Study And Implementation Of Face Detection Based On Adaboost Algorithm

Face detection use for searching automatically in an input image and judging whether or not faces exist in the image, if faces were found, giving the location and size of each faces. Face detection is treated as a fundament and important research problem in Computer Vision and Pattern Recognition; it can be used in many fields, such as video surveillance, face recognition and human computer interface, etc. In the last few years, with the rapid development of large amounts of computer technology, new technologies and new methods are widely used, which get more vitality to the research of human face detection, and the result of detection become more and more better and faster than before. The primary works and contributions in this thesis are showed as follows:(1) This thesis analyzes the AdaBoost algorithms systematically according to the principle of AdaBoost; the research start from the classical adaBoost algorithm, which was optimized from training process. Based on the idea, this thesis proposed the concept ’Auxiliary Classifier’ which was used to improve the detection rate of face detection, this method can be used to improve dection rate through experiment result.(2) Weak classifier, strong classifier are introduced in detail in this thesis, based on the understanding of AdaBoost algorithem, this article builds a face detection system while in-depth research on face detection algorithem, it can be used for verification of methods;(3) For the disadvantages of classical AdaBoost algorithems:for the problem of the training process of the algorithm consumed such a long time, the number of features and the category of training set were optimaized respectively, the training time was reduced dramatically;(4) In the classical AdaBoost algorithems, training consume a lot of time, in order to solve this issue, after a round of training, it does not update the weights of sample but does training of next round based on current classifer, Test result show that this algorithm gains higher training efficiency and saves a lot time;To verify the proposed method, experiments are committed on common face database. The test result show the classification accuracy of the proposed methods is better than classical adaBoost algorithm. It’s worth mentioning that the concept ’Auxiliary Classifier’ which was used to improve the detection rate of face detection, this method can be used to improve dection rate, at the same time, the speed was down a little. This article will be a good foundation for further research and application.