Research On Large-scale Palm Recognition Technique Based On A Cloud Computing Environment

As an emerging biometric technology, palmprint recognition has formed a relative mature theoretical system in recent years through continuous research and accumulation. Palmprint recognition system is now approaching the market stage for the reason of great market prospect. But the existing framework only suits to small size of user applications. With the emergence and development of cloud computing, it provides a platform for large-scale palmprint identification.This paper has built the cloud-based palmprint recognition system framework and system prototype on the basis of cloud computing technology, which can solve the efficiency problem for processing large data sets with a long time, the results and the difficulty of management and maintenance problems. The main research contents are as follow.The palmprint recognition system includes two key parts: feature extraction and feature recognition. This paper realizes palmprint feature extraction algorithm based on 2D Gabor filter and palmprint feature matchi ng algorithm based on Hamming distance. The feature extraction algorithm combines Map Reduce and 2D Gabor filter to extract feature of the ROI of palmprint in the cloud environment; and the feature matching algorithm uses Hamming distance as the matching criteria to achieve multi-machine matching function in a cloud environment.Upon completion of these two core algorithms, this paper also implements a sequential file for palmprint image data storage module and a strong system security SSH public key-based authentication module. The sequential file storage module solves the problem that the cloud platform performs poor efficiency while running on a large number of small files using sequential file, while public key-based authentication module enhancs the security of the system. And this paper designs and completes a complete palmprint recognition system based on cloud-framework.The performance of the cloud architecture based palmprint identification system has significantly improved compared with the traditional system. For more, the performance improves as the amount of data increases, which is appliable in large-scale data. Experimental results show that this system spends 0.011 ms to match a palmprint in the case of data size of 5 million, which is 8 times higher than the traditional palmprint identification system.