<p>Chickpea flour is used in a variety of culinary preparations to manufacture foods with high protein content. Chickpea flour’s protein content is commonly measured using the Dumas method, but this process is time-consuming, expensive, and labor-intensive. This study employed near-infrared (NIR) hyperspectral imaging to predict the protein content of chickpea flour. Eight different chickpea varieties with different protein contents were processed into chickpea flour. Chickpea flour samples were subjected to NIR reflectance hyperspectral imaging in the 900–2500&#xa0;nm spectral region. Using the Dumas combustion method, the protein content of twenty-four samples of chickpea flour (8 var. × 3 replications) was determined. The spectral data of the chickpea flour samples and the observed reference protein content (dependent variables) were correlated. Out of a total of 24 samples, 16 powder samples were used to build the calibration model, and 8 powder samples were used to build the prediction model. While using the full spectrum, the optimum protein prediction model was obtained using Partial least square regression (PLSR) and orthogonal signal correction (OSC) + standard normal variate (SNV) preprocessing, which resulted in correlation coefficient of prediction (R²p) and root mean square error of prediction (RMSEP) values of 0.934 and 1.006, respectively. Additionally, 11 feature wavelengths were chosen from the investigated spectrum using competitive adaptive reweighted sampling (CARS), which also yielded the best PLSR model with R²p and RMSEP of 0.944 and 0.889, respectively. By combining PLSR, OSC + SNV, and CARS chosen wavelength, the optimum model for protein content prediction in chickpea flour was established.</p>