Deep learning-based algorithm for recognition of graduated cylinder in physical and chemical experiments

As pivotal instrument in physical and chemical experiments, graduated cylinders are typically used to dispense specific volume of liquid. Accurate readings of graduated cylinders is critical to the success of the experiment and the reliability of the data. However variations in liquid volume lead to...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Proceedings (International Conference on Computer Engineering and Applications. Online) s. 1211 - 1216
Hlavní autor: Gao, Chenxuan
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 12.04.2024
Témata:
Accuracy
deep learning
Distortion
graduated cylinder
image processing
Instruments
liquid volume
Liquids
Metrology
Noise
scale
Stability analysis
ISSN:2159-1288
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:As pivotal instrument in physical and chemical experiments, graduated cylinders are typically used to dispense specific volume of liquid. Accurate readings of graduated cylinders is critical to the success of the experiment and the reliability of the data. However variations in liquid volume lead to the effective regions of the liquid levels and scales manifest as curves or semi-ellipses with distinct curvatures. This paper introduces an image intelligence algorithm with graduated cylinders as primary subject of study. It employs a combination of YOLOv8 and U-Net models to address the intricate challenge of precisely recognizing graduated cylinder metrology in physical and chemical experiments. The verification results show that the average accuracy of the cylinder readings output by the algorithm reaches 0.3% of the total range. It also performs in scale and liquid level recognition within complex scenes commendably and the accuracy rate reached 98.95% within a 1.5 milliliter margin, 96.34% within a 1 milliliter margin and 93.73% within a 0.5 milliliter margin.
ISSN:2159-1288
DOI:10.1109/ICCEA62105.2024.10604242