基于稀疏化CNN在Cortex-M视频的身份识别研究  

Research on Identity Recognition Based on Sparse CNN on Cortex-M Video

作　　者：李晓琳 庞保孟 朱青青[1] 耿相珍 LI Xiao-lin;PANG Bao-meng;ZHU Qing-qing;GENG Xiang-zhen(Intelligence and Manufacture College,Qingdao Huanghai University,Qingdao 266500,China;Shandong Hi-Speed Maintenance Group Co.,Ltd.,Jinan 250014,China;Liaocheng Branch,China Mobile Communications Group Co.,Ltd.,Liaocheng 252059,China)

机构地区：[1]青岛黄海学院智能制造学院,山东青岛266500 [2]山东省高速养护集团有限公司,山东济南250014 [3]中国移动聊城分公司,山东聊城252059

出　　处：《机械工程与自动化》2023年第3期16-18,共3页Mechanical Engineering & Automation

基　　金：山东省高等教育研究项目(21HER056);青岛黄海学院科研立项项目(2021KJ16)。

摘　　要：目前市场上存在的门禁、考勤等身份识别系统,基本上是利用算法直接在基于操作系统的专用设备上进行身份的验证,大量视频图像直接上传到中心服务器进行数据运算处理,造成服务器的压力过大。针对这一问题,提出一种基于稀疏化CNN在微控制器芯片上实现监控视频人员身份识别的方法。首先在不依赖于操作系统的微控制器芯片上对小区、商场及校园等公共场所的监控视频进行结构处理,采用CMSIS-NN在微控制器芯片上搭建与PC训练机(服务器)相同的稀疏化改进的AlexNet模型,然后将利用训练机获取的经验值移植到微控制器芯片上,实现对结构化视频的人员身份识别。经实验证明,在微控制器和训练机上识别准确率基本保持一致。因此,在系统前端实现视频结构化处理以及对其搭建稀疏化神经网络实现对人员身份识别是可行的。At present,the entrance guard,attendance on the market such as identity recognition system,most use algorithms to verify identity directly on dedicated devices based on operating systems.A large number of video images are directly uploaded to the central server for data calculation and processing,resulting in excessive pressure on the server.Therefore,this paper proposes a research on the realization of surveillance video personnel identification on microcontroller chip based on sparse CNN.First on micro controller chip is not dependent on the operating system to the community,public places such as shopping malls and campus monitoring video structure processing,the CMSIS-NN on microcontroller chip set up with PC training machine(server)the same sparse improved AlexNet model,then using training machine to obtain the experience value of transplanted into micro control chip,to realize the person identity recognition of structured video.The experimental results show that the recognition accuracy of the Microcontroller and the trainer is basically the same.Therefore,it is feasible to realize the video structured processing in the front end of the system and to build a sparse neural network to realize the personnel identity.

关 键 词：身份识别 稀疏化CNN 微控制器芯片 视频结构化 Cortex-M 

分 类 号：TN911.73[电子电信—通信与信息系统]