Research On Identification Authentication Technology Based On Dental Images

It's really a hot research topic that use biometrics to human identification,because it can confirm a person's true identity quickly and expediently.Identification techniques such as face recognition,palmprint recognition,and iris recognition are enforceable on the premise that the corresponding biometrics are intact,so the above-mentioned identification techniques do not work to the corpses made by fire,flood,explosion,or crime scene.The tooth is a biological feature with corrosion resistance,high melting point and high hardness.Like human face,it also has the personality and can assist the forensic doctor to identify the unknown body.Dental images has become more and more common in dental treatment,and computer vision technology has now achieved unprecedented breakthroughs.Its algorithmic concept and digital image processing technology have laid the foundation for the technology of identity authentication through dental images.In this paper,the research on identity authentication technology is carried out with dental images(teeth X-ray and dental model images)as research objects.The steps of identity authentication based on dental images can be roughly divided into location segmentation of interested area,feature extraction and feature matching.The main research contents of this paper are as follows:1)Segmentation the location of interest in the dental X-ray imagesThe grayscale of the position of interest in the dental X-ray images is very similar to the gray-scale of the background,and there is adhesion in the crown of the adjacent teeth.It is difficult for the existing segmentation algorithm to completely separate these positions.In order to solve these problems,this paper proposed a segmentation algorithm based on full threshold for dental X-ray images.Experiments show that the proposed algorithm in this paper can overcome the inadequacies of the dental X-ray images themselves and the segmentation deviation rate is only 0.1489.2)Dental model images' segmentationThere are lots of personality characteristics such as contour area sequences,tooth structure curvature and the like can be extracted from the contour information of the dental model images.The U-net neural network can be applied to image segmentation tasks with only a small number of training samples.This paper improves the U-net neural network model and applies it to the contour segmentation of dental model images.The experimental results show that the improved u-net model can roughly segment the contour of tooth marks.3)Feature extraction and recognition scheme of dental model imagesFeature extraction is not necessarily obtained by contour information.This paper proposed an algorithm for extracting features from complete dental model images.The algorithm will calculates and analyzes the similarity ratio features of the dental model images based on the SIFT.Then transform the recognition task into binary classification task.And use three machine learning classifiers to train and test the sample features,compare and analyze the results separately.Experiments show that the feature extraction scheme proposed in this paper has strong personality and can distinguish different individuals.The recognition scheme of this paper also obtains a good recognition result,the verification accuracy is 91.60%,the test accuracy is 90.83%,and the average recognition accuracy rate when applying the model designed by the paper is 93%,and the best case is 100%.This paper researched identification technology through two kinds of dental images,and has achieved good results in the segmentation of dental X-ray images and the identification based on dental model image features.In the future research,this paper will consider combining image texture and other feature information to further develop the research on the identification technology based on tooth image.