Connect your Account to access your data

Language Switcher

Workshop Report WHI 28-11-2022 On site

| Julien Vandanjon-Rancoule | War Heritage Institute Digitization Benchmark
Report File
  • Location type:
    on site
  • Location:
    BRUSSELS (Exhibition Hall)
  • Date / Time:

    2022-11-28 / 9h to 14h (5h)

  • Crew:
    Robby HOUBEN (WHI), SolèneLELIEVRE (Objet Témoin), Julien VANDANJON-RANCOULE (Objet Témoin)
  • Item(s) Shooted (Reference Code(s)):
    Mark IV LODESTAR III (Vehicule#1110216)
  • Section (if relevant):
    Vehicule, WW1 section

Shooting Summary

Preproduction actions performed (1h) : Making OFF Pictures link here - Early tests Uploads link here

  • Setup Unplugging of internal exhibition lighting, replacement with locally disposed dimmable LED Panels. Sponsons plexiglass, already removed, no modification
  • Pollution and Safety measures : Crew protective equipment shoes, gloves, suit

Production

For this last workshop based on techniques for 3D reconstruction we gathered 3 groups of data :

  1. Missing data for previous photogrammetry experiments (Reality Capture for huge photography sets)
    • Left and right sponsons underlayers missing photographs shots were done for more accuracy in reconstruction - Ongoing Processing
    • idem with rooftop hatch door (closed)
  2. Comparative data from Apple LiDar scanning but with 3 different 3rd party software applications : Scanning with IPad Pro M1 LiDar of the same 3 vehicle features for comparable quality results : Right Sponson, Rooftop Hatch, Rear Exhaust output :
    1. test with SCANNIVERSE : good UX scanning visual assitance, slower post-processing, best post-processing integrity results - ACCEPTABLE for quick conservation notes, SUITABLE for quick Museography and communication assets production
    2. test with POLYCAM : best UX scanning visual assistance, faster post-processing but lower integrity results - ACCEPTABLE for YOUNG AUDIENCES public demos
    3. test with XXX : scans faster, mediocre results - NOT ACCEPTABLE
  3. High resolution 3D workflow scanning with handheld retro-engineering precision laser scanner.
    • Scanning 0.2mm resolution of left sponson internal graffiti wall - FAIL
    • Scanning 0.5mm with markers - BAD QUALITY
    • Scanning 2mm resolution of left sponson 6pounder canon turret internal features - SUCCESS

Total raw pics produced: 222

Successful Panoramic Views Rebuilt: 2

360° Equirectangular shots/rebuilds: 2

Total pics delivered: Work in Progress

Nb of Cameras: 3

Camera Types: Handheld 3D Scanner Einstar Shining3D, iPad Pro M1, Canon 5Ds

Issues

Corrections required on previous models to compensate : sponsons down faces were missing data hence were messy for reconstruction.

3D Scanner is very low texture sensitive on homogeneous surface with satin paint or repetitive patterns such as metal bolted plates of the tank sponson internal surface. Many tracking losses occured even with the help of marking stickers all around, causing the digitization to become extremely slow and painful.

Also, the pilot computer station somehow was too short in memory and data processing power, causing the firsts scans and recording time to be evenly painful until we downgraded the 3D scanner resolution. This is by far the major issue to adress for futureproofing of the workflows.

Fast Conclusions Debrief

The scanning benchmark is a success, even considering the difficulties encountered. First observation is that even if the hardware may have become way more accessible than in the recent past, the required processing hardware attached still cost high value and need training to be properly used. For reference, when writing this report : Einstar Shining 3D opening price is 1K€ instead of usual =10k€ for this type of devices, but need at least a +3k€ computer laptop / mobile desktop workstation to operate properly with acceptable processing and saving delays.

Compared to Apple iPhone/iPad Pro handheld standalone devices, the accessibility's gap stays a huge step in productivity :

  • APPLE DEVICE : 1-2K€ and 10-20mn scanning feasible by almost anybody (alone) with less than 10mn training : centimeter to decimeter accuracy
  • Pro Laser 3D SCANNING : +4k€ (1k€ + 3k€ minimum) and 1-4h scanning feasible by a team of 1 or 2 depending the object size + mirrored multiple screensets (for control station + operator) + at least 2h postprocessing : millimeter to centimeter accuracy

The Mark IV vehicle has been a perfect sample for meeting with the challenges of this type of digitization, as the satin yellow-ish semi-reflective paint, large surfaces of the same color with no discernable patterns for the machine to compose with as the guns portholes and bolts are basically undiscernable from one to another by the machine. Therefore the operators are to be extremely careful not creating "ghost features" baked into the model which are extremely slow and painful to remove in post-production and may affect the overall quality of the scan.

At the end, to be usable in daily museography or online transmission digital creations, models have to be drastically downscaled with high complexity decimation since the original file may store trillions of points, where usual web-compliant 3D models should only count millions. This processing also adds time-to-readiness and reduces user-friendly workflow, added with necessity of texture recompilation at each steps.

The good point here is from a scientific perspective the experiment still confirms hardware and workflows are globally far more accessible than just a few years earlier though... And starting to be worthit for a cultural heritage institution to be owned and used on a daily schedule.

Workshop production samples

Here are some samples of the actual raw productions of the Workshop :

Objet Témoin ASBL website is proudly curated by J. Vandanjon Consulting - Our work is documented under creative commons license, but some items mentioned and illustrated may be regulated by specific conditions of private collectors, NGOs, foundations and public museums. Please advise. thx.