DETECTION OF FUNGI-CONTAMINATED PEANUTS USING MULTISPECTRAL IMAGING TECHNOLOGY

Peanuts can be affected by fungi, especially Aspergillus flavus and Aspergillus parasiticus, which form carcinogenic Aflatoxin B1 (AFB1) in peanuts during harvest and storage. Early intervention of these infections is critical towards food security and economic protection. This paper has created a non-destructive technique based on visible multispectral imaging and machine learning to detect fungal infection in peanuts at an early stage. Six spectral bands were used to capture images and analyse differences between healthy and slightly infected samples. K-means clustering and Otsu thresholding, among others, were applied to 23 samples to subdivide peanuts and determine areas with fungus affected. The Random Forest (RF) regression model was able to detect differences in spectra related to fungal colonisation, and thus to effectively differentiate between healthy and infected peanuts. The findings suggest that this multispectral imaging system, together with image processing and regression analysis, provides a fast and efficient mechanism of early detection of fungi, which helps to prevent AFB1 development during storage.