A Novel Hyperspectral Imaging System for Whole Corn Ear Analysis and Aflatoxin Detection

Highlights Development of a whole corn ear hyperspectral imaging system. Field experiment found 25% inoculation with highest aflatoxin contamination. 80% detection accuracy for whole corn ear aflatoxin detection with NDFI method. Fluorescence peak shift observed for fungus infected and aflatoxin contaminated samples. ABSTRACT. Corn is the most widely cultivated crop in the United States and accounts for approximately one-third of total world production. Inspection of whole corn ears under field conditions is important during preharvest scouting because it is crucial for farmers to assess the cleanliness of the crop at harvest and for corn breeders to evaluate postharvest traits. This paper reports on aflatoxin contamination detection in whole corn ears using a novel hyperspectral imaging system for whole corn ear imaging. The hyperspectral imaging system employs pushbroom line-scanning uniquely with a rotational stage, enabling the acquisition of a flattened, two-dimensional spectral image of the entire ear surface from the 3D ear body. Inoculation experiments were conducted to obtain corn ears contaminated with aflatoxin in the field. Specifically, four treatments were implemented including Control, Control Pierced, inoculation of 25%, and 100% of Aspergillus flavus (AF13) inoculum concentrations. Fifty ears were harvested for each treatment for analysis. Each ear was imaged under ultraviolet light (365 nm) for excited fluorescence hyperspectral data acquisition and subsequently analyzed chemically for aflatoxin concentration. Image analysis was based on a rapid spectral index method, Normalized Fluorescence Difference Index (NDFI), using wavelengths at 537 and 437 nm for NDFI calculations. It was found that both inoculated treatments were severely contaminated with aflatoxin, with the 25% inoculation treatment having the highest aflatoxin concentration. A fluorescence peak shift was observed for the contaminated pixels. The clean control pixels had a fluorescence peak at 478 nm. The fluorescence peaks for the contaminated pixels were at 506-516 nm. When using 20 ppb (parts per billion) as a threshold for aflatoxin detection, the overall detection accuracy reached 80%, with the Control Pierced group achieving the highest accuracy of 86%. Although the study found no direct correlation between the number of detected “hot pixels” and measured aflatoxin concentration, as few as ten pixels would classify an ear as contaminated at the 20-ppb threshold using the NDFI algorithm. This research is the first to report on the use of the kernel crown fluorescence for aflatoxin detection and highlights the potential of whole-ear hyperspectral imaging for non-invasive contamination screening and morphological traits analysis in corn breeding. Keywords: Aflatoxin detection, Aspergillus flavus inoculation, Detection accuracy, Fluorescence peak shift, Whole corn ear hyperspectral imaging.