Research On Facial Skin Quality Detection And Evaluation System

As an important sign of personal image and temperament,along with the development of people's consciousness of beauty and skin care,various products related to the testing of human face skin quality have also started to be gradually accepted by the market.Aiming at the problem that the function of skin testing instrument based on a single sensor design on the market is relatively limitd,and that the large-scale skin testing equipment with complete functions is costly and bulky,this paper,on the basis of he idea of "hardware + software",designs a personal skin testing and assessment system of daily skin beauty,and realizes the target of accuracy of testing and evaluation of seven skin quality indicators including common skin color,oil,moisture,texture,pores,pigmentation,and skin inflammation.Firstly,the structural characteristics of skin tissues and common skin texture detection algorithms has been studied,and the development of the human face skin texture detection system has been introduced,and then the specific implementation of the system has been described,including image acquisition hardware integrated with the induced light source,and the server Skin detection algorithm running on the client and client software design.Secondly,combined with the different image collection modes of the front end,the skin quality indicators are summarized into three categories,which are surface features,pore features and bottom features.The algorithm of skin texture detection based on the collected skin images is studied.Specifically,the surface skin feature detection is performed using statistical methods,which solves the problem that no significant features of the surface skin image can be extracted as an evaluation basis;Besides,a skin pore feature value calculation method based on an adaptive area growth algorithm has been designed.The self-adaptive area growth separated the pore area in the image,then the circularity judgment,and finally the number of pores and the average radius and radius,the variance of which provided accurate classification features for subsequent skin texture evaluation.A skin bottom image recognition method based on deep residual network is designed,which uses simplified batch normalization to accelerate the convergence of network weights during the training process and is used by skin care experts.The Face Data Set containing the underlying skin image data set was established with the help of the skin care specialist.The transfer learning method was used to train the network model and compare the residual network performance at different depths to determine the RN-50 network as the final network structure.After the feature detection algorithm has extracted the eigenvalues of each skin quality indicator,three common classifiers are used to design skin quality evaluation models.Corresponding data sets are used to train and test various classifiers,and compare and analyze the classification results.Finally,the comparison model of classification performance is used to determine the evaluation model composed of SVM classifier.Finally,the main functional test of the client was completed in combination with specific use cases,and the performance of the system's index measurement and evaluation was completed based on the results of manual classification.The experimental results show that this system meets the daily requirements for beauty skin care testing and can help users accurately grasp their own skin condition.