Research On Palmprint Preprocessing Algorithms And The Development Of Palmprint Verification System On Android Devices

Android operating system has several advantages,such as resource openness and application diversification,so mobile devices with Android operating system have the maximum market share.With the widespread use of Android system,the identity authentication on this system has gradually become people's focus.Currently,neither numeric password nor graphic password is strongly secure.In addition,passwords with high entropy are difficult to remember.The biometric-based verification technologies enhance the security of mobile device drastically;however,the current biometrics,like face,fingerprint,etc.,for the authentication on mobile devices inevitably suffer some certain problems.Palmprint-based verification has a lot of remarkable strengths.It is less subject to interference factors,has high accuracy and rich discriminant features.Palmprints are not easily replicated in non-perceived situations.The contactless mode can avoid the problems of hygiene and disease transmission.Moreover,palmprint systems commonly have low cost and high acceptance in collection stage.Thus palmprint verification has bright development prospect and potential.Palmprint verification of mobile devices is commonly disrupted by several severe challenges,such as illumination variation,background complexity,uncontrolled palm position and posture.Furthermore Android mobile devices,like smart phone,have less computing capability and storage space than PCs.These factors definitely bring severe difficulty to palmprint verification,in which the accurate palmprint feature extraction depends heavily on palmprint preprocessing.In order to solve the aforementioned problems,the segmentation and localization algorithms of palmprint in mobile environments are studied in this dissertation.A palmprint verification APP is designed and implemented on Android operating system.The main contributions and research contents of this dissertation are as follows:(1)According In the acquisition of top-allocated finger-valley,the region of the interest(ROI)of palmprint is localized based on the bisector of the angle between the boundaries of the two sides of valley-between-finger.First,color and brightness are combined to extract the complete palm edge.In YCbCr color space,the edges of Cb channel and Cr channel are fused with certain weights.Then,the morphological operations are performed to repair the contour of valley-between-finger,and fill each valley as a closed region.A straight line fitting is performed on the midpoints of the paired edge points at the same horizontal line.Each pair of edge points are on the both sides of valley-between-finger,respectively.On the basis of the first straight line fitting,the outliers are removed for the second straight line fitting.The key point is the intersection of the fitted line and the bottom contour of valley-between-finger.In the developed method,the key point localization depends less heavily on the region segmentation.Angle bisector is more robust to the variation of palmprint direction.The two-time line fitting can remove more singular point interference than single-time line fitting.(2)Intersecting circle is employed for palmprint preprocessing.First,the palm is segmented from the background by using the modified Gaussian skin model.The contour curvatures at valley-between-finger are greater than those at the other edge locations.The edge contour points are set as the centers to set up their corresponding radii.The central angle of the intersecting circle,which is generated with the two intersections between the circle and the contour,is used to measure the curvature of the intersecting circle center approximately.The points on the contour of valley-between-finger are judged and selected according to the angle size and the direction of central angle of the intersecting circle.Finally,the bottom of valley-between-finger is considered as one key point.Intersecting circle method can accurately restrict the contour points of valley-between-finger,and depends less heavily on the extracted edge contour.(3)A palmprint verification system on Android platform is designed and implemented.The environment setup and variable configuration on Android operating system are carried out according to the demand analysis.The PC and test mobile device are selected according to the requirements of various technical indices.In software subsystem,the main interface and image capture interface are designed to meet the system requirements.All kinds of classes and threads are programmed in Java language.Finally the software and hardware are jointly debugged.To sum up,the challenges of palmprint verification on Android devices are analyzed systematically in this dissertation.The focus is the research on palmprint preprocessing algorithms that are suitable for the corresponding assistant acquisition methods.The designed methods take full advantage of the geometry of finger profile,the distance between the fingers,palm width,to name a few.In addition,the methods can better resist the interference of the skin-like color and complex background.Therefore,the accuracy of the developed palmprint verification system is high,which meets the secure authentication requirements of Android operating system.