Automated Human Identification through Dental Image Enhancement and Analysis

Abstract

Every individual has a unique identity like dental feature in which variations in tooth morphology, size, shape, missing tooth, and other characteristics are known. Hence, when uncovering a person’s identity, developing a dental profile based on the body at hand is an accurate start. However, dental profiling requires the use of radiography, which is proven to be harmful to the forensic dentist (and a living patient). This study attempted to develop a complete identification system that will aid in dental forensics without the use of radiation. The researchers used the Recurrent Neural Network (RNN) for machine learning wherein a feature vector is to be classified. An imaging device was also developed to capture the teeth of a person with the use of USB Camera, and a database of stitched, enhanced, and analyzed dental images using Image Stitching with OpenCV, Contrast Limited Adaptive Histogram Equalization and Smith Waterman Algorithms. The images were analyzed using Adaptive Harris Corner Detection Algorithm, which in turn identifies a person through the enhanced and analyzed top view images of the teeth in upper and lower jaw. The computed reliability from the matched and mismatched data is 83.33% and the computed accuracy for the ability of the program to compute the correct area is 94.8393%. Future studies may explore other factors as key points aside from tooth area.