Color Recognition And Browning Prediction Of Roast Mutton Based On Machine Vision Technology

Roast mutton is one of the traditional meat products in China,which is favored by consumers because of its long history and unique flavor.The color of mutton will change when it is roasted at high temperature,and the Browning degree and physicochemical indexes of mutton will change with the prolonging of roasting time.Therefore,the color of roast mutton can be identified,the external color can be used to correlate the internal attributes,and the relationship between the color change in the process of mutton roasting and other internal indicators can be established,and the changes can be analyzed.To roast mutton as the research object,this study using machine vision technology to quickly identify the color of the roast mutton,and based on three kinds of algorithms for the roast mutton mapped the recognition than color card,color finally use the roast mutton color parameters and the correlation of Browning products to establish the regression equation,to realize the roast mutton Browning product rapid judgment.The main research results are as follows:1.Explore machine vision models that can recognize all colors of lamb during cooking.After image preprocessing,the color recognition effects of Xception-CNN(Convolutional Neural Networks)model were compared with those of Res Net-50,Inveption-V3 and traditional CNN models.The results showed that the color recognition accuracy of the four models were 87.00%,80.00%,81.00%and 78.00%,respectively,among which Xception-CNN model had the highest recognition accuracy,and the test time was 1.22s shorter than that of the traditional CNN model.2.Based on machine vision technology,combined with Mean algorithm,K-Means algorithm and K-Means+image denoising algorithm,three kinds of color recognition colorimetry cards of roast mutton were constructed,and K-medoids algorithm combined with sensory experiment was used for verification.The results showed that the three colorimetric cards could improve the recognition accuracy of mutton cooking time,and the algorithm verification accuracy and sensory verification accuracy of the colorimetric card made by K-Means algorithm were 95.70%and73.71%,respectively,which were higher than the other two algorithms.3.Maillard reaction degree index of color change during mutton roasting was determined and its correlation was analyzed.The results showed that the contents of A₄₂₀,A₂₉₄,fluorescence intensity and 5-HMF were all increased during the roasting process,and there was a significant positive correlation between them.A color-based predictive regression model(R²=0.820-0.992)for 5-HMF content of roasted mutton was established by using machine vision technology to identify the color parameters of the surface of roasted mutton.The content of 5-HMF of roasted mutton could be quickly determined by measuring the color.