An Automated Two‐Dimensional Form Registration Method for Osteological Pair‐Matching |
| |
| Authors: | Jeffrey James Lynch M.Sc. |
| |
| Affiliation: | Defense POW/MIA Accounting Agency, Omaha, NE |
| |
| 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. |
| |
| Keywords: | forensic science osteometric sorting pair‐matching Hausdorff distance shape analysis two‐dimensional form registration |
|
|
