Identification Of Damaged Potatoes Based On Hyperspectral Imaging Techniques And Classification Of Damage Degrees

Damage potato is difficult to be detected in the process of damage identification and is perishable in storage,thus leading to a serious problem of food safety and economic issue.Therefore,a nondestructive detection method,based on hyperspectral imaging technique,has been proposed in this study to realize the identification of damaged potatoes and classification of damaged level.The hyperspectral images of healthy and damaged potatoes were taken as experiment objects,where damages were induced by a device for quantitative damage and these samples were divided in 5 levels(level I,II,III,IV and V damages)according damaged degree.First of all,some pre-processing methods of hyperspectral images were performed,including the correction of hyperspectral images,background segmentation and image cropping.Then,linear discriminant analysis(LDA),support vector machine(SVM)and adaptive boosting(AdaBoost)models were built respectively based on fullwavelength for damage identification.The results showed that AdaBoost had the best damage recognition accuracy.Next,Savitzky-Golay(S-G),first derivative(D1),second derivative(D2),standard normal variable(SNV),Multiscatter Scatter Correction(MSC)and their combination method were respectively used to pre-process spectral data.The optimal methods corresponding to five models used to discriminated the damage,level I damage,level II damage,level III damage,and level IV damage samples(called A1?A2?A3?A4?A5)were D1,SNV,no preprocessing,no preprocessing and SNV.In addition,the optimized simulated annealing algorithm based on correlation coefficient algorithm was applied to select characteristic wavelengths,the classification rates of A1,A2,A3,A4,and A5 models achieved to 99.12%,97.89%,96.05%,100% and 100% respectively.The results indicated that potato with damage could be identified accurately and effectively,and the different bruised levels could be classified by the hyperspectral imaging technique,which provided an idea for on-line and non-destructive testing of potato.