邓芋蓝  林飞振  马婷婷  麦志颛  孙涛 

（广州计量检测技术研究院，广东 广州 510663）

摘要：针对人工测量螺纹几何参数存在的精度低、操作困难、易磨损等问题，提出一种面向螺纹精密测量的多模态机器视觉系统，实现螺距、牙型角和外径等几何参数的自动化测量。首先，通过多模态图像采集装置的微距相机和短焦距相机分别采集螺纹工件的杆部和头部图像；然后，基于OpenCV几何图像算法和卷积神经网络设计一种双链路测量方法，用于测量螺纹的螺距、牙型角和外径；最后，通过改进型注意力集中机制将两条链路的测量结果动态融合，使测量结果在准确性和稳定性之间达到平衡。实验结果表明，该系统能够有效地提升螺纹几何参数测量的准确性、稳定性和效率，具有工程应用和推广价值。

关键词：多模态机器视觉；螺纹精密测量；螺距；牙型角；外径；双链路测量；注意力机制

中图分类号：TP391            文献标志码：A          文章编号：1674-2605(2025)04-0006-08

DOI：10.12475/aie.20250406                                 开放获取

Multimodal Machine Vision System for Precision Thread Measurement 

DENG Yulan  LIN Feizhen  MA Tingting  Mai Zhizhuan  SUN Tao

(Guangzhou Institute of Measurement and Testing Technology, Guangzhou 510663, China)

Abstract: To address the issues of low accuracy, operational difficulty, and susceptibility to wear in manual thread geometric parameter measurement, a multimodal machine vision system for precision thread measurement is proposed. This system automates the measurement of geometric parameters such as pitch, thread angle, and major diameter. First, images of the thread workpiece's shank and head are captured using the macro camera and short-focal-length camera of a multimodal image acquisition device. Then, a dual-path measurement method is designed based on OpenCV geometric image algorithms and convolutional neural network to measure the thread's pitch, thread angle, and major diameter. Finally, an enhanced attention mechanism dynamically fuses the measurement results from the two paths, achieving a balance between accuracy and stability in the final results. Experimental results demonstrate that this system can significantly improve the accuracy, stability, and efficiency of thread geometric parameter measurement, possessing substantial practical engineering value and potential for widespread adoption.

Keywords: multimodal machine vision; thread precision measurement; pitch; thread angle; major diameter; dual-path measurement; attention mechanism