The Application of Traditional Hakka Clothing Cultural Elements in Modern Clothing Design in Western Fujian Based on Image Feature Extraction Technology

Abstract

Abstract In this paper, by applying the SIFT algorithm to the color category, texture and shape features of the dress pattern image for feature extraction and eliminating the low contrast texture feature points of the Taylor series expansion, using the statistics of the grayscale covariance matrix as the texture features. Based on the SIFT algorithm, the SURF feature extraction algorithm is proposed. The ideas of the Hessian matrix, determinant value approximation (DoH) and integral image (IGI) classifier are integrated to construct the feature fusion algorithm. Finally, the experimental results of several algorithms are compared and analyzed by taking the traditional dress pattern of the western Fujian Hakka family as the experimental object. The results show that among all the fusion algorithms, the SURF+IGI algorithm achieves an average recognition rate of 0.8914, which is the best performance. And the AUC value of its test set is 0.9445 when the learning rate is 0.005. The research in this paper meets the demand of extracting and applying the cultural elements of the traditional dresses of the Hakka family in western Fujian in modern dress design.