摘要
针对图像中服装廓形的自动识别问题,选取近8年女西装秀场图片共2 045个样本进行服装轮廓的提取与筛选。通过建立坐标系,研究服装外部轮廓与廓形支撑数据之间的定量关系;分析服装外廓形的宽度、相对角度和直角坐标值数据,从客观的角度重新量化廓形,并建立女装廓形数据样本库;通过支持向量机SVM算法训练出针对X廓形的分类及预测模型。一共进行了6次试验验证,精度均在89%以上。
Aiming at the problem of automatic recognition of clothing silhouettes in images, this paper extracted and filtered clothing silhouettes with a total of 2 045 samples based on pictures of women’s suit shows in the past eight years. By constructing a coordinate system, it studied the quantitative relationship between the outer contour of the garment and the silhouette support data. By analyzing the width, relative angle and rectangular coordinate values of the outer shape of the garment, it redefined the outline from an objective and quantitative perspective and established a sample database of women’s clothing outline data. The classification and prediction model for the X profile was trained through the support vector machine SVM algorithm. A total of six verification tests were carried out, and the accuracy was all above 89%.
作者
段澍湉
崔明海
陈怡帆
汪娜
杨如意
DUAN Shu-tian;CUI Ming-hai;CHEN Yi-fan;WANG Na;YANG Ru-yi(School of Fashion Arts and Engineering,Beijing Institute of Fashion Technology,Beijing 100029,China)
出处
《北京服装学院学报（自然科学版）》
CAS
北大核心
2022年第2期46-52,共7页
Journal of Beijing Institute of Fashion Technology：Natural Science Edition
基金
北京服装学院2021年研究生科研创新项目(120301990131/001)。
关键词
服装廓形
特征部位
位置关系
廓形分类
clothing silhouette
characteristic part
positional relationship
profile classification
