上海交通大学学报 ›› 2025, Vol. 59 ›› Issue (5): 703-710.doi: 10.16183/j.cnki.jsjtu.2023.397

• 机械与动力工程 • 上一篇    

足部特征点准确定位及形态参数自动测量方法

纪冕1, 林艳萍1(), 王冬梅1, 陈立2, 马昕2   

  1. 1.上海交通大学 生物医学制造与生命质量工程研究所,上海 200240
    2.复旦大学附属华山医院 足踝外科,上海 200040
  • 收稿日期:2023-08-16 修回日期:2023-10-05 接受日期:2023-10-26 出版日期:2025-05-28 发布日期:2025-06-05
  • 通讯作者: 林艳萍,副研究员,博士生导师;E-mail: yanping_lin@sjtu.edu.cn.
  • 作者简介:纪 冕(2000—),硕士生,从事医工交叉研究.
  • 基金资助:
    国家重点研发计划(2022YFC2009504)

Precise Foot Feature Point Localization and Automatic Parameters Measurement

JI Mian1, LIN Yanping1(), WANG Dongmei1, CHEN Li2, MA Xin2   

  1. 1. Institute of Biomedical Manufacturing and Quality of Life Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
    2. Department of Foot and Ankle Surgery, Huashan Hospital, Fudan University, Shanghai 200040, China
  • Received:2023-08-16 Revised:2023-10-05 Accepted:2023-10-26 Online:2025-05-28 Published:2025-06-05

摘要:

为了快速获得足部形态相关的参数信息并量化足部变形程度,提出一种能够准确定位足部特征点并自动计算足部形态特征参数的算法.首先,使用UPOD激光扫描仪获取93名受试者的足部三维模型.然后,采用随机采样一致性算法和主成分分析方法对模型坐标系进行对齐,结合足部形态特征识别并定位特征点,实现对足部长度、角度和围度形态参数的自动测量.为评估测量结果的准确性、重复性和一致性,使用平均绝对误差(MAE)、平均相对误差(MAPE)、组间相关系数(ICC)和Bland-Altman图进行分析.结果显示,足长和足宽的MAE小于2 mm,跖围、跗围和兜跟围的MAE小于4 mm,MAPE小于2%;3次重复测量的ICC均大于0.99,Bland-Altman图显示95%以上的散点位于一致性界限内.结果证明,该算法可以实现站立姿态下足部模型的坐标系自动对齐、特征点准确定位以及参数的精确测量;测量精度满足临床测量需求,具有较高的测量精度和可靠性;测量结果可为足型分类、智能辅具适配与个性化辅具设计提供数据支持,具有重要的临床应用价值.

关键词: 足部测量, 坐标系对齐, 特征点定位, 三维点云

Abstract:

In order to quickly obtain foot parameters and quantify the degree of foot deformation, an algorithm that can accurately locate foot feature points and automatically calculate foot parameters is proposed. First, a total of 93 patients participate and their foot models are obtained using the UPOD laser scanner. Then, the random sampling consensus algorithm and principal component analysis are used to align the foot coordinate system. The algorithm utilizes foot features to identify and locate feature points, enabling the parameter calculation of length, angle, and girth. The accuracy, repeatability, and consistency of the measurements are evaluated using mean absolute error (MAE), mean absolute percentage error (MAPE), interclass correlation coefficient (ICC), and Bland-Altman plots. The MAE of foot length and width is less than 2 mm, and for ball girth, instep girth, and heel girth, it is less than 4 mm. The MAPE is less than 2%, and the ICCs for the three replicates exceed 0.99. More than 95% of the scattered points in the Bland-Altman plots are within the consistency boundary. The results show that the proposed algorithm can automatically align the coordinate system, accurately locate feature points, and accurately measure foot parameters in the standing posture. The measurement accuracy meets clinical needs with high accuracy and reliability. The findings provide valuable data support for foot classification, intelligent assistive device adaptation, and personalized assistive device design, showing important clinical application potential.

Key words: foot measurement, coordinate alignment, feature point localization, 3D point clouds

中图分类号: