Research Of Non-contact Palmprint Preprocessing Algorithm Based On Complex Background

In recent years, with the constant advance of science and technology, the security of the personal identify information has been receiving much attention. Especially in bank accounts, online payment platform and other areas to involve the application field of financial transactions. The traditional methods storing information by using simple password has become increasingly obvious disadvantage. They are easily forgotten or stolen. The current biometric identification technology is very good to solve these problems. Biometric identification technology mainly refers to identifying individuals by using the inherent physiological characteristics. These characteristics usually have some unique advantages, such as uniqueness, stability, safety and so on. Palmprint identification technology as a kind of biometric identification technology. It has not only the basic advantages of biometric identification technology, but also has the advantages of easy acquisition and high user acceptance. So the study of palmprint identification technology will has very broad application prospects. At present, the palmprint identification precision is very high in the lab, since background of palmprint image that we acquire is very single, but in the real scene the palmprint image that we acquire has larger difference than in the lab. In a real scenario, the background of palmprint image may be more complex, and the angle of palm that we photograph may be uncertain, it is very difficult to accurately extract palmprint interested area(ROI) only using the palmprint preprocessing algorithm used in the laboratory. It is also a key factor which limits the marketization of palmprint identification technology. In order to make the palmprint identification technology is applied to the People's Daily life. The research under the complex background of palmprint preprocessing technology is particularly important.This paper mainly studies the contactless palmprint preprocessing method based on complex background, there are three main point: firstly, through analysis of the more popular palm segmentation algorithm based on complex background, this paper proposes a method of artificial neural network which combine single gaussian skin color model with light weight in the complex background of palm foreground segmentation. This method solved the inaccurate segmentation problem when plam is in the case of uneven illumination; secondly, through analysis of shortcoming of the existing classical palmprint ROI extraction algorithm,this paper invents an interested region extraction method based on the calibration of palmprint direction. The method can quickly extract interested region in any direction of palm area by correcting the palm direction, and solve the inaccurate extraction problem of palmprint interested area in the real scene; thirdly, for the palmprint interested area by using preprocessing algorithm proposed in this paper, this paper extracts ULBP feature and proceeds one-to-many matching recognition, then it assess the contactless palmprint preprocessing algorithm performance in based on the complex background this paper.In order to verify the validity of the algorithm proposed in this paper, we set up the acquisition of palmprint database under the environment of complex background, The database that phoning palm images by using Iphone6 in complex background is composed of 60 different palm, every palm collect 5 images, it is a total of 300 copies of palm images. In the database, experiment show that error rate using the ULBP feature matching in palmprint interested area based on the contactless palmprint preprocessing algorithm in complex background has reached 1.65%.Therefore, this article proposed algorithm is more applicable to palmprint image of the real scene, to promote marketization of palmprint identification technology has certain reference significance.