基于GoogLeNet的秦绣纹样的识别与分类研究

Recognition and Classifcation of Qin Embroidery Patterns Based on GoogLeNet

秦绣作品具有极强的地域特征和文化风格,其表现出的独特的艺术色彩相比于其他绣种更加强烈与鲜明。但是秦绣纹样的分类仍然依赖人工,这不利于秦绣文化的研究。针对这一问题,本文提出了一种基于GoogLeNet的秦绣纹样的识别与分类算法。首先,采用展览收集和走访个人收藏家收集到的大量秦绣作品照片进行处理所得到的数据集作为研究对象,基于神经网络算法对秦绣纹样的识别与分类进行研究。其次,选取GoogLeNet作为分类网络,提取丰富的秦绣图像特征信息,用于秦绣纹样分类。通过对神经网络的多次训练,降低了秦绣纹样分辨错误率。最后,构建了便捷的秦绣纹样分类系统,可以根据用户上传的图片实时输出识别出的秦绣纹样种类。本研究简化了秦绣纹样分类流程,对秦绣纹样的学术研究与秦绣文化的社会推广有着重要意义。

Qin embroidery works have strong regional characteristics and cultural styles,and their unique artistic value and practical value are more intense and vivid than other kinds of embroidery.Nevertheless,the classification of patterns in Qin embroidery still relies on manual work,which is not conducive to the study of Qin embroidery culture.To solve this problem,a recognition and classification algorithm for Qin embroidery patterns based on GoogLeNet is proposed in this paper.Above all,countless photographs of Qin embroidery works are collected by exhibition collections and individual collectors,and these pictures are used as the research object to study the recognition and classification of Qin embroidery patterns based on a neural network algorithm.Second,GoogLeNet is selected as the classification network to extract rich feature information from Qin embroidery images for the classification of Qin embroidery patterns.By training the neural network numerous times,the error rate of Qin embroidery pattern discrimination is reduced,so as to better realize the purpose of studying and inheriting Qin embroidery pattern types.Ultimately,a convenient classification system for Qin embroidery patterns is constructed,which can output the types of Qin embroidery patterns identified according to the pictures uploaded by users.This study simplifies the classification process of Qin embroidery patterns,which is of great significance to the academic research of Qin embroidery patterns and the social promotion of Qin embroidery culture.