摘要
针对大多数评价织物缝纫平整度等级方法对实验条件均有较高的要求,且织物种类及环境等因素对实验结果均有较大影响的现状,提出利用卷积神经网络分析织物缝纫平整度等级的方法,以提高等级分类的准确率和效率。设计了一个基于残差卷积神经网络的织物缝纫平整度客观评价模型,该模型以1000个普通织物的缝纫图像作为训练样本输入,得到缝纫平整度的分类结果,所选织物包含10种常见服装面料品类(塔夫绸、塔斯隆、雪纺、顺纡绉、尼丝纺、麂皮绒、天丝斜纹、真丝缎面、平布、交织绸)。研究结果表明:经200个测试集样本的验证,该模型的评价准确率达96%,与智能化评价以及建立预测模型方法相比,利用卷积神经网络分析织物缝纫平整度等级的方法,具有较好的准确率,且具备获取样本图像流程简单、效率高的优势。
At present,the requirements on the experimental conditions are high for most methods when evaluating the flat level of fabric sewing,and the factors such as the type of fabric and the environment have a higher impact on the experimental results.In view of this situation,a method for analyzing the level of fabric sewing flatness was proposed in this paper by using a convolutional neural network to improve the accuracy and efficiency of classification.An objective evaluation model of fabric sewing flatness based on residual convolutional neural network was designed.The sewing images of 1000 ordinary fabrics were taken as input for training samples for the model,and the classification results of sewing flatness can be obtained.The selected fabric contains 10 kind of common clothing fabrics(Taffeta,Taslon,Chiffon,Downy Crepe,Nylon,Suede,Tencel Twill,Silk Satin,Plain Cloth and Interwoven Silk).The research results showed that,after verified by 200 test set samples,the evaluation accuracy of the model was 96%.Compared with the intelligent evaluation and the prediction model method,the convolutional neural network method has higher leveling accuracy,simpler and higher accuracy on sample images obtaining.
作者
王萌萌
刘成霞
WANG Mengmeng;LIU Chengxia(School of Fashion Design&Engineering,Zhejiang Sci-Tech University,Hangzhou,Zhejiang 310018,China;Zhejiang Province Engineering Laboratory of Clothing Digital Technology,Zhejiang Sci-Tech University,Hangzhou,Zhejiang 310018,China)
出处
《毛纺科技》
CAS
北大核心
2020年第5期87-91,共5页
Wool Textile Journal
基金
浙江省自然科学基金项目(LY20E050017)。
关键词
织物缝纫图像
平整度等级
卷积神经网络
客观评价
fabric sewing image
flatness grade
convolutional neural network
objective evaluation
