Bananito (Musa acuminata, AA) fruit in three maturity stages (stages 2, 4 and 6) were investigated in this study. Correlation analysis was conducted between fruit firmness, soluble solids content (SSC) and several colour parameters. Point-measured spectroscopy (Vis-point) and spatial-measured hyperspectral imaging (Vis-HSI) were applied to collect visible spectra (400–740 nm) from the fruit peel. Three classification methods, k-nearest neighbour (k-NN), soft independence modelling of class analogy (SIMCA) and partial least square discriminate analysis (PLSDA), were applied for maturity stage classification. Results showed that a strong correlation was found between SSC and peel yellowness index (r = 0.92). Ripeness classification models developed using Vis-HSI data performed better than using Vis-point data. The best model based on PLSDA achieved a total correct classification rate of 93.3%. A simplified PLSDA model established on three wavelengths (650 nm, 705 nm and 740 nm) derived from the regression vector provided an equivalent model performance. This study demonstrated the use of hyperspectral imaging for accurate and non-destructive ripeness classification of bananito fruit based on the visible peel spectra, and the potential of using the three feature wavelengths to develop a multispectral imaging system for industrial application.