Improving the accuracy of spectroscopic identification of geographical origins of agricultural samples through cooperative combination of near infrared and laser-induced breakdown spectroscopy
- Authors
- Eum, Changhwan; Jang, Daeil; Kim, Jonghyun; Choi, Sanghoi; Cha, Kyungjoon; Chung, Hoeil
- Issue Date
- Nov-2018
- Publisher
- PERGAMON-ELSEVIER SCIENCE LTD
- Keywords
- Laser induced breakdown spectroscopy; Near-infrared spectroscopy; Discrimination of geographical origin; Milk vetch root; Support vector regression; Sensitivity Analysis
- Citation
- SPECTROCHIMICA ACTA PART B-ATOMIC SPECTROSCOPY, v.149, pp.281 - 287
- Indexed
- SCIE
SCOPUS
- Journal Title
- SPECTROCHIMICA ACTA PART B-ATOMIC SPECTROSCOPY
- Volume
- 149
- Start Page
- 281
- End Page
- 287
- URI
- https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/149054
- DOI
- 10.1016/j.sab.2018.09.004
- ISSN
- 0584-8547
- Abstract
- As a versatile strategy to improve accuracy for identification of the geographical origin of agricultural samples, milk vetch root samples in this study, both near-infrared spectroscopy (NIRS) and laser-induced breakdown spectroscopy (LIBS) have been cooperatively combined. The motivation was based on the potential of accuracy improvement by utilization of these two methods providing complementary spectral information, compositions of organic compounds and elements. For initial evaluation, NIRS and LIBS were separately employed to discriminate domestic milk vetch root samples from imported ones using support vector machine (SVM). The near infrared (NIR) information in a full spectral range was used for the analysis, while in LIES spectra, the intensities of 35 selected discrete element peaks were used. The use of NIR information providing organic compositions of the samples resulted in a discrimination accuracy of 91.5%, better than that of using LIBS elemental peak intensities (73.1%). Next, to utilize both sets of spectral data for discrimination, support vector regression (SVR) was used to represent NIR spectral feature of a sample as a SVR coefficient, and then it was merged with the existing discrete LIBS intensity data; accuracy was improved to 95.8%. The cooperative combination of information on organic and elemental composition of the samples was the root of improvement.
- Files in This Item
-
Go to Link
- Appears in
Collections - 서울 자연과학대학 > 서울 화학과 > 1. Journal Articles
- 서울 자연과학대학 > 서울 수학과 > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/149054)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.