自动化与信息工程2024，Vol.45Issue(4)：1-9,17,10.DOI:10.3969/j.issn.1674-2605.2024.04.001
基于HSV+Canny模型、HED网络模型的衣物轮廓提取算法研究
Research on Clothing Contour Extraction Algorithm Based on HSV+Canny Model and HED Network Model
摘要
Abstract
In response to the problem of edge detection algorithms being easily affected by noise in the process of clothing contour extraction,traditional image processing methods and deep learning methods are used to study the clothing contour extraction algorithm.Traditional image processing methods are mainly based on the HSV model and Canny algorithm.Firstly,foreground images are segmented using the HSV model;Then,perform morphological processing on the binary image;Finally,the Canny algorithm is used to extract clothing contours,which can accurately extract clothing contours of most colors.The deep learning method is mainly based on the HED network model.To address the problems of missing and rough edge localization in the output of the HED network model,improvements are made to the HED network model.Firstly,the pooling layers in stages 3 and 4 are removed;Then,introduce attention mechanism in the feature fusion stage;Finally,the Canny algorithm is integrated for edge refinement.The comparative experimental results show that the HSV+Canny algorithm has improved the ODS and OIS values by 13.16%and 14.72%respectively compared to the Canny algorithm,with a slight increase in detection speed;The improved HED network model improves the ODS and OIS values by 4.84%and 3.97%respectively compared to the HED network model,while maintaining the same detection speed.关键词
衣物轮廓提取/HSV模型/Canny算法/HED网络模型/注意力机制Key words
clothing contour extraction/HSV model/canny algorithm/HED network model/attention mechanism分类
信息技术与安全科学引用本文复制引用
黄启华,杜玉晓..基于HSV+Canny模型、HED网络模型的衣物轮廓提取算法研究[J].自动化与信息工程,2024,45(4):1-9,17,10.基金项目
国家自然科学基金(61976059,61640213) （61976059,61640213）
