当今人脸识别技术的现状与未来研究方向

Current Situation and Future Research Direction of Face Recognition Technology

近些年来,人脸识别成为了热门领域。从最初的传统人脸识别方法,到现在的基于深度学习的人脸识别模型,人脸识别技术的性能得到了巨大提升。然而,在图像信息受噪声或者受光照、模糊、侧脸、遮挡等现象影响较严重的情况下,其图像处理效率以及人脸匹配正确率依然有较大的提升空间。重点综述了当今基于深度学习的人脸识别模型,报告了当今人脸识别技术的现状。最后,对人脸识别技术未来研究方向进行了展望,认为未来应着重解决卷积神经网络的局部最优问题、训练数据集的“深”“宽”问题和最佳网络结构问题。

In recent years,face recognition has become a hot field.From the original traditional face recognition methods to the present deep learning based face recognition model,the performance of face recognition technology has been greatly improved.However,in the case where the image information is seriously affected by noise or illumination,blur,side face,occlusion,etc,the image processing efficiency and the correct rate of face matching still have a large room for improvement.This paper focuses on the current face recognition model based on deep learning and reports on the current situation of face recognition technology.Finally,the future research direction of face recognition technology is prospected,and it is believed that the local optimum problem of convolutional neural network,the"deep"&"wide"problem of training data set and the optimal network structure problem should be focused on in the future.