Finger Vein Recognition Based On Genetic Algorithm

Compared with biometric technologies such as fingerprint recognition and facial recognition,finger vein recognition technology has the characteristics of internal characteristics,living body recognition and high security level,and has greater technical advantages and development prospects.Existing finger vein recognition methods usually use grayscale images containing vein distribution as the object for algorithm design,but due to the limitations of the collection device,the uncertainty of light intensity,and the complexity of the tissue around the finger blood vessels,etc.Even after a series of image processing operations are performed on the obtained image sample,there will still be irregular shadows and non-vein features in the resulting grayscale image.Therefore,this paper proposes threshold segmentation based on genetic algorithm and algorithm design based on binary images containing texture distribution of finger veins,so as to reduce the influence of interference factors as much as possible in the matching and recognition process.The specific work is as follows:This article first introduces the relevant situation of biometric technology.By comparing the advantages and disadvantages of different biometric technologies,the advantages and development prospects of finger vein recognition technology are discussed.At the same time,the research status of finger vein recognition method and genetic algorithm in the field of image processing at home and abroad was explored.Then,through the edge detection operator and non-maximum suppression method to accurately detect the finger edge,the ROI(Region of Interest)of the finger is extracted from the acquired image.In order to solve the problem of low contrast between the texture of the finger veins and the background,the CLAHE(Contrast Limited Adaptive Histogram Equalization)algorithm is used to enhance the spatial domain of the image.The Gabor filter is then used to enhance the vein texture in the frequency domain,and finally the finger vein image is obtained by fusing multiple angle filtering results.Secondly,the problems that need to be solved when applying genetic algorithm to image threshold segmentation tasks are studied.The genetic algorithm is used to optimize the neighborhood size parameter and the adjustable parameter 6)in the local threshold segmentation method.The 14-bit binary numbers are used to encode and 6),and the value range is transformed into a feasible solution space that can be iteratively optimized.In addition,based on the method of solving the entropy of the target and background in the maximum entropy threshold segmentation method,it is proposed to define the distance between the target and the background as a fitness function to determine the evolution direction of the population in each iteration.After 100 generations of inheritance,the finger vein image is threshold segmented using the obtained optimal neighborhood size parameters and adjustable parameters to obtain a finger vein binary image.Then,we studied and compared the commonly used feature point(corner point)extraction algorithms,and selected the FAST(Features from Accelerated Segment Test)algorithm that is most suitable for the finger vein binary map to extract feature points.On this basis,it is proposed to use the pixel point amplitude and angle information to determine the main direction of the feature point and vector description,so that it has a fixed dimension and direction invariance.At the same time,based on the circular neighborhood of the feature points,by combining the number of feature point pairs successfully matched between the images and the average Euclidean distance,a matching model to evaluate the similarity between the two images is proposed.Based on the matching model and the K nearest neighbor classification Method to judge the type of test image.Finally,the performance of the proposed algorithm was tested using the finger vein database published by Shandong University,and compared with several similar feature extraction and recognition algorithms.The results show that the method proposed in this paper has great advantages in key performance indicators such as recognition rate and EER(Equal Error Rate).At the same time,after the field of view area is reduced by 40%,the performance indicators of the method proposed in this paper basically remain unchanged,which is still significantly better than other algorithms.