Development Of A Rapid Detection System For Intramuscular Fat Content In Pork Based On Computer Vision Scoring

The intramuscular fat content(IMF)is one of the important factors affecting the quality of pork,and it has important significance for pork grading and pig breeding.The traditional method for IMF determination is the Soxhlet extraction method,which has high accuracy but requires complex pretreatment steps,has a long measurement cycle,and poses certain safety risks,making it unable to meet the demand for rapid determination of intramuscular fat content in modern meat industry.In this paper,a rapid detection method for intramuscular fat content in pigs was developed by combining computer vision technology and machine learning technology in the field of artificial intelligence to overcome the shortcomings of the traditional method.A total of 200 pigs were slaughtered and 1481 transverse section images of the pig eye muscle were collected(6-10 images per pig).Computer vision technology was used to obtain the individual intramuscular fat grading(IIMF)of each pig.The actual IMF,meat color,marbling score,and backfat thickness of each pig were measured to obtain meat quality and carcass trait information.Using IIMF and conventional meat quality and carcass trait information as independent variables,stepwise regression(SR)and gradient boosting machine(GBM)were used to construct IMF estimation models.The optimal model was selected to determine the rapid detection method for intramuscular fat content in pigs.The developed computer software was based on the Qt framework and Open CV library,using C++ programming language.The main research results are as follows:(1)By analyzing the characteristics of the longest muscle on the back of a pig and its cross-sectional images,we determined the relevant hardware facilities for image acquisition and constructed a computer vision image acquisition room that meets the requirements for image acquisition.(2)We studied the uneven illumination of the image of the longest muscle on the back of the pig and used the adaptive histogram equalization method to process the normalized and gamma-corrected images of the longest muscle on the back,effectively solving the problem of uneven illumination in the image.(3)We slaughtered 200 experimental pigs,determined their meat quality and carcass performance,and obtained a pig information dataset for IMF rapid estimation using IIMF.The correlation coefficients between IMF and computer vision score,marbling score,backfat thickness,moisture content,and p H value were 0.68,0.64,0.48,-0.45,and 0.25,respectively.(4)We constructed SR and GBM models to estimate intramuscular fat content.The results showed that the accuracy of the SR and GBM models based on residual distribution were 87.5% and 89%,respectively.This indicates that the method of rapid detection of pig intramuscular fat content based on computer vision score is reliable.(5)We developed a computer vision score software and an image acquisition room to form a rapid detection system for pig intramuscular fat content.This system can complete the measurement of pig intramuscular fat content within 24 hours by quickly identifying pig marbling images.In conclusion,this study developed a rapid detection system for pig intramuscular fat content by integrating computer vision technology and machine learning technology with visual score and conventional meat quality information.The research results provide a new possible way for the rapid and intelligent detection of pig intramuscular fat content.