The Research Of Face Recognition Technology Based On DCNN In Authentication

In order to prevent and investigate the corrupt practices in the entrance examination,we will further ensure the fairness and impartiality of the examination.Based on the analysis of the existing traditional authentication system of the higher education admission office of Henan Province,this paper proposes a solution of authentication about the face recognition technology based on DCNN and focus on two key issues: optimal design of DCNN model and threshold determination.Based on the authentic data sets of merit candidate by analyzing the scale and characteristics and practical application of the candidate data set,this paper designs a more expressive and applicable deformation network Structure GoogLeNet-D(Deformation)by means of the design idea of GoogLeNet and the related optimization method of DCNN.The model consists of 34 layers and training parameters about 8.9M,and finally outputs 128-dimensional vector as the candidate's face feature vector.In order to evaluate the GoogLeNet-D model and set a reasonable threshold to determine whether the candidate is the same person,the paper which is on the basis of the number of positive samples in the candidate data set being much larger than the negative samples number,makes use of the real positive samples directly in the evaluation model.The precision of face search / classification is used as the basis for the model evaluation.However,the evaluation method is unable to determine the threshold,this paper further proposes a direct,simple and effective algorithm to quantify the threshold,and determine the threshold when calculating the precision.According to the experiments results,Goog LeNet-D has a good ability to express facial features,which used 10,406,024 face data of 1.7 million candidates in 2014-2016,and achieved 98.87% of the face classification accuracy in the test set which includes 200 thousand candidates with 1,022,031 face data.Finally,the optimal threshold of the model is 0.35.At present,the model has been successfully applied in college enhance examination and other entrance examination in Henan Province.