Research And Application Of Key Technology Of Image Processing For Minority Costumes

The traditional culture of minorities has played a very important role in the historical process of the cultural development of the Chinese nation.However,with the acceleration of social modernization,the traditional culture of minorities is gradually disappearing,which makes more and more scholars begin to pay attention to the protection of traditional minority cultures.As an important part of minority culture,the digital processing and application of minority costumes is undoubtedly one of the effective ways to realize the protection and inheritance of minority costume culture.This paper mainly researches the key technologies of image processing of minority costumes,and aims to explore effective ways of digital application of minority costumes,and to provide technical support for research in related fields.The research content of this article mainly includes the following aspects:This paper studies the pattern detection of minority costumes based on HOG +SVM.First,the traditional HOG feature extraction algorithm is improved,and the pattern features of minority costumes patterns are extracted.Then,the SVM algorithm is used to train the HOG features of the extracted minority costumes pattern and finally realize the detection of minority costumes pattern.Research on the image matching method of minority costumes.This paper proposes a K-Means+CPD-based image matching algorithm for minority costumes,and uses the detected patterns as matching images.First,the K-Means algorithm is used to cluster image feature point sets for preprocessing,and then the CPD algorithm is used.The images to be matched in the minority costume data set are matched,and finally the effectiveness of the proposed method in the image matching of minority costumes is verified through multiple sets of comparative examples and analysis.This paper studies the image recognition of minority costumes.First,the 7,000 training and test image sets containing the Wa and Yi nationalities were constructed by image scaling the collected images of minority costumes,and then a TensorFlow deep learning framework was built to migrate the deep convolutional network VGGmodel to the task of identifying minority costumes.The parameters of the network model are optimized by repeatedly iteratively comparing the model's loss and recognition rate,so as to achieve the classification and recognition of minority costumes images.