[Identification of spilled oil by NIR spectroscopy technology based on sparse nonnegative matrix factorization and...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning

Context:

• Accurate identification of spilled oil is crucial for effective environmental remediation and forensic analysis.
• Traditional methods for oil spill identification can be time-consuming and may lack specificity.
• Developing rapid and reliable methods for distinguishing between different oil types is an ongoing challenge.

Purpose:

• To propose and validate a novel method for the rapid and accurate identification of different types of spilled oil.
• To leverage Fourier transform near-infrared (FT-NIR) spectrophotometry for spectral data acquisition.
• To employ Sparse Nonnegative Matrix Factorization (SNMF) for feature extraction and Support Vector Machine (SVM) for classification.

Summary:

• A novel approach combines FT-NIR spectrophotometry with SNMF for feature extraction and SVM for classification to identify spilled gasoline, diesel, and kerosene.
• The SNMF-SVM model was trained on 210 samples and validated on 90 samples, achieving an identification accuracy of 97.783%.
• The study investigated the influence of feature number and sparseness factor on model performance, demonstrating the method's robustness.

Impact:

• Provides a new, rapid, and highly accurate method for identifying spilled oil types.
• Enhances capabilities for environmental oil spill response, treatment planning, and source tracing.
• Demonstrates strong generalization ability, making it a valuable tool for real-world applications in oil spill management.