Face Position Detection And Application Research Based On Static Image

Biological recognition technology is a kind of intelligent machines to simulate a technique to detect authentication, including facial recognition technology can use face of physiological or behavioral characteristics to detect faces in the images of the position or identify the identity of the person. Because of the erratic face shape and the influence of external factors such as light, environment, and from the point of the research status of face recognition technology at present, there are still shortcomings within existing algorithms, because of depending on the purpose and background adjustment of the existing algorithms. Human face is an important biological characteristics of human body, so face detection plays a very important role in face recognition, and the accuracy of face recognition can be effectively improved by correct face positioning. With the development of information technology, biological recognition technology application field is wider and wider, so face recognition algorithms need to do in-depth research.A variety of face recognition algorithms are studied in this paper, mainly focused on the AdaBoost algorithm and 2DPCA algorithm. The advantages and disadvantages of AdaBoost algorithm is analyzed first, and improved the weak classifier of two different samples. And then combined the 2DPCA algorithm with AdaBoost algorithm,with AdaBoost learning algorithm which makes more weak classifiers form a strong classifier.AdaBoost algorithm is quite popular in machine learning, and is a kind of Boosting algorithm of polynomial level. Implemented by changing the data distribution according to the classification of the training set for each sample if it is correct, and the last time the overall classification accuracy, and change the weights of each sample to determine the next step. Due to the image in the face of torsion angle is different, the complexity of the image background is different, and the results of detection rate and false detection rate will be very different. In this article, the third chapter is put forward based on the Haar-like AdaBoost face detection algorithm,which combines improved characterization of Haar-like templates, the main purpose of which is to represent a certain degree of traverse of the face, combining AdaBoostalgorithm, training the classifier, and then through weighted cascade classifier,stronger combined into classification ability of cascade classifier.Principal Component Analysis(PCA) as one of the most successful linear differential method, is still widely applied to the face process, such as digital image processing field. AdaBoost is a kind of machine learning algorithm with adaptive learning ability. In this paper, the fourth chapter puts forward the combination algorithm is a kind of face detection based on the 2DPCA-AdaBoost algorithm, and combination of two-dimensional principal component analysis with AdaBoost improved the original algorithm. The 2DPCA algorithm is used to training the face images first in order to form the training characteristics face space, and then combined with AdaBoost learning algorithm which makes more weak classifiers form a strong classifier. The experimental results indicate that the reformed algorithm improves the detection rate and reduces the number of error detection as well.