Continuous Video Stream Of Human Face Detection And Recognition Research

With the rapid development of computer and network technology in the modern society, information security shows an unprecedented importance. Identification is an important prerequisite to ensure security of information systems , people's identity and identification need for national security, e-government, security checks, security monitoring and access control systems. Face detection and identification technology with its unique advantages in the field of computer vision has been thoroughly studied and widely used. However, face detection and identification to move towards practical application, detection accuracy and detection rate is an urgent need to resolve two key issues. After nearly ten years of development, human face detection and recognition accuracy has been greatly improved, but the speed problem has to be improved.This paper focuses on the video image of people on the face detection and recognition, in the course of the study achieved the static images and dynamic video images of the human face detection, face recognition remains to be further studied.Reading at the graduate level by a large number of face detection and recognition of technical aspects of literature, this paper first introduces the face detection and recognition of technology development, current situation and development trend of a brief overview, followed by pairs of face detection and recognition algorithm for classical analysis and research, in order to improve the face detection and recognition of the efficiency of detection of this paper to treat the image histogram transformation enhanced lighting of compensation pre-processing, and then graphics of edge detection and color segmentation, and further improve image quality. On this basis, the Adaboost algorithm is described in detail the entire process, including the Harr features selected, the calculation of rectangle features, integral image calculation, and then introduced the weak classifier, strong classifier and the cascade classification device structure, with the problem based on the algorithm, this paper put forward to improve the algorithm, mainly to improve the classifier construction process, the probability update rules were changed so that the new cycle of more emphasis on misclassification of samples, thus constructed classifier is more efficient and practical, and to improve the image of the human face detection rate and reduce the false detection rate, then this paper to achieve the improved algorithm, and the results of images and dynamic video images's face detection were analyzed. Experiments show that the algorithm achieved a better face detection, and has a higher detection rate and low false detection rate, while detection of rotation angle of the human face has a certain degree of robustness. In the face recognition research, this paper mainly studies the face recognition based on Hidden Markov Model, HMM introduces the basic ideas, including the basic concepts of hidden Markov, and then analysis of the hidden Markov basic algorithms, and then analyse continuous hidden Markov model, the hidden Markov model for the selection of initial and training are searched in the final.