Research And Application Of Edge Extraction Of Qinghai Embroidery Pattern

As one of the national intangible cultural heritage,Qinghai embroidery has a long history and a wide variety of varieties.The special geographical and cultural environment of Qinghai Province has created a unique production form of Qinghai embroidery,pattern drawing is the core craft of embroidery.Traditional embroidery patterns still rely on hand drawing,which is labor intensive and requires the producer’s fine brushwork skills,therefore,it is very necessary to use digital technology to extract patterns of embroidery images,while ensuring the quality of edge extraction,it also improves efficiency,it has important academic research value and social and cultural significance for realizing the digital protection of embroidery images in Qinghai.At present,with the development of artificial intelligence algorithms in the field of image processing,convolutional neural networks have become a research hotspot in the field of image processing,gradually replacing traditional edge detection algorithms,and have achieved good results in the field of edge extraction,there are few studies on the application of edge extraction technology to embroidery.Traditional edge extraction algorithms only study local changes and cannot reflect more complex scenes,they are usually affected by noise and texture clarity,resulting in discontinuity and incompleteness in the extracted edge image.However,Qinghai embroidery images have the characteristics of fineness and complex patterns,experiments show that traditional edge extraction algorithms can hardly meet the needs for edge extraction of embroidery patterns,traditional algorithms need to be improved and combined with convolutional neural network methods for in-depth research on edge extraction,in order to obtain more high-level semantic information of embroidery images,the patterns are clearer and more complete,the main research contents and innovations of this thesis are as follows:(1)Aiming at the problem that the traditional Canny algorithm has limited noise smoothing ability,this article improves the algorithm,using adaptive median filtering instead of Gaussian filtering,which can effectively suppress noise while retaining more pattern textures;for traditional Canny algorithm,the amplitude is calculated This paper uses the Sobel template to replace the first-order difference template,and the calculated amplitude is not easy to miss the detection.For the traditional Canny algorithm high and low threshold selection is easily affected by human factors,this paper adopts the Otsu algorithm to set adaptively.Set high and low thresholds.The experimental results prove that the improved Canny algorithm is clearer and more complete than the original algorithm.(2)Aiming at the problem that traditional algorithms cannot cope with the complex patterns of Qinghai embroidery,this thesis adopts the HED network model improved by the VGG16 convolutional neural network model,the network model adds a side-output after each convolutional layer,the output result is deconvolved to reach the same size as the original image size,and the predicted edge image is output through the weight fusion layer.Experiments show that the objective evaluation index of Pratt’s quality factor and signal-to-noise ratio is higher than that of traditional edge extraction algorithms,and good results have been achieved on Qinghai embroidery images.(3)Based on the design and implementation of traditional algorithms,improved Canny algorithm and HED network model,the Qinghai embroidery image edge extraction system was developed,this system is easy to operate and has a simple interface,which is conducive for craftsmen to use the system to extract Qinghai embroidery patterns.