SVM-based feature extraction and classification of aflatoxin contaminated corn using fluorescence hyperspectral data

Support Vector Machine (SVM) was used in the Genetic Algorithms (GA) process to select and classify a subset of hyperspectral image bands. The method was applied to fluorescence hyperspectral data for the detection of aflatoxin contamination in Aspergillus flavus infected single corn kernels. In the band selection process, the training sample classification accuracy was used as fitness function. Two aflatoxin thresholds, 20 ppb and 100 ppb, were used to divide the single corn kernels into clean and contaminated samples. The validation accuracy was 87.7% for the 20 ppb threshold and 90.5% for the 100 ppb threshold. The results were generated from the GA selected 36 bands and 11 bands, respectively. Compared to the full wavelength classification, the subset of image bands had slightly better or similar performance. A reduced image space could save time both in spectral data acquisition and analysis, which is crucial in the development of rapid and none invasive methods for contamination detection.