动态人脸监控识别系统的研究与实现
文献类型：学位论文
| 作者 | 倪鑫 |
| 学位类别 | 工程硕士 |
| 答辩日期 | 2014-05-27 |
| 授予单位 | 中国科学院大学 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 李子青 |
| 关键词 | 视频监控 人脸识别 WCF Socket 数据同步 Video Surveillance Face Recognition WCF Socket Data Synchronization |
| 其他题名 | Research and Implementation of Dynamic Face Surveillance Recognition System |
| 学位专业 | 计算机技术 |
| 中文摘要 | 随着网络信息化的不断发展，视频监控系统已经成为安防领域的主流产品，但现行系统只着眼于监控场景的记录，缺乏对人脸图像的进一步分析与理解，并且对视频内容只能靠人来判断，这样便不能充分发挥视频监控系统的主动性。基于人脸识别技术的视频监控系统为此提供了一种更加方便、可靠和安全的身份认证手段。 本课题经过详细地市场调研及需求分析，接着进行了功能和流程设计，最后基于Windows平台以VC++语言和C#.NET语言实现了一套动态人脸监控识别系统。其主要研究内容有：（1）前端人脸图像采集模块：从多路监控摄像机不断地获取视频图像，然后对图像进行人脸检测、质量评估等相关预处理，将处理后的人脸数据通过WCF（Windows Communication Foundation）或Socket技术发往后台服务器进行特征抽取和比对识别；（2）后端人脸数据接收模块：通过WCF或Socket技术从前端多路采集通道接收人脸数据包，并将其放入数据缓冲区中；（3）后端人脸数据处理模块：通过轮询的方式从数据池中获取人脸数据，然后对其中归一化后的人脸图像进行特征抽取，接着与人脸特征库中的特征模板进行比对识别，最后将超过阈值的比对结果封装成报警数据包；（4）后端报警人脸信息发送模块：通过WCF或Socket技术将报警数据包发送至报警终端进行显示，并且将相关报警信息写入后台数据库以便事后查询分析；（5）报警模块：主要进行报警结果展示，并实时显示前端多路视频画面；（6）数据同步服务模块：根据人脸库类型从后台数据库中导出特征数据，并将特征数据自动分发至比对服务器上，保证人脸数据处理模块所使用的人脸特征库与后台数据库中的数据完全一致。 最后通过大量测试表明，本系统能对图像中的人员进行自动检测和实时分析，发现监控画面中的异常情况，能以最快和最佳的方式发出预警信息，从而能够更加有效地协助现场工作人员处理危险境况。 |
| 英文摘要 | With the continuous development of network information technology, video surveillance systems have become the mainstream products in the field of security and protection. But the existing systems only focus on recording the scenes, lacking further analysis and understanding for the facial images. The systems rely on people heavily to analyse the video content, so they can’t work automatically. The video surveillance system based on face recognition provides us with a more convenient, reliable and secure way to recognize the identities of human in the video. Firstly，this thesis conducted market research and requirement analysis in detail. Then, it designed the function and process of the system. Finally, this thesis implemented a dynamic face surveillance recognition system based on the Windows platform with VC++ and C#.NET language. The research contents mainly are as follows. (1) Front-end face image acquisition module. It continuously acquired images from multiplex network cameras. Then it conducted a series of image preprocessing operations, including face detection, quality assessment, etc. Finally it sent the preprocessed face data to the back-end server, via WCF (Windows Communication Foundation) or Socket technology, for feature extraction and face recognition. (2) Back-end face data receiver module. It received lots of face packets from multiple front-ends via WCF or Socket technology. Then, it placed these face packets into a buffer pool. (3) Back-end face data processing module. It polled the face data from the buffer pool, and then extracted the feature of the normalized face images. Next, the feature of captured image is compared to the templates stored in the feature database of the system. Finally, the comparison results which exceeded the threshold were encapsulated into alarm packets. (4) Back-end alarm information sending module. It sent alarm information to the alarm terminal for display via WCF or Socket, and some relevant alarm information were stored in the database for querying and analysis later. (5) Alarm module. It mainly showed the alarm results and displayed the multiple images from front-ends in real-time. (6) Data synchronization service module. It exported specified feature data from the database, and then automatically distributed the feature data to the comparison module. It made that the features used by face processing module and the features in the database were in the same. Finally, a lot of tests sho... |
| 语种 | 中文 |
| 其他标识符 | 2011E8009061074 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/7718]  |
| 专题 | 毕业生_硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 倪鑫. 动态人脸监控识别系统的研究与实现[D]. 中国科学院自动化研究所. 中国科学院大学. 2014. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。