Moldy peanut identification based on hyperspectral imaging technology and spatial-spectral extreme learning machine model

Peanut is an important oil crop with rich nutritional value and occupies an important position in food industry. When peanuts go through transportation and storage, it is very easy to mold and produce aflatoxin, a strong carcinogen, which is a serious food safety hazard to human health. Therefore, it is of great social significance to design a more rapid and accurate detection method for the identification of moldy peanuts. In this study, hyperspectral imaging technology was used to detect moldy peanuts. A new classification method combining principal component analysis (PCA), local binary model (LBP) and extreme Learning Machine (ELM) (PCA-LBP-ELM) is proposed. Firstly, PCA dimensionality reduction was performed on peanut data, and then LBP was used to extract spatial features, and spatial-spectral features were fused as input data of ELM to identify mildewed peanuts. The experimental results show that this method is faster and more accurate than other algorithms, and the accuracy rate of identifying moldy peanut is 99.91%, which provides another technical support for real-time monitoring of moldy peanut.