Application Of Machine Learning In Egg Yield Prediction And Quality Detection

As a high-protein,low-fat nutritious food,eggs are one of the most important sources of nutrition for people.Although China's egg resources are abundant,the industrialization of eggs is slower and the level is low.Whether it is egg production,processing,storage or circulation,the average level of production is behind the developed countries.The specific performance is inaccurate in the supply of eggs,low level of egg processing and lack of testing equipment for the industrial chain.Therefore,reliable prediction of egg yield and practical detection of egg quality are of practical significance.This paper summarizes and analyzes the research status and existing deficiencies of egg production prediction and quality testing at home and abroad.The machine learning algorithm is applied to egg production prediction and quality inspection,and the following three aspects of research work are completed:(1)Research on egg yield prediction model based on machine learning.In this paper,the egg yield prediction model is studied.Based on the prediction index of egg production rate,two different egg production rate prediction models based on machine learning are proposed,which are the time series prediction model of egg production rate based on extreme learning machine and based on Multi-parameter prediction model of egg production rate based on support vector machine regression.Among them,the extreme learning machine prediction model is to select the batch of historical data of the laying rate of Hailan brown hens to construct a time series model,and use the first batch of data samples for the network training,complete the extreme learning machine prediction model,and other batch data.Used for prediction of the egg production rate model.Based on the support vector machine,the egg production rate prediction model is to deal with the six types of impact factors such as feed intake,laying hen age,body weight,temperature,duration of illumination,and stress,as input data of support vector machine.Modeling and predicting the laying rate of laying hens.The experimental results show that the established extreme learning machine prediction model andsupport vector machine prediction model can accurately predict the egg production rate.The predicted results are consistent with the actual egg production rate of laying hens.(2)Research on egg dark spot detection method based on convolutional neural network.Aiming at the problem that the detection of dark spot eggs is mainly done by manual work,which is labor intensive and inefficient,a method of detecting dark spot eggs based on convolutional neural network GoogLeNet model is proposed.In this method,the Inception module is stacked repeatedly to construct the neural network architecture,and the multi-scale convolution kernel is used to extract the features of egg spots and to cascade the fusion.In order to obtain enough image samples to verify the validity of the model,an egg translucent image acquisition device was designed.A total of 1200 dark spot egg images and 8850 normal egg images were obtained,and 1200 samples were selected for network modeling.The experimental results show that the detection accuracy of dark spot eggs based on Google LeNet model is 98.19%.In order to further verify the GoogLeNet model,this paper repeats the above experiments using the VGG16 and VGG19 models of CNN,and compares the accuracy.The results show that the three CNN models have higher detection accuracy,and the GoogLeNet model has better effect.Compared with the HOG-SVM method,the results show that the accuracy of detection based on the GoogLeNet model is higher than that based on the HOG-SVM model by more than 10 percentage points.The results show that the detection method based on GoogLeNet model is feasible and has high detection accuracy,which provides a new method for egg quality detection.(3)Research on egg freshness detection method based on convolutional neural network.In this paper,a method of egg freshness detection based on CNN is proposed.In the past,the research on egg freshness detection has the problems of too few samples,uneven sample distribution and low model precision,the idea of "individual reflecting population" and multi-angle image sample collection is designed.The idea of sample collection expands the sample size,balance the number of egg samples in different categories,and obtains 6444 total samples.In order to be able to grade the experimental freshness,this paper uses 4 convolutional layers,4 pooling layers,and 2 fully connected layers.The convolutional neural network with 6 active layers,2 Dropout layers and a classifier is used to conduct independent learning and classification of eggs.The whole detection process has no excessive pre-processing steps.The experimental results show that the method established in this paper is accurate.The rate is high,and the detection accuracy is 94.63%.Compared with the previous detection methods,the model has higher generalization ability and higher accuracy.