Research On Finger-knuckle-print Recognition Based On Fusion Of Basis And Depth Convolution Network And Local Texture Mapping

In the past 20 years,other technologies related to biometric identification have been widely used in various industries.Among many biological features,some biological features of human hands,such as fingerprints,finger veins,palm shapes,palm prints,palm veins,etc.,have attracted extensive attention and research from scholars at home and abroad due to their strong stability,difficult imitation and easy availability.Finger-Knuckle-Print(FKP)refers to the veins of the second and third segments in the middle of human fingers,also known as "finger knuckles" and is one of the biological characteristics of people's hands,which can be applied in biological feature recognition technology.Compared with other biological features,such as fingerprint,palm print,face,iris and other biological feature recognition technologies,knuckles feature recognition started late,the research level is not deep,and there are still some limitations in practical production.Firstly,fingerprint image recognition is realized by using traditional recognition technology,which has low recognition efficiency and redundant feature data.Secondly,it needs to be recognized quickly in actual production and has high accuracy for fingerprint image recognition.In order to solve the above problems,this paper studies the method of improving LBP algorithm by combining convolution neural network to realize fingerprint image recognition.The specific research is as follows:(1)Normalization,image enhancement,binarization and other processing are carried out on the image through traditional methods to reduce the influence of noise,improve the robustness of the image and improve the recognition rate of the image.(2)Recognition of knuckles is realized.In this paper,convolution neural network is used to replace the traditional recognition method for knuckles image recognition,which can effectively solve the problem of feature redundancy.In this paper,Resnet network is used to simulate knuckles image,and good recognition effect is achieved.(3)It is proposed that due to the influence of uneven illumination and image noise on the features learned by the neural network,the depth convolution network fuses local texture descriptors to identify knuckles.LBP features after fingerprint image preprocessing are used as network input to reduce the influence of noise.The input data are the improved LBP feature experimental results.Compared with the experimental results directly input to the original image,the recognition rate ofknuckles is effectively improved.(4)The defect of fingerprint feature extraction by traditional algorithm is that LBP algorithm with neighborhood correlation is proposed.The improved algorithm is to extract texture features by using the correlation between adjacent pixels of image pixels.The improved LBP algorithm is used to process the fingerprint image and input it into the convolution neural network.The experimental results show that compared with the original algorithm,the improved algorithm can improve the recognition rate.