Versatile Recognition of Graphene Layers from Optical Images Under Controlled Illumination Through Green-Channel Correlation Method

Abstract

The proposed method for identifying the number of exfoliated graphene layers on an oxide substrate from optical images is both simple and versatile. It involves using a limited number of input images for training, paired with a larger set of well-published Github images for testing and prediction. This study has employed a linear regression-based method in executing its thresholding process. This method leverages the red, green, and blue color channels of image pixels and establishes a correlation between the green channel of the background and the green channel of various graphene layers. The method is positioned as an alternative to both deep learning-based graphene recognition and traditional microscopic analysis. Notably, the proposed methodology performs well under conditions where the influence of surrounding light on the graphene-on-oxide sample is minimal. It enables the rapid identification of various graphene layers, showcasing its feasibility for non-destructive identification. The study also addresses the functionality of the methodology under nonhomogeneous lighting conditions, demonstrating successful predictions of graphene layers even in lower-quality images compared to those typically published in literature. In summary, the proposed methodology offers a quick, inexpensive, and effective means of non-destructively identifying graphene layers from optical images. Its versatility and performance under varying conditions make it a promising approach for practical applications in graphene research. Additionally, and critically: the methods highlighted in this research can be utilized across a multitude of disciplines (from bioengineering to electrical, materials, nanoengineering, etc.) for one of the most fundamental areas of experimental research in STEM at the undergraduate level: accurately identifying multiple systems from optical images. A broad, relevant, and timely curriculum can be built around data analytics and application to solving STEM problems - including components such as data mining, cleaning, wrangling, and analysis, and critically, tying in these processes in solving real experimental challenges in a laboratory setting.