Research On Facial Biometric Recognition Based On Partial Information

With the development of the computer science and the popularity of the Internet technology,the requirements for verifying or recognizing the individual's accurate identification and judging differentacitions become more and more emergency.Therefore,there are more and more needs for the human biometric identification technology,and higher and higher requirement for the precision of identification.Among the various successful biometric identification technologies,such as face recognition,iris recognition,fingerprint recognition,speech recognition and DNA recognition,face recognition has its several outstanding advantages,: the simple and convenient acquisition of the face image,no constraint for the acquistion distance,no need for the user's cooperation,and the collecting environment is natural.Furthermore,face can provide abundant information of the individual identity such as race,gender,age and expression.Thus,the face recognition has extensive demands and prosperous application in real life.In general,with aforementioned advantages and broad application prospect,face recognition has always been a hot research topic in the field of pattern recognition.Lots of effects have been devoted to the basic theory and real applications of face recognition in the last several decades.The dissertation mainly studies on the followings problems of the face recognition technology.The main contents of this dissertation are summarized as follows:1.The proportion between the facial features is analyzed statistically.In order to explore the realtions of facial features in human face organs,this dissertation firstlocates the key points of the facial features,and then figures out the proportions of the width of the coners of the eye,nose and mouth respectively.It is found that the proportion is relatively fixed for each individual.To be more specific,the ratio of the distances between the eyes,nose and mouse of each endividual is always stable.This would make the reference for the facial recognition technology in different settings such as the pose-varied face recognition,partial-occlusion face recognition,the light effects and the impacts of the plastic surgery.2.A new face recognition method based on facial texture feature aided deep learning feature is proposed.To combine the distinctive feature of the facical information and the outstanding adavantage of the deep learning in the visual perception system,this dissertation fuses the texture features abstracted from the regions of the eye,nose and mouth with the deep learning features,to feed them into the target function layer of Convolutional Neural Network,which makes the features of the deep learning self-regulated to integrete with the texure features of the facial organs,and then to construct the more efficient facial features.The experiment results demonstrate the effectiveness of the proposed features.3.A new deep learning based age estimation method aided by the facial features is prsented.In the context of cross temporal face recognition,the age is the most influential facor for the recognition effect.In recent years,the research of the face-image based age estimation gains attentions of more and more researchers.This dissertation makes the patially aligned face image block generated by the key points of the face as the input of the Convolutional Neural Network,and fuse the classifier design method with the traditional facial regional features abstracted and the end-to-end sorting method based on the deep Convolutional Neural Network,directly to deal with the age estimation from pixel points of an image,and to reinforce the generalization ability of system model.For adopting the multi-scale network structure,it highly enhances the performance and obviously promotes the estimation effects.4.The idea of identify a person by the nose region based on Gabor Kernel Fisher Discriminant Analysis method is presented.The dissertation analyzes the physiological structure of the nose and points out that the nose is the slightest influenced facial region for the factors of expression,age,shapes,cosmetic and shadow.The dissertation puts forward the feature and avialibility of the facial recognition by the nose region,and simplifies the Gabors' parameters of the nose features.Further,the recognition effects of the nose regions are experimented in different data.The results show that the recognition effects of the nose region is better that the parts of the facial recognition,which provides the guidance and reference for the research on performance enhancement under the situations of the changes in facial expression,posture and age.