Location: Insect Behavior and Biocontrol Research
Title: A new EEMD-based scheme for detection of insect damaged wheat kernels using impact acousticsAuthor
GUO, MIN - Shaanxi Normal University | |
MA, YUTING - Shaanxi Normal University | |
MA, MIAO - Shaanxi Normal University | |
WU, XIAOJUN - Shaanxi Normal University | |
Mankin, Richard |
Submitted to: Acta Acustica
Publication Type: Peer Reviewed Journal Publication Acceptance Date: 9/22/2016 Publication Date: 11/1/2016 Citation: Guo, M., Ma, Y., Ma, M., Wu, X., Mankin, R.W. 2016. A new EEMD-based scheme for detection of insect damaged wheat kernels using impact acoustics . Acta Acustica. 102(6):1108-1117. https://doi.org/10.3813/AAA.919022. DOI: https://doi.org/10.3813/AAA.919022 Interpretive Summary: Small insects feeding inside wheat kernels cause significant but unseen damage to stored grain. Researchers in the School of Computer Science at Shaanxi Normal University, China, and the USDA, Agricultural Research Service, Center for Medical, Agricultural, and Veterinary Entomology, Gainesville, Florida, collaborated to investigate mathematical methods to distinguish sounds produced by undamaged wheat kernels dropped onto a steel plate from sounds by insect-damaged kernels using an inexpensive impact acoustics measurement and analysis system. New analyses were developed that considerably improved the accuracy of insect detection and can assist warehouse managers in targeting hidden infestations of insects in stored grain. Technical Abstract: Internally feeding insects inside wheat kernels cause significant, but unseen economic damage to stored grain. In this paper, a new scheme based on ensemble empirical mode decomposition (EEMD) using impact acoustics is proposed for detection of insect-damaged wheat kernels, based on its capability to process non-stationary signals and its suppression of mode mixing. The intrinsic mode function (IMF) kurtosis, IMF form factors, IMF third-order Rényi entropies, and the mean of the degree of stationarity were extracted as discriminant features used as the inputs to a support vector machine (SVM) for non-linear classification. In these experiments, 98.7% of undamaged wheat kernels and 93.3% of insect-damaged ones were correctly detected, which indicated the effectiveness of the proposed method for categorizing undamaged wheat kernels from insect damaged kernels. |