基于人脸目标的监控系统

发布时间：2018-12-23 18:26

【摘要】：监控系统是一种涉及模式识别、图像处理、计算机视觉和人工智能等多学科的前沿研究领域,它存在着巨大的经济价值和广泛的应用前景,广受国内外研究人员的关注,因此监控系统技术也是国内学者热衷于研究的热点话题之一。本文避免过于依赖软技术造成的算法复杂性,采用软硬件结合的方式,展开了对监控系统的研究。为了能够更准确的对人脸定位,对人脸检测结果进行了分析;为了能够跟踪更多的人脸信息,针对不同结果采取了不同的定位算法。在对监控系统理论分析的基础上,确定了监控系统的组成,根据系统各部分的功能和连接关系,建立了系统的总体结构。借助VS2008环境与Open CV辅助库,通过摄像头采集现场图像,利用Haar-like特征法寻找出图像中的人脸特征矩形,通过Ada Boost算法生成级联分类器,进行人脸检测,根据不同检测结果,进行不同的人脸定位算法进行人脸定位,将图像显示窗口中心选为坐标原点,用线性变换的方法获取人脸位置偏差,采用映射的方法把二维像素偏差转换成二维角度偏差,并发送给Mega2560单片机,单片机根据接收到的角度偏差信息,控制摄像头使人脸始终在坐标原点位置,对人脸进行实时跟踪。基于Open GL搭建监控装置三维模型,并与装置本体级联,两者统一动作,随时根据三维模型观察本体的转动角度。最后,单片机根据接收的信息,通过控制舵机组去控制摄像头,使摄像头实时跟踪人脸,基于单片机搭建外围语音控制电路,通过语音芯片,控制系统的工作状态。对人脸检与跟踪结果采样并分析,实验表明系统能完成单人脸和多人脸的检测,同时可以准确跟踪人脸。
[Abstract]:Monitoring system is a kind of multi-disciplinary research field involving pattern recognition, image processing, computer vision and artificial intelligence. It has great economic value and wide application prospect, and has been paid much attention by researchers at home and abroad. Therefore, monitoring system technology is also one of the hot topics that domestic scholars are keen to study. In this paper, the algorithm complexity caused by soft technology is avoided, and the monitoring system is studied by combining software and hardware. In order to more accurately locate the face, the face detection results are analyzed. In order to track more face information, different localization algorithms are adopted for different results. Based on the theoretical analysis of the monitoring system, the composition of the monitoring system is determined, and the overall structure of the system is established according to the functions and connections of each part of the system. With the help of the VS2008 environment and the Open CV assistant library, the scene images are collected by the camera, and the face feature rectangle is found out by using the Haar-like feature method. The cascade classifier is generated by the Ada Boost algorithm, and the face detection is carried out according to the different detection results. Different face localization algorithms are used to locate the face. The center of the image display window is chosen as the origin of the coordinate. The face position deviation is obtained by linear transformation, and the two-dimensional pixel deviation is converted into the two-dimensional angle deviation by mapping method. It is sent to Mega2560 singlechip. According to the received information of angle deviation, the single chip computer controls the camera to keep the human face at the origin of coordinates and tracks the human face in real time. The 3D model of monitoring device is built based on Open GL and cascaded with the device body. The two movements are unified and the rotation angle of the body is observed at any time according to the 3D model. Finally, according to the received information, the single chip microcomputer controls the camera by controlling the steering gear group to make the camera track the face in real time. The peripheral voice control circuit is built based on the single chip computer, and the working state of the system is controlled by the voice chip. The results of face detection and tracking are sampled and analyzed. The experimental results show that the system can detect single face and multi-face and track face accurately.
【学位授予单位】：哈尔滨理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP277

【相似文献】