Detection of Grain Pest Insects in Cereals: The Potential of Hyperspectral Imaging

The presence of insects in commercial batches is economically detrimental to operators. Rapid detection of insects is therefore of major interest. The potential of the IHS has been tested to quantify live or dead adults and larvae of S. oryzae, R. Dominica and T. castaneum in wheat and barley samples. Prediction errors (SEP) range from 0.7 to 3.96 insects/kg for live adults and are greater than 5 insects/kg for dead adults. For larvae, performance is poorer and varies depending on the species: from 6.93 (S. oryzae) to 67.36 larvae/kg (R. dominica). Specific work carried out on images of dead and live insects shows a possible distinction with an accuracy of over 88 % for the three species.