가시화 기술을 이용한 식물플랑크톤의 탐지에 대한 연구
A Research on Phytoplankton Identification Using Visualization Technology
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
We analyzed the intrinsic wavelengths of two phytoplankton, Cochlodinium polykrikoides and Karenia mikimotoi, under specific external wavelengths. Spectral analysis showed that C. polykrikoides had a maximum radiance of 148, an effective range of 104–183, and an effective width of 69, with a narrow, tall cylindrical spectral shape. K. mikimotoi showed a maximum radiance of 147, an effective radiance range of 101–199, and an effective width of 84, with a narrow, steeply sloped conical spectral shape. Although both exhibited similar maximum brightness values, their spectral widths and shapes differed. An image deep learning experiment for red tide identification was conducted. The recognition rate for C. polykrikoides ranged from 52% to 93%, with an average of 82% and a standard deviation of 6.58%. For K. mikimotoi, the recognition rate ranged from 50% to 91%, with an average of 81% and a standard deviation of 7.23%. Validation experiments using a red tide sensor developed with 520 nm green and 620 nm red wavelengths confirmed its high potential for foreign substance detection at green wavelengths. The red tide detection sensor demonstrated maritime operational stability. Further experiments could enable the development of an artificial intelligence-based classification model to enhance the measurement accuracy.
Abstract
We analyzed the intrinsic wavelengths of two phytoplankton, Cochlodinium polykrikoides and Karenia mikimotoi, under specific external wavelengths. Spectral analysis showed that C. polykrikoides had a maximum radiance of 148, an effective range of 104–183, and an effective width of 69, with a narrow, tall cylindrical spectral shape. K. mikimotoi showed a maximum radiance of 147, an effective radiance range of 101–199, and an effective width of 84, with a narrow, steeply sloped conical spectral shape. Although both exhibited similar maximum brightness values, their spectral widths and shapes differed. An image deep learning experiment for red tide identification was conducted. The recognition rate for C. polykrikoides ranged from 52% to 93%, with an average of 82% and a standard deviation of 6.58%. For K. mikimotoi, the recognition rate ranged from 50% to 91%, with an average of 81% and a standard deviation of 7.23%. Validation experiments using a red tide sensor developed with 520 nm green and 620 nm red wavelengths confirmed its high potential for foreign substance detection at green wavelengths. The red tide detection sensor demonstrated maritime operational stability. Further experiments could enable the development of an artificial intelligence-based classification model to enhance the measurement accuracy.