Study Of Pork Quality Based On Spectral Imaging Technology

Pork qualities mainly contain safety quality and eating quality. This study designed a near-infrared multispectral imaging(MSI) system for quantitative analysis of pork quality,hoping to realize the application of emerging imaging technique from the laboratory to practical applications like real-time monitoring of meat quality. And then, we carried the preliminary study on pork organization structure combining the microscopic imaging and spectral imaging technologies, aiming to make a further understanding to the process of pork spoilage. The main research points were summarized as follows:(1) Study of detection for pork eating quality indicators by near infrared multispectral imaging technique. Firstly, to use the HSI system capturing hyperspectral images of pork, we selected characteristic wavelengths by PCA. Characteristic filters of 1280±10nm, 1440±10nm and1660±10nm were fabricated for developing of MSI system. This system was used for acquisition of multispectral images of pork. To extract the texture feature variables based on GLCM. Then,the correlation between multispectral images data and tenderness of pork, the correlation between multispectral images data and water-holding capacity(WHC) of pork, were respectively established by models. In this work, the Warner-bratzler shear force(WBSF) and cooking loss rate were respectively represented the tenderness and WHC of pork. Ant colony optimization(ACO) combined with back propagation artificial neural network(BP-ANN), namely ACO-BPANN, was used for modeling, and classical PLS and BP-ANN were studied systematically and comparatively. The results showed that, the performances of nonlinear models were significantly better than those of PLS models for the two indicators. Specially, the performance of ACO-BPANN was made great progress. The ACO-BPANN model for evaluating pork tenderness by WBSF was achieved with the correlation coefficient of the prediction set(R p)= 0.8451, root mean square error of prediction(RMSEP) = 0.9087 kgf. The ACO-BPANN model for evaluating pork WHC by cooking loss rate was achieved with the correlation coefficient of the prediction set R p = 0.9116, RMSEP = 1.5129 %. Finally, APa RPs combined with Runs Test was used to validate the relationship between multispectral images data and the WBSF value of pork samples, the relationship between multispectral images data and the cooking loss rate of pork samples, respectively. Results showed that their relationships were nonlinear. From the results, we concluded that MSI technique combined with proper chemometric methods could be applied to evaluate the eating quality indicators of pork.(2) Study of detection for pork freshness indicator by near infrared multispectral imaging technique. Firstly, the MSI system in chapter 2 was used for acquisition of multispectral images of pork. To extract the texture feature variables based on GLCM. Then, the correlation between multispectral images data and pork safety quality indicator- freshness was established by models.In this study, total volatile basic nitrogen(TVB-N) content was measured for characterization of pork freshness. Linear and nonlinear algorithms were comparatively used in model calibration.The model result showed that the performances of nonlinear models were significantly better than that of PLS model for the freshness indicator. In addition, the performance of BP-Ada Boost model was made a great progress. The BP-Ada Boost model for evaluating pork freshness by TVB-N was achieved with the correlation coefficient of the prediction set R p = 0.8325, RMSEP =6.9439 mg/100 g. Finally, APa RPs combined with Runs Test was used to validate the relationship between multispectral images data and the TVB-N content of pork samples. Result showed that their relationship was nonlinear. This work sufficiently demonstrated that the MSI technique has a high potential in non-destructively sensing pork freshness, and the nonlinear BP-Ada Boost algorithm has a strong performance in solution to a complex data processing.(3) Study of pork internal organization structure by hyperspectral microscopic imaging(HMI).Our work developed a HMI system for acquisition of hyperspectral microscopic images of pork under micro scale. PCA was then performed for extraction of characteristic images. 521.08 nm,589.69 nm, 636.88 nm, 687.58 nm and 738.66 nm were selected as the characteristic wavelengths for subsequent analysis. Seen from the microscopic images of pork under different storage times,we found that images of pork have the obvious difference between different storage times. The damage degree of pork structure increases with the increase of storage times. Then, texture analysis was used for extracting characteristic variables from the region of interest(ROI) of characteristic image by Gray level co-occurrence matrix(GLCM). Linear discriminant analysis(LDA), BP-ANN and support vector machine(SVM) were used for freshness classification.These achieved classification rate of 91.67%, 95.00% and 98.33% respectively in calibration set,and the classification rates of 93.33%, 100.00% and 93.33% respectively in prediction set. From the results, the overall classification rates of nonlinear methods were higher than the overall classification rate of linear method. From the study, it can be concluded that it is feasible to describe the changes of pork micro-structural using HMI technique, and HMI technique has the potential in classification of meat’s freshness in the microscopic field combined with proper algorithm.