Detection and analysis of sweet potato defects based on hyperspectral imaging technology

• Spectral analysis was conducted to healthy and defective sweet potatoes. • Characteristic wavelengths were extracted from SNV spectra with MCUVE, RF and SPA. • Healthy, frostbitten and diseased sweet potatoes were classified by PLS-DA and LDA. In order to identify the defective sweet potato quickly, this study used hyperspectral imaging technology to classify healthy, frostbitten and diseased sweet potatoes. The training set and prediction set were divided according to the ratio of 3:1 by sample set partitioning based on joint x-y distances (SPXY) algorithm, and then the standard normal variable (SNV) pretreatment was carried out for the spectral data. In order to improve the running speed of the model and eliminate redundant variables, Monte Carlo uninformative variables elimination (MCUVE), Random frog (RF) algorithm and successive projections algorithm (SPA) were used to extract characteristic wavelengths. Finally, partial least square discrimination analysis (PLS-DA) and linear discrimination analysis (LDA) were used to establish classification models. The final results showed that the classification performance of the SPA-LDA model was the best, and the total accuracy of the prediction set reached 99.52%. In summary, hyperspectral imaging technology can accomplish the accurate detection of sweet potato defects, and provides feasible ideas for the automatic classification of sweet potatoes.