The Research On Face Recognition Of Sichuan Golden Monkey

As a biometric recognition technique,face recognition has been studied over 40 years.Initially,it is studied based on human face.With the development of facial recognition techniques and expansion of its application field,people have gradually applied the techniques to other animals,such as chimpanzee,golden monkey etc.Using face recognition technique to differentiate individual within specific animals can significantly help biologist promoting their research by collecting information more accuracy.It's also very useful to help people to protecting endangered species of animals.Sichuan golden monkey is a rare animal unique in China,and a primary protected animal.In the paper,take the Sichuan golden monkey as research object,we will analyze the difference between human face and Sichuan golden monkey face,and design appropriate face recognition algorithms according to different practical scene and its texture feature,geometry feature and structure feature etc.1.Face recognition of Sichuan golden monkey based on Locally Enhanced LBP In this paper,the Locally Enhanced LBP(LE-LBP)is proposed under the situation that the quantity of face image of Sichuan golden monkey is usually small,the texture feature of face hair is changeful due to illumination and wind,and the high similarity of facial appearances between individuals,especially those of the same age of the same sex.The technique mainly focused on the facial skin region of Sichuan golden monkey to avoid interfering of variable hair texture and reduce the overall feature dimensionality.For the skin region with similar appearance,Faster R-CNN is firstly used to detect and segment the facial regions.The segmented result is regarded as the interesting regions that the proposed method focused on.Secondly,the texture complexity of each region of interesting is calculated,and the result is as the basis whether the regions could be divided.Finally,we extract the LBP features of all the regions and cascade it.Compared with the traditional LBP feature,this feature can better reflect the detailed features of the face of Sichuan golden monkey,and more accurately complete the facial recognition of Sichuan golden monkey.2.Face recognition of Sichuan golden monkey based on SPL-BCNN We proposed a SPL-BCNN algorithms applied to face recognition of Sichuan golden monkey in the situation that the size of image dataset is large,and recognition accuracy rate is low using the traditional methods because of variations of pose,expression and lighting etc.By combining the Self-Paced learning and Bilinear Convolutional Neural Network,the proposed algorithm can be trained gradually in the order from easy to complex sequence based on Bilinear Convolutional Neural Networks analogous to human learning process from simple to complex knowledge.So the resulting model can avoid that the training process fall into local optimum,and achieve better recognition accuracy rate and generalize performance.