摘要
市场主流基于传统图像处理的足部尺寸测量算法存在工作运行效率低、测量结果误差大等问题.针对该问题,本文提出一种基于深度学习与传统图像处理相结合的足部尺寸测量系统.首先,使用YOLOv5(you only look once v5)检测出A4纸的有效区域,再采用RCF(richer convolutional features for edge detection)边缘检测方法提取A4纸及足部的边缘信息,通过A4纸的边缘估计出4个顶点来进行图像矫正.最后,对已矫正的图像使用降噪、滤波与二值化处理,从而计算出足部尺寸.在自行收集的足部样本数据集上进行测试,准确性为97.36%.实验结果表明,本文提出的基于YOLOv5和RCF的足部尺寸测量系统,提高了足部测量的精度以及运行效率,可帮助用户更便捷更高效更精确地获取个人足部数据,具有较好的实用价值和较广的应用前景.
The mainstream foot size measurement system based on traditional image processing in the market have problems such as low operational efficiency and large measurement error.To solve the above problems,this paper proposes a foot size measurement system based on the combination of deep learning and traditional image processing.Firstly,YOLOv5 was used to detect the effective area of A4 paper.Then,RCF edge detection method was used to extract the edge information of A4 paper and feet,and four vertices were estimated based on the edges of A4 paper for image correction.Finally,the rectified image was denoised,filtered and binarized to calculate the foot size.The results showed that the accuracy of the foot size measurement system proposed in this paper based on YOLOv5 and RCF was 97.36% when tested on a self-collected foot sample dataset.This system improved the accuracy and operational efficiency of foot measurement,and could help users obtain personal foot data more conveniently,efficiently,and accurately,which provided good practical value and broad application prospects.
作者
温铭浩
苗谦盈
陈祥豪
刘复昌
王竹萍
WEN Minghao;MIAO Qianying;CHEN Xianghao;LIU Fuchang;WANG Zhuping(School of Information Science and Technology,Hangzhou Normal University,Hangzhou 311121,China;Alibaba Business School,Hangzhou Normal University,Hangzhou 311121,China)
出处
《杭州师范大学学报（自然科学版）》
CAS
2023年第4期389-396,共8页
Journal of Hangzhou Normal University(Natural Science Edition)
基金
国家级大学生创新创业训练计划项目(202110346041)。
