Abstract: For more reasonable and effective classification of eggs, a method for non-destructive identification of egg varieties based on dielectric properties was developed. In this experiment, four groups of eggs (caged eggs from Zhenjiang, Jiangsu province, and free-range eggs from Zhenjiang, Jiangsu province, from Laonangou, Anhui province, and from Dongtai, Jiangsu province) were measured for dielectric properties in the frequency range of 10–200 kHz by the parallel plate method. A classification model for egg varieties by the support vector machine (SVM) algorithm was established. The effects of different kernel functions (linear, polynomial, RBF, and Sigmoid) and different parameter optimization algorithms (grid search, genetic algorithm, and particle swarm optimization) on the accuracy rate of the classification model were analyzed. The results showed that the performance of the SVM classification model based on linear kernel function and particle swarm optimization was the best, giving a prediction accuracy of 95.83% and 95.83% for the training and test sets, respectively. The non-destructive testing technology based on SVM algorithm using dielectric properties achieved good classification results. This study has provided a new effective method for the identification of egg varieties.

Key words: dielectric properties, egg, variety, non-destructive testing, support vector machine (SVM)