Clothing Style Recognition by Constructing the Chord-Features Matrices

摘要：

　　With the development of computer image processing technology,many fields are developing towards intellectualization and integration.The recognition of garment style has become the focus of current research in the field of garment.It is very necessary for us to find a way to quickly and accurately identify the style of the clothing because of the diversity and complexity of the clothing style.This paper introduces a novel approach for shape characterization,which creates chord feature matrix as shape feature description operator.This method combines the average projection length of the arc to the chord of the target contour and the outer and inner chord lengths of the target contour to constructs the CFM descriptor (Chord-Features Matrices) which characterizes the contour features of the garment style components,and uses the combination of support vector machine and nearest neighbor to recognize and classify the garment styles.As will be shown in this paper,the method has good robustness against noise and can accurately represent the shape characteristics of the target.And the accuracy of algorithm recognition is 95.6%,which is suitable for clothing style recognition.The constructed CFM descriptor can fully describe the geometric characteristics (concavity and convexity) of the target contour,and has good shape representation ability,which can be used for the characterization and recognition of image shape features.

关键词： Fashion style recognition Feature extraction Image segmentation Chord-Features Matrices SVM

作者: Yan-Hong Zhang Gui-Qing Chen Zeng-Bo Xu Si-Yang

作者单位: Shanghai University of Engineering and Technology, No.333 Longteng Road, Shanghai, Songjiang Distinct,201620, China

会议类型: 国际会议

会议名称: 2019上海纺织服装创意创新研究生学术论坛暨第十三届纺织服装创新国际论坛

会议地点: 上海

会议语种:英文

页码: 124-129

在线出版日期: 2019-10-23（万方平台首次上网日期，不代表论文的发表时间）