A Method Based On Coherence Operation And Denoise Auto Encoder For 3D Palmprint Verification

PalmPrint identification is an important biological feature recognition technology, and palmprint image contains a wealth of line features and texture so that it has a very high value in the authentication. But traditional 2D PalmPrint recognition methods havesome issues difficult to be resolved:2D Palmprint is easily affected by illumination,occlusion and other external factors;Image extractions are easily affected by finger posture changes;The 2D palmprints are easily copied.With the proposed 3D palmprint, a new solution is provided to solve the shortcomings of 2D palmprint.3D palmprint is not affected by light, and not easy to imitate.3D PalmPrint recognition as a robust identity verification technology,will become an important research field. However currently,3D PalmPrint identification process still requires precise registration. Therefore, this paper presents amethodbased on coherence operation and denoiseauto encoder for 3D Palmprint verification.(1) Propose a method of palmprint authentication based on coherence operation. Firstly, the method which base on coherence operation extracted the feature vectors of two palmprint similarity, then using pattern recognition method to achieve authentication. This method does not need to make accurate registration of palmprint in advance, which has good robustness.(2) In the authentication algorithm,we useddenoise autoencoderto reduce the dimensionality of the coherence operation characteristicsandimprove the efficiency of the algorithm and the anti-noise ability.Furthermore we took use of real adaboost classifier to classify the feature which reduce dimensionality,and applied dimensionreduction characteristics to distinguishing that whether the two images from the same hand. Experimental results show the availability and robustness of the proposed method.