Research On Key Technologies In A Large-scale Distributed Fingerprint Recognition System

In the changing information age,the emergence of big data has brought new problems and challenges to the data storage and processing.In the field of biometric identification technology,fingerprint recognition technology is becoming more and more important,and it plays a more and more prominent role in identity recognition and information security.The problem of big data in fingerprint recognition technology is mainly reflected in the storage and processing of massive fingerprinting.The traditional stand-alone fingerprint recognition system is unable to solve this problem.The extremely low efficiency and the low matching accuracy in the fingerprint recognition process have brought about the emergence of the distributed fingerprint recognition system.Depending on the parallel processing capability of the distributed cluster,the distributed fingerprint recognition system can effectively solve the problem of massive data processing.However,how to carry out fingerprint recognition more efficiently and how to optimize the fingerprint recognition algorithm have become a hot spot and a difficult problem in the field of fingerprint recognition.And for the storage of fingerprint,with the increasing of the database type and the more powerful function,the selection for the database storing fingerprints is also diversified.For the storage and processing of massive fingerprint,this paper simply introduces the existing distributed fingerprint recognition system.The system is built on the Redis database.In the design of storage method for the fingerprint data,a smart and efficient encoding way saves the storage space and the parallel decoding strategy accelerates the calculation speed.The fingerprint matching process is divided into rough match and accurate match,which makes the process of fingerprint recognition more efficient.In this paper,we have mainly studied the system's matching strategy and put forward to two kinds of innovative matching strategies,namely the accurate matching strategy based on Quick-Return and the strategy optimization based on the memory read and write in the Redis.For the research of accurate matching strategy,this paper makes a detailed analysis of the original accurate matching strategy of the fingerprint recognition system,and describes the effect of OpenMP parallel technology in the process of accurate matching.Aiming at the advantages and disadvantages of OpenMP technology,this paper puts forward to an accurate matching strategy based on Quick-Return.Compared with OpenMP,Pthread parallel technology is more flexible for the control of threads,so it is applied to the implementation of the new matching strategy.The accurate matching strategy based on the Quick-Return can greatly reduce the matching time and improve the matching efficiency of the distributed system.For the research of rough matching strategy,this paper presents a new matching strategy based on the memory read-write in the Redis and mixes rough match together with accurate match.In this paper,we use the hash data structure in the Redis.In the process of searching the hashtable in the rough match,we can carry out the real-time accurate match for the found similar fingerprint.If the matching is successful,we will read it out the memory and stop the rough match at the same time,which means ending the hashtable search.The optimization strategy takes good advantage of high speed of memory read-write and high query efficiency of hashtable,greatly reducing the number of retrieved fingerprints and the matching time.The matching strategy can effectively improve the performance of the system.