Soil organic matter (SOM) is one of the most important indexes to reflect the soil fertility, and soil moisture is a main factor to limit the application of hyperspectral technology in monitoring soil attributes. To study the effect of soil moisture on the accuracy for monitoring SOM with hyperspectral remote sensing and monitor the SOM quickly and accurately, SOM, soil water content (SWC) and soil spectrum for 151 natural soil samples in winter wheat field were measured and the soil samples were classified with the method of traditional classification of SWC and Normalized Difference Soil Moisture Index (NSMI) based on the hyperspectral technology. Moreover, the relationship among SWC, SOM and NSMI were analyzed. The results showed that the accuracy of spectral monitor for SOM among the classifications were significantly different, its accuracy was higher than the soils (5%～25%) which was not classified. It indicated that the soil moisture affected the accuracy for monitoring the SOM with hyperspectral technology and the study proved that the most beneficent soil water content for monitoring the SOM was less 10% and higher 20%. On the other hand, the four models for monitoring the SOM by the hyperspectral were constructed by the classification of NSMI, and its accuracy was higher than the classification of SWC. The models for monitoring the SOM by the classification of NSMI were calibrated with the validation parameters of R2, RMSE and RPD, and it showed that the four models were available and reliable to quickly and conveniently monitor the SOM by heperspectral. However, the different classifiable ways for soil samples mentioned in the study were naturally similar as all soil samples were classified again with another way. Namely, there may be another optimal classifiable way or method to overcome and eliminate the SWC effect on the accuracy for monitoring SOM. The study will provide some theoretical technology to monitor the SWC and SOM by remote sensing.