Research On Techniques Of Palm Image Defocusing Analysis And Recognition

With the arriving of Internet and smart home era,an improved style of identity authentication is expected.Biological identification is getting popular for its convenience and security.There is a huge market for palmprint identification due to its high accuracy,receptively and copy complexity.Touching palm image equipment is used in traditional palmprint identification system,which is unsanitary,inflexible and provides poor user experience.Touchless equipment is thus in urgent need to solve the problem.However,because of the uncertainty of environment and user behavior,factors like defocus,illumination,palmprint translation,rotation,tilt and distortion may exist and affect the system accuracy.Based on this point,a strategy is made in this thesis,to take the corresponding image processing method according to its definition,and relevant research and experiment is carried.To measure the degree of image defocus,a method of assess it through image definition is proposed.This method considers the factors of optical defocus principle,imaging result and palmprints' texture property,and used the edge sharpness value method.Verified by experiment,the method is effective in monotonicity,susceptibility and content independence.According to the system accuracy in each definition section,different image process method is taken.A trend is fond in the analysis of the relation between image definition and system accuracy,which is when the image definition reduce,the system accuracy first rise and then decrease.Based on this,the system use 4methods,namely image blurring identification,direct identification,deblurring identification and abandon,according to the palm image's definition section.Verified by experiment,the system accuracy is improves through this strategy.In the blurring identification method,a method of adding Gaussian blur is carried.According to the relation between image definition and system accuracy,we can add blur to an image to increase the accuracy.Verified by experiment,Gaussian blur can increase the accuracy of system.In the deblurring identification method,a method of extracting images' stable feature to recognize is proposed.According to the features of palmprint images within class,the stable feature template is got.When new images sign in,the system first extract its stable feature through the template before recognition.Compared with the classic Gabor code method,our method performers better.