Rice blast (Magnaporthe grisea) is an epidemic rice disease that reduces rice yield and quality worldwide. The objective of this study was to present and evaluate a data reconstruction method for assessment of rice leaf blast severity using hyperspectral imaging technology at the late vegetative growth stage. Experiments were carried out on Mongolian rice, which is susceptible to the disease. To carry out the study under natural conditions, rice was cultivated without any disease control measures. We obtained hyperspectral images of rice leaves and extracted average spectral reflectance data for entire leaves and undiseased leaf regions. To analyze the hyperspectral data, we presented a spectral reflectance ratio (SRR) data reconstruction method. A support vector machine model was constructed to identify five infection severities based on the transformed data. The classification accuracy of the model at jointing, booting and heading stages was 83.33%, 97.06% and 83.87%, respectively. According to our results, the SRR data reconstruction method presented here can be used to assess rice leaf blast severity during late vegetative growth.