Berlin 2024 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 3: Poster

AKPIK 3.2: Poster

Donnerstag, 21. März 2024, 11:00–14:30, Poster B

Estimating sliding drop width using recurrent neural networks — •Sajjad Shumaly¹, Fahimeh Darvish¹, Xiaomei Li¹, Oleksandra Kukharenko¹, Werner Steffen¹, Yanhui Guo², Hans-Jürgen Butt¹, and Rüdiger Berger¹ — ¹Max Planck Institute for Polymer Research, Ackermannweg 10, D-55128, Mainz, Germany — ²Department of Computer Science, University of Illinois Springfield, Springfield, IL, USA

Recording videos serves as a technique for monitoring objects and researching physical phenomena through image processing. Challenges emerge when dealing with soft matter objects such as sliding drops, which exhibit variations in size. Adding additional cameras or mirrors to track drop size variation from the front view can be inconvenient and limit the field of view. This limitation can impede a comprehensive analysis of sliding drops, especially when dealing with scenarios that entail surface defects. Our study explores the use of various regression and multivariate sequence analysis models to estimate drop/solid contact width (drop width) solely from side-view videos. The long short term memory (LSTM) model obtains an RMSE value of 67 um. Within the spectrum of drop widths in our dataset, ranging from 1.6 mm to 4.4 mm, this RMSE indicates that with our approach we can predict the width of sliding drops with an error of 2.4%.

Keywords: Sliding drops; Drop width estimation; Multivariate sequence analysis; recurrent neural network (RNN),; Long short-term memory (LSTM)