Research On Classification Of Pork Freshness Based On SSA-ELM Algorithm

The freshness of meat is closely related to the physical and mental health of consumers.Therefore,reasonable and accurate detection of the freshness of pork is of far-reaching significance to guarantee the quality of pork and ensure the legitimate rights and interests of consumers.This dissertation uses machine vision technology in non-destructive testing methods to collect and analyze pork images,in order to judge the freshness level of pork.Firstly,according to the features of the pork surface,a machine vision acquisition system was built to obtain the original image information of the pork.After that,the original image of pork was pretreated.The original image was used the weighted average method for grayscale processing.The image noise is suppressed by adaptive median filtering.Selecting the local threshold method to segment the muscle and fat in the image,and using the open operation to remove small burrs and smooth the contour edges.The target image that can be used to extract features was obtained through the image mask.Finally,extracting pork features.Pork color features were extracted from RGB,HSV and L*a*b* color space,and pork texture features were extracted from gray level co-occurrence matrix.Marble features of pork were extracted from marbling image,and these three kinds of features were fused into 53 dimensional pork features for subsequent experiments.When constructing the pork freshness classification model,the classification performance of support vector machine(SVM)and extreme learning machine(ELM)is compared,and the classification accuracy is 76% and 78% respectively.It is concluded that ELM is more suitable for pork freshness classification.After that,a new type of swarm intelligence algorithm was proposed,Sparrow Search Algorithm(SSA).In order to highlight the advantages of SSA algorithm in optimizing the pork freshness model,it was combined with Particle Swarm Optimization(PSO)algorithm,Differential Evolution(DE)algorithm and Beetle Antennae Search(BAS)algorithm,and the results show that the accuracy,stability and convergence of SSA are the best.Finally,by comparing and discussing the SSA-ELM,PSO-ELM,DE-ELM and BAS-ELM pork freshness classification accuracy rates,repectively82%,88%,92% and 98%,we conclude that the SSA-ELM pork freshness classification model has the highest accuracy rate,which is more suitable for the application of pork freshness classification recognition.