Study On The Prediction Method Of Beef Physiological Maturity Based On Image Processing

In the last few years,the structure of domestic food consumption has undergone significant changes in China.With the pursuit of high-quality life,the demand for high-quality beef is increasing.Physiological maturity is an important indicator when assessing the quality of slaughtered calves or divided pieces.Therefore,a method for judging the physiological maturity of beef during slaughtering or processing is effective for improving the quality rating system for beef.After analyzing relevant researches about beef maturity,it is reported that beef in different physiological maturity has different feature.In this study,the color image of beef and the microscopic image of muscle fiber were collected.Image processing was used to obtain the feature of beef color image and muscle fiber image.The prediction model of beef maturity was established and the model was compared under different parameter optimization algorithms.The main content and significant results of this study are as follows:(1)Microscopic images of muscle fibers in four different physiological maturity samples were collected.These images were processed by using Image-Pro Plus software to extract the diameter,perimeter and density of beef muscle fibers.The parameters of muscle fibers were statistically analyzed.The results showed that the diameter and perimeter parameters of beef muscle fibers were positively correlated(0.8967 and 0.9891)with their physiological maturity.The density parameter of muscle fibers and its physiological maturity were negatively correlated,having a correlation coefficient of 0.9972.The result of the Pearson correlation coefficient analysis showed that there was a negative correlation between the muscle fiber density and perimeter as well as diameter,and the correlation coefficients were-0.750 and-0.604 respectively.A linear model of single input based on the diameter,perimeter and density parameter of beef muscle fibers was established,and the prediction accuracy of physiological maturity was 66.67%,83.33%,and 87.50%,respectively.(2)Obtained color images of beef samples in different physiological maturity by using online image collection platform.After image preprocessing including graying,filtering,noise reduction,image sharpening enhancement,threshold segmentation,binarization,corrosion and expansion processing,the extraction and edge detection of the bovine eye muscle,then the area and perimeter parameters of the beef marbling were extracted.The parameters of marbling were statistically analyzed.The results showed that the area and perimeter of the beef marbling were positively correlated with physiological maturity,and the correlation coefficients were 0.9985 and 0.9396 respectively.A linear model of single feature input based on the marbling area and perimeter was established.The prediction accuracy of physiological maturity was 20.83%and 25%,respectively.This model effect was weak.The result of the Pearson correlation coefficient analysis showed that there was a little correlation between beef marbling and beef muscle fiber parameters.(3)A support vector machine model(SVM)for predicting physiological maturity of beef was established based on multiple parameters inputs.To optimize the penalty parameter C and g of kernel function parameter of the model,using the grid search method,the improved grid search method,genetic algorithm and particle swarm algorithm to realize parameter optimization under the cross validation method,respectively.According to the results,the improved grid search method has best prediction accuracy and optimal time in general,and it is suitable for the prediction model of beef physiological maturity in this study.(4)Based on the improved grid search method,the SVM model was established based on marbling area and perimeter parameters.The prediction results of the test set showed that the prediction accuracy was 54.17%,and the model effect was better than that single feature input linear model;The diameter,perimeter and density parameters of beef muscle fiber were used to establish a SVM classification model.The prediction results showed that the prediction accuracy was 83.33%,and the model effect was slightly better than that the linear model based on muscle fiber single feature;Furthermore,all the parameters of beef marbling and the muscle fiber are used to establish a multi-feature input SVM model.The prediction accuracy of the physiological maturity of this model reaches 95.83%,and the model is better than the other models mentioned above.