基于机器视觉的透明软管内微量液体体积测量  

Machine vision-based method for measuring micro-volume liquid in transparent tubes

作　　者：王昊 李星辉 肖巍 钱翔 王晓浩 WANG Hao;LI Xinghui*;XIAO Wei;QIANG Xiang;WANG Xiaohao(Tsinghua Shenzhen International Graduate School,Tsinghua University,Shenzhen 518005,China)

机构地区：[1]清华大学深圳国际研究生院数据与信息研究院,广东深圳518005

出　　处：《光学精密工程》2025年第5期763-776,共14页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.62275142);深圳市稳定支持计划资助项目(No.WDZC20231124201906001)。

摘　　要：针对工业生产中透明软管内微量液体(约0.1~0.6 mL)体积快速、精准计量的需求,提出了一种基于机器视觉的计量方法。通过构建特定成像系统,采用自适应邻域加权亮度分析算法,稳定提取液段像素长度,有效克服气泡干扰、光照不均及管壁反光等不利因素的影响。在此基础上,提出两种计量策略:第一种方法基于静力称重的拟合标定测量法,通过拟合液段像素长度与实际体积的关系建立定量模型;第二种方法基于单应矩阵的图像特征坐标变换测量法,将像素坐标映射至物理空间并结合管道内径实现体积计算。实验结果表明,在复杂环境中,两种方法的计量精度分别达到98.3%和98.4%,能够有效满足工业现场对微量液体体积快速测量的需求,具有显著的应用价值和推广潜力。This study addresses the necessity for rapid and precise measurement of micro-volume liquids(approximately 0.1 mL to 0.6 mL)within transparent tubes in industrial production settings.A machine vision-based measurement methodology is proposed,which integrates a specifically designed imaging system alongside an adaptive neighborhood-weighted brightness analysis algorithm to reliably extract the pixel length of liquid segments.This approach effectively mitigates challenges such as bubble interference,uneven lighting,and reflections from the tube's surface.Two measurement strategies are consequently introduced:the first is a calibration-based method utilizing static weighing to establish a quantitative model that correlates the pixel length of the liquid segment with the actual volume;the second is a homography-based coordinate transformation method,which translates pixel coordinates into physical space and calculates volume by incorporating the tube's inner diameter.Experimental results indicate that,under complex conditions,the proposed methods achieve measurement accuracies of approximately 98.3%and 98.4%,thereby satisfying the demands for rapid micro-volume liquid measurement in industrial applications.This methodology demonstrates significant potential for application and offers prospects for widespread adoption.

关 键 词：机器视觉 目标分割 曲线拟合 体积测量 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术] TH691.9[自动化与计算机技术—计算机科学与技术]