Research On Auxiliary Techniques For Palmprint Segmentation And Localization On Mobile Terminals

With the enhanced compatibility and processing ability of mobile devices,their functionalities are increasingly diversified,which involves more and more personal privacy information.The traditional digital password or Sudoku pattern authentication methods on mobiles,which are with low security,and easily peeped or cracked,commonly lead to property or spiritual loss.In addition,the users,especially the elder people,usually suffer from a larger burden of memory caused by the traditional authentication methods,Palmprint features are with rich discrimination power,highly stable and unique.Furthermore,palmprint systems are with low cost and good user acceptance,accordingly palmprint has been the rising star of the family of biometrics.Therefore,the research of palmprint authentication algorithm on mobile terminals has important theoretical significance and practical value.In mobile environments,the background is unanticipated and the interference,as the serious challenge of palmprint preprocessing,is drastically strong.Moreover,due to the limitation of hardware resources of mobile devices,the traditional palmprint image preprocessing algorithms are difficult to be transplanted directly.To solve the aforementioned problems,in this dissertation,palmprint recognition techniques are intensively studied,and palmprint preprocessing algorithms are improved or modified.The main research contents and innovations of this paper are summarized as follows:(1)The relevant theoretical knowledge of palmprint recognition is summarized.First of all,the author introduces the background and significance of the research,analyzes the feasibility of palmprint authentication on mobile terminal.Secondly,the current states and progress of palmprint recognition are explained in great detail,and the difficulties to be solved are pointed out.Finally the evaluation indexes of palmprint authentication system are introduced in order to pave the way for the performance test of the algorithm in the following chapters.(2)A new palmprint image segmentation method is proposed.Firstly,a novel I-shaped guide technique is developed to guide the users to place their hands,based on which a palmprint database is established to verify the performance of the proposed method.Secondly,the parameters of Gaussian skin-color model are trained in sub-image automatically,and the non-complete segmentation result is obtained.Region growing starts from the non-complete region and accurately segments complete palmprint region.The proposed method resists the skin-color-like interference regions in background,which are not connected to the hand,and the efficiency of region growing is improved by choosing skin color segmentation results as the initial seed.It overcomes the failure of the fixed threshold for gradual change in hand region.The experimental results confirm that the proposed method can achieve higher correctness and verification accuracy in mobile environments.(3)A new method for locating the region of interest(ROI)is proposed.Since the traditional methods are heavily depend on the quality of image segmentation,and sensitive to the angle between fingers,this dissertation develops a scheme based on the bisector of the angle between finger valley edges.Firstly,the edges of three components in YCbCr space are fused according to certain rules,and then the edges of finger valley area need to be retrieved.The loop edges are transformed to regions following the close and filling operation.Reference points are selected to fit the edges of regions on both sides as straight lines.There is an angle between two straight lines in one region,so the angle bisector can be calculated,and the lowest intersections between angle bisectors and regions are the desired key points.ROI can be acquired based on the coordinate system established according to the key points.The proposed method not only reduces the computational complexity through searching the key points on loop edges,but also ensures that the key points only vary in small range near the middle points of finger root even if the hand placement angle deflects.As a result,the accuracy of the authentication system is greatly improved.In summary,this dissertation studies the palmprint recognition under auxiliary conditions.The proposed methods for palmprint segmentation and localization improve the verification accuracy.The research on auxiliary techniques for palmprint segmentation and localization algorithms on mobile terminals is of great theoretical significance and engineering application value.