0512-52441555 
18915638155Pork is rich in nutrients and is an essential source of food in human life. In recent years, China's pork production has continued to increase, people's demand for pork products has increased, and higher quality requirements have been put forward. However, due to the rich fat and protein content in meat products, the water activity is high. In the process of processing, storage and sale, it is easily contaminated by microorganisms and affected by environmental factors, causing the products to spoil and deteriorate, thus losing their edible value. Fresh pork is susceptible to spoilage and deterioration during the circulation and storage process due to the action of endogenous enzymes, external environment, and microorganisms. The protein in raw pork is decomposed by enzymes and bacteria to produce basic nitrogenous substances such as ammonia and amines, and combines with acidic substances in tissues to form salt-based nitrogen. The TVB-N content in fresh pork showed three processes of slow increase, steady increase and large increase with the extension of the standing time.
Spectral imaging technology is an emerging platform technology. It is a combination of spectral analysis technology and image analysis technology. It has spatial resolution and spectral resolution, and can simultaneously obtain spatial and spectral information of objects. Spectral imaging technology is widely used in the detection of food and agricultural products. Peng and Lu et al. used spectral scattering characteristics to predict the pH, tenderness and color of beef. Multispectral imaging technology is superior in portable, small, and mass production applications.
The goal of this paper is to explore the feasibility of using multispectral imaging techniques combined with mathematical modeling methods to predict the shelf life of fresh pork.
1 Materials and methods 1.1 Multi-spectral imaging system Received H: revised period: Item 9 Gold: Public welfare industry (agriculture) research special funding (project number: 201003008).
Introduction: Li Cuiling (1985), female (Chinese), new, master, is mainly engaged in the research and development of key technologies for quality, safety and rapid detection of raw meat. No. 17 East Road, Haidian District, Beijing, 100083.Email: Corresponding author: Peng Yankun (1960), male (Han), quality and safety of agricultural and livestock research, especially destructive testing of Shandong, professor, cattle to be engaged. No.7, Tsinghua East Road, Haidian District, Beijing, China, 丨00083.丨1: Sichuan 6881.6 (.
The multispectral imaging system used in the multispectral imaging system is as shown. It mainly consists of a light source unit, an image acquisition unit and a data processing unit. The light source unit includes a regulated power supply, a bromine tungsten light source (HL2000, USA), and an optical fiber; the image acquisition unit includes a high performance visible/near infrared CCD camera (UM301, USA), a capture card (CronosPlus, Canada), and a filter. Referring to the relevant, a total of 7 narrow-band filters are selected, the center wavelengths are 551, 560, 580, 600, 625, 760, 810rnn9, and the half-height bandwidth is 1015nm; the main function of the data processing unit is to receive and save multi-spectral image data. Extract effective information and build predictive models. The black box is used to isolate interference from outside light and noise. Turn the bracket knob to adjust the height of the stage. The meat sample is placed on the stage. When the light is irradiated on the surface of the meat sample, the diffuse light of the meat sample passes through the filter, and a multi-spectral image is formed by the CCD camera, and an 8-bit image data file is generated by the image acquisition card.
In July 2011, the fresh pork tenderloin from the same pig was purchased from the supermarket 24 hours after slaughter, and the meat was divided into pieces of meat about 2 cm thick. 21 pieces were selected as effective samples, packed with plastic wrap and placed. The gradation value of the pixels on the image transported back to the image indicates the reflected light intensity of the image acquisition area of the sample.
2.2 Determination of volatile base nitrogen According to the national standard GB/T5009.44-2003, the standard method was improved appropriately. The KDY-9820 semi-automatic nitrogen analyzer was used instead of the semi-micro nitrogen method to determine the volatile salt base in pork. nitrogen. The TVB-N reference values of the 21 pork samples measured are shown in Table 1.
No. 2.3 Multispectral Scattering Image Feature Extraction The data was processed using MATLAB 7.11.0 (R2010b).
Find the average value of the gray value of all the pixels on the image with the incident point of the beam as the center of the beam and the width of the pixel (approximation) with a pixel size (8.6 nm x 8.3). The radius of the ring is increased by one pixel from small to large. The average value is used as the gray value of the pixel on the corresponding ring to reduce the error. Taking the radius of the circle as the abscissa and the gray value of the ring pixel as the ordinate, the scattering curve of the image at each wavelength can be made.
