Research On Feature Fusion And Recognition Method Of Hand Shape And Finger Phalangeal Prints Under Non-contact Imaging

Nowadays,information security has attracted widespread attention,and biometrics is gaining more and more attention due to its safe and reliable characteristics.At present,biometrics technology is in a large-scale development stage,such as face recognition,iris recognition,fingerprint recognition,etc.At the same time,the researchers found that multi-biometrics can improve the accuracy and robustness of biometrics.In the field of multi-biological identification,the identification technology based on the multi-feature of the hand has been playing an important role.Compared with fingerprints,hand shapes and finger phalangeal prints are more stable,not easy to forge and have a high degree of user acceptance.The details and singularities of fingerprints are easily disturbed.This paper presents a dual-modal recognition method for extracting hand shape features and finger phalangeal prints features.The new crown epidemic has increased the demand for non-contact identity authentication.Traditional fingerprint and palmprint recognition usually require contact with the hand to collect images.This topic uses a smart phone to take non-contact hand images in different backgrounds,requiring four fingers to be close together and large.The thumb is open,and the distance between the smartphone and the subject is about 20 cm when shooting.This research first uses Deep Lab v3 to segment the palm image,remove background interference,and provide preconditions for subsequent hand shape recognition and finger phalangeal prints recognition.In the process of hand shape recognition,this article presents a method of using the contour shape of the hand to identify the user's identity.The segmented palm image is binarized,and then the binarized palm image is input into the network for user identification.In the finger phalangeal prints recognition,this paper mainly considers the inner finger phalangeal prints.First,the inner finger phalangeal prints area is marked.Because the inner finger phalangeal prints is small,not clear enough,and low in contrast,the MSRCR algorithm is used to enhance the segmented hand image.Then input the marked pictures into the Res Net network for training to establish a recognition model,and then use the inner finger phalangeal prints for user recognition.Finally,this paper conducts experiments on the self-built hand image library,and performs experiments on hand-shaped features,finger phalangeal prints,and images with common hand-shaped and finger phalangeal prints features.The dual-modal recognition accuracy rate is up to 94%.,The experiment verifies the feasibility of the method proposed in this paper.At the end of the article,the experimental results are given and analyzed,and further research work is discussed.