Chicken embryo fertility detection based on PPG and convolutional neural network

Abstract

Listen

Translate article

At present, when specific pathogen free(SPF) egg culture of viruses, a lot of manpower is needed to detect the fertility of chicken embryos, distinguish the three states of live, dead and weak, and then treat them separately. Moreover, due to personnel fatigue and other reasons, it is easy to cause misjudgment and omission, especially weak embryo is most prone to wrong judgment, resulting in resource waste. In this paper, a method based on PhotoPlethysmoGraphy(PPG) and convolutional neural network(CNN) is presented to determine the fertility of chicken embryo automatically, quickly and accurately, and the detection of weak embryo is proposed for the first time. In the experiment, 5000 samples are used as the training set and 1000 samples as the test set. After preprocessing, data are inputted into the independently designed convolutional neural network model to obtain classification results. The accuracy of chicken embryo classification reached 97.3 percent, including 100 percent live embryo, 98.33 percent dead embryo and 92.67 percent weak embryo. The experiment results show that the method of chicken embryo fertility detection based on PPG and convolutional neural network has high application value in the process of vaccine production.