Please wait a minute...
Chinese Journal of Engineering Design  2021, Vol. 28 Issue (4): 495-503    DOI: 10.3785/j.issn.1006-754X.2021.00.056
Whole Machine and System Design     
Research on micropipetting technology based on image monitoring
SHANG Zhi-wu, ZHOU Shi-qi
Tianjin Key Laboratory of Modern Electromechanical Equipment Technology, Tianjin Polytechnic University, Tianjin 300387, China
Download: HTML     PDF(3191KB)
Export: BibTeX | EndNote (RIS)      

Abstract  In order to improve the reliability of micropipetting of biochemical analyzer, a pipetting fault diagnosis and pipetting volume detection system based on image segmentation method was proposed. Based on STM32 microcontroller, with the stepping motor of micropipetting system controlled by speed-position double closed-loop PID (proportion-integral-derivative) algorithm, the automatic pipetting control system of biochemical analyzer was designed. The pipetting process image was collected as the sample image,and the pipetting data set was established. The U-Net neural network models before and after pruning were trained, and the calculation, parameters and mean intersection over union (MIOU) of the models were compared. The pipetting area was segmented and processed by the pruned U-Net model to realize the fault diagnosis of the pipetting system. Combined with the geometric characteristics of the pipetting suction head, the pipetting volume model was established to calculate the pipetting volume. The micropipetting experiment based on image segmentation method was carried out to analyze pipetting errors, and the errors were compensated by least square method, sub-pixel corner point detection and limit learning machine. The results showed that after pruning, the calculation, parameters and MIOU of U-Net neural network model were reduced by 47.30%, 93.99% and 0.61% respectively, and the operation efficiency of the model was significantly improved. For the verification points of 10, 50, 100 μL, the pipetting accuracy of the micropipetting system reached 1.72%, 1.36% and 1.39% respectively, meeting the accuracy design requirements of micropipetting system. The pipetting monitoring system based on image segmentation method can effectively judge the pipetting fault and detect the pipetting volume. The research result promotes the development of micropipetting technology.

Received: 09 October 2020      Published: 28 August 2021
CLC:  TH 77  
  TP 29  
Cite this article:

SHANG Zhi-wu, ZHOU Shi-qi. Research on micropipetting technology based on image monitoring. Chinese Journal of Engineering Design, 2021, 28(4): 495-503.

URL:

https://www.zjujournals.com/gcsjxb/10.3785/j.issn.1006-754X.2021.00.056     OR     https://www.zjujournals.com/gcsjxb/Y2021/V28/I4/495


基于图像监测的微量移液技术的研究

为了提高生化分析仪微量移液的可靠性,提出了一种基于图像分割法的移液故障判断和移液量检测系统。基于STM32微控制器,利用速度位置双闭环PID (proportion-integral-derivative,比例-积分-微分)算法控制微量移液系统的步进电机,设计了生化分析仪的自动移液控制系统。采集移液过程图像作为样本图像,建立了移液数据集。分别对剪枝前后的U-Net神经网络模型进行训练,并比较模型的计算量、参数量和平均交并比。通过剪枝后U-Net模型对移液区域的分割和处理,实现移液系统的故障判断,并结合移液吸头的几何特征建立移液体积模型,以计算移液体积。进行了基于图像分割法的微量移液实验,对移液误差进行分析,并利用最小二乘法、亚像素角点检测和极限学习机进行误差补偿。结果表明,剪枝后U-Net神经网络模型的计算量、参数量和平均交并比分别下降了47.30%,93.99%和0.61%,模型运行效率显著提升。针对10,50,100 μL的检定点,微量移液系统的移液精度分别达到1.72%、1.36%和1.39%,满足微量移液系统的精度设计要求。基于图像分割法的移液监测系统能够有效判断移液故障并检测移液体积。研究结果对微量移液技术的发展起到一定的推动作用。
No related articles found!