Abstract: The feasibility of using near-infrared spectroscopy (NIR) to measure the basic density of wood in the field was studied in this paper.The discs of elm wood were used as samples to simulate the sawn surface of the felled tree,NIR were collected from the surface of the disc.The NIR prediction model of basic density of elm trees was developed by partial least squares (PLS).The determination coefficient (R2) of the calibration and validation models was 0.845 6 and 0.801 1,respectively.The root mean square error of calibration (RMSE) was 0.023 1 and 0.026 6 with standard error of 0.023 2 and 0.026 8.To improve the accuracy of the prediction model,savitzky-golay,wavelet transform and other four pretreatment methods were applied to the raw spectrum.The best model was achieved by wavelet denoising.The determination coefficient (R2) was 0.899 6 and 0.866 2 for the calibration and validation model,respectively.The RMSE and SE both went to the minimum value.The results showed that NIRS can be used in predicting wood basic density in the field.