CAJ | 학술논문

传统的AdaBoost人脸检测算法训练时间长对设备要求高,在复杂背景下存在漏检误检.提出一种基于小波重构和特征提取的AdaBoost人脸检测算法,并应用到身份验证中.采用小波重构方法,实现对人脸信息有用特征的重构,进行去噪处理提高人脸识别的准确性.采用人脸特征关联性方法将不同的人脸特征子集进行分类处理,采用特征提取算法实现对AdaBoost算法的改进.仿真结果表明,采用改进的人脸检测算法进行身份验证,能检测到一定范围内的非正脸图像特征,有效提取人脸的局部信息特征点,提高身份验证对象的检测精度和正确识别率.
전통적AdaBoost인검검측산법훈련시간장대설비요구고,재복잡배경하존재루검오검.제출일충기우소파중구화특정제취적AdaBoost인검검측산법,병응용도신빈험증중.채용소파중구방법,실현대인검신식유용특정적중구,진행거조처리제고인검식별적준학성.채용인검특정관련성방법장불동적인검특정자집진행분류처리,채용특정제취산법실현대AdaBoost산법적개진.방진결과표명,채용개진적인검검측산법진행신빈험증,능검측도일정범위내적비정검도상특정,유효제취인검적국부신식특정점,제고신빈험증대상적검측정도화정학식별솔.
The traditional face detection algorithm needs much training time, and has high requirements for equipment,and there are false detection in complex background. An AdaBoost face detection algorithm is proposed based on wavelet recon-struction and feature extraction, and it is applied in identity verification. By using the wavelet reconstruction method, the re-construction of useful face information features is obtained, denoising processing is used to improve the accuracy of face recognition. The facial feature relevance method is used, and the different features of face are classified in different sets. The feature extraction algorithm is used to improve the AdaBoost face detection algorithm. Simulation results show that,the new method can detect the non frontal face image features within a certain range, and the detection precision and accurate recognition rate are improved.