Optimization Research And Implementation Of Traditional Pattern Multi Label Learning Algorithm Based On Multilayer Fusion Dictionary Learning

In order to better interpret,protect,and inherit the traditional culture of the Chinese nation,and make it easier for more people to learn traditional culture,it is important to study the multi label labeling of traditional clothing.This paper takes traditional ethnic costume images as the research content,and uses multi-layer dictionary learning as the main research method to achieve multi-label labeling of traditional costume images.Comparative experiments were conducted on the Ming and Qing costume patterns and museum heritage image datasets.The main work and research content can be summarized as follows:(1)A multi-level fusion dictionary learning algorithm was proposed.Currently,only one algorithm is used to implement multi-layer dictionary learning,and multiple algorithms can achieve better performance.To address this problem,multiple algorithms are proposed to implement multi-level dictionary learning,which improves the performance of the algorithm.The algorithm performance has been verified on traditional clothing image datasets,which has better value for multi-label labeling of traditional clothing images.At the same time,comparative experiments have been conducted on other cultural images,verifying the robustness of the algorithm.(2)A deep fusion dictionary learning algorithm is proposed.Aiming at the high resolution of high-quality images,which leads to excessively high dictionary atomic dimensions,a neural network is proposed to extract traditional clothing image features and reduce dimensions based on a multi-layer fusion dictionary learning algorithm,which solves the problems of excessive training time and memory required for dictionary learning algorithm training.This paper also uses traditional feature extraction algorithms,and carries out comparative experiments to verify that the neural network feature extraction effect is better.(3)Design and implement an online multi label labeling system for traditional clothing images.The system is built using Python language and the implemented algorithm is integrated into the system.Users can access the system using a browser and upload traditional clothing images.The system will return the image label.Users can also manually annotate traditional clothing images on the system to further enhance their understanding and representation of traditional clothing images.This paper proposes and implements a multi-level fusion dictionary learning algorithm based on the idea of multiple algorithm fusion,and uses neural networks to reduce the dimension of traditional clothing images.Good results have been obtained in the multi-label labeling experiment.