The scattering curves of the images of the 21 samples at the 760 nm band are as shown.
A nonlinear regression method was used to fit the scattering curve at each wavelength with a Lorentz function containing four parameters. The Lorentz function with four parameters is expressed as follows: where: R is the reflected light intensity (gray value) at any point on the scattering curve: z is the scattering distance (circle radius) of the point from the incident point of the beam; a is The asymptotic value of the scattering curve; b is the peak value of the scattering curve: c is the half-wavelength bandwidth of the scattering curve; and d is the slope at the inflection point of the scattering curve.
The scatter function is fitted to the scattering curve at each wavelength, so that the scattering image at each wavelength can be described by the four parameters of the LD function. Further analysis of the parameters can predict the TVB-Na of the fresh pork. A fitting plot of the scattering curve of the image is as follows. 4 The establishment of the volatile salt-based nitrogen prediction model The partial least squares regression (PLSR) method usually has a higher prediction accuracy because it takes into account the correlation with the component to be analyzed when extracting the principal component. In this paper, the prediction model of fresh pork TVB-N was established by PLSR method. The samples were divided into two groups. Seven samples with a sample number of 3 were used as the prediction set, and the remaining 14 samples were the calibration set. The TVB-N reference value for each sample corresponds to 4 parameters of the LD function of its scatter image. The TVB-N reference value of the 14 samples of the calibration set was subjected to partial least squares regression with 392 (14 x 7 x 4) parameter values. The model prediction correlation coefficient is 0.87, and the prediction standard error is 2.50mg/100g, which is the prediction effect map modeled by PLSR method.
2.5 The establishment of the shelf-life prediction model The change of the reference value in the observation can be found that the TVB-N changes logarithmically with time in the study, and it is in the stage of rapid increase and change. Therefore, this paper uses the logarithmic function to fit TVB-N. The change curve, based on this, establishes a shelf life prediction model for fresh pork.
Rate parameter; t is time (d). The transformation of the formula (2), such as shifting and logarithm, can be obtained as follows: Equation (3) is derived on the basis of the formula (2), and the shelf life of the fresh pork can be calculated.
The variation law of TVB-N is logarithmic, and the correlation coefficient of the fitting is 0.93, and the standard error is 1.76mg/100g. The fitting effect diagram is as shown. Three parameters A=29.1076, B=19.8618, k=0.1824 were fitted. The three parameters were substituted into equation (3) to establish a predictive model of shelf life (Shelflife, SL, d) of fresh pork: TVB-N The 21 TVB-N values predicted by the volatile salt-based nitrogen prediction model are brought into equation (4) over time, and the shelf-life values of the corresponding meat samples can be obtained, which is the predicted effect map of the shelf life of fresh pork. The overall trend of the forecast is correct, but the prediction accuracy is poor.
3. Conclusion The shelf life of fresh pork is one of the important basis for measuring the value of its products. Therefore, it is very important to obtain a rapid, non-invasive and reliable detection method for the shelf life of fresh pork. Although the traditional chemical laboratory analysis and other methods have the advantages of high accuracy and reliability, the analysis process is cumbersome, time-consuming, destructive to the sample, and gradually fails to meet the rapid development of production and economic requirements. Spectral image technology combines spectral and image analysis techniques to provide fast, non-destructive features, among which multispectral imaging technology is superior in portable, small, and mass production applications. Based on the multi-spectral imaging technology, this paper explores a feasible and effective detection method for the shelf life of fresh pork for small sample nonlinear problems. The results show that it is feasible to predict the shelf life of fresh pork by using multi-spectral imaging technology combined with corresponding mathematical modeling methods.
Some problems and deficiencies were found in the course of the research and need to be improved: (1) the selected samples are too small, and the extensiveness of the samples should be increased to establish the appropriate increase in the number of samples; (2) due to the acquisition of multi-spectral images. The filter is manually changed, and the thickness of the sample is not uniform, so it is difficult to ensure that the distance from the upper surface of the sample to the filter is equal, the multi-spectral imaging system used needs to be improved; (3) the error is transmitted from the volatile There are errors in the prediction model of the salt-based nitrogen to the shelf-life prediction model. The error should be minimized and the accuracy should be improved: (4) The evaluation index is single, and some other suitable indicators can be selected from the factors affecting the shelf life. They assign weight to the extent of their impact on the shelf life.
