In order to visualize the spatial distribution of potato dry matter, the internal dry matter content of potato was studied by using visible/near-infrared hyperspectral imaging (HSI) and a detection model of dry matter of potato was established. The reflectance spectra of sliced potatoes which were extracted from the regions of interest of HIS were performed with different pretreatments. The standard normal variable (SNV) combined with Savitzky-Golay smoothing (SG) and the first derivative (SNV-SG-FD) was the optimal pretreatment. Based on optimal pretreatment, competitive adaptive reweighted sampling (CARS) combined with successive projections algorithm (SPA) was used to select variables of the spectrum and obtained 22 variables. Three regression models based on principal component regression (PCR), support vector regression (SVMR) and partial least squares regression (PLSR) were established. The best performance was achieved by PLSR model, its determination coefficient (R2P), root mean square error for prediction and relative percent difference were 0.849, 0.878% and 2.312, respectively. The PLSR model based on 22 variables was superior to the full-spectrum model. An imaging processing algorithm was developed to transfer each pixel in potato dry matter content with the SNV-SG-FD-CARS-SPA-PLSR model. The imaging showed the distribution of dry matter within the potatoes. It showed that the potato dry matter was mainly distributed between the inner pith and vascular bundle and the inner pith had the lowest dry matter content. It was gradually increased from the inner pulp to the outer. Dry matter content was 12.16% in inner pith and the outer layer reached up to 24.62%. The results show that the visible near infrared hyperspectral imaging is a useful tool for rapidly and effectively visualizing detecting spatial distribution of potato dry matter.