Digital Archaeology and Database Initiatives in 2015

3D digital model of Smallpox Bay ruin and excavations,
 based on 373 photos and with a dense cloud of 21.6 million points.

Stemming from Principal Investigator Michael Jarvis’s recent appointment as Director of Digital Media Studies at the University of Rochester and his increased access to new technologies, software, and devices, this season’s work included training students in a variety of new cutting-edge digital archaeology initiatives. Based on the 2014 season’s success in developing an interactive 3D model of Oven Site using Agisoft’s Photoscan software, the P.I. and Digital Archaeology specialist trainees Miriam Beard and Alice Wynd conducted photogrammetry surveys of all five sites, with Oven Site and Smallpox Bay being surveyed at the beginning, middle, and end of the season.

  
Although highest resolution processing awaits a return to the University of Rochester and access to the BlueGene supercomputer, we were able to generate excellent medium-resolution point cloud and textured models during the field school. Subsections of sites (the fallen wall scatter of Smallpox Bay; the interior of Oven Site’s oven, the Oven Site cistern) were also modelled separately.

Sparse Point Cloud preliminary model of Oven Site, including the Cistern in the upper left corner.
 Model derived from 722 photographs. View faces east.

 The 3D models derived from the photogrammetry series for all sites will be archived at the University of Rochester, provided to the Bermuda National Trust and Bermuda Government (if requested) and made available to interested scholars and the general public through Michael Jarvis’s professional webpages.

During the final days of the 2015 field season, we also attempted a fine-resolution 3D surface capture of Oven Site using a Kinect for Windows sensor, laptop running Skanect software, and a boat battery and AC power inverter to make it work in the field. Although a trial scan of the interior of the Oven Site’s ovens was successful, attempts to scan the rest of the site (open to natural daylight) failed, due to the reflective nature of white Bermudian limestone. 

The 3D Kinect laser scan trial proved a failure, but there are several alternatives that may yield better results. The Kinect may have greater success with another program, such as Artec Studio 9.2 or FARO SCENE or SCENE LT. Also, a higher resolution 3D scanner such as the FARO Focus X130 or X330 promises to be able to capture white surfaces in direct sunlight.

Finally, after reading a review in Make magazine, the P.I. obtained a Seek Thermal infrared camera compatible with IPads and smartphones capable of detecting slight temperature changes within rooms and environments. If suitable, the Seek Thermal promised to potentially identify buried features visible through heat signature differentials a little after dawn and dusk, since this would constitute the same dynamics upon which infrared photography remote sensing site detection is based. A partial survey was conducted at 6:00 am (just before sunrise) to 6:30 am on Saturday, June 27 in the wooded Cottonhole Bight valley area, with no discernible results (other than detecting the movements of wild brown rats) for site identification. Additional trials may yield better results, however, at cooler times of the year when night-to-day temperature swings are more pronounced.

Finally, Lab Supervisor and Data Manager Leigh Koszarsky oversaw the entry of nearly all artifacts recovered to date into ArtiBase, a Microsoft Access database of her design, based on established SGARP context sheet protocol and artifact inventory standards established by Dr. Brent Fortenberry. Veteran students Alice Wynd and Miriam Beard did much of the data input and also trained first-year students in proper entry techniques. Although the artifacts in 200 contexts had been entered in an Excel datasheet in 2014, a quality control audit found numerous inaccuracies, prompting the wholesale re-entry of these contexts afresh. The data architecture Koszarsky designed maintains separate forms and tables for each Smiths Island site and cross-references context site information (layer/feature fill/cut, unit size, Munsell color, soil type, etc.) with artifact inventory information. As a result in the future, site supervisors can have immediate access to profiles of artifacts recovered in all previously excavated areas as they proceed, allowing better real-time interpretation of new patterns of material in new contexts as they are recovered. With the exception of totals for contexts excavated between June 26 and July 4, ArtiBase is complete for all contexts excavated in every site since 2010.

Site	Excavated	Artifact Count
Oven Site	2010-15	25,137 (incl. 12,383 bone, 3992 metal & 3434 charcoal)
Smallpox Bay	2013-15	3,772
Cottonhole Bight	2012, 2014	1,732
Cave Site	2014-15	843
Limekiln	2015	134

Artifact Totals, by Site