Tooth identification based on teeth structure feature

Abstract

Massive disasters, like earthquake and fire, can easily damage most features of the human body, which bring difficulties in cadaver identification. In this paper, to improve the efficiency and accuracy of forensic autopsy, a novel tooth identification strategy based on teeth structure features from dental impression images is proposed. The strategy achieves identification including tooth mark segmentation, teeth structure feature extraction, and matching. To resolve the problem of too much noise, we propose a locating segmentation method based on object detection to segment every single tooth mark contour. Finally, to evaluate the proposed strategy more objectively, we propose an indicator named rate of filtration based on the indicator accuracy. The experimental results show that using only upper-jaw teeth, our method achieves a rate of filtration of 90.45% for 47 persons comprised of 311 images, and an accuracy of 86.17% for top-5 (top 12%) image matching, which are highly encouraging. To the best of our knowledge, the proposed method is the first to use dental impression images to achieve tooth identification.