The Research On Face Detection And Recognition Technology Based On Adaboost Algorithm

Modern society’s fast development requires identification more stringent on the rapidity, accuracy, safety and other aspects. While the rapid development of the biometric identification technology makes it all possible. In these technologies, the face detection and recognition technology develops fastest. Relatively speaking, this method is more direct, friendly, convenient and more widely used in production practice.Nowadays, identification system plays an important role in many fields, and its market prospect is very wide. The research on face detection and recognition system which is more quickly and effectively has very important significance to the development of the society, public security, national security and other aspects.On the basis of summarizing and analyzing domestic and oversea relevant research fruits, this paper does a research which goes deep into the algorithm of face detection and recognition and continues to improve, so that the face detection and recognition system has the corresponding improvement on the accuracy and speed. The focus of this paper is divided into the following three parts.(1) In order to achieve better image quality, the image needs to be preprocessed before detection and recognition.Here the image enhancement technology is mainly used. The image to be detected is used a method of histogram equalization for processing, which makes the contrast enhanced, local details more prominent for subsequent detection.The facial image to be recognized is used a method of illumination compensation for processing, so the too dark or bright facial image’s quality can be improved obviously for subsequent recognition.In addition, the image segmentation technology commonly used in image processing is briefly introduced, and at the same time, the OpenCV technology used in this paper is also detailed.(2) This paper studies the face detection algorithm of Adaboost based on static and video images, which mainly includes the selection and calculation of Haar features, and the design and construction of classifier. In the process of algorithm implementation, update the rules of samples’s probability, which makes the training pay more attention to the error samples. In order to make the detection’s effect better, this paper proposes an improved face detection algorithm of Adaboost based on skin color segmentation. Firstly, the image is converted to YCbCr color space. Segement the region including skin color using a good clustering of skin color in this space. Then, get the skin region after morphological and smoothing processing. Finally, use Adaboost algorithm for face detection. After verification, the improved algorithm has a good effect.The detection rate and speed is improved, and the false detection rate is reduced effectively.(3) On the face recognition stage, this paper studies the face recognition method based on Fisherfaces and LBP. The method of Fisherfaces is composed by the method of principal component analysis (PCA) and linear discriminant analysis (LDA), which is on recognition from the perspective of the overall image. The algorithm of local binary pattern (LBP) is from the perspective of the local image, which constitutes the feature vector for recognition through the features of each sub block image. A face recognition algorithm combined with LBP and Fisherfaces is given in this paper after summarizing the advantages and disadvantages of Fisherfaces and LBP algorithms.This algorithm not only keeps the advantanges of LBP’s less training samples and strong robustness to illumination change and rotation, but also uses Fisherfaces’s thoughts of dimensionality reduction and classification to reduce the amount of computation and complexity and improve the accuracy of recognition.After verification, the improved algorithm compared with other algorithms has a higher recognition rate and better robustness.