Estimation of surface roughness for transparent superhydrophobic coating through image processing and machine learning. Issue 1 (10th January 2022)

• Record Type: Journal Article
• Authors: Hooda, Amrita; Kumar, Adesh; Goyat, Manjeet Singh; Gupta, Rajeev
• Is Part Of: Molecular crystals and liquid crystals. Volume 726:Issue 1(2021)
• Page Start: 90
• Page End: 104
• Publication Date: 2022-01-10
• Subjects: Digital image processing -- electron microscopy -- K-means clustering -- machine learning -- superhydrophobic coating -- surface roughness; Molecular crystals -- Periodicals; Liquid crystals -- Periodicals; Liquid crystals; Molecular crystals; Periodicals; 548
• DOI: 10.1080/15421406.2021.1935162
• Languages: English
• ISSNs: 1542-1406

Abstract

Abstract: In the current era, superhydrophobic surfaces/coatings have gained significant attention worldwide due to their exclusive features such as self-cleaning, anti-corrosion, anti-adhesion, anti-reflection, and anti-icing, etc. The idea of the self-cleaning mechanism of superhydrophobic coatings has emerged from the self-cleaning effect of lotus plant leaves. The superhydrophobic surfaces have a great ability to eliminate dust, bacteria, and viruses due to the very large contact angle (> 150°) between the surface and the water droplets. The present study is based on the surface roughness estimation of field emission scanning electron microscope (FESEM) images of the developed superhydrophobic coatings via image processing and machine learning approach. Transparent superhydrophobic coatings of functionalized SiO2 nanoparticles embedded polystyrene (PS) and dual functionalized ZnO nanoparticles embedded PS were prepared using a modified sol-gel approach. The superhydrophobicity of the synthesized coatings was realized by the large contact angles of more than 150 ° between water droplets and the coatings. The FSESM images of the superhydrophobic coatings were processed using MATLAB 2018 image processing and machine learning tool to compute the roughness by computational algorithms. The discrete wavelet processing was used for image segmentation, and k-means clustering was applied for predicting the roughness score against different compositions of the coatings. The computational methods exhibited ∼ 91.70% accuracy of the surface roughness estimation of the coatings.