Abstract: Brassica napus L leaf diseases could cause seriously reduction in crop yield and quality. Early diagnosis of Brassica napus L leaf diseases plays a vital role in Brassica napus L growth. To explore an effective methodology for diagnosis of Sclerotinia infected Brassica napus L plants, healthy Brassica napus L leaves and Brassica napus L leaves infected by Sclerotinia were prepared in a controlled circumstance. A visible/short-wave near infrared hyperspectral imaging system covering the spectral range 380-1030 nm was set up to identify healthy and infected Brassica napus L leaves. The clear and non-deformable hyperspectral images were captured and the spectral information was exacted from the hyperspectral images according to the predefined region of interest (ROI). The spectral ranges of 380-439 nm and 951-1030 nm which contained obvious noises were removed. Moving average was used as the pretreatment method for spectra to remove noises. Chemometric methods were applied to build classification models for healthy and infected Brassica napus L leaves identification, including Principal Component Analysis (PCA), Partial Least Squares Discriminant Analysis (PLS-DA) and Support Vector Machine (SVM) models. Weighted regression coefficient (Bw) was applied to select sensitive wavelengths, and PLS-DA model and SVM model were also built on the basis of the sensitive wavelengths for healthy and infected leaves identification. For the infected leaves, the healthy part, the infected part and the joint part of the healthy part and the disease part were analyzed separately to investigate which part was most efficient for disease identification by PLS-DA, and the results showed that the infected part obtained better identification results than the other parts. For both full spectra and the sensitive wavelengths, PLS models and SVM models showed good performances with identification rate over 85%, and PLS models using the spectra extracted from the infected part obtained the identification rate over 90%. The classification models using sensitive wavelengths showed similar or better performances compared with the classification models using full spectra, indicating that selected sensitive wavelengths could be used for Brassica napus L disease diagnosis with fewer input variables. The overall results indicated that Brassica napus L leaf diseases could be early diagnosed by hyperspectral imaging and multivariate techniques effectively, which would be helpful for the prevention and treatment of Brassica napus L diseases. The results were gained in laboratory, and to obtain more accurate and practical results, the hyperspectral imaging system for field application should be developed, and more experiments should be conducted to select more accurate and stable sensitive wavelengths and build more robust identification models.