基于机器视觉的微孔板液位检测研究
Research on Microporous Plate Liquid Level Detection Based on Machine Vision
樊驰豪 1张晓青 1郭阳宽 1端木正 1庄传领2
作者信息
- 1. 北京信息科技大学仪器科学与光电工程学院
- 2. 北京哈美顿生物科技有限公司
- 折叠
摘要
在使用人工检测微孔板的多孔液位时耗时费力、准确率低下,针对此问题提出一种基于机器视觉的微孔板液位检测方法.通过设计相机采集角度、打光方式等搭建视觉检测系统并采集到清晰的单排液位图像;基于零均值归一化互相关的匹配原理,设计算法对中值滤波后的图像进行位姿校正,结合液位图像重心定位8 个单孔液位位置;设定相似度双阈值实现对液位的检测,并对不规则液位图像加入掩码进行二次匹配检测.通过实验验证,该方法对溶液容积变化范围为-10～10 μL及以上的液位检测准确率达100%,满足检测微孔板液位的准确性需求.
Abstract
The manual detection of liquid levels in porous microplate is time-consuming,laborious,and prone to inaccura-cies.To address this issue,a machine vision-based method for detecting liquid levels in microplate was proposed.This method in-volved the construction of a visual detection system by designing camera acquisition angles,lighting methods and so on to capture clear single-row liquid level images.Utilizing the principle of zero-mean normalized cross-correlation matching,an algorithm was developed to perform pose correction on median-filtered images and locate the positions of eight individual well liquid levels using centroid positioning of liquid level images.Liquid level detection was achieved by setting dual similarity thresholds,with irregular liquid level images undergoing secondary matching detection using masks.Experimental validation demonstrates that this method achieves a 100%accuracy in detecting liquid levels within a range of-10～10 μL of solution volume variation or more,meeting the accuracy requirements for microplate liquid level detection.
关键词
机器视觉/液位检测/图像处理/微孔板/零均值归一化互相关匹配Key words
machine vision/liquid level detection/image processing/microplate/zero mean normalized cross-correlation matching引用本文复制引用
出版年
2024
NETL
NSTL
万方数据