Research On Skin Quality Recognition System Based On Multi-Parameter

Consumers are paying more and more attention to their skin condition and skin type classification.In this project,a skin quality recognition system is designed based on five characteristic parameters of oil,pigment,pores,roughness and wrinkles,including facial image collection and skin feature recognition.The main work and innovations of this subject are as follows:1.Design the face acquisition system based on FPGA architecture: The FPGA ZYZQ7020 N chip is used as the foundation of the acquisition system,and the OV5640 camera is used to complete the acquisition system,so that the system can complete the face skin image acquisition under natural light.2.A facial skin pore detection algorithm based on filter banks and improved maximum entropy is proposed.First,an ICA-Homo filter bank was constructed.S-space was separated from the original image before homomorphic filtering was performed,and then the filtered low-pass and high-pass components were extracted.At the same time,the melanin layer in the image was separated by ICA.The processed data were processed by the improved maximum entropy method,and the final detection result was obtained by logical operation.The experimental results show that the effective detection rate of the proposed algorithm is 96% and the error range of pore size is 0.0304,which proves the effectiveness of the proposed algorithm.3.A genetic algorithm-based feature selection of skin roughness is proposed.Extract 24 attributes that affect skin roughness from the collected skin images to form the original feature space,use genetic algorithms to optimize the original feature space,and retain appropriate dimensions to form a new feature space.The two feature spaces are verified by SVM classification,and the accuracy,precision and recall rate of the space after screening have been improved to a certain extent.4.Separating the oil component and the dye feature in a manner based on the color feature.The melanin layer and the hemoglobin layer were separated by HSV space channel separation and ICA.Compared with the professional equipment,the average error of dyes is 6.03% and that of oil is 7.04%.5.Wrinkle feature extraction based on dimensionality reduction algorithm of Lab space.The L space in the Lab space is used for dimension reduction by PCA,followed by filtering and morphological processing,and finally the segmented wrinkle image is obtained.The algorithm is compared with the GF filter bank algorithm.The algorithm can extract wrinkles well under the interference information of facial pores,hair,stains and uneven illumination.In this paper,the improved particle swarm optimization SVM method is used to establish the classification model.Compared with the BP classifier and the KNN classifier,the average recognition rate of the five skin characteristics reaches 89.32% in the case of small average standard deviation of the classification model in this paper,which can better identify and classify the skin.