Research On Mobile Palmprint Authentication Algorithm Based On Double-line-single-point

Nowadays more and more activities,such as mobile payment,are performed on mobile Internet,which has put forward higher requirements for mobile access security.Different from traditional knowledge passwords(such as numbers,graphic and grids),biometric features have higher uniqueness,security,stability and other favorable characteristics.Therefore,biometric recognition is desired to be widely used in the field of mobile security.Biometric recognition identifies/verifies individuals based on their physiological or behavioral characteristics.Physiological characteristics include fingerprint,facial feature,iris,palmprint,etc.,while behavioral characteristics include gait,voice,signature,etc.In recent years,more and more attention has been paid to palm-based biometric recognition due to its multiple advatanges,including high recognition accuracy,good availability,high acceptability,etc.Palmprint can be captured with the built-in camera of the mobile device,so it does not require additional sensors,and consequently can be used as a secure and practical mobile biometric feature.However,palmprint acquisition in mobile environments suffers from several severe challenges,including complex and changeable background,remarkable variant illumination,uncontrollable hand postures and positions,limited hardware resources(limited storage capacity and computation capability)and so on,which upgrade the difficulty of palm localization in image preprocessing stage.The traditional methods based on simple skin-color model are difficult to be directly transplanted.The state-of-the-art deep learning algorithms are restricted by the limited computation resources of mobile devices,so they are difficult to achieve practical performance.In this dissertation,an assistance localization technology and a calibration algorithm are proposed based on the morphological features of palm.The palmprint authentication system is implemented on the mobile terminal to verify the accuracy and practicability of the proposed schemes.The main contributions and works are summarized as follows:(1)Background,research significance,and relevant knowledgeThe background and research significance are introduced for the analysis on the feasibility and difficulties of palmprint recognition on the mobile platform.Theresearch status of the pre-processing and feature extraction of palmprint recognition is elaborated,and the unsolved critical problems are further pointed out.In addition,secure palmprint recognition and related evaluation indices are briefly introduced.(2)Assistance localization for palmprint acquisition on mobile terminalAn assistance localization method dubbed "double-line-single-point"(DLSP)is developed to effective limit the postures and positions of palm.The graphic view is displayed on the image preview interface,which contains two parallel assistance line segments with certain spacing,and the upper endpoint of the short line segment is marked as the assistance point.In the preprocessing stage,the palm is directly localized and the region of interest is cropped,which reduces the localization error and computation complexity.Furthermore,in order to prevent inappropriate registration due to improper operation,the palm existence criteria are developed to judge whether a user places his/her palm correctly according to the separation distances and lengths of the fingers.(3)Calibration algorithm robust to palm translationThe proposed calibration algorithm includes two parts: double boundary line calibration and inter-finger valley point calibration.The double boundary line calibration localizes the position of the palm boundary in the rectangular region cropped by the assistance graphics,which reduces the translation error perpendicular to the orientation of the assistance line.In order to reduce the translation error along the orientation of the assistance line,the accurate abscissa of the inter-finger valley point is localized in the rectangular region cropped with the assistance point as the reference.The sophisticated calibration algorithm remedies the deficiency of DLSP assistance localization,and improves the accuracy of palmprint recognition.(4)Design and development of palmprint recognition system on Android platformThe designed and developed palmprint recognition system can be executed in two modes,namely registration and verification.The feasibility and practicability of the proposed algorithms,including DLSP assistance localization method,double boundary line calibration and inter-finger valley point calibration,are tested and verified on the implemented system.Furthermore,a palmprint database is collected and established with DLSP.To sum up,through the analysis on the palmprint preprocessing algorithm in mobile environments,DLSP is proposed as a novel assistance localization forpalmprint preprocessing.The well-designed assistance localization scheme and the corresponding calibration algorithms with few constraint conditions remarkably improve the recognition accuracy and the robustness,and they provide important theoretical significance and engineering application value.