Relationship between wood and moisture content has always been a focus in wood science research field. Shrinkage or swelling that affects dimensional stability of wood occurs while moisture content changes. The problem is closely related to the end use of wood. Generally wood deformation is mainly depend on polysaccharides of wood chemical composition containing hydroxyl and water to form hydrogen bonding. Near-infrared spectroscopy (NIR) with a high sensitivity to organic materials containing hydrogen groups will be a useful tool to realize on-line rapid detection on shrinkage or swelling of wood. Relationships between moisture contents and size changes of wood were investigated in this research by using NIR technology, furthermore the predicted models of size changes induced by moisture content were built. The near infrared spectra coupled with chemometric techniques were collected on three sections of wood. Prediction models of size change of wood were constructed on base of partial least squares and corresponding cross-validation. The results showed that size changes of wood in tangential and radial section under several different moisture contents had a high correlation with the corresponding near infrared spectra. It is feasible to detect size change of wood by NIR. On the other hand determine coefficients (R2) of prediction models in tangential and radial direction are over 0.90, which are satisfactory to predict wood deformation under different moisture content conditions. Additionally the prediction model of size change in tangential is better than that of in radial direction. The above results showed size changes induced by moisture content could be predicted quickly and accurately by NIR technology.