| With the rapid development of computer vision,the application of computer vision technology becomes more and more important in daily life.Biometric technology has become an important way of personal identification,and it is also a very active research topic.As an important branch of biometric,face verification is easy to be accepted by users and has a wide range of applications in smart office and criminal detection and other fields.Learning the current situation of face verification at home and abroad,We have studied the face verification technology under the face recognition based on the structure of the convolutional neural network under the framework of deep learning.The main works have been done in the thesis are as follows:(1)We have studied the structure of convolutional neural networks based on the framework of deep learning.A method mixed characteristics of convolutional neural network with Gabor is proposed.Abstract feature of the image is extracted gradually when it is input into the trained network model.At the same time,the image is processed to extract feature by Gabor transform.Two sets of feature maps are combined into a new feature vector.And the feature vector is calculated by cosine similarity.Then,the value is input into the trained SVM to classify.If the output is 1,the person is considered to be the same person,otherwise;the person is different person.(2)A Siamese network model is proposed for the problem of face verification with two class.In general,a model of deep learning is used as classifier to recognize the picture.The application of the Siamese network is to input a pair of face images and measure the similarity of the two images.After inputting the two pictures into the network model,the network is transformed into a nonlinear and more easily differentiated two-dimensional space from the original spatial distribution mapping to the similarity measurement.It can be classified the two pictures belong to the same or different class by the similarity of vectors.Compared with other algorithms,the results show that the algorithm has a higheraccuracy.Meanwhile,the above method is applied to the real Environment.Matlab is used as the simulation platform to implement a face verification simulation system based on cascade deep neural network.Meanwhile,the DNN module is compiled in OpenCV to use convolutional neural network model.Using the Qt as the development platform and C++as the development language,the application of face verification system based on cascade deep neural network is realized. |
