亿级人像比对系统架构设计及优化
|
摘要
为满足公安机关大规模人像数据库的比对应用需求,提高人像比对系统应用规模和可用性,采用结构化开发方法,从应用层、数据库层、主机层、存储层等方面对人像比对系统进行模块化架构设计和优化。系统能够支撑15亿以上人像比对库容量,实现了全国范围的人像自动比对和纠错,提高了身份辨认速度。系统平均建模速度、平均比对速度和查重速度分别提升38%、41%和55%。优化后的人像比对系统在保证识别率高的同时能够快速处理海量数据,满足公安机关应用需求,有效解决了户口登记管理中的"错、重、假"问题。
In order to meet the needs of the public security organs for large-scale portrait database application,and to improve the application scale and scene and high availability of the face recognition system,we use the structured development methodology,modular architecture designs and optimize the face recognition system from application layer,database layer,host layer,and storage layer,etc. The system can support more than 1.5 billion portrait matching library capacity,which realizes nationwide automatic comparison and correction of portraits,speeds up the identification process of relevant personnel in relevant departments. The mold speed,average comparison speed,and check speed were increased by 38%,41%,and 55%,respectively. After optimization,the face recognition system can quickly process massive data while ensuring high recognition rate,meet the application requirements of public security agency,and effectively solve the problem of wrong,repeated and fake information in the registration management of household registration.
In order to meet the needs of the public security organs for large-scale portrait database application,and to improve the application scale and scene and high availability of the face recognition system,we use the structured development methodology,modular architecture designs and optimize the face recognition system from application layer,database layer,host layer,and storage layer,etc. The system can support more than 1.5 billion portrait matching library capacity,which realizes nationwide automatic comparison and correction of portraits,speeds up the identification process of relevant personnel in relevant departments. The mold speed,average comparison speed,and check speed were increased by 38%,41%,and 55%,respectively. After optimization,the face recognition system can quickly process massive data while ensuring high recognition rate,meet the application requirements of public security agency,and effectively solve the problem of wrong,repeated and fake information in the registration management of household registration.
引文
[1]林晓东,刘心松.高可用性系统的研究与实现[J].电子科技大学学报,1997(5):82-87.
[2]TORKAMANI M A,BAGHERI H,SHALTOOKI A A,et al.Architectural solution for reaching high availability[J].Global Journal on Technology,2013(1):105-109.
[3]罗福强,熊永福.基于分层的Web系统的性能优化研究与探讨[J].计算机技术与发展,2012(11):85-88.
[4]DIMITRIAIVALIOTIS,艾维利,陶利军.精通Nginx[M].北京:人民邮电出版社,2015.
[5]JOACHIM GOLL.软件架构与模式[M].北京:清华大学出版社,2017.
[6]胡迎松,彭利文,池楚兵.基于.NET的Web应用三层结构设计技术[J].计算机工程,2003,29(8):173-175.
[7]曾晖泽.公安人像比对系统的设计与实现[D].厦门:厦门大学,2014.
[8]林萍,朱婵.基于Ajax技术和JAVAEE的分页查询优化[J].计算机系统应用,2017(8):916-1001.
[9]THOMAS KYTE.Oracle Database 9i/10g/11g编程艺术:深入数据库体系结构[M].第2版.北京:人民邮电出版社,2011.
[10]周志德.Oralce数据库的SQL查询优化研究[J].计算机与数字工程,2010,38(11):173-178.
[11]査盛音,王玉玫,聂颖.OpenMP在图形数据文件加载中的应用[J].软件导刊,2013,12(10):159-162.
[12]OpenMP application program interface version 4.0.0[EB/OL].http://www.openmp.org.
[13]李艳鹏,杨彪.分布式服务架构:原理、设计与实战[M].北京:电子工业出版社,2017.
[14]ILYA GRIGORIK.Web性能权威指南[M].北京:人民邮电出版社,2014.
[15]张杰,刘琦.多级联动的人像比对应用平台研究与设计[J].电脑知识与技术,2016,12(31):177-178.
[16]刘永梅.基于人脸识别身份认证系统数据库的设计与研究[D].呼和浩特:内蒙古大学,2014.
[17]刘合星,刘琦.人像识别系统的设计与研究[J].河南科技,2016(21):62-65.
[18]关乐,刘磊,朱鹏,等.动态人像比对研究及应用[J].甘肃科技,2017(12):87-89.
[19]杨洋,杨义光,范东亮.人像比对系统建设中几个关键问题研究[J].信息技术与信息化,2017(5):86-87.
[20]黄静,李炳.基于Nginx的Web服务器性能优化研究[J].浙江理工大学学报,2016,35(7):600-606.
[21]蒲晓阳.基于Nginx和Redis高并发Web服务负载均衡的研究[J].内江科技,2016,37(1):40-41.
