|Cetin, Enis - BILKENT UNIVERSITY|
|Tewfik, Ahmed - UNIV OF MINNESOTA|
Submitted to: Digital Signal Processing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: September 2, 2005
Publication Date: May 1, 2007
Repository URL: http://naldc.nal.usda.gov/download/23605/PDF
Citation: Pearson, T.C., Cetin, E.A., Tewfik, A.H., Haff, R.P. 2007. Feasibility of impact-acoustic emissions for detection of damaged wheat kernels. Digital Signal Processing. 17(3):617-633. Interpretive Summary: A system was built that is able to distinguish good wheat kernels from a variety of damaged kernels by dropping kernels, one at a time, onto a steel plate and digitally analyzing the resulting sounds from the impact. The types of damage studied were insect damaged kernels with exit holes, hidden insect damaged kernels without exit holes, sprout damage, and scab damage. It was found that 98% of the good kernels and 87% of the insect damaged kernels with exit tunnels can be distinguished from each other. Accuracy for scab and sprout damaged kernels was 70% and 45% for hidden insect damaged kernels. The device should be capable of inspection rates exceeding 40 kernels/s, or ~70g/min. It is non-destructive and can be made to sort kernels into one of three different groups. This technology should help grain inspectors and millers better ascertain the quality of a wheat load under consideration.
Technical Abstract: A non-destructive, real time device was developed to detect insect damage, sprout damage, and scab damage in kernels of wheat. Kernels are impacted onto a steel plate and the resulting acoustic signal analyzed to detect damage. The acoustic signal was processed using four different methods: modeling of the signal in the time-domain, computing time-domain signal variances and maximums in short-time windows, analysis of the frequency spectra magnitudes, and analysis of a modified cepstrum. Features were used as inputs to a stepwise discriminant analysis routine, which selected the best subset of features for classification using a neural network. For a network presented with only insect damaged kernels (IDK) with exit holes and undamaged kernels, 87% of the former and 98% of the latter were correctly classified. It was also possible to distinguish undamaged, IDK, sprout-damaged, and scab-damaged kernels.