Method Research Of Dual-model Recognition Of Touchless Fingerprint And Finger-vein Based On Hash

Multi modal biometric recognition technology,overcomes many disadvantages of the single modal biometric recognition technology,can also have a variety of biological features were compared with higher recognition performance,safety and convenient performance to meet the actual demand,it has a good application prospect.Non contact fingerprint recognition and finger vein recognition is the current research focus,with some research results accumulated,have high security and good recognition effect.At the same time,fingerprint recognition and finger vein recognition have many advantages over other multimodal biometrics methods.Contactless fingerprint identification has convenient collection,feature rich information advantages,but easy to be stolen;finger vein recognition has the advantages of high security,living in nature,but the feature information is relatively small,the defect of the high cost of acquisition.In this paper,the advantages and disadvantages of two kinds of biological features of non contact fingerprint and finger vein are fully considered.According to the characteristics of non contact fingerprint image and finger vein image,the key technology of image preprocessing,feature extraction and matching,and dual mode biometric fusion method is studied.Firstly,according to the characteristics of non-contact fingerprint and finger vein image,the corresponding preprocessing method is adopted.In the non contact fingerprint,extracted by finger region YCb Cr space and Otsu based on the method,then by high-frequency emphasis filtering and iterative adaptive histogram equalization for fingerprint image enhancement,the correcting method of finger caused the midline of the image acquisition method based on finger rotation and migration,finally extract the region of interest.In the finger vein,the extraction method of super pixels and finger regions based on Sobel operator,and then using the method of adaptive histogram equalization and median filtering combination of image enhancement,using rotation correction method to correct the finger vein image,and then refers to the image region of interest extraction vein.Through the experimental comparison,the non-contact fingerprint and finger vein image preprocessing methods have good robustness and improve the recognition rate.Secondly,the methods of feature extraction and matching for non-contact fingerprint and finger vein image are analyzed.In the aspect of feature extraction of contactless fingerprint image,this paper adopts the block discrete cosine transformalgorithm is used on the pre processed image Perceptual Hashing feature extraction,and then the obtained hash sequence,feature matching based on Hamming distance.Discussed the block number,size,size and other factors on the influence of the low frequency part of the performance,the performance comparison of various algorithms from the aspects of accuracy,computing speed,storage space,and analysis in image quality is good and bad in two cases the algorithm recognition effect.In the finger vein image feature extraction,using discrete cosine transform algorithm for Perceptual Hashing characteristics on the preprocessed images extracted,then the obtained hash sequence,feature matching based on Hamming distance.The influence of related factors on the recognition effect is analyzed,and the recognition effect of the algorithm is analyzed in two cases of good and bad image quality.Compared with the traditional algorithm,the algorithm proposed in this chapter greatly reduces the computation time and occupies the size of storage space,but the recognition effect is not satisfactory.In addition,in the case of poor image quality,the recognition effect is more obvious decline.Again,the traditional algorithm has good recognition effect,inspiration and hash algorithm has advantages in processing speed,occupy space,this chapter combines the advantages of the two methods,preprocessing of the image before feature extraction operator,then the response map perception hash.In the non contact fingerprint,using AR-LBP operator for processing prominent features in the image preprocessing,and then the AR-LBP response map using block discrete cosine transform algorithm is used on the pre processed image Perceptual Hashing feature extraction,and then the obtained hash sequence matching method based Yu Hanming distance matching characteristics,and analysis the image quality is good or not under the two conditions of the recognition effect.In that vein,the GLGC operator for processing prominent features in the image preprocessing,then the response map using discrete cosine transform to extract the hash sequence,the multiple hash sequence series into a hash sequence,then the matching method based on Hamming distance matching characteristics,and analysis in image quality good and bad in two cases the algorithm recognition effect.The proposed algorithm,the identification effect is directly using discrete cosine transform method to improve processing speed,while retaining fast,small footprint advantages,but the image quality is not good,the recognition effect will still decline to a certain extent.Finally,traditional unimodal biometric systems are vulnerable to light andpersonal characteristics or other factors,resulting in single modal biometric systems are restricted in practice use,and as a result of recognition using single modal biometric identification in theory could reach the rate there is a limit,even by improvement of image acquisition,preprocessing,feature extraction and matching algorithms cannot fundamentally solve these problems.However,combining multiple biological features can provide more identity information about individuals,making up for the defects that single modal biometrics can not fully describe personal identity.This chapter presents a data fusion method based on the perception of Hashi's response,more effective use of contactless fingerprint and finger vein feature,and solved before coupling method based on distance metric learning requires a lot of sample questions.Considering the influence of poor image quality on recognition effect,a feature level fusion method based on image quality evaluation is proposed.The experimental results show that this method achieves satisfactory recognition effect and provides an effective solution for multimodal identification.