Research On Finger Vein Multi-feature Extraction And Matching Algorithm Based On Maximum Curvature

With the development of science and technology,people have higher and higher requirements for the security of identity authentication.Finger vein biometrics,as a type of biometrics,has gradually become a research hotspot in the field of biometrics due to its unique advantages.However,there are certain bottlenecks in improving the performance of the finger vein recognition system.For example,the finger vein extraction algorithm is not robust when dealing with background noise and low contrast,which makes it difficult to improve the finger vein recognition rate.Most finger vein systems use a single feature for matching.There is also a problem of insufficient utilization of information contained in the finger vein image.Therefore,this thesis studies the deficiency of the algorithm for feature extraction and matching of finger vein biometrics and makes the following work:(1)Through the study of the finger vein database images,it was found that some images in the database have "low quality" problems such as uneven exposure,blurred images,and low discrimination between the finger veins and the background.To solve this problem,this thesis proposes an adaptive finger vein pattern extraction algorithm based on maximum curvature.The algorithm adopts operations such as adaptive weakening of finger vein background and removal of pseudo-vein points,which improves the robustness of the extraction algorithm.Simulation experiments show that this method improves the accuracy of finger vein pattern extraction and improves the performance of finger vein recognition system.(2)Aiming at the low performance of the finger vein recognition system using a single feature,this thesis proposes a weighted mean multi-feature fusion matching algorithm based on particle swarm calculation weights at the matching layer.This method combines three types of features and a matching matching method.For finger vein grayscale images,LBP(Local binary pattern)is used to extract feature vectors for matching.A template matching method is adopted for the features of finger veins.For feature points,a feature matching method of finger vein direction information based on IHD(Improved Hausdorff Distance)is designed for matching.Experimental results show that the multi-feature fusion matching algorithm proposed in this thesis can improve the accuracy of the finger vein recognition system.