Ethnic Face Image Retrieval Method Based On Facial Feature Enhancement And Convolutional Neural Network

China is a multi-ethnic country.The facial features of different ethnic groups will have certain differences due to different regions,climates and other environments.However,with the change of time,the multiplication of ethnic groups and the exchange and integration of cultures,the facial features of ethnic groups gradually weakening or even disappearing,so the use of computer technology to study and apply ethnic facial features has far-reaching significance.This article aims to explore effective methods for extracting facial features of ethnic minorities,and researches on the national facial image retrieval method based on facial feature enhancement and Convolutional Neural Network(CNN)to provide new ideas and references for research in related fields.The main research work is as follows:(1)Construction of ethnics face resource library.This paper collects and preprocesses ethnic face images of 282 people from two ethnic minority areas in Dehong and Honghe Prefectures of Yunnan Province,and establishes a library of ethnic faces covering different expressions,postures,angles,lighting,and background changes.(2)The face features of ethnic groups are enhanced.Aiming at the difference of human face ethnic features,this paper proposes a method of enhancing ethnic face features based on regions of interest by detecting key feature points of ethnic faces based on the comparative analysis of two ethnic face feature areas The ability to express and discern the characteristics of ethnic faces.(3)CNN-based facial image retrieval.Based on the enhancement of ethnic features,this paper builds a CNN-based ethnic face retrieval framework to realize the extraction and retrieval of ethnic facial depth features,and through experimental comparison and analysis of multiple sets of feature enhancement facial retrieval effects under different deep network structures.The effectiveness and superiority of the ethnic face retrieval method based on feature enhancement and CNN proposed in this paper is presented.(4)Design and implementation of a prototype system for ethnic face image retrieval.This paper uses PyQt5 and TensorFlow to design and implement a prototype system for ethnic face image retrieval.The system includes four modules: data preprocessing,ethnic face feature enhancement,CNN-based ethnic face image retrieval and retrieval result visualization,which can realize a variety of Visualization of image retrieval mode results.