Gradient Boosted Non Rigid Registration for Finger Vein Authentication

Abstract

In the present era of development in electronic world it is reasonably difficult to secure the personal information. Considering the limitation of the uni-modal biometrics authentication the appropriate selection of modality plays important role for authentication. With the advancement and growth of modern society, recognizing identity of humans and protecting their information security is a social key problem, which needs to be addressed in the current era of information technology. Traditionally there are two methods of identifying humans - based on known contents (passwords, secrets codes, etc.) and processing (certificates, credentials, smart cards, keys, etc.). Many abusing methods were practiced in the recent past due to the development of infrastructural equipment, fraud, misuse, cracks and hackings. Individual s passwords are often cracked, misused by others and forgotten which adds further to the complication of traditional methods. Thus, there is an emerging need for better recognition system which is fool proof, secure and easy to operation and this urge instigated development of modern technology for personal recognition. The presence of oxygenated and deoxygenated hemoglobin enables the visibility of finger veins and its subcutaneous structures that develops inside a finger randomly. The visibility is enabled due to reflection of light and its absorption of near infrared lights and helps as strong force to avoid theft and forgery. newlineThis thesis work proposes a topical authentication system using finger vein. The vein pattern present underneath the skin of finger is idiosyncratic and stable. Finger Vein structure, transformation of image, smattering and orientation will naturally affects the vein images and also influenced by its illumination, posture and background. The performance of identification process may degrade due to these factors and implies the need of defining Region of Interest (ROI). It is proposed in this thesis, extraction of features of finger vein based on deep learning technique, RROI localization which proves to be much effective and forcefulness. The paper emphasized the finger-vein framework, its recognition performance parameter i.e. false acceptance rate (FAR) and false rejection rate (FRR). The characteristics of finger vein authentication shows that it is more secure than the other correlated modalities. newlineFinger Vein Recognition System (FVRS) ideally suffers from various factors which are external in nature the models used for imaging and uneven illumination. The internal factors which influence are scattering of images and tissues of finger. These external and internal factors creates considerable impact on images of finger vein having characteristics of unstable and squat contrast makes FVRS a challenging process for achieving an precise and dependable performance in factual scenario. newlineThe finger vein based identification system is a recently evolving biometric technology and had been considerably focused in the field of biometric based recognition systems. The miniature size of image capturing devices which is used to capture vein images helps to achieve miniaturizing the finger vein authentication technology. ROI based finger vein authentication system as proposed in this work will have high robustness as compared to old traditional methods. With the consideration of experimental results, it can be proved that ROI based authentication method will be one among the best in terms of recognition accuracy. newline newline newline newline