1：Experimental Design and Method Selection:

A fluorescence hyperspectral detection system was designed and fabricated, utilizing a 405 nm line laser for excitation and an in-house imaging spectrometer with a prism-grating-prism structure for detection. The system includes a motorized linear stage for scanning samples. Methods include spectral calibration, data preprocessing (background subtraction, stripe noise elimination, Savitzky-Golay smoothing), and data analysis using PCA and K-means clustering for oil type classification and linear regression for thickness estimation.

2：Sample Selection and Data Sources:

Three types of crude oil and their mixtures (as listed in Table 1) were used for component analysis. Diesel oil samples with thicknesses from 100 μm to 400 μm at 25 μm intervals were prepared by adding precise volumes to water in a Petri dish, mimicking oil spill conditions.

3：List of Experimental Equipment and Materials:

Equipment includes a 405 nm line laser (200 mW), dichroic beam splitter (DMLP425R, Thorlabs), longpass filter (FELH0450, Thorlabs), imaging lens (AC254-030-A, Thorlabs), slit, aspherical achromatic lenses (#49-665, Edmund), prism-grating-prism structure with grating (GT25-03, Thorlabs), CMOS camera (ASI74MM, ZWO), motorized linear stage, sample slot, calibration source (HG-1, Ocean Optics), and pipette for volume measurement. Materials include crude oil samples, diesel oil, and water.

4：Experimental Procedures and Operational Workflow:

The system was calibrated using a calibration source. For oil samples, hyperspectral data were acquired by moving the motorized linear stage at 20 mm/s, capturing images every 50 μm with 100 ms exposure. For thickness estimation, diesel oil samples were scanned with 50 images per sample at 0.5 mm intervals. Data preprocessing involved background subtraction, noise elimination, and smoothing. PCA reduced dimensionality, followed by K-means clustering for oil type distribution and linear regression for thickness-intensity relationship.

5：5 mm intervals. Data preprocessing involved background subtraction, noise elimination, and smoothing. PCA reduced dimensionality, followed by K-means clustering for oil type distribution and linear regression for thickness-intensity relationship. Data Analysis Methods:

5. Data Analysis Methods: PCA was implemented using nonlinear iterative partial least squares for dimensionality reduction. K-means clustering categorized spectra into oil types. Linear regression analyzed the relationship between fluorescence intensity and oil film thickness, with coefficient of determination R2 calculated.