Research And System Design Of Finger Dual-modal Fusion Recognitio

In today’s information society,biometric recognition has been integrated into various fields.At the same time,information security has attracted more and more people’s attention.Among all biological features,human’s finger features are the easiest one to collect,and fingerprint and finger vein are the most widely used.Nowadays,fingerprint recognition has a mature application system,but there are problems of wear and theft because it is located on the surface of the human body.And finger vein is a internal characteristic of the body,which can be collected while the sample is alive.Therefore,this paper takes into account the convenience of collection and the security of recognition,and constructs a bimodal fusion recognition system based on fingerprints and finger veins,focusing on solving the problem of single-mode feature extraction and bimodal feature fusion.The main work of this paper is as follows:(1)In response to the characteristics that fingerprint images are usually dense and easily rotated,this paper combined Gabor filters with morphological operations to preprocess the fingerprint image,improving the quality of the fingerprint image through background segmentation and burr removal;in view of the characteristics that the background area of the finger vein is prone to noise and has exposure problems,this paper proposed a method of extracting the region of interest of the finger vein image based on rotation offset correction to effectively extract the region of interest of the image.On this basis,bilateral filter was introduced to enhance the finger vein image,and fast fourier transform was used to accelerate the bilateral filter to effectively protect the key information of vein texture,which highlighting edges and small information,these works lay the foundation for subsequent recognition work.(2)Aiming at the characteristics of finger veins which were sparse but prone to joint and edge overexposure,this paper used the Binary Pattern of Phase Congruency which can extract vein features from both spatial and frequency domains and the Pyramid of Histograms of Orientation Gradients which can extract local and global features of the image to extract features from finger vein images.These two texture operators can extract multi-dimensional features of veins and weaken the impact of noise,thus improve the recognition effect of digital veins;on this basis,in view of the limitations of single-mode biometric recognition,this paper proposed fusion operator that fused the Binary Pattern of Phase Congruency and the Pyramid of Histograms of Orientation Gradients,the proposed operator has a good feature extraction effect and performance for detailed texture features of finger images,and it can significantly improve the recognition system performance.(3)In order to solve the problem of high feature dimensions and computational complexity that may occur in bimodal fusion recognition,and to further improve the performance of fingerprint and finger vein bimodal fusion recognition,this paper proposed a bimodal fusion recognition network based on Dense Net.The network used channel fusion technology and attention mechanism to improve the ability to distinguish samples between classes,and further improved the performance of multi pattern recognition through excitation and weighting operations.In addition,this paper changed the original loss function to cross entropy loss plus circle loss,reducing the problems of over fitting and poor convergence.In addition,this paper has developed an integrated recognition system that includes functions such as collection,recognition,and management,improving the applicability of the finger bimodal recognition algorithm.