摘要
Abstract
A recognition method based on self-organizing map(SOM)is proposed to improve the performance of face recognition in severe environments such as complex background,changeable illumination,occlusion and posture change in real-time video surveillance.Firstly,the AdaBoost algorithm is used to accurately locate face regions in the complex backgrounds,reducing background interference.Secondly,the face region is corrected with the local binary pattern(LBP)to eliminate posture bias.Finally,the SOM is used to map high-dimensional LBP features to a low dimensional space,and the face recognition is achieved fast and accurately by unsupervised learning.The experimental results show that the false recognition rates of the proposed method is significantly lowered than that of the comparison methods in normal lighting,low lighting,occlusion,posture changes,and complex backgrounds.In the case of low lighting and occlusion,its false recognition rates are reduced to 0.06 and 0.08,respectively,demonstrating stronger robustness and adaptability.This method can effectively recognize human face in complex backgrounds and provide reliable technical support for intelligent management of universities.关键词
视频监控/复杂背景/人脸定位/人脸校正/人脸识别/AdaBoost/SOMKey words
video surveillance/complex background/face localization/face correction/face recognition/AdaBoost/SOM分类
电子信息工程
