基于训练集重组的人脸识别算法研究

Based on the Training Set of Recombinant Human Face Recognition Algorithm Research

传统的人脸识别算法都要依赖于面部特征的严格配准来归一化人脸以便提取人脸特征,而在非限定条件系统中,传统识别算法提取特征偏差较大,且容易受姿态、表情、光照等噪声干扰。基于此,提出先对训练图像进行特征配准,再重建训练样本用于识别。实验结果表明,本方法较传统的识别算法具有较高的识别性能。

The traditional algorithms for face recognition is aim on face images that are accurate alignment and have been normalized, unfortunately, face features are extracted inaccurately in non-qualification system by the traditional algorithms. And the traditional algorithms are sensitive to noise such as posture, facial expressions, lighting etc. This paper presents the training face images should be registered firstly and then used to feature extraction. Experimental results show that the method presented in this paper performs better than the traditional algorithm.