To detect the concentration of Sudan red No. 1 (C) in chili powders non-destructively, the hyper-spectral imaging technique was used to extract 120 chili powder images that had different C, including 60 for calibration set and rest for prediction set. Successive Projections Algorithm (SPA) was used to extract 45 effective wavelengths from large amount of spectra data in calibration set. Partial Least Squares Regression (PLSR), Multiple Linear Regression (MLR) and Principal Component Regression (PCR) were applied to set prediction models. As a result, better model to detect the C in chili powders was based on MLR. The Correlation Coefficient Calibration (Rc) and Root Mean Square Error of Calibration (ERMSC) were 0.998 and 0.737 3 μg/ml. The Correlation Coefficient Prediction (Rp) and Root Mean Square Error of Prediction (ERMSP) was 0.987 and 1.921 3 μg/ml. This demonstrates that the method can be successfully used for detecting the C in chili powders non-destructively.