Research On Face Detection And Recognition Technology Based On Adaboost Algorithm

In many biometrics,face recognition technology is more friendly,simple,accurate and fast.Face recognition technology involves areas,and has great research value.Face recognition technology is also widely used,such as video surveillance,corporate access control and so on.Therefore,it is.very necessary to invest in the research of face recognition technology.Face recognition technology is divided into two stages of face detection and face recognition.This paper makes an in-depth study on the two stages of the algorithm,and improves it,and has achieved good results.In face detection stage,I studied face detection based on Adaboost algorithm.Before face detection,face image is preprocessed.Histogram equalization is used to process the image,improve the image quality and filter the processed image.Adaboost algorithm extracts features of Haar,while Haar features can well describe facial features.During the training process of Adaboost algorithm,the training of erroneously judged samples is increased,which improves the detection ability.The last cascade classifier is a cascade classifier which is cascaded from small to large according to the ability of classification.The area that is easily classified as non face will be eliminated quickly,only a few face areas are left,which greatly improves the detection speed.The experimental results show that,in general,the detection effect of face detection based on Adaboost algorithm is ideal,but in some cases,the problem of misdetection and missing detection will appear.So the face detection based on Adaboost algorithm is improved,and the skin color area is segmented from the background before face detection.Because skin color has good clustering in YCbCr color space,the image is transformed to this space.The segmented skin area is processed by morphological processing to remove the black box and connect the disconnected connected area.Finally,face detection is carried out on the skin area.Experimental results show that the improved algorithm improves detection rate and reduces false detection rate.In the face recognition stage,we improved the local two valued pattern(LBP).LBP can describe texture features well,and has good robustness to illumination change and image rotation.However,a large amount of computation slows down the speed of recognition.PCA can reduce the dimension of the feature,but does not consider the category information,easily affected by the illumination;LDA considers the category information,but can not be able to find the mapping space.Fisherfaces first uses PCA to reduce the dimension of features,and then maps to LDA space to achieve face recognition.It not only reduces the amount of computation,but also greatly reduces the influence of illumination.It is based on this idea that after extracting LBP features,we use the PCA method to reduce the dimension of features to achieve face recognition.Experimental results show that the improved LBP algorithm preserves inherent advantages and reduces computation.