[22]赵峡策.基于Nginx和Memcache的负载均衡集群架构设计[J].电子技术与软件工程,2014(5):39-40.
[23]蔡文海,陈洺均.基于OpenMP并行编程模型与性能优化的稀疏矩阵操作研究[J].软件导刊,2016,15(3):27-29.
[24]张辉,赵郁亮,徐江,等.基于Oracle数据库海量数据的查询优化研究[J].计算机技术与发展,2012,22(2):165-167.
[25]陈南.ASP.NET中大数据量分页技术的研究与实现[J].计算机应用与软件,2011,28(4):205-207.
[26]OPENMP C.OpenMP C and C++application program interface draft[EB/OL].http://www.openmp.org.
[27]BHATTACHARJEE A,CONTRERAS G,MARTONOSI M.Parallelization libraries:characterizing and reducing overheads[J].ACMTransactions on Architecture&Code Optimization,2011,8(1):1-29.
[28]SUPINSKI B R D,SCOGLAND T R W,DURAN A,et al.The ongoing evolution of openMP[J].Proceedings of the IEEE,2018(6):1-16.
[2]TORKAMANI M A,BAGHERI H,SHALTOOKI A A,et al.Architectural solution for reaching high availability[J].Global Journal on Technology,2013(1):105-109.
[3]罗福强,熊永福.基于分层的Web系统的性能优化研究与探讨[J].计算机技术与发展,2012(11):85-88.
[4]DIMITRIAIVALIOTIS,艾维利,陶利军.精通Nginx[M].北京:人民邮电出版社,2015.
[5]JOACHIM GOLL.软件架构与模式[M].北京:清华大学出版社,2017.
[6]胡迎松,彭利文,池楚兵.基于.NET的Web应用三层结构设计技术[J].计算机工程,2003,29(8):173-175.
[7]曾晖泽.公安人像比对系统的设计与实现[D].厦门:厦门大学,2014.
[8]林萍,朱婵.基于Ajax技术和JAVAEE的分页查询优化[J].计算机系统应用,2017(8):916-1001.
[9]THOMAS KYTE.Oracle Database 9i/10g/11g编程艺术:深入数据库体系结构[M].第2版.北京:人民邮电出版社,2011.
[10]周志德.Oralce数据库的SQL查询优化研究[J].计算机与数字工程,2010,38(11):173-178.
[11]査盛音,王玉玫,聂颖.OpenMP在图形数据文件加载中的应用[J].软件导刊,2013,12(10):159-162.
[12]OpenMP application program interface version 4.0.0[EB/OL].http://www.openmp.org.
[13]李艳鹏,杨彪.分布式服务架构:原理、设计与实战[M].北京:电子工业出版社,2017.
[14]ILYA GRIGORIK.Web性能权威指南[M].北京:人民邮电出版社,2014.
[15]张杰,刘琦.多级联动的人像比对应用平台研究与设计[J].电脑知识与技术,2016,12(31):177-178.
[16]刘永梅.基于人脸识别身份认证系统数据库的设计与研究[D].呼和浩特:内蒙古大学,2014.
[17]刘合星,刘琦.人像识别系统的设计与研究[J].河南科技,2016(21):62-65.
[18]关乐,刘磊,朱鹏,等.动态人像比对研究及应用[J].甘肃科技,2017(12):87-89.
[19]杨洋,杨义光,范东亮.人像比对系统建设中几个关键问题研究[J].信息技术与信息化,2017(5):86-87.
[20]黄静,李炳.基于Nginx的Web服务器性能优化研究[J].浙江理工大学学报,2016,35(7):600-606.
[21]蒲晓阳.基于Nginx和Redis高并发Web服务负载均衡的研究[J].内江科技,2016,37(1):40-41.
[22]赵峡策.基于Nginx和Memcache的负载均衡集群架构设计[J].电子技术与软件工程,2014(5):39-40.
[23]蔡文海,陈洺均.基于OpenMP并行编程模型与性能优化的稀疏矩阵操作研究[J].软件导刊,2016,15(3):27-29.
[24]张辉,赵郁亮,徐江,等.基于Oracle数据库海量数据的查询优化研究[J].计算机技术与发展,2012,22(2):165-167.
[25]陈南.ASP.NET中大数据量分页技术的研究与实现[J].计算机应用与软件,2011,28(4):205-207.
[26]OPENMP C.OpenMP C and C++application program interface draft[EB/OL].http://www.openmp.org.
[27]BHATTACHARJEE A,CONTRERAS G,MARTONOSI M.Parallelization libraries:characterizing and reducing overheads[J].ACMTransactions on Architecture&Code Optimization,2011,8(1):1-29.
[28]SUPINSKI B R D,SCOGLAND T R W,DURAN A,et al.The ongoing evolution of openMP[J].Proceedings of the IEEE,2018(6):1-16.