Submitted to: Aflatoxin Elimination Workshop Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: October 28, 2004
Publication Date: N/A
Technical Abstract: Reflectance spectra (550 - 1700 nm), visible color reflectance images, x-ray images, multi-spectral transmittance images, and physical properties (mass, thickness, cross-sectional area) were analyzed to determine if they could be used to detect fungal infected corn kernels. Corn ears were inoculated with one of several different common fungi several weeks before harvest, then collected at harvest time. It was found that two spectral bands centered at 715 nm and 965 nm can correctly identify 98.1% of controls and 96.6% of the kernels severely infected with Aspergillus flavus, Aspergillus niger, Diplodia maydis, Fusarium graminearum, Fusarium verticillioides, or Trichoderma viride. Histogram features from three transmittance images (blue and red components of visible images and another at 960 nm) can distinguish 91.9% of kernels with severe fungal infection from 96.2% of un-infested kernels. A neural network was trained to classify kernels by species as well as into one of three damage categories (control, minor damage, or severe damage) with good accuracy using principle components of the reflectance spectra. Spectra of single corn kernels can be measured automatically and kernels sorted into different mold species categories using commercial instruments.