Intelligent Security Surveillance System Based On Computer Vision

With the development of the era, people improve their requirements for security andintelligence of living environment. The security surveillance system plays an important role in it.Nowadays, it mainly includes two kinds of security surveillance system: the infrared alarm systemand the video surveillance system. The infrared alarm system is vulnerable to the interference ofenvironmental complexity, and the video surveillance system needs people to sit in front of thescreen and observe. So it not only consumes storage resources, but also has low efficiency. Thisessay focuses on image preprocessing, face detection and face recognition, and the intelligentsecurity surveillance system based on face recognition has been proposed.In this thesis: For image preprocessing, based on the research of relevant papers, a series ofimage filtering experiments have been done to improve the image quality; For face detection, onthe basis of the AdaBoost algorithm, we fully study the extraction of face feature and the training ofcascade classifier. Meanwhile, some related experiment has be done to prove the theory; For facerecognition, on the basis of the Eigenface algorithm, K-L transform and PCA algorithm arediscussed in detail. By analyzing the Fisherface method, the two algorithms are tested against twodatabases: UMIST and ORL. Furthermore, we put forward the distributed face recognition solutionbased on Hadoop which accelerates the speed for identification; On the basis of the above study, wedesign the intelligent security surveillance system including connection of webcam, face detection,data storage, face recognition and remote alarm.The system performances well in this essay. On one hand, face can be detected correctly frominput image, and stored in database. On the other hand, the system can accurately identify the face,and activate the alarm when the stranger appeared.