Abstract:
Dental biometrics is used in the forensic odontology to identify individuals based on their
dental characteristics. Forensic odontology is a branch of forensic medicine that deals
with teeth and marks left by teeth (as in identifying criminal suspects or the remains of a
dead person). In this project, we have designed a system for person identification using
dental data. The system used two types of dental images for person identification i.e.
dental radiographs and colored dental images. The radiograph dataset consists of antemortem and post-mortem radiographs of same person. For radiograph images, the
implemented system first extracts a single tooth manually from dental radiograph and
then performs binarization using adaptive thresholding. The horizontal and vertical
profiles are extracted along with the orientation of tooth as features to construct a feature
vector. The post-martem images are used for matching purpose which is based on simple
distance based matching. For colored dental images, HSV and LAB color spaces are used
and teeth are extracting by using k-means clustering. The system applies morphological
operations as post processing steps to remove undesirable regions. The feature vector
consists of mean, min and max saturation values, Euler number and area. Matching is
again performed using distance between training and test feature vectors. The validity of
implemented system is tested using accuracy (identification rate) which is the ratio of
correctly matched images to total number of images in dataset.
Keywords: dental biometrics, biometrics, forensic odontology, dental radiograph, k
means clustering, distance based matching
