关键词: 禽蛋裂纹检测, 磁致伸缩, 功率谱, 主成分分析, 广义回归神经网络

Abstract: To establish a method for nondestructive testing of cracked eggs, we firstly developed an egg cracks detection system based on the magnetostrictive vibrator sweep frequency vibration. The system based on the acoustic characteristic, through collecting and using Welch power spectrum analysis of eggs vibration audio signal, extracting useful information in the feature vector through the principal component and constructed based on generalized regression neural network analysis (GRNN) of the egg cracks detection model. Among them, to detect 290 eggs (the training set of 200 eggs, the test set of 90 eggs), the result shows that the recognition rate of intact eggs and cracked eggs reached 96.7% and 98.3% respectively in the test set. The research shows that using the magnetostrictive vibrator sweep and Welch power spectrum analysis, extracting useful information in the feature vector through principal component analysis method combined with egg cracks detection GRNN neural network model is feasible.

Key words: Egg cracks detection, Magnetostriction, Welch, The principal component analysis, GRNN