Study On Finger Vein 3D Point Cloud Acquisition And Infornmation Extraction

Finger vein biometric technology is contactless for living identification,which has higher security and anti pseudo ability compare to the traditional biometric identification technology,so it is a high demand and practical application value for the places with high safety performance requirements.Traditional finger vein recognition technology obtains 2-dimensional images of finger vein through a monocular camera with a near-infrared acquisition device,then performs feature extraction and recognition.Two-dimensional image extraction has a small number of finger vein and the image quality is easily influenced by external factors such as finger posture and illumination environment,therefore,cause the problems including high misrecognition rate,low identification rate,constraints on user posture and inability to support large-scale data set.Aiming at the above problems,we put forward the method of finger vein recognition based on 3-dimensional point cloud,and reconstruct the 3-dimensional point cloud of the finger vein through the binocular stereoscopic 3-dimensional reconstruction system,then,carried out the recognition experiment using the method of the local feature similarity measurement on the 3-dimensional point cloud.The advantages of the method are as following :1)to use the 3-dimensional point cloud as the feature information of the vein;2)to solve the problem that the user's attitude constraint;3)and inaccurate description of vein characteristics cause by 2-dimensional images which lead to high misrecognition rate and low recognition rate.Our study provides a practical and applicable solution for 3-dimensional finger vein recognition research.The main work accomplished in this paper include:1.We design the hardware platform of finger vein 3-dimensional point cloud acquisition device.Firstly,design the light source and imaging method,use the wavelength of 850 nm near-infrared array LED as the transmission light source,then use a parallel optical axis binocular stereo vision mode to collect images.The three-dimensional model of acquisition device is designed by Solidworks2016,and finally the first generation prototype is maded by Westcom 3-dimensional printer.2.We complete the binocular stereo calibration and correct the parameter.We Make a checkerboard target with near-infrared backlight,calculate the calibration and distortion parameters of monocular camera according to camera imaging model and calibration principle.Then,carry out the calibration of stereo camera and correct the polar line to obtain the internal and external parameters.3.We complete the image preprocessing and reconstruct the finger vein cloud structure.Design four-point method to acquire the finger vein area,filter and enhance the image,then design 9 ?9 direction template operator to extract finger vein patterns.Use the triangulation principle and SIFT algorithm to calculate the vein pattern feature depth information,then complete the 2-dimensional reconstruction of the finger vein.4.Subspace feature extraction of finger vein 3-dimensional point cloud and calculate the similarity to complete the finger vein 3-dimensional point cloud recognition.Firstly,normalize the point cloud,we use two-level segmentation method make the point cloud into several sub-space,then calculate the angle and feature vector of each sub-space.5.We set up a 30-digit finger vein 3-dimensional point cloud database and perform the identify experiments.The original database included 15 males,15 females and 180 fingers consist of the index finger,middle finger and ring finger respectively of the left and right hands.Through the analysis of experimental results,the recognition rate is 96.7%,the false rejection rate is 0%.