In order to quickly and nondestructively detect the content of whole potato powder in potato noodles, the hyperspectral imaging technology was used to quantitatively detect the content of whole potato powder in noodles. Totally 120 noodle samples with a total potato flour content of 0~35% were selfmade, and hyperspectral images of the noodles were collected in the 900~2500nm spectral range. Totally 80 samples were randomly selected as the calibration set, and the original spectra and the spectra preprocessed by moving average, smoothing S-G, baseline, normalize, standard normalized variate, and multiplicative scattering correction were used to establish the partial least squares regression model, principal component regression model and support vector machine regression model. The results showed that the partial least squares regression modeling effect was the best after the standardized preprocessing method. The coefficient of determination of the calibration set (R2C) was 0.8653, the coefficient of determination of the cross validation set (R2CV) was 0.6914. The characteristic wavelength was extracted from the spectral data preprocessed by normalize by regression coefficient method, and a simplified model of potato powder content PLSR was established. The coefficient of determination of the calibration set (R2C) was 0.8685. The validation set determination coefficient (R2CV) was 0.8021. The results showed that the model based on the characteristic wavelength was better than the fullband model. Using the remaining 40 samples as the prediction set, the NormalizePLSR simplified prediction model was established based on the characteristic wavelength. The coefficient of determination of the prediction set (R2P) was 0.8456. The model had good prediction ability. The results showed that it was feasible to use hyperspectral imaging technology to detect the total potato flour content in noodles.