Research On Orientation Distance Feature For Palmprint Identification

As a kind of biometric technology,palmprint recognition technology has attracted more and more researchers'attention in recent years.Palmprint recognition technology not only meets the universality,uniqueness,and collectability of biometrics,but also has unique advantages such as rich features,easy collection,and user-friendliness.With the increasing demand for biometrics technology in contemporary society,how to take advantage of palmprint recognition,build a robust and efficient palmprint recognition system is the main research direction of palmprint recognition research,today.The recognition method based on palmprint orientation features is one of the popular palmprint recognition methods.However,in the selection of palmprint orientation features,the use of single orientation features will make the features less robust,and the use of multi-orientation features often fails to meet the real-time requirements due to excessive computation.At the same time,in the palmprint matching algorithm,the traditional one-to-one palmprint verification algorithm has been unable to meet the one-to-many palmprint identification task.The sparse representation based method has proved to be effective in solving such one-to-many recognition problem,and it has also become a breakthrough for solving the palmprint recognition task.Aiming at the above problems of robustness,recognition speed and recognition rate in palmprint recognition,the main work of this paper includes the following:First,an orientation distance based palmprint feature is studied and proposed.Based on the rotation invariance of distance,by using a set of linear filters in different orientations to convolve with the palmprint image,and then calculate the distance between the strongest response orientation and the weakest response orientation,the orientation coding distance and orientation response distance are obtained.Next,the obtained orientation coding distance and the main orientation coding feature are fused separately,and finally it is expressed as a palmprint feature in the form of a histogram together with the orientation response distance,HOD_l _m andHOD_r.Experimental results show that the palmprint distance feature is more robust to the palmprint image than the traditional palmprint orientation feature,and the recognition performance is better.Secondly,a multi-feature palmprint recognition algorithm based on sparse representation is studied and improved.The traditional two-step sparse representation method has good recognition accuracy and speed,but it can also be studied in depth,and make it more suitable for the application scenarios of palmprint recognition.In this paper,through the reasonable configuration of the orientation distance feature under the framework of the two-phase method,a MTPTSR method is proposed,The MTPTSR method can better match the orientation distance feature and meet its application scenario.Finally,palmprint verification and palmprint identification experiments are conducted in the three popular palmprint databases Poly U,IITD,GPDS.Experimental results show that the proposed method improves the recognition accuracy and ensuring the recognition speed compared with the traditional sparse representation method.