1：Experimental Design and Method Selection:

The study used hyperspectral imaging in the VIS/NIR region (400-1100 nm) to acquire images of pork samples. Two regression models, PLSR and SVR, were built and compared for predicting TVC, with various preprocessing methods applied to improve accuracy. An image processing algorithm was developed for real-time visualization.

2：Sample Selection and Data Sources:

Fifty pork tenderloin samples were purchased from a local supermarket, sliced into pieces, and stored at 4°C for up to 15 days. Samples were selected randomly over time for analysis.

3：List of Experimental Equipment and Materials:

Hyperspectral imaging system including CCD camera (Sensicam QE), imaging spectrograph (ImSpector V10E), illumination unit with quartz tungsten halogen lamp, laser displacement sensor (CD33-120N-422), computer with software (Camera Control Kit V2.19, Microsoft Visual Studio 2010, OpenCV 2.4.10), translation stage (AH-STA03300), and materials like pork samples, phosphate buffer solution, culture medium, etc.

4：19, Microsoft Visual Studio 2010, OpenCV 10), translation stage (AH-STA03300), and materials like pork samples, phosphate buffer solution, culture medium, etc. Experimental Procedures and Operational Workflow:

4. Experimental Procedures and Operational Workflow: Samples were scanned line by line with the hyperspectral system to acquire hypercubes. Spectral data were extracted, corrected for dark and white references, and processed using MATLAB and ENVI. TVC was measured using the standard microbiological plating method. Models were built and validated with calibration and prediction sets.

5：Data Analysis Methods:

Statistical analysis included correlation coefficients (R) and standard errors (SE) for model evaluation. Preprocessing methods like SNV, MSC, derivatives, etc., were applied. Visualization was done by applying the model to each pixel in the image.