一种舰船人员人脸检测及识别的智能系统设计  

Design of an Intelligent System for Ship Personnel Face Detection and Recognition

作　　者：徐草草[1] 杨启明 张镘 徐晶 XU Caocao;YANG Qiming;ZHANG Man;XU Jing(College of Engineering&Technical,Chengdu University of Technology,Leshan 614000;Leshan Branch of China Mobile Communications Group,Leshan 614000;Zhuhai Bowen Education Technology Development Co.,Ltd.,Zhuhai 519031;Neijiang Normal University,Neijiang 641112)

机构地区：[1]成都理工大学工程技术学院,乐山614000 [2]中国移动乐山分公司,乐山614000 [3]珠海博闻教育科技发展有限公司,珠海519031 [4]内江师范学院,内江641112

出　　处：《舰船电子工程》2023年第12期128-131,共4页Ship Electronic Engineering

基　　金：国家自然科学基金项目(编号:61201397);乐山市科技局科技支撑计划(编号:19GZD055)资助。

摘　　要：舰船舱室授信人员授信,是舱室安全管理的重要组成部分。由于舱室存在光线暗、背景复杂、且授信人员存在部分遮挡等可能性,给舱室安全授信带来了极大的困难。为了应对复杂环境中人脸监测困难,识别精度低等一系列问题,论文提出了一种基于深度学习的应用于舰船人员人脸检测及识别的MTCNN+FaceNet算法,通过数据降维的方式将系统人脸数据降至128维,经过在公开数据库测试,该算法不论是关键点检测还是执行效率均高于其他算法。通过实验验证发现,该算法识别反应速度约为5ms,识别精度高达98%。The Credit granting to personnel in ship cabins is an important component of cabin safety management.Due to the possibility of dark lighting,complex backgrounds,and partial shielding by credit granting personnel in the cabin,it has brought great difficulties to the cabin security credit granting.In order to address a series of issues such as difficult face monitoring and low recognition accuracy in complex environments,this paper proposes a deep learning based MTCNN+FaceNet algorithm for face detec⁃tion and recognition of ship personnel,which reduces the system's face data to 128 dimensions through data dimensionality reduc⁃tion.After testing in an open database,the algorithm in this paper is superior to other algorithms in terms of key point detection and execution efficiency.Through experimental verification,it is found that the recognition response speed of this algorithm is about 5ms,and the recognition accuracy is as high as 98%.

关 键 词：人脸识别 MTCNN FaceNet 深度学习 识别精度 

分 类 号：TN911.73[电子电信—通信与信息系统]