摘要
人脸识别技术常遇到姿态、光照影响等问题,针对测试样本的姿态变化对人脸识别的影响,主要的研究工作如下:利用多项式变换增加虚拟样本,通过增加训练样本提高识别率。在增加虚拟样本后,使得基于线性判别准则的方法对单训练样本的人脸识别问题也可以使用。
Facial recognition often meets with such problems as pose and illumination variation.The paper explores the effect of pose change on facial recognition.Polynomial transformation is used to generate virtual samples,and the recognition rate can be improved through increasing training samples.With the increase of virtual samples,the subspace algorithms can be effective for facial recognition based on single training sample.
出处
《金陵科技学院学报》
2010年第4期27-31,共5页
Journal of Jinling Institute of Technology
关键词
多姿态人脸识别
人脸姿态校正
虚拟样本
pose-varied facial recognition
facial pose correction
virtual samples
