Hyperspectral Analysis of Plastic Particles in the Ocean
MetadataVis full innførsel
Huge amounts of plastics end up in the ocean each year. Some of the plastics are eaten by fish and birds, and can cause death. It is important to find an easy, fast and reliable method to detect the plastics, as a first step towards a solution for removing these from the ocean. Hydrocarbons (C-H bindings) are found in all types of plastics, and their absorption properties are significant in the short wave infrared (SWIR) region between 1000nm to 2300nm. A hyperspectral camera operating in the SWIR region was chosen to record these optical properties. Samples of plastic have been collected and imaged in a laboratory setup. The plastics were imaged alone and together with seawater. The hyperspectral images have been inspected and analyzed to find the main spectral features and to identify the plastics and the water correctly. The analyzing methods used were spectral angle mapper (SAM), principal component analysis (PCA) and partial least squares-discriminant analysis (PLS-DA). The absorption peaks of the plastic types were found to match the absorption peaks for C-H-bindings reported in the literature. The absorption peaks were found at 1100-1225nm, 1300-1420nm, 1650-1800nm, and 2200-2450nm. SAM proved to be the most consistent method for identification of the different types of plastics, and was able to correctly identify the materials in the images. The analysis using PCA showed an inconsistency between the images, and was not always able to distinguish between plastic and water. PLS-DA worked well for some samples. The data basis acquired was too small to achieve a training set for PLS-DA that were big enough to get consistently good predictions. To make hyperspectral imaging of plastics in the ocean more applicable, the mentioned analysis methods were also used to analyze images made up of three spectral bands (instead of the original 256). Four bands were chosen from the mean spectral features observed for the different plastics, and were found at 1162nm, 1392nm, 1684nm, ans 2281nm. Using the band at 1392nm and two of the mentioned bands, the analysis showed promising results to identify the plastics from the water. In the ocean and on the shoreline, other materials like seaweed, vegetation, glass and cardboard are found. These spectra have been compared to the plastics, which have significantly different features than the other materials. Therefore, hyperspectral imaging is an applicable method to detect plastics in the ocean.