5:15 PM - 6:45 PM
[MIS09-P14] Preliminary Result in Classifying Plastic Marine Debris in Ocean Using Spectral Information
Keywords:Plastic marine debris, Spectropolarimetric, Polarization, Support Vector Machine, Classification
The annual production of plastic worldwide has reached a staggering 400 million tons, with a significant amount reaching the marine environment due to human activity. Plastic marine debris has become an essential global concern that requires immediate attention. Although visual inspections and in situ measurements using calibrated equipment are adequate, they are time-consuming and require a substantial workforce. Remote sensing through air- or space-borne spectral imaging is an alternative; however, it requires a comprehensive understanding of the spectral properties of plastics in the ocean. This is a significant challenge because of the camera and light angle, water conditions, plastic debris characteristics, and nonplastic materials.
Spectropolarimetric measurements were conducted to classify plastic samples, nonplastics, and marine debris. The samples were collected from different locations in Japan: (a) Nokobiura Bay, Goto Islands, Nagasaki; (b) Aso-beach Ishikari, Otaru, Hokkaido; and (c) Amami Oshima, Kagoshima. A portable spectrometer camera with a wavelength range of 406-1147 nm and a 1.8 nm resolution was used for the measurements. A polarizer was attached to the camera, manually rotating from 0-90°±45°. Different sample conditions were included in the measurements to simulate the lifetime of the samples. Finally, the spectral data were assigned to a spectral library.
The spectral dataset was subjected to a classification algorithm (RF and SVM) using a feature selection procedure that employed the recursive feature elimination (RFE) method. The RFE method was used to determine n features (n = 4,5,6,7,8) based on the number of wavelengths. The final goal of this study is to incorporate a polarization and bidirectional reflectance distribution function (BRDF) using wavelength information to simplify the camera system and enable rapid identification of objects in future drones or satellite installations. This research was supported by Dr. Shin’ichiro Kako of Kagoshima University, Dr. Daisuke Matsuoka of the Japan Agency for Marine-Earth Science and Technology (JAMSTEC), and Mr. Shinya Kimura.
Spectropolarimetric measurements were conducted to classify plastic samples, nonplastics, and marine debris. The samples were collected from different locations in Japan: (a) Nokobiura Bay, Goto Islands, Nagasaki; (b) Aso-beach Ishikari, Otaru, Hokkaido; and (c) Amami Oshima, Kagoshima. A portable spectrometer camera with a wavelength range of 406-1147 nm and a 1.8 nm resolution was used for the measurements. A polarizer was attached to the camera, manually rotating from 0-90°±45°. Different sample conditions were included in the measurements to simulate the lifetime of the samples. Finally, the spectral data were assigned to a spectral library.
The spectral dataset was subjected to a classification algorithm (RF and SVM) using a feature selection procedure that employed the recursive feature elimination (RFE) method. The RFE method was used to determine n features (n = 4,5,6,7,8) based on the number of wavelengths. The final goal of this study is to incorporate a polarization and bidirectional reflectance distribution function (BRDF) using wavelength information to simplify the camera system and enable rapid identification of objects in future drones or satellite installations. This research was supported by Dr. Shin’ichiro Kako of Kagoshima University, Dr. Daisuke Matsuoka of the Japan Agency for Marine-Earth Science and Technology (JAMSTEC), and Mr. Shinya Kimura.