Research Of Biometric Identification Method Based On 3D Palatal Rugae

Forensic identification is a type of biometric identification technology that compares and identifies the similarities and differences between two or more features of the known and unknown objects,and infers whether the studied object originates from a given object or not.The biometric indicators commonly used in this technology mainly include face,fingerprint,palmprint,iris,teeth,craniofacial morphology and DNA,etc.However,in extreme environments,indicators such as face,fingerprint,palmprint and iris may be restricted by environmental factors such as fire,chemical corrosion or external trauma,resulting in information loss.And environmental and economic factors often limit the determination of DNA from mass victims.Therefore,it has become one of the urgent problems in forensic identification to study a kind of human biometric index which is not easy to be destroyed and forged and has a low cost.Palatal rugae,as a new biological feature,has the characteristics of resistance to corruption and high temperature.It has become the gold index of identification in forensic identification,and also provides a new means of identification for forensic identification.At present,the research of palatal rugae recognition mainly adopts manual method to calibrate the morphology,quantity,length and direction of palatal rugae,and carries on the statistical classification research.Moreover,there is a lack of research on 3D palatal rugae recognition using digital image processing.In this paper,the key technologies of digital 3D palatal rugae recognition system are studied from the aspects of data set creation,feature extraction,feature reduction and classification..The main research work is as follows:(1)Considering that the previous study of palatal rugae recognition relied too much on the manual participation of forensic experts and the degree of intelligence was low,this paper proposes a 3D palatal rugae automatic recognition method based on cyclic spectrum analysis..Specifically,according to the characteristics of palatal rugae data,the cyclic spectrum is introduced as the recognition feature,and the K-nearest neighbor classifier is used to realize palatal rugae recognition.In addition,in order to solve the problem of information redundancy caused by the large dimension of palatal rugae data,two strategies are used in this paper to simplify the complexity of 3D palatal rugae features.On the one hand,isometric section is proposed to reduce the dimension of 3D palatal rugae data.On the other hand,the block scheme of cyclic spectrum features is used for feature dimension reduction.The method proposed in this paper is the first to realize palatal rugae recognition using information technology completely,which is helpful to promote the digitization and intelligence of the forensic identification system.The experimental results show that the recognition accuracy of the proposed method is more than 95%.(2)Considering that palatal rugae data has rich features both in time domain and frequency domain,this paper proposes a 3D palatal rugae recognition method based on Fractional Fourier transform,which extracts palatal rugae feature information from time domain and frequency domain,and uses SVM classifier to realize palatal rugae recognition.In addition,in order to prevent the feature dimension explosion and reduce the influence of irrelevant information on palatal rugae recognition,this paper used RFE to carry out feature selection on palatal rugae Fractional Fourier transform features,and selected the optimal combination of feature vectors.The experimental results show that the recognition accuracy of the 3D palatal rugae recognition system can reach 97.8%.(3)A 3D palatal rugae recognition method based on deep learning is proposed.Compared with the manually designed features,deep learning can quickly learn more effective feature representations from training data.Deep learning model extracts feature information layer by layer from pixel level original data to abstract semantic level,which makes it have outstanding advantages in extracting global feature and context information of image and improves the accuracy of classification.In this paper,the pre-trained VGG19 and Res Net18 networks in deep learning were used to extract and classify palatal rugae features,and a 3D palatal rugae recognition system was constructed.Experimental results show that the recognition accuracy of the constructed palatal rugae recognition system can reach 97%.