Deciphering Aspergillus Flavus-Peanut Kernel Complex Micro-Interaction Mechanism Through Hyperspectral Imaging Fusion: Enhanced Nutrient Composition and Fungal Contamination Detection with Bi-Dimensional Focus Ripple Module-Attentive Spatial-Spectral Synergy Network

Abstract

Peanut kernels are susceptible to contamination by Aspergillus flavus, necessitating efficient detection methods to ensure food safety. In this study, hyperspectral imaging was employed to detect nutrient composition and fungal contamination in peanut kernels. The micro-interaction mechanism of the Aspergillus flavus-peanut kernel complex was elucidated, revealing spatio-temporal nutrient degradation and aflatoxin B1 accumulation. It also identified a phased nutrient utilization strategy that supported fungal growth and toxin production, with critical contamination points on day 3 and day 5. An attentive spatial-spectral synergy network (AS3Net) significantly improved prediction accuracy for moisture content, protein content and oil content, achieving R²V values of 0.932, 0.859 and 0.786, respectively. Bi-dimensional focus ripple module (BFRM) accelerated model convergence and reduced computational time, enabling AS3Net to effectively capture complex spatial-spectral interactions. This algorithm achieved 100% classification accuracy for contaminated kernels, offering a efficient solution for food safety monitoring and aflatoxin management in agricultural products.

Keywords: Information fusion, hyperspectral imaging, Generalized two-dimensional correlation spectroscopy analysis, Fungal contamination, deep learning