<p>This study presents a rapid, non-destructive methodology for both small-scale geographic traceability and adulteration detection of camellia oil from Jiangxi Province, China, by coupling Fourier-transform near-infrared (FT-NIR) spectroscopy with orthogonal partial least squares-discriminant analysis (OPLS-DA). A total of 63 camellia oil samples (distributed as 10–12 samples per city) were collected from six major production cities (GanZhou, ShangRao, YiChun, JiAn, JiuJiang, and FuZhou) and tested in the range of 4000–10,000&#xa0;cm⁻¹ at 4&#xa0;cm⁻¹ resolution. Spectral data were preprocessed using Savitzky-Golay (SG) smoothing (window size was 7, SG-7) and evaluated via principal component analysis (PCA) for exploratory clustering, followed by partial least squares-discriminant analysis (PLS-DA) and orthogonal partial least squares-discriminant analysis (OPLS-DA) for supervised classification. To ensure model reliability and prevent overfitting, a sevenfold cross-validation strategy was employed. The SG-OPLS-DA achieved excellent fitting and predictive performance (R²X, R²Y, and Q² were 1.000, 0.970, and 0.949, respectively) and delivered 100% classification accuracy across all 189 validation spectra. Variable importance in projection (VIP) analysis identified two characteristic spectral intervals (6400–7300&#xa0;cm⁻¹ and 9500–10,000&#xa0;cm⁻¹) that, when combined, yielded a simplified model with equivalent discriminative power. Moreover, the proposed FT-NIR-OPLS-DA framework could sensitively detect adulteration levels as low as 10% when binary mixtures of GanZhou oil with oils from neighboring cities were evaluated.</p>