基于深度卷积网络算法的人脸识别方法研究

Based on the Depth of Theconvolution Network and Local Binary Pattern of Face Recognition

由于人脸图像在采集过程中,容易收到光照等环境影响,使得人脸特征存在突变性。传统的识别方法主要通过采集人脸特征进行人脸识别,对图像清晰度要求很高,针对模糊图像不能及时进行人脸特征采集,导致人脸识别不准确的问题。提出深度卷积网络算法的人脸识别方法。首先要用局部二值算法提取人脸局部纹理特征,对深度卷积网络模型进行构建,并利用卷积网络共享权值和池化、下采样等降低模型的复杂度。在模型的顶层形成人脸图像特征分类面,得到完成好的深度卷积网络模型,利用该模型对人脸图像进行特征提取,有效的完成了人脸的识别。实验结果很好地证明了利用深度卷积网络算法的人脸识别方法对人脸特征表达效果良好,显著提高了人脸识别的准确率。

Facial image is easily affected by light and other environmental impacts during the process of gathering. It makes face feature have mutability. Traditional method identifies face by gathering facial feature. It has high requesting to image resolution. It results in the face recognition inaccurate because it cannot gather fuzzy image timely. This paper proposes a face recognition method based on the deep convolution network algorithm. Firstly, the partial textural feature is extracted to build deep convolution network model. The shared weight, pooling and down-sampling of the convolution network are used to reduce the model complexity. The feature classification surface is formed at the top of the model Then the deep convolution network model is obtained. Finally the model is used to extract the feature of face image and the face recognition is effectively completed. The simulation results show that the modified method has better effect of expression than traditional method. It can improve accuracy rate of face recognition apparently.