Hierarchical classification pathway for white maize, defect and foreign material classification using spectral imaging

Abstract This study aimed to present the South African maize industry with an accurate and affordable automated analytical technique for white grading using near infrared (NIR) spectral imaging. The 17 categories sub-categories stipulated in legislation were simultaneously classified (1044 samples; 60 kernels of each class) 25 partial least squares discriminant analysis (PLS-DA) models. models assembled a hierarchical decision pathway that progressed from most easily classes difficult. full NIR spectrum (288 wavebands) model performed overall accuracy 93.3% main categories. Three waveband selection techniques employed, namely windows (48 wavebands), variable importance projection (VIP) (21 covariance (CovSel) (13 wavebands). Overall, VIP set based on only 7.3% original variables was recommended as best trade-off between performance expected cost reduced system.