| In order to meet people’s demand for higher quality pork and reduce or even eliminate the entry of fresh and rotten pork into the market,endangering people’s health,it is of far-reaching significance to explore a method that can scientifically and reasonably detect the freshness of pork.In this paper,the combination of computer vision technology and p H(acid-base value)detection is used to conduct in-depth research on pork freshness grading.According to the actual situation of the site environment,a computer vision acquisition system was built for real-time pork images.The preprocessing of the acquired pork image mainly includes grayscale processing of the original image,image noise suppression,segmentation of image muscle and fat parts,removal of image burrs and contour edge processing,image masking and other steps.Finally,the unique color characteristics of pork surface were extracted from three common and different color space models: red(R),green(G),blue(B),brightness(L*),red-green value(a*),yellow-blue value(b*),hue(H),saturation(S),and brightness(V).At the same time,the grayscale symbiosis matrix of pork image is obtained by calculation formula,and the unique texture characteristics of pork surface are extracted from it,the texture characteristics in pork are studied by using the extracted pork marbling image,the p H value of pork is measured by p H meter,and the representative pork features with different attributes of these four categories are fused into multiple sets of 54-dimensional pork image feature data as subsequent experimental data.In the process of selecting the classification model of pork freshness,the support vector machine(SVM)and the extreme learning machine(ELM),which are commonly used in the experiments of meat freshness classification,were used to predict the freshness of pork,and the classification accuracy was 78% and 80%,respectively,and the results showed that the classification recognition performance of the ELM model was better.In this paper,a new swarm intelligence algorithm,Harris Hawks Optimization(HHO),is proposed to optimize the input weights and thresholds of ELM.At the same time,the effects of Particle Swarm Optimization(PSO)and Longhorn Beetle Whisker Algorithm on pork freshness classification were explored,and it was concluded that HHO had the best grading accuracy,model stability and convergence.Finally,six different pork freshness classification models of HHO-ELM,PSO-ELM,BAS-ELM,BAS-SVM,PSO-SVM and HHO-SVM were compared and analyzed,and the classification accuracy was 92%,82%,88%,84%,86% and 90%,respectively,indicating that the HHO-ELM pork freshness classification model proposed in this paper can accurately identify the pork freshness grade and can be used to study the freshness of pork. |
