Feature Extraction And Recognition Of Ethnic Costumes

The traditional culture of minority costumes is the precious wealth of the entire Chinese nation.In recent years,with the development of economic globalization and social modernization,the traditional culture of minority costumes in China is on the verge of extinction,and even some has disappeared.So the protection and inheritance of minority costume culture has become a new social issue today.At present,many scholars have conducted extensive research on the inheritance and protection of intangible cultural heritage,and have also collected a large number of national costumes images.However,the processing of these costume images still remains in the traditional manual processing stage.Few of them work on automatically recognition and classification of the national costumes images.From the perspective of computer vision,this paper analyzed the visual features of minority costumes images.We extracted totem regions on national costumes and extracted their features.A classifier was used to identify various national costumes of totem image.So we completed the automatic recognition and classification of the minority costumes images.This laid a good foundation for the national clothing information processing and had made certain contribution in the inheritance and protection of national costume culture.The main work of this article is as follows:(1)The color adaptive threshold segmentation algorithm was used to achieve a rough segmentation of the costume totem region.Then the morphological noise reduction and area threshold filtering methods were used to refine the rough segmentation results.Finally,the target totem region was completely segmented from the entire image.(2)This paper proposed a method of shape feature function to judge the similarity of the shapes between patterns,and solves the similarity judgment problem of the patterns due to the different sizes and tilt angles.(3)Based on the analysis of the Hu invariant moment feature algorithm and HOG feature algorithm,a feature fusion algorithm based on HOG feature was proposed to integrate the Hu invariant moment feature with geometric invariance for costume totem image feature extraction.Compared to the single HOG feature or Hu invariant moment feature,the fusion feature has a stronger ability to represent national costume totem images.(4)We established a classification model to classify the extracted HOG + Hu fusion features with support vector machine SVM,ompared with a single feature model,the fusion feature classification recognition model proposed in this paper has higher recognition accuracy in the identification and classification of national costume totem.(5)Under the MATLAB environment,a classification and recognition system for ethnic minority costume images was established.