Research On Face Recognition In Complex Scene Based On Deep Learning

In recent years,with the improvement of people's living standards,people's awareness of security has become higher and higher.However,since most of the monitoring equipment in our country is the monitoring equipment installed a few years ago,the resolution of the recorded video images is very low,and it is difficult to distinguish the identity of the criminals through it,which gives many criminals the opportunity to carry out illegally activity.It is really need a system which can easily identify the position of the face and lock identity at low resolution.Therefore,this paper introduces technologies related deep learning into the monitoring system,including super-resolution image reconstruction based on GAN network,face detection based on convolutional neural network,and face recognition based on capsule neural network.And the related key technologies are further studied in depth.Super-resolution reconstruction is performed based on the SRGAN network,and the discriminator is redesigned for the SRGAN network,which retains the spatial position information for the reconstructed image,which lays a solid foundation for the subsequent processing of the image.Based on the Faster R-CNN network,drawing on the improved ideas of the DSFD network,the Faster R-CNN network is optimized to improve the detection effect of the face detection system on the blocked face,and tested on the widerface of the blocked data set.Verify the effectiveness of the designed system.For complex scenes,it is difficult to extract faces with more than one picture,and traditional deep learning methods are all supervised learning algorithms,which require a large number of data sets with labeled photo faces,but the actual conditions cannot be met.This article uses the third generation of stacked capsule self-encoders developed by capsule neural networks,combined with the most popular Transfomer to develop an unsupervised face recognition system.This article uses the Tensorflow deep learning framework,after verifying the rationality of the model,it is packaged to the Flask side,and the research results are demonstrated with the help of the We Chat applet platform.