a dome like the one I had would be incredibly hard to upgrade, diffuse and polarize, so to fix this I redesigned the basic structure to be more consistent and go with a semicircle. Plans are as follows (very scribbly, sorry!)


  • Build a mount for a ring of lights, with a polarizing sheet placed over a diffusing material, behind the lights is a large reflector.
  • the whole unit is held together and tilts upwards/downwards, so that the camera is always focused on the subject which is sitting on the turntable.
  • You change the angle of the lights for each revolution.
  • Since the light is uniform, flat and polarized we can safely extract crisp images of the subject matter, Its also going to be so much more stable compared to the flexible PVC pipe and number 8 wire system I had going before.

Again, building on the cheap and rapid prototyping is my goal, so I should have this done by the end of the week

Thanks for reading, I hope to push out some updates every evening this gets worked on.

Sunday, 6 November 2016

A Few weeks back it was a diffuse day, Which means I had the chance after work to scan some of the heaver rocks outside my Workplace on a park bench.

Got a fairly alight scan, 107 images. I may bring this one into my dome once I have the turntable ready!

This also gives me a chance to test out Sketchfab integration inside Reality Capture.

Warning before clicking, the mesh is 80k polygons.



Details:


Software: Reality Capture
  • Normal Alignment
  • Standard Reconstruction Normal
  • Excess Points Filtered
  • Total Run-time: 31:07
PC Specs:
  • Intel i7 4790k @4Ghz
  • Nvidia GTX 1080
  • 16GB Ram (I know. first thing on my shopping list!)
  • Running off SSD


Hey all.

My name is Erik Christensen, I am an Environments modeler at Weta Digital and photogrammetery is a massive part of it, I mainly use Reality Capture for my scans, due to the cheap but powerful nature of the program, it has its kinks but the results are awesome.
I have gotten quite far so there will be a few images in this initial post, however I will try my best to answer any questions you may have about photogrammetery or my methods in subsequent replies.

Anyway, enough of an intro.


The idea:

I want to build a small photogrammetery studio for shooting small to medium sized objects (Rocks), with the goal of a fully automated turntable and camera shutter and a really nice rock library.
My idea is based around a half sphere of lights, producing enough light to uniformly cover my target.

The photos will be shot on either a black or solid blue background (for plants, leaves and branches for easy masking)

I ultimately want to rebuild the dome to use polarized lights, however that will require a lot more thought.

My gear:

I am using a Sony RX100 IV, A really tidy little camera, great for me as I can carry it everywhere even when not intending to photoscan (Gotta live your hobbies, right?)

I made a simple model of my idea, which was thrown out of the window as soon as I stepped into a hardware store.


I worked out that it would be more efficient to purchase some cheap PVC pipe and strap LED strips to them, connect them at the top with some form of cylinder and mount the camera in the center. A trip to the hardware shop got me all the equipment needed at under 35NZD



I got 5m rolls of 60LEDs/m for roughly 22USD, Powering them with a old computer PSU I had lying around



Around this point I realized that my LEDs were cool white and not from the set warm white ones I got, However this turned out OK as I need more light anyways



First test of the dome netted alright results, light coming from the sides could of been better, however all that was required was some cheap reflector material made out of a bedroll with a silvery side, diffuses the light and gives a great amount of bounce!



Test with a friends leather bag, still need to shift the area of the turntable closer to the center of the dome.


Latest image as of 26th October, Getting a better idea of where the light hits are, then comes the diffuser and automated turntable!



Thanks for reading, hope you stick around for when I start producing 3d results!