Research On The Detection Method Of Embryonic Egg Formation Activity Based On CNN And Heartbeat Signal

Vaccine-virus immunization is the main therapy for avian flu pandemic preparedness.The main way to cultivate influenza vaccine is achieved by injecting the virus strain into hatching eggs to reproduce the virus.However,in the process of culturing viral strains,dead eggs must be removed to prevent contamination of normal embryos.Therefore,in the production of avian influenza vaccine,it is important to detect and classify the fertility of hatching eggs inoculated with avian influenza virus strains.At present,the method of manual detection is the main method to detect the fertility of hatching eggs.This method has many disadvantages,such as low detection efficiency,high labor intensity,easy to cause false detection and missed detection,which does not meet the needs of high-efficiency and high-quality automated production in factories.The combination of deep learning method with image features and other physiological parameters of hatching eggs has gradually become the trend of current research.In order to detect and classify the fertility of hatching eggs effectively,this paper takes 9-day-later hatching eggs as the research object.PhotoPlethysmoGraphy is used to collect the heartbeat signals of hatching eggs as discriminant features,and a method combining CNN and heartbeat signals is proposed to detect the fertility of hatching eggs.Firstly,a Butterworth high-pass filter is designed to filter eggs heartbeat signal,which is in accordance with the characteristics of the heartbeat signal of hatching eggs Secondly,for the collected heartbeat signal data of 9-day-later hatching eggs,a fertility classification algorithm of hatching eggs based on heart rate threshold is proposed and the validity of the algorithm is verified.Finally,aiming at the threshold sensitivity of hatching eggs classification algorithm based on heart rate threshold in processing heartbeat data of eggs 9-days-later,a method of combining CNN with heartbeat signal of eggs for fertility detection is proposed.Sequence convolutional neural network E-CNN used for analyzing heartbeat sequence of hatching eggs and two-dimensional convolutional neural network SR-CNN used for processing heartbeat wavefonrm of eggs are designed respectively.SR-CNN is capable of effectively compensating for the deficiency of E-CNN in processing noisy signal sequence by extracting features of heartbeat waveform.In addition,SR-CNN not only plays a role in feature screening by combining channel weighting and residual structure,but also makes the network convergence better and achieves improving network performance.The experimental results show that it is reasonable and feasible to use the convolutional neural network E-CNN and SR-CNN to detect the fertility of hatching eggs.The convolutional neural network E-CNN and SR-CNN achieve 99.50%and 99.62%detection accuracy respectively on the 9-day-later hatching eggs heartbeat dataset.