detection and classification of potato defects using multispectral imaging system based on single shot method

Detection and classification of potato defects are of great significance to ensure food safety and improve product quality. this study investigated the potential of a novel multispectral imaging system based on single shot method for detection and classification of potato defects. a total of 417 potato samples were used in the experiment. the 25 spectral images with spatial resolution of 409 × 216 pixels over the spectral region between 676 and 952 nm were acquired for each potato. after improving the image contrast between the defect and defect-free regions by band math method, the defect regions were segmented from whole samples by using simple threshold. the spectral and textural features of the segmented regions were calculated and used for classification. a model for classifying different defects of potato was developed using least squares-support vector machine (ls-svm) based on all feature set. the ls-svm model achieved the classification accuracy of 90.70% for the test set. this research demonstrated that multispectral imaging system based on single shot method is a potential tool for online detection and classification of potato defects. the proposed data processing algorithm can be used for non-destructive testing of potato.