基于ZYNQ的智能监控系统关键技术的研究

发布时间：2018-05-13 15:09

本文选题：ZYNQ + 人脸检测　；参考：《西安邮电大学》2017年硕士论文

【摘要】：随着我国经济的快速发展,居民生活水平的不断提高和安全意识的加强,人们对安防及安全防范系统的认识也越发深入。现有安防系统主要以视频监控为主,通常是将摄像头的输出信息进行保存,然后人工对视频信息进行排查和浏览。这种人工处理的方法必然会引发系统异常,导致系统无法及时发出预警信息。这就使得监控系统迫切需要一种快速、非接触式的身份识别技术,该技术能够对采集到的视频图像进行自行分析,确认图像中人员的身份信息,并将异常及时发送给安保人员,从而避免非安全事件的发生。本文结合人脸识别技术,研究基于ZYNQ的智能监控系统关键技术,为实现智能监控系统奠定理论基础。对于智能监控系统中关键技术,本文主要研究内容如下:(1)提出基于LBP和AdaBoost的人脸检测算法。针对背景复杂、光照不均匀等外界因素所导致的人脸检测率较低的问题进行了改进,实验结果显示该算法对外界因素具有很强适应性;(2)提出基于 LBP(Local Binary Pattern,局部二值模式)和 PCA(Principle Compon-entAnalysis,主成分分析法)的支持向量机人脸识别算法。为了尽可能保留人脸图像的细节信息,在提取人脸特征之前对人脸图像进行分块处理。随着人脸分块数的增加,人脸特征向量的维数将会逐步提升,这将会导致后期训练和识别时间过长,针对这一问题本文采用PCA算法对特征向量进行降维处理,从而降低训练和识别时间。本文采用LBP算子提取人脸特征,使得所提算法对光照和复杂背景等外界因素有很强的鲁棒性;(3)针对在嵌入式平台运行人脸检测算法实时性差的问题,提出一种硬件加速方法,将图像预处理和人脸检测算法在ZYNQ架构的FPGA中进行硬件加速,结果显示经过硬件加速后人脸检测的平均时间可以达到12ms。本文主要研究了人脸检测和人脸识别算法及在智能监控系统中的应用,针对人脸识别算法实时性差和识别率低的问题进行了深入研究,并提出新的方法。本文采用软硬件协同的设计思路对智能监控的原型系统进行了实现,经实验验证,该原型系统达到预期目标。
[Abstract]:With the rapid development of economy, the improvement of living standard and the strengthening of safety consciousness, people have a deeper understanding of security and safety prevention system. The existing security systems mainly focus on video surveillance, usually save the output information of the camera, and then manually search and browse the video information. This manual processing method will inevitably cause system anomalies, resulting in the system can not send out early warning information in time. This makes the monitoring system urgently need a fast, contactless identification technology, which can analyze the collected video images, confirm the identity information of the people in the images, and send the abnormal information to the security personnel in time. In order to avoid the occurrence of unsafe events. In this paper, the key technologies of intelligent monitoring system based on ZYNQ are studied in combination with face recognition technology, which lays a theoretical foundation for the realization of intelligent monitoring system. For the key technology of intelligent monitoring system, the main contents of this paper are as follows: 1) A face detection algorithm based on LBP and AdaBoost is proposed. Aiming at the problem of low detection rate caused by external factors such as complex background, uneven illumination and so on, the paper improves the performance of face detection. Experimental results show that the algorithm has a strong adaptability to external factors. (2) A face recognition algorithm based on LBP(Local Binary pattern (local binary mode) and PCA(Principle component analysis (principal component analysis) is proposed. In order to preserve the details of the face image as much as possible, the face image is processed in blocks before extracting the features of the face. With the increase of the number of face blocks, the dimension of face feature vector will be gradually increased, which will lead to the late training and recognition time is too long. In this paper, PCA algorithm is used to reduce the dimension of the feature vector. Thus reducing the training and identification time. In this paper, LBP operator is used to extract face features, which makes the proposed algorithm robust to external factors such as illumination and complex background. Aiming at the problem of poor real-time performance of face detection algorithm running on embedded platform, a hardware acceleration method is proposed. The hardware acceleration of image preprocessing and face detection algorithm in FPGA based on ZYNQ architecture shows that the average time of face detection can reach 12 Ms after hardware acceleration. In this paper, the algorithms of face detection and face recognition and its application in intelligent monitoring system are studied. The problems of poor real-time and low recognition rate of face recognition algorithm are studied deeply, and a new method is proposed. In this paper, the prototype system of intelligent monitoring is implemented by using the idea of hardware and software co-design. The experimental results show that the prototype system achieves the expected goal.
【学位授予单位】：西安邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP277

【相似文献】