Research On Palmprint Preprocessing In Mobile Environment

Compared with other biometric technology,palmprint recognition has the characteristics of high recognition accuracy and low hardware requirements.The existing palmprint recognition system mostly applies the contact acquisition equipment,with features of high clarity,uniform illumination,and single background.However,with the widespread use of mobile smart devices,information security and identity authentication of the mobile terminal are becoming increasingly significant.Thus,palmprint recognition technology in a mobile environment has become an important research direction in recent years.Among all the areas,the first priority is to develop image preprocessing which directly affects the accuracy of recognition in the entire process.So far,in palmprint recognition under a mobile environment,there are two main unsolved issues:one is that the variety of palmprint image illumination in open environments.With the complex background and hand contact position changing,palmprint region of interest(Region of,Interest,ROI)positioning is not accurate which has a great influence on the recognition accuracy.On the other hand,due to the speed of the mobile computing platform and application scenarios,some algorithms who have a large amount of calculation is not suitable to transplant to the mobile device.To solve the above problems,the constrained local model is introduced into the palmprint preprocessing.On the basis of a comprehensive analysis of the constrained local model,two improved methods are proposed in this paper.The main work of this paper is shown as follows:(1)due to no public palmprint database is available in the mobile environment,we apply smartphones to capture and establish our own palmprint database in the mobile environment.(2)the constrained local model is applied to palmprint preprocessing,and a new palmprint preprocessing algorithm in a mobile environment is proposed.The results is able to prove the feasibility of the algorithm.(3)a constrained local model based on Haar-random forest is proposed.The experiment's result illustrates that the proposed algorithm is more robust to complex background compared with the local model using PCA or SVM.(4)since the constrained local model fitting algorithm based on Gaussian mixture model is very accurate but slow,it is not suitable for mobile devices.In this paper a constrained local model fitting algorithm based on kernel density estimation is proposed.In this algorithm,the kernel density estimation is used to fit the response image of the local key points,and then the parameters of the model are estimated by maximizing the global response.The experimental results show that the algorithm is able to greatly improve the fitting speed without losing the accuracy of the key points.