Research On Contactless Palmprint Image Enhancement And Image Recognition Algorithm

Palmprint recognition has become a new and effective technology in the field of biometrics.The existing palmprint recognition is mainly contact-based palmprint recognition technology,while the hygienic problems of the public environment in the contact palmprint collection process are easy to cause the users’ exclusion,which is not conducive to the development of the palmprint recognition system.The contactless palmprint recognition method improves the acceptability of the user,extends the application of the palmprint recognition technology,and gradually becomes the mainstream identification method.However,in the process of contactless palmprint recognition,the captured palmprint image is not only susceptible to the noise pollution and light,but also has a lot of deformation,such as translation,transformation and rotation,which largely affect the palmprint recognition efficiency.In view of these problems,the main contents of this paper are as follows:(1)Aiming at the problem that palmprint image is susceptible to noise pollution and illumination.The paper proposes a palmprint image preprocessing algorithm based on low-pass filtering and fuzzy domain.The algorithm combines the typical low-pass filter and the improved fuzzy image enhancement algorithm.Firstly,the low-pass filter is used to process the region of interest of the palmprint,and then the improved fuzzy image enhancement algorithm is used to transform the image into the fuzzy domain,eliminating the palmprint image in the light and noise,and enhancing the palmprint image contrast.Compared with other palmprint image enhancement algorithms,the proposed algorithm has the lowest AMBE value and the maximum PSNR value under the evaluation criteria of Absolute Mean Brightness Error(AMBE)and Signal to Noise Ratio(PSNR),and achieved good results.(2)The contactless palmprint image deformation problem is the key issue of palmprint recognition rate.Aiming at the deformation problem existing in palmprint image,this paper mainly designs the algorithm of contactless palmprint recognition based on fine-tuning convolutional neural network.Compared with the traditional palmprint feature extraction algorithm,the feature learning ability of CNN and the translation,rotation or other deformation of the image during the feature recognition process can maintain a high degree of invariance.In this paper,the method of fine-tuning convolutional neural network AlexNet model is used to train and classify palmprint images by adjusting parameters.The simulation experiment is mainly carried out in two kinds of public palmprint database,and the recognition rate of the fine-tuning algorithm is more than 98%,compared with contactless palmprint recognition algorithm without fine-tuning convolutional neural network and other contactless palmprint recognition algorithms.At the same time,the contrast between the enhancement image and the original image is compared,and the validity of the algorithm is further verified.