1：Experimental Design and Method Selection:

The study used hyperspectral imaging (HSI) in the 900-1700 nm range to evaluate impact damage in mangoes caused by dropping from heights of 0.5, 1.0, and 1.5 m. Prediction models for quality attributes (pulp firmness, total soluble solids, titratable acidity, chroma) were developed using partial least squares (PLS) regression, with spectral preprocessing methods including Savitzky-Golay smoothing, standard normal variate transformation, and multiplicative scatter correction. Competitive adaptive reweighted sampling (CARS) was used for wavelength selection, and discriminant analysis (DA) for classification based on ripening index (RPI).

2：5, 0, and 5 m. Prediction models for quality attributes (pulp firmness, total soluble solids, titratable acidity, chroma) were developed using partial least squares (PLS) regression, with spectral preprocessing methods including Savitzky-Golay smoothing, standard normal variate transformation, and multiplicative scatter correction. Competitive adaptive reweighted sampling (CARS) was used for wavelength selection, and discriminant analysis (DA) for classification based on ripening index (RPI). Sample Selection and Data Sources:

2. Sample Selection and Data Sources: 240 unripe mangoes from the same batch were purchased from a local market in Tianjin. They were divided into groups: 20 undamaged and 60 damaged (20 per dropping height). Hyperspectral images were acquired 24 hours after damage and every other day for three days.

3：List of Experimental Equipment and Materials:

Hyperspectral imaging system (Imspector N17, Spectral Imaging Ltd.), drop test machine (PD-315A, Suzhou New District Dongling Vibration Testing Instrument Co., Ltd.), texture analyzer (TA.XT plus, Stable Micro Systems Ltd.), portable refractometer (PAL-1, Atago), colorimeter (Ultrascan Pro, Hunter Associates Laboratory, Inc.), and standard chemicals for titration (e.g., NaOH, phenolphthalein).

4：Experimental Procedures and Operational Workflow:

Mangoes were dropped from specified heights, stored at 20°C and 42% RH. Hyperspectral images were captured after calibration (black and white). Quality attributes were measured: pulp firmness via puncture test, TSS via refractometer, TA via titration, chroma via colorimeter. Spectra were preprocessed, key wavelengths selected, and PLS models developed for prediction. Classification was done using DA based on RPI.

5：Data Analysis Methods:

PLS regression for prediction models, with performance evaluated using coefficient of determination (R2) and root mean square error of prediction (RMSEP). Discriminant analysis for classification accuracy. Statistical analysis included mean and standard deviation calculations.