The Research And Implementation Of Face Detection And Location Algorithm In Face Recognition System

Face recognition is a subject that have great value of academic research and extensive application . It involved in multi-disciplinary knowledge, for example, image processing, pattern recognition, cognitive science. There are a wide applications,such as identity, social security, e-commerce, etc.The main contents of this paper is the face detection and localization algorithm in the face recognition system. It is the early stage of face recognition systems. A set of algorithms based on Adaboost face detection and localization were designed and achieved. The algorithm is machine learning algorithm which based on the statistical feature of the face's gray-scale distribution. The pigment between face's organ and the surrounding skin are extracted as feature. The combinations of these difference can be used to distinguish between face and non-face. These difference can be described with several rectangular characteristics. The algorithm is divided into two phases: the mission of training phase is the training samples (face samples and non-face of the sample) and formating the classifier in certain conditions of the probabilistic distribution. These characteristics were translated into a weak classifier after the rectangular characteristics of the sample were extracted in the process. Some optimal weak classifier were choosed as a strong classifier through the learning algorithm, and then a number of strong classifier were combined into a cascade classifier; The mission of detection phase is to use the cascade classifier to classify face images and non-face images. The region of the face were located if the face be found in the image.In the paper, the process of the design and implementation based on face detection algorithm have been discussed, and the process that how to design classifier which is from weak classifier to strong classifier to cascade classifier based on the training samples have been explained. The test process of cascade classifier have been discussed, too. Ultimately, the effective cascade classifier were achieved. When we use the detection system to detect a large number of real images, we found the detection speed is very fast and high efficiency. The face detection system has certain practical value. It has been able to meet the needs of the next step of the face recognition system.