基于改进边缘检测算法的服装款式识别  

Clothing style identification based on improved edge detection algorithm

作　　者：庹武[1] 王哓玉 高雅昆 于媛媛[1] 郝潇潇 刘永亮 郭鑫 TUO Wu;WANG Xiaoyu;GAO Yakun;YU Yuanyuan;HAO Xiaoxiao;LIU Yongliang;GUO Xin(College of Fashion,Zhongyuan University of Technology,Zhengzhou,Henan 451191,China;College of Electrical Engineering and Automation,Henan Institute of Technology,Xinxiang,Henan 453003,China)

机构地区：[1]中原工学院服装学院,河南郑州451191 [2]河南工学院电气工程与自动化学院,河南新乡453003

出　　处：《纺织学报》2021年第10期157-162,共6页Journal of Textile Research

基　　金：河南省高等学校重点科研项目(19A540004,21B413002)。

摘　　要：为快速识别服装款式类型,提高生产效率,针对现有传统边缘检测算法难以准确提取轮廓特征序列的不足,设计一种改进的边缘提取算法。通过定义一种新的优化卷积核,在使用传统边缘检测算法提取训练样本的服装轮廓基础上,将该卷积核与目标矩阵进行卷积得到新的外轮廓,将新轮廓序列的傅里叶描述子作为特征向量,进一步利用BP神经网络模型完成服装款式的自动分类与识别。为验证改进方法的有效性,建立一个包含4类服装500个不重复服装图像的样本库,选取281个作为训练样本,对剩余219个样本进行测试,测试识别准确率最低为93.48%,最高达到了100%。该改进算法提高了服装款式识别率,对服装智能化生产具有借鉴意义。In order to quickly identify clothing styles and improve production efficiency,an improved edge extraction algorithm was designed to solve the problem that the existing traditional edge detection algorithm was difficult to accurately extract contour feature sequence.By defining a new optimization convolution kernels,the use of traditional edge detection algorithm based on clothing contours of the training sample,will the convolution convolution kernels and target matrix are new outer contour,a new contour sequence of Fourier descriptor as a feature vector,to increase the use of BP neural network model to complete the design of automatic classification and recognition.In order to verify the effectiveness of the improved method,a sample library containing 500 non-repeated clothing images of four categories of clothing was established.281 samples were selected as training samples and the remaining 219 samples were tested.The recognition accuracy of the test was as low as 93.48%and as high as 100%.It is of reference significance to the intelligent production of clothing.

关 键 词：服装款式识别 边缘检测算法 傅里叶描述子 BP神经网络 

分 类 号：TS941.26[轻工技术与工程—服装设计与工程]