Spatial Recording via Structure-from-Motion (SfM)

The primary method of spatial recording of excavated contexts at Kaymakçı is through a type of image data capture and “structure-from-motion” (SfM) processing—colloquially referred to as (but technically different from) “photogrammetry”—in which a sequence of overlapping photographs are used to generate 3D models. Each spatial context is represented by a model that captures the bottom surface of the context following excavation. This “bottom” model can then be combined with one or more “top” models to create a continuous envelope, producing a “context volume,” that is, a 3D digital surrogate of the volume of the context. These models are georeferenced from the moment of their generation and therefore—once combined into “context volumes”—record the real locations of contexts in 3D, volumetric space.

A collection of data used to create each model is referred to as a “photobatch” and consists of a set of overlapping photographs and coordinate data that are used to construct subsequent digital outputs. Excavation area supervisors are responsible for collecting the photos and coordinates (a process casually referred to as “photobatching”), which are then passed along for processing to the 3D Spatial team. Every spatial context is thus recorded with one photobatch, the name of which is based on the date and time it was collected. If multiple sets of photos and coordinates are collected to record a single context—for example, when the top and then bottom model of an in situ vessel, find, or small feature can be captured quickly during the course of excavation and removal of that vessel, find, or feature—the data associated with each model (photos and coordinates) should be stored in separate “top” and “bottom” folders within a parent folder that bears the photobatch number.

Each season begins with the production of initial 3D models after light cleaning—either the removal of surface vegetation for new excavation areas or the removal of sediment and vegetation that has accumulated since previous excavations that are being reopened. Subsequently, routine recording will include capturing the “bottom” model of each excavated context; the “top” model of each context is composed by the “bottom” of the context(s) above it. The only exception to this rule is the recording of spatial contexts that represent special, discrete objects or features recovered during the excavation of a larger context. As described above, for example, if a complete vessel is exposed within a fill deposit, take the following steps:

  1. remove as much of the fill as possible from around the vessel without disturbing it;
  2. record the “top” photobatch of the vessel;
  3. remove the vessel, disturbing as little of the fill around it as possible; and
  4. record the “bottom” photobatch of the vessel.

As much as possible, the volume between these two photobatches should represent ONLY the object that was removed itself. If additional earth is removed with the vessel in step 3, it will be recorded in the second (“bottom”) photobatch and will thus be erroneously included in the volumetric model of the vessel.

Powered by BetterDocs