Analysis And Research Of Facial Attractiveness Based On Facial Structural Features And Skin Texture Features

Facial attractiveness is an important proposition in the field of genetic psychology and cognitive psychology.The research results have great scientific significance for face evolution,and even human evolution.This paper combines traditional cognitive psychology with artificial intelligence technology to conduct a cross-research on facial attractiveness,which can improve the shortcomings of the traditional research work such as strong subjectivity and single features.This paper uses existing machine learning methods to obtain the quantitative standard of facial attractiveness,and conducts in-depth research and analysis on facial attractiveness evaluation methods based on facial structural features and skin texture features.The specific research work and achievements are as follows:(1)A block adaptive weighted LBP histogram matching method is proposed to predict the facial attractiveness based on skin texture features.This method matches the test sub-block with the training sub-block according to the similarity of the LBP histogram,and the weight value of each test sub-block is determined by the information entropy contained in it.On the face image database used in this paper,it is experimentally demonstrated that the optimal number of blocks of this method is 64.Under this parameter,it can obtain more advantageous prediction performance of facial attractiveness.(2)The 79-point facial contour feature and the 24-dimensional geometric feature are designed to represent the facial structural features,and lasso algorithm is used to select the two features and their combination features.Machine learning algorithms are used to evaluate the performance of facial attractiveness prediction of extracted facial structural features.The experimental results show that these two features designed in this paper show better prediction performance compared with other features,and the combination of facial structure features after feature selection can further improve the prediction performance.(3)The facial attractiveness evaluation method using facial structural features and skin texture features are combined.The specific method is to fuse the scores of the machine scores obtained using these two facial features.The experimental results show that scores fusion calculation model based on facial structural features and skin texture features can greatly improve the prediction performance of facial attractiveness.In the self-collection student face database and the middle-aged and elderly face database established in this paper,the best prediction performance is obtained when the fusion weight ratio is 6: 4 and 3: 7,the Pearson correlation coefficient is 6.62% and 7.35% higher than the experiment performed by the two features separately.