基于改进高斯模型的人脸识别

Face Recognition Based on Improved Gaussian Mixture Model

针对当前人脸识别算法对光照变化、姿态变化鲁棒性差的局限性,提出了一种改进混合高斯模型的人脸识别算法。首先对采集的人脸图像进行预处理,消除光照等外因素的不利影响,然后分别提取多种人脸图像的特征,最后采用混合高斯模型对人脸进行识别,并采用多个人脸数据库进行验证性实验,结果表明,本文算法对光照变化、姿态变化具有较强的鲁棒性,获得比经典人脸识别算法更优的识别结果,在人脸识别中具有广泛的应用前景。

In the view of the limitations of the current face recognition algorithm for illumination changes and pose variation robustness,an improved face recognition algorithm based on hybrid Gauss model is proposed. Firstly,the face image acquisition is pretreatment to eliminate the adverse effects of external factors such as illumination,then the characteristics of a variety of face images are extracted,finally,the Gaussian mixture model is used to carry out face recognition,and validation experiments are carried out by using a lot of standard face database. Results show that the proposed algorithm has stronger robustness changes in illumination and pose variation and recognition results are better than the classical face recognition algorithm,has broad application prospects in face recognition.