Study On The Rapid Detection And Discrimination Of Lamb Freshness Based On Optical Information Detection Technology

lamb is nutritious,delicious,and popular among consumers.Freshness is an important standard to measure the edible value of lamb.It has great significance for rapid,accurate detection for lamb to promote the healthy and rapid development of lamb industry.The traditional sensory evaluation,physical and chemical detection or microbiological experimental methods can't meet the requirements of rapid,accurate and nondestructive detection for freshness in lamb circulation.Optical detection technology is a promising method among the many fast nondestructive testing methods.In this study,chilled lamb with different freshness was taken as the research object to analyze the freshness change rule in the process of lamb deterioration.The key freshness indexes were excavated and the best visible near infrared(350?1050nm)spectral detection model for each index were studied.The lamb freshness classification model were established using the optimal multisource spectral characteristics of fusion key indexes.On the basis of the above,further expand the research spectrum of lamb freshness.The near-infrared optical information of 935-2539nm was obtained by hyperspectral imaging system,and the lamb freshness was further studied using TVB-N as the main research index.To explore the optimal spectral and image characteristics method for TVB-N prediction,the spectral and spatial image features including color and texture represented internal chemical components of lamb were mined and effectively fused,and the more stable prediction model for lamb freshness prediction was developed.study on rapid detection of lamb freshness based on optical information detection technology with multiple perspectives and methods,so as to provide a theoretical basis for accurate,fast and nondestructive discriminate and analysis for mutton freshness.The specific research content and results are as follows:(1)The spoilage mechanism for chilled lamb in different storage time was studied and the physical and chemical indicators,microbial indicators and sensory indicators affecting the lamb freshness were analyzed.The changes of each index in the lamb spoilage process and the correlation between the indexes were studied.L*,pH value,TVB-N and TVC were identified as the key research indexes of lamb freshness.(2)The effect of different spectral pretreatment methods on accuracy of lamb freshness prediction model was analyzed and best spectral detection model of each key freshness index was selected.The RBF kernel function of support vector machine(SVM)model was optimized by the grid search method of "rough" and "fine",and the best spectral characteristics and prediction model for the key freshness indexes were selected by comparing the prediction effect of the optimal SVM and PLSR model.(3)Lamb freshness discrimination models of CART classification tree were established based on the spectral characteristics for TVB-N and the fusion characteristics for key freshness indexes,and prediction accuracy of Single-CART and Combination-CART model were verified.The results show that the average classification accuracy of the Single-CART and Combination-CART model was 100%for correction sets respectively,and was 83.33%and 95.83%for validation set,respectively.Compared with the Single-CART model,the Combination-CART model is more accurate and stable.The results show that the lamb fresh degree could be discriminated accurately by optimizing and using multi-source characteristic variables to establish the classification model.(4)The prediction method of lamb freshness has been studied in depth with TVB-N as the main research object.An modified binary discrete particle swarm optimization algorithm((MDBPSO))was proposed to select the near infrared characteristic wavelength of lamb TVB-N.The traditional discrete particle swarm algorithm was optimized in the two aspects of particle update method and inertia weight.The comparison between the MDBPSO method and the conventional feature wavelength extraction method was used to analyze the prediction effect of the PLSR model.The results show that MBPSO-PLSR prediction model achieved the highest accuracy.The Rc2 and RMSEC were 0.82 and 3.61 for calibration set,respectively,and the Rp2 and the RMSEP were 0.81 and 3.68 for prediction set,respectively.the MBPSO-PLSR model has significantly improved convergence efficiency and prediction accuracy compared with other models.(5)Spectral characteristics and spatial image characteristics including color and texture were deeply excavated.The TVB-N content prediction model for lamb based on RFR and BPANN was established though the optimization spectral characteristics by MDBPSO.The TVB-N prediction model based on BPANN algorithm was established by using the image features optimized by PCA and GA.The results show that MDBPSO-RFR was the best spectral prediction model.The Rc2 and RMSEC were 0.87 and 3.12 for calibration set,respectively,and the Rp2 and the RMSEP were 0.85 and 3.56 for prediction set,respectively.GA-BPANN was the best spectral prediction model.The Rc2 and RMSEC were 0.81 and 3.71 for calibration set,respectively,and the Rp2 and the RMSEP were 0.80 and 4.20 for prediction set,respectively.It indicate that the spectral features prediction model for lamb freshness was better than that of image feature model.(6)The prediction models for lamb freshness based on spectral and image features were compared and analyzed,and the best TVB-N content feature information was optimized.The prediction model for lamb freshness was established by effectively fusing spectral and image features with BPANN model.The results show that the Rc2 and RMSEC were 0.81 and 3.71 for calibration set,respectively,and the Rp and the RMSEP were 0.80 and 4.20 for prediction set in fusion model,respectively.It indicate that the prediction effect of the fusion model was better than that of the single sensor model of spectrum or image,and it can reflect the lamb freshness more comprehensively and accurately.The above results show that the rapid detection for external sensory quality and internal physical and chemical quality for lamb can be realized by optical information detection technology,and the quantitative analysis of lamb freshness and qualitative discrimination of freshness grade were realized simultaneously.The research can provide a good theoretical foundation for developing a rapid detection system for lamb freshness based on spectral and image information.