288 rice straw samples collected from different regions and varieties were used to study the feasibility of rapidly detecting the content of soluble sugar in rice straw with the near infrared reflectance spectroscopy (NIRS) technique.The near infrared spectral information of samples were collected within the near infrared wavelength range (10 000-4 000 cm-1).The models of quantitative analysis based on near infrared spectra of stepwise multiple linear regression (SMLR),partial least squares regression (PLS) and principal component regression (PCR) were established using stoichiometric algorithm SMLR，PLS and PCR，combined with different spectral pretreatments including multiplicative scatter correction (MSC),standard normal variation transformation (SNV),derivative,S-G smoothing and their combinations.Through comparison and analysis, the optimal effect of PLS model which established by using first derivative spectra pretreatment had a determination coefficient R2C between the calibration set chemical analysis values and predicted values of 0.880 6， the determination coefficient R2CV and R2V of 0.771 1,0.857 8, and root mean square difference RMSEC, RMSECV, RMSEP of 0.318%，0.440%，0.404%, respectively. Both of the relative analysis error RPDC and RPDV are greater than 2.5. The results show that establishing the model by using near infrared spectroscopy, combined with PLS modeling method can quickly detect the content of soluble sugar in rice straw.