In this paper， the changes of polysaccharide content in the ultrasonic extraction process were studied. At the same time， the spectral information of the extraction process was collected by near-infrared in situ real-time monitoring technique. The mathematical model was established by the relationship between spectral information and chemical value. The results showed that the content of polysaccharide increased in the process of ultrasonic extraction and then fluctuated in 60 min. After pretreatment of the acquired primitive spectra， SG was the most efficient pretreatment method with the highest Rc， Rp value and the least RMSECV， RMSEP in all pretreatments. After modeling with PLS， iPLS and si-PLS methods， the model established by si-PLS method is the best， which combined four optimal wavelength ranges are 1 258.591-1 316.735 nm， 2 027.121-2 084.226 nm， 2 153.93-2 210.888 nm and 2 337.239-2 387.696 nm. Under the optimum model， the Rc， Rp were 0.9621， 0.9667 and RMSECV， RMSEP were 0.0116， 0.0137， respectively. In conclusion： the near-infrared spectroscopy can be used to predict the content of polysaccharide in the ultrasonic extraction of maca polysaccharide and the in-situ monitoring during the extraction process is realized.