基于改进SRCNN算法的人脸识别技术研究

Research on face recognition technology based on improved SRCNN algorithm

针对因硬件或者网络传输导致图片质量不佳的情况下的人脸识别方法及其应用。以传统的超分辨卷积神经网络(SRCNN)算法为基础,针对原算法存在的差值误差计算量大的问题做出了改进,提出了改进的SRCNN网络模型,并引入了General-100训练集对改进的网络模型进行训练。最后通过在不同测试集上和其他算法进行对比,得到了改进的SRCNN算法在不同的上采样倍率条件下性能都优于双三次插值法和SRCNN算法的结果,证明了算法的适用性和优越性。

The face recognition method and its application in the case of poor picture quality due to hardware or network transmission.Based on the traditional super-resolution technology and the SRCNN algorithm based on convolutional neural networks,this paper has made improvements to the problem of large amount of calculation of the difference error in the original algorithm,proposed an improved SRCNN network model,and introduced General-100 training set to train the improved network model.Finally,by comparing with other algorithms on different test sets,the performance of the improved SRCNN algorithm is better than the results of the bicubic interpolation and SRCNN algorithm under different upsampling magnification conditions,which proves the applicability and superiority of the algorithm.