This paper investigates the effect of support vector machine (SVM) for the classification of intact and cracked eggs. The four frequency features of the sound impulse resonance of an egg excited with a light mechanical impact on the equator of the eggshell are extracted, including the normalization average of the frequency domain, the first dominant frequency, and the average x - and y – coordinates of the centroid for the frequency domain. These features and also the various combina-tions of them are used to construct SVM classifiers. It is shown that the SVM-PFXY classifier based on all the four frequency features gives the best classification effect with 98% testing accuracy, 98.18% crack detection and 2.11% false reject, and that the SVM-P, SVM-PF and SVM-PFY are respectively the best single-feature, binary-feature and three-feature SVM classifiers. It is also revealed that the SVM classifier associated with more features generally gives a better classification effect. For evaluating the effects of SVM classifiers for actual crack detection, this paper proposes a detection scheme of eggshell cracks based on four measurements, and the experimental example achieves the highest crack detection of 98.77% and the smallest false reject of 1.87%.
Key words: Eggshell crack, detection, acoustic impulse response, frequency feature, support vector machine.
Copyright © 2022 Author(s) retain the copyright of this article.
This article is published under the terms of the Creative Commons Attribution License 4.0