In order to investigate the feasibility of visible/Near Infrared(Vis-NIR)spectroscopy to predict the surface roughness of natural polymer material(wood) as a non-contact measurement method,the correlations between Vis-NIR spectroscopy and surface roughness measured by contact(stylus) instruments from three different sections of wood samples were analyzed. The results showed that the surface roughness parameters, arithmetical mean deviation of profile (Ra), ten-point height of irregularities (Rz) and the maximum height of profile (Ry), of wood samples were successfully predicted by using Vis-NIR (400~2 500 nm) models from the three sections of the samples. The correlations between values measured by the stylus instruments and the values predicted by the models were good. The correlation coefficients of Rz reached up to 0.92. Compared to the models based on the Vis-NIR from the radial section and tangential section of the samples, the predictive effect of the model based on cross section was the best. The correlation coefficients between the values measured by the stylus instruments and the values predicted by the models based on different spectrum wavelength range, 400~780, 780~1 100, 1 100~2 500, 780~2 500 and 400~2 500 nm, were generally above 0.80. The prediction results of the model based on spectrum wavelength range 400~2 500 nm was better than the models based on the other spectrum wavelength ranges. The results showed that the predictive effect was not improved by pretreatment of the spectrum. It is proposed to use the original spectrum to predict the surface roughness of natural polymer material.