Thesis
HYPERSPECTRAL IMAGE ANALYSIS TO IDENTIFY OF COMPONENTS IN MUNICIPAL SOLID WASTE
Washington State University
Master of Science (MS), Washington State University
05/2024
DOI:
https://doi.org/10.7273/000006921
Abstract
This thesis explores the application of hyperspectral imaging techniques for the classification of plastic pollution in water. The study utilizes a double-camera setup covering the electromagnetic spectrum from visible (VIS) to shortwave infrared (SWIR) ranges. In particular, the Specim FX17 camera captures images in the NIR-SWIR range, offering a high-resolution spectral view of the samples.
The preprocessing of the acquired data involves several steps, including the combination of white and dark reference images, masking to isolate the target plastics, and patch extraction to extract spectral information. These processes are crucial for enhancing the quality of the dataset and preparing it for further analysis.
One of the significant challenges encountered in the study is the presence of background noise in the images. To address this issue, manual masking techniques are employed to isolate the plastic samples from other materials present in the water. However, improvements in automated masking techniques are identified as an area for future research.
The classification process involves training a convolutional neural network (CNN) on the preprocessed data. The network is trained to classify different types of plastics, including PET, PP, and others. Results indicate the effectiveness of the CNN in accurately identifying and classifying plastic samples based on their spectral signatures.
Further experiments involve the analysis of plastic samples obtained from a recycling company using infrared spectroscopy. The dataset consists of spectral data collected from various plastic samples, each represented as a 6 x 9 x 690 image. The study explores different patch extraction techniques to maximize the information extracted from the dataset.
Overall, this thesis highlights the potential of hyperspectral imaging combined with machine learning techniques for the classification of plastic pollution. It identifies key challenges and proposes avenues for future research, including improvements in data collection methods, automated masking techniques, and the development of more sophisticated classification algorithms.
Metrics
11 File views/ downloads
25 Record Views
Details
- Title
- HYPERSPECTRAL IMAGE ANALYSIS TO IDENTIFY OF COMPONENTS IN MUNICIPAL SOLID WASTE
- Creators
- Macy Christianson
- Contributors
- John Miller (Chair)Luis De La Torre (Committee Member)Scott Hudson (Committee Member)Kaushik Thallapally (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- School of Engineering and Applied Sciences (TRIC)
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University
- Number of pages
- 69
- Identifiers
- 99901125140401842
- Language
- English
- Resource Type
- Thesis