Detecting moldy peanuts via moldiness index and kernel features by hyperspectral imaging

Foods such as peanuts are prone to mold and contamination by the deadly carcinogenic, aflatoxin (AF), which can have harmful effects on human and animal health. Therefore, the rapid and nondestructive detection of moldy foods through Hyperspectral Imaging (HSI) is crucial. In this study, a novel method to classify moldy peanuts was developed using a combination of the Moldiness Index (MI) and the distribution features of pixels in peanut kernels using HSI. Samples of 3000 + peanuts from five different varieties with varying degrees of moldiness (healthy, mildly moldy and severely moldy) were prepared for the test, and their AF content was measured. The formula for calculating the MI was first obtained by regressing the spectra difference and labels. Short-wave Infrared hyperspectral images (930–2500 nm) were used to generate an MI distribution map of peanuts, and 27 features were extracted from the map to classify the peanuts. The proposed method achieved classification accuracies of up to 92% and 88% for the Cai and Hei peanut varieties, respectively, outperforming the control method in all classification tasks. These results suggest that the proposed method can be employed in the development of moldy peanut sorting equipment, thereby ensuring a healthy food environment for humans and animals by eliminating harmful AF-contaminated food sources.