Seminar face recognition system began in the 1960s, after 80 years of follow conduct computer skills and optical imaging skills improved, and true into the primary use of the stage in the late 1990s, and the United States, Germany and Japan, skills the completion of the main; the key to success lies in whether face recognition system has the top center of the algorithm, and the results have to identify practical identification rate and identify speed; "face recognition system" integrates artificial intelligence, machine recognition, machine learning, model theory, expert systems, video picture processing and other professional skills, together with the need to combine theory and complete processing of intermediate values, is the latest use of biometric identification, completing its center skills, it demonstrated weak to strong AI artificial Intelligence transformation.
Face recognition system comprises four main components are: collection and drawing the face detection, face picture preprocessing, feature extraction, and drawing the face matching and recognition.
Face painting collection and testing
Face painting collection: different face pictures can be collected down through the camera lens, for example aspects of static pictures, dynamic pictures, different orientations, different facial expressions, etc. can get a good collection. When a user within range to take pictures of the collection device, the device will automatically find and collect user's face photographed picture.
Face Detection: Face Detection in practice for pre-primary face recognition, that is in the picture accurately calibrate the face orientation and size. Formal features face painting contains very rich, such as histogram features, color characteristics, template feature, structure and Haar features and so on. Face detection is to put useful information which singled out, and use these features complete face detection.
River face detection approach chosen Adaboost learning algorithm based on the above characteristics, Adaboost algorithm is a classification approach used, it put some weak classification together, the combination of strong new classification.
Face detection process using Adaboost algorithm to pick out some of the most representative features a rectangular face (weak classifiers), in accordance with the weighted voting approach of the weak classifiers into a strong classifier structure, and then train some strong classifier obtained series to form a cascade structure laminated classifier effectively improve the detection speed of the classifier.
Face painting pretreatment
Face painting Pretreatment: Pretreatment on drawing the human face is based on the results of face detection, processing and final picture of the service process for feature extraction. Because the system was acquired original picture and randomly disturb constraint conditions often can not be directly used, it must be a gray calibration, noise filtering, etc. at an early stage pretreatment picture drawing processing. On the face picture, the pre-treatment process which includes a human face picture primary light compensation, gradation conversion, histogram equalization, normalization, dash proofing, filtering, and sharpening.
Face picture feature extraction
Facial feature extraction drawing: face recognition system can use visual features into the feature general, pixel calculation features, facial features picture transform coefficients, facial features, etc. algebraic drawings. Facial feature extraction is for certain facial features carried out. Facial feature extraction, also known as the face representation, it is the human face feature modeling process. Facial feature extraction approach summed up into two categories: one is based on common sense approach to characterize; the other is based on algebraic features characterize the way of learning or computing.
The primary basis for characterizing common sense approach is based on the shape of human faces depicting characteristics of the organ and the spacing between them to get the help of face characteristic data classification, which generally includes the weight of the Euclidean distance, curvature and viewpoints between feature points, etc. . Face consists of eyes, nose, mouth, chin portion, these portions of the structure depicted in dash and links between them, can be identified as an important feature of the human face, the features referred dash feature. Based on knowledge of the primary face representation basis dash features include approaches and template matching.
Face picture identification match
Face painting to match the recognizable: the characteristic data extracted face features template with pictures stored in the database to find matches, after setting a threshold value, when the degree of similarity exceeds the threshold value, then the output match to get results. Face recognition is the facial features to be identified have been compared with people face features template, face identity information to discriminate according to a similar degree. This process is divided into two categories: one is to admit that the process of drawing is one to one comparison, the other is to identify, carried a picture-many matching process cf.