The Figure Vein Recognition Algorithm Study Based On Multi-scale Line Detection

The sophisticated Biometric is the main factor of the 21 Century. So far there have been fingerprint identification,fingerprint identification ,iris identification, palmprint identification, vein identification, face identification, voice identification and so on. The vein recognition is a rising way of Biometric in recent years. There are some advantages of vein recognition compare as other Biometric ways.Almostly there are two people who has the same vein feature. Because vein extisted in the internal of human body that the characters wouldn't lose and forget. It wouldn't be from external contamination and minor injuries and had fastly recognition speed. Non-invasive and non-contact infrared imaging technology used to ensure that the user's convenience and cleanliness. Using infrared light didn't need any auxiliary light,locations and weather didn't have any impact.The main study in this paper including:We had studied wavelet transform focus on multi-scale of continuous wavelet transformd which used Gauss-Laplace second-order differential function as wavelet basis function For the finger vein image characteristics of this paper,We had studied image pre-processing method. We propoused a feature extract method based on multi-scale maximum integration of image filtering. Though many of multi-scale Gauss-Laplace second-order differential function to inhance the image and compared to the inhance results under diffferent scale. The biggest response and relative scale according to the same point in different image were recorded and according to the biggest response to reconstruct image.We studied the method of extracting ridge direction using Hessian matrix and local non-maxima suppression of Canny algorithm. Combining the two methods,we could get the centerline of vein by extrat the directional local maxima.According to the tracking algorithm in this paper the centerline of vein was tarcked and connect to extrct the whole centerline. At the meantime we could get the directon and relative biggest response scale of centerline.We studied the feature extract-matching methods and propused a new method of using the direction and relative biggest response scale of centerline as features. The preliminary tests demonstrate the algorithm in this paper had more reliability and accuracy than traditional algorithm.