Cumin is a valuable spice for medical and food applications. Nevertheless, reports of the presence of undeclared adulterants in cumin require the development of analytical methods for its authentication. The goal of this study was to develop a new analytical method based on NIR-HSI (900–1710 nm) and chemometrics to detect low-cost adulterants peanut shell, pecan shell and walnut shell in cumin. PCA was applied to investigate the spectral features of pure samples and mixtures. Soft Independent Modelling Class Analogy (SIMCA) was applied to classify pure cumin and adulterated samples, achieving an accuracy of 95 % for test samples. Partial Least Square Regression (PLSR) model based on selected variables using iPLS or GA showed a similar (for walnut shell) or better (peanut and pecan shell) performance than PLSR models based on a full wavelength with detection limits above 1% and RPD (Residual Prediction Deviation) values higher than 5, indicating excellent predictive ability. Chemical maps allow visualization of shell concentration and distribution in cumin samples. This work demonstrated the potential of NIR-HSI and chemometrics to detect and quantify nutshells in cumin powder.