An Automated Two‐Dimensional Form Registration Method for Osteological Pair‐Matching

Abstract

This study introduces an automated method for osteological pair-matching using two-dimensional outline form data extracted from photographs. A procedure for acquiring photographs that improve the differentiation of specimens from the background is presented along with an extraction procedure that allows the capture of form data from photographs. The raw form data are used in a two-dimensional registration procedure, which combines iterative closest point, K-nearest neighbor search, and iterations around an estimated mean. Form data are used in optimized distance calculations that minimize true-pair difference and maximize false-pair difference. The sample consists of 122 calcanei and 110 tali from the UI-Stanford collection. Performance statistics are provided for the maximum and average Segmented-Hausdorff, Hausdorff, and Procrustes distances to show the comparative statistical results for matching. Results indicate 98.36% and 98.2% accuracy in pinpointing true-pairs for the calcanei and tali, respectively, using a shortlist of one-lowest-distance.

An Automated Two‐dimensional Pairwise form Registration Method for Pair‐matching of Fragmented Skeletal Remains

“…Further, the sensitivity (exclusion power) is >99%. This is in contrast to the previous study showing 98% accuracy and >99% exclusion power (10). While the two analyses on complete specimens are virtually identical, it should be noted that the pairwise procedure increases the time of analysis 100-fold due to the number of registrations required and larger number of coordinates available.…”

“…In common with the previous two‐dimensional study by the author , this method does not confirm the lowest distance is the correct match but has the benefit of the true‐pair being the lowest distance >90% of the time. Each lowest distance match should be visually verified by analysts prior to accepting it as a match.…”

“…However, the methodology itself is calculation intensive in that there are a larger number of coordinates in the outlines and it relies on pairwise registration rather than a single form space. Due to these constraints, if an assemblage is nonfragmentary then the other methods published by the author should be used to increase time efficiency and to allow for a greater variety of Hausdorff distances . Further, it should be noted that higher resolution photographs introduce a greater number coordinates, which significantly increases the time required for registration.…”

“…Shape and form analyses have become more popular in physical and forensic anthropology (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15) with a few studies focused on matching algorithms (6,(10)(11)(12)(13)(14)(15). With the adoption of shape analysis, the literature on osteological sorting has started to shift from linear metric approaches (16)(17)(18)(19)(20)(21)(22)(23) to shape and form approaches (10)(11)(12). However, the current literature on shape and form osteological pair-matching has exclusively focused on complete skeletal elements (10)(11)(12).…”

“…With the adoption of shape analysis, the literature on osteological sorting has started to shift from linear metric approaches (16)(17)(18)(19)(20)(21)(22)(23) to shape and form approaches (10)(11)(12). However, the current literature on shape and form osteological pair-matching has exclusively focused on complete skeletal elements (10)(11)(12). While some of these methods appear easily extendable to fragmentary elements, they all have procedural drawbacks, such as nonautomated registration or significant pre-and postprocessing of data, that prevent them from being fully automated.…”