Recognition of Transparent Image of Cracked Egg Based on Ordered Wave Signal Extraction and Discrimination Algorithm

Consumers have higher and higher requirements for egg quality, so it is more and more important to systematically carry out nondestructive testing of egg freshness. The color transparent image of eggs was obtained by cold light source illumination. The L component of the transparent image in lab space was extracted and the binary image of egg was obtained by gray threshold segmentation, The binary image of the air chamber is obtained. The G component of RGB space of transparent image is extracted for adaptive gray adjustment and linear space filtering of user-defined template. The yolk image is extracted by morphological method and threshold segmentation method. The edge of egg yolk is extracted by Obel operator and the yolk ellipse is fitted by least square method. Three kinds of characteristics related to freshness, such as air chamber size, yolk size and ovality, were extracted from the translucent image, and the single factor linear regression model of them and egg freshness was established respectively. The unit freshness model of egg transparent image was obtained by training the three characteristic factors by gradient descent method. The model can be used to test the reliability of the model.