Non-destructive detection and classification of in-shell insect-infested almonds based on multispectral imaging technology

Abstract The feasibility of non-destructive detection and classification of in-shell insect-infested almonds was examined by using multispectral imaging (MSI) technology combined with chemometrics. Differentiation of reflectance spectral data between intact and insect-infested almonds was attempted by using analytical approaches based on principal component analysis and support vector machines, classification accuracy rates as high as 99.1% in the calibration set and 97.5% in the prediction set were achieved. Meanwhile, the in-shell almonds were categorized into three classes (intact, slightly infested and severely infested) based on the degree of damage caused by insect infestation and were characterized quantitatively by the analysis of shell/kernel weight ratio. A three-class model for the identification of intact, slightly infested and severely infested almonds yielded acceptable classification performance (95.6% accuracy in the calibration set and 93.3% in the prediction set). These results revealed that MSI technology combined with chemometrics may be a promising approach for the non-destructive detection of hidden insect damage in almonds and could be used for industrial applications.