Research On Edge Extraction Of Traditional Mongols Patterns Based On Improved Canny Operator

With rich artistic features,Mongols patterns have been widely used in Mongolian architectural utensils and clothing.Traditional Mongols decorative patterns and patterns are complex,and manual drawing requires not only a lot of time but also high professional skills,which seriously affects the efficiency of work and learning.In order to better preserve the traditional Mongols decorative patterns and improve the work and learning efficiency of relevant industrialists,this thesis aims to vectorize these patterns and extract edges,so as to promote the inheritance and development of Mongols decorative patterns.In view of the problems such as the complexity of Mongols pattern sketch and the long time required,this thesis mainly studies the edge extraction method of artificial Mongols decorative patterns.First,a traditional edge extraction method of Mongols decorative pattern based on deep learning LDC model is proposed to train and test the data.When the data set is only 270 pieces,image enhancement processing is used to expand the data set to 4050 pieces and 30 rounds of training are carried out.The edge of Mongols decorative pattern is extracted on the trained model.Then,an improved Canny operator is proposed to vectorize the edge of Mongols decorative patterns.Compared with the traditional Canny algorithm,the improved algorithm uses adaptive median filter instead of Gaussian filter,and automatically adjusts the size of the filter according to the noise intensity,which can better preserve the details of the image while removing noise;In the process of calculating image gradient values,the Sobel template is used to replace the firstorder differential template,which improves the accuracy of edge extraction;In addition,using an adaptive threshold based on iteration and Otsu to replace the dual threshold improves the computational efficiency of the Canny operator.Compare the improved Canny operator with traditional Canny operators,LDC models,and other traditional edge detection operators.The experimental results show that the improved Canny operator not only extracts complete and clear edges,but also improves the structural similarity(SSIM)by an average of 6.37%,30.9%,and 24.5% compared to traditional Canny operators,LDC models,and other traditional edge detection operators,respectively;Compared to other traditional edge detection operators such as the Canny operator,LDC model,Sobel operator,Roberts operator,Prewitt operator,and others,the Peak Signal to Noise Ratio(PSNR)is improved by an average of 7%,167%,and 175%,respectively.The improved Canny operator improves the accuracy of the algorithm and the edge extraction efficiency of Mongols traditional patterns.Finally,the "Muhe" APP for traditional Mongols decorative pattern processing was designed and developed,and the improved Canny operator was applied to the APP,which maximized the use of the existing Mongols decorative pattern image library to provide users with a Mongols decorative pattern sharing platform,greatly improving the work efficiency and enthusiasm of relevant practitioners.