| 摘 要: 电磁选针器的刀片运动位移是评估其性能的关键参数,但现有测量方法难以实现多个刀片同步、高效且通用的轨迹获取。为解决该问题,提出一种基于图像处理的多刀片运动轨迹测量方法。该方法首先融合稀疏光流与图像差分技术,自适应定位刀片并动态裁剪ROI区域,提升计算效率。采用自适应Otsu阈值分割提取刀片连通域,结合最小外接矩形拟合与灰度重心法确定亚像素级特征点。最终通过图像序列中特征点的时序变化实现位移测量。实验表明,该方法在测量精度上与激光传感器相当,最大RMSE为0.042 mm,R2 为0.99,且能同步高效处理多个刀片的运动轨迹,具备良好的通用性、鲁棒性与适用性。 |
| 关键词: 选针器 轨迹测量 自适应图像裁剪 稀疏光流 灰度重心法 特征点提取 |
|
中图分类号:
文献标识码:
|
| 基金项目: 国家重点研发计划项目：纱线生产关键工序质量在线检测传感器及系统应用(SQ2023YFB3200093) |
|
| Vision-based Motion Measurement of Multiple Selector Blades Using Image Processing |
|
liusong, Peng Laihu, Qi Yubao
|
Zhejiang Sci-Tech University Zhejiang Key Laboratory of Intelligent Manufacturing Equipment for Flexible Functional Materials
|
| Abstract: The displacement of blades in electromagnetic needle selectors is a critical parameter for performance evaluation. However, existing measurement methods struggle to achieve synchronous, efficient, and generalizable trajectory acquisition for multiple blades simultaneously. To address this challenge, this paper proposes a machine vision-based method for measuring multi-blade motion trajectories. The method first fuses sparse optical flow with frame differencing to adaptively locate the blades and dynamically crop regions of interest (ROIs), thereby improving computational efficiency. An adaptive Otsu thresholding technique is then employed to extract connected components of the blades. Subsequently, sub-pixel feature points are determined by combining minimum bounding rectangle fitting with a gray-level centroid algorithm. Finally, blade displacements are obtained by tracking the temporal evolution of these feature points across image sequences. Experimental results demonstrate that the proposed method achieves measurement accuracy comparable to that of laser sensors, with a maximum RMSE of 0.042 mm and an R2 of 0.99. Moreover, it enables synchronous, efficient, and robust trajectory measurement for multiple blades, exhibiting strong generalizability, robustness, and practical applicability. |
| Keywords: Needle selector Trajectory measurement Adaptive ROI cropping Sparse optical flow Gray-level centroid method Feature point extraction |
