Spatial graph structure estimation of nanoparticles using centroid-to-contour distance analysis and deep encoder framework
- Authors
- Jang, Jaeuk; Lee, Hyunsoo
- Issue Date
- Jun-2023
- Publisher
- SPRINGER
- Keywords
- Nanoparticle structure estimation; Spatial graph; Deep learning-based encoder; Centroid-to-contour distance analysis; Deep learning
- Citation
- JOURNAL OF NANOPARTICLE RESEARCH, v.25, no.6
- Journal Title
- JOURNAL OF NANOPARTICLE RESEARCH
- Volume
- 25
- Number
- 6
- URI
- https://scholarworks.bwise.kr/kumoh/handle/2020.sw.kumoh/21823
- DOI
- 10.1007/s11051-023-05772-9
- ISSN
- 1388-0764
- Abstract
- Nanotechnology is increasingly being used in several application areas for deriving material properties through nanostructure control. Most nanoparticle images are obtained using high-magnification electron microscopy. However, processing numerous nanoparticle images and improving accuracy requires considerable time and labor. A number of studies focused on precisely assessing the overlapping features of nanoparticle images. For correlation and reliable discrimination of overlapped nanoparticles, this study proposed a framework using a deep-learning-based encoder and centroid-to-contour distance analysis. In the proposed framework, specialized preprocessing was used to divide an image into many images, and a convolutional neural network (CNN)-based encoder was used to derive the properties of nanoparticles included in the image. The resulting characteristics were used to estimate the center point of each nanoparticle through the proposed centroid-to-contour distance approach. Then, a nanoparticle spatial structure graph was generated using this analysis. The nanoparticle structure graph of the proposed framework clearly maps the overlapping relationship between nanoparticles. Experiments proved that the proposed framework more accurately assesses the nanoparticle structure than conventional algorithms do.
- Files in This Item
- There are no files associated with this item.
- Appears in
Collections - School of Industrial Engineering > 1. Journal Articles
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.