Channel Face Recognition Surveillance Alarm System Design And Implementation

With the researches on automatic face recognition technology continues to increase, automatic face detection, practicality growing recognition system for automatic face recognition systems of various specific environments gradually practical. Application of these systems in urban security played a growing role. Combining face recognition and video surveillance, can be widely used in airports, train stations, border crossing, large-scale exhibition, a high level of security in public places. In this paper, the channel type face recognition surveillance alarm system has been designed and implemented, mainly inside include:1. for the development of face recognition technology has been studied. Summarizes the current status of the main methods of face recognition technology and face recognition technology at home and abroad, for face recognition system in the field of security classified, face recognition technology in the future foreseen broad market prospects, and lists of people Face recognition technology issues and challenges facing the stage.2. studied in this paper Face-channel monitoring alarm system overall design. Combined with practical application environment, using the embedded DSP devices proposed for face detection to improve the usefulness of the system, as a basis for the system architecture design, system requirements for analyzes, lists the system performance and design functional modules of the system.3. face detection and location technology design and implementation. According to the basic principles of recognition, proposed using AdaBoost face detection algorithm, the first use of gray size normalization and normalization of image pre-processing, generate classifier through face samples and non-face samples for training, Then according to the classifier to detect the input image, the output of the detected face photos, and conducted AdaBoost algorithm; proposed based on random forest face critical point positioning method, the Random Forest classifier design face critical point and implement this algorithm.4. The method of face recognition based on multi-feature local and global integration were studied. Proposed face recognition method using multi-feature fusion, image preprocessed using Gabor filters to extract facial feature grayscale, using HOG operators face shape feature extraction, using LBP operator to extract facial skin texture, then sub-region of space using direct reuse existing global characterize, after falling by PCA+ LDA dimension, in the final score layer fusion method, and tested on FRGC v2.0 database.5. the channel-type face recognition technology to monitor critical alarm system has been studied and realized. Describes the development environment used by Visual C++ 6.0, OpenCV2.4.10 and ccs5.2, DSP devices and hardware configuration, system implementation, using grayscale image, image histogram equalization for image pre-processing, design face detection core function, call the function to calculate the eigenvalues and so achieve the core code, and finally the realization of the system was tested. Illumination for 300-900Lux cases, face detection rate reached 80%, targets to correctly identify 69%, meets the design requirements and proposed analysis improvements.