China has huge rice reserves. In order to develop a rapid and accurate method for harmful mold infection detection in rice, near infrared (NIR) spectroscopy was applied for qualitative and quantitative analysis of the process of rice mildew in this study. Sterilized rice samples were firstly inoculated with four mold Aspergillus spp. species (A. flavus 3.17, A. flavus 3.3950, A. parastiticus 3.3950, A. glaucus 3.0100), respectively. Then the rice samples were stored under appropriate conditions (28 ℃, 80% RH) for mould growth. NIR spectra of samples were collected during the storage on different days (0, 2, 4, 7 and 10 d). Analysis models of mold infection in rice were developed by principal component analysis (PCA), discriminant analysis (DA) and partial least squares regression (PLSR), respectively. The results indicated that rice samples infected by different mold species could be effectively distinguished by NIR spectroscopy, and the average classification accuracy was 87.5%. The degree of mildew intensified during storage. The average correct classification accuracy of storage time (mildew degree) was found to be 92.5% for samples infected by one mold species, and 87.5% for samples infected by the four mold species. The PLSR prediction results of mould cell concentration in samples was: R2P=0.882 3, root mean square error of prediction (RMSEP)=0.339 Log (CFU·g-1) and residual predictive deviation (RPD)=2.93. Overall, the results demonstrated that the NIRS can be used as a rapid and non-destructive method for harmful mold infection detection in rice, ensuring the safety of grain storage and transportation.