Research On Application Of Face Recognition Technology In Video Surveillance Systems

Video surveillance system can be seen everywhere in our daily life, but most of the systems are based on PC-multimedia platform and can not be real intelligent. With the rapid development of computer vision, many individuals and institutions focus on how to process video images in real time and get valid information.This thesis focuses on the application of face recognition technology in video surveillance systems, and compares it with other biological characteristics such as fingerprints, eye mask. Faces are more easily accessible than other biological characteristics. In recent decades, many effective face detection and face recognition algorithms have been presented, and they have been applied into real-time monitoring successfully. This thesis describes the video image acquisition, image preprocessing, face detection, face recognition, and phone alarm based on the SMS cat. The whole thesis eventually accomplishes the prototype video surveillance alarming system based on face recognition.The main tasks of this thesis are:(1) Image preprocessing, including image denoising, image enhancement, image normalization. Because the key noise caused by camera is Gaussian noise and salt noise, we use median filtering and wavelet analysis to finish image denoising in this thesis. Image enhancement is accomplished by fractional differential technology. This thesis uses the nearest neighbor interpolation and bilinear interpolation methods to finish image normalization.(2) Image acquisition. This thesis uses the stream media processing development kit, which is Direcshow provided by Microsoft, to carry out image acquisition platform.(3) Face detection. This thesis use Adaboost algorithm which is popular with researchers to finish face detection. After using this algorithm in the CMU face database, online collection of images and images on video capture separately, we get that this algorithm has a high accuracy. (4) Face recognition. This thesis do a comprehensive analysis about the PCA algorithm and the SIFT algorithm of the face recognition. Through using the two algorithms on ORL face database and laboratory database, this thesis compares their advantages, disadvantages and practicality.(5) Phone alarm based on the SMS cat. This thesis introduces the principles and usage of SMS cat, and then use face recognition to control the phone alarm.(6) Achievement of the video surveillance alarm prototype system based on face recognition. This system uses OpenCV to achieve the function of automatic alarm eventually. The whole design and achievement of this system has a certain research and practical value.