Method for determining freshness of different grades of salmon meat based on colorimetric array sensor

Abstract

The invention relates to a colorimetric array sensor for determining different grades of freshness of salmons and application of the colorimetric array sensor, the colorimetric array sensor comprises a sensitive material layer composed of a solid-phase carrier and a sensitive material, the solid-phase carrier is a nitrocellulose membrane, the sensitive material is a composition consisting of eight indicators, namely bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3, 3, 5, 5-tetraiodo phenolsulfonphthalein, pararosaniline and bromothymol blue. The colorimetric array sensor is used for carrying out real-time nondestructive testing on different grades of freshness of the refrigerated salmon, a new technical means is provided for quality control of the salmon, and beneficial reference is also provided for freshness detection of other aquatic products.

Description

Technical Field

[0001] The invention relates to the field of food detection, and in particular to a method for determining the freshness of salmon meat at different levels. Background Art

[0002] As an important raw material for sashimi, salmon is rich in protein, brain gold, dimethylaminoethanol and other nutrients, and its meat is delicious and tastes good, so it is favored by more and more consumers. However, due to its soft texture, high protein content and rich unsaturated fatty acids, salmon muscle tissue is easily contaminated by microorganisms after death, leading to corruption and deterioration, which not only significantly reduces the economic value of salmon, but also poses a potential threat to consumers' health. The "shelf life" label on the food packaging alone cannot accurately reflect the actual remaining edible time of salmon.

[0003] Common methods for evaluating salmon freshness include sensory evaluation, physical and chemical indexes, and microbiological testing. Although these methods can provide accurate quantitative results, they often require expensive equipment, cumbersome operating steps, long testing time, and are destructive. Therefore, there is an urgent need for a method that can perform real-time, non-destructive testing of salmon freshness.

[0004] In recent years, research on smart packaging has gradually become a hot topic. It aims to provide consumers with more accurate product information by monitoring changes in the internal quality of the product to provide real-time feedback on the packaging status. The colorimetric array sensor detects different freshness levels of salmon through multiple reaction points. It has the advantages of high sensitivity, high precision, rapid response and non-destructive detection, and has become an effective tool for monitoring the freshness of salmon.

[0005] The colorimetric sensor is mainly composed of a fixed carrier and a sensitive material. As the storage time of salmon increases, short-chain alcohols of C6-C9, ammonia, biogenic amines such as trimethylamine, and other sulfur-containing volatile substances will be produced in the fish meat. These compounds can be used as indicators of freshness. Certain sensitive materials can react with these volatile substances, causing color changes, thereby converting the chemical changes in the spoilage process into visible color changes. Therefore, by analyzing the difference in the image before and after the exposure of the colorimetric array sensor, real-time monitoring of the freshness of salmon can be achieved.

[0006] Solid phase carrier is an important component of colorimetric array sensor, which directly affects the performance of sensor. Cellulose is the most commonly used carrier material, which has the advantages of low cost, wide source and simple preparation, and has become a hot topic in the research of freshness indicator.

[0007] At present, there are no reports on the method of measuring the freshness of salmon meat at different levels based on colorimetric array sensors. Summary of the invention

[0008] To solve the above technical problems, the present invention includes the following aspects:

[0009] A first aspect of the present invention provides a colorimetric array sensor for determining different levels of freshness of salmon meat, wherein the colorimetric array sensor comprises a sensitive material layer composed of a solid phase carrier and a sensitive material, wherein the solid phase carrier is a nitrocellulose membrane, and the sensitive material is selected from one or more of bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline and bromothymol blue indicators.

[0010] Preferably, the sensitive material is a composition consisting of three indicators: bromophenol blue, thymol blue and 3,3,5,5-tetraiodophenolsulfonphthalein.

[0011] Preferably, the sensitive material is a composition consisting of eight indicators, namely, bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline and bromothymol blue.

[0012] Preferably, the colorimetric array sensor further comprises an optical adhesive layer, and the optical adhesive layer is located below the sensitive material layer.

[0013] Preferably, the material of the optical adhesive layer is LG optical adhesive.

[0014] Preferably, the colorimetric array sensor further comprises a PVC layer, wherein the PVC layer is on the upper layer of the sensitive material layer, and the PVC layer serves as a backing layer of the solid phase carrier.

[0015] Preferably, the colorimetric array sensor further comprises a polypropylene (PP) board layer, and the polypropylene board layer is located under the optical adhesive layer.

[0016] Preferably, the colorimetric array sensor comprises, from top to bottom, a PVC layer, a sensitive material layer, an optical adhesive layer and a polypropylene (PP) board layer.

[0017] Preferably, the colorimetric array sensor comprises 1-10 indicators, and the indicators are composed of the above-mentioned PVC layer, the above-mentioned sensitive material layer and the above-mentioned optical adhesive layer from top to bottom.

[0018] Preferably, the indicator is square, rectangular or circular. More preferably, the indicator is circular.

[0019] Preferably, the colorimetric array sensor comprises 1-10 of the above indicators and a polypropylene plate.

[0020] More preferably, the colorimetric array sensor comprises 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 of the above indicators and a polypropylene plate.

[0021] Further preferably, the colorimetric array sensor comprises 8 indicators and a polypropylene plate, the indicator consisting of a PVC layer, a sensitive material layer and an optical adhesive layer from top to bottom, the sensitive material layer consisting of a solid phase carrier and a sensitive material, the solid phase carrier is a nitrocellulose membrane, and the sensitive material is a combination of bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline and bromothymol blue indicators.

[0022] Preferably, the top PVC layer of the colorimetric array sensor is torn off before use.

[0023] A second aspect of the present invention provides a method for preparing the above-mentioned colorimetric array sensor, the method comprising the following steps:

[0024] (1) soaking the nitrocellulose membrane with PVC backing into different sensitive materials respectively, taking out the nitrocellulose membrane and drying it at room temperature to prepare a sensitive material layer with a PVC layer;

[0025] (2) attaching the sensitive material layer with the PVC layer obtained in step (1) to the optical adhesive layer respectively, and then cutting it into circular indicators;

[0026] (3) The circular indicators obtained in step (2) are sequentially pasted on a polypropylene plate to obtain a colorimetric array sensor.

[0027] Preferably, the sensitive material is selected from one or more of bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline and bromothymol blue indicators.

[0028] More preferably, the sensitive material is a composition consisting of three indicators: bromophenol blue, thymol blue and 3,3,5,5-tetraiodophenolsulfonphthalein.

[0029] Further preferably, the sensitive material is a composition consisting of eight indicators including bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline and bromothymol blue.

[0030] Preferably, the concentration of the sensitive material in step (1) is 100-300 ppm. More preferably, the concentration of the sensitive material in step (1) is 200 ppm.

[0031] Preferably, in step (1), the nitrocellulose membrane is immersed in the sensitive material for 5-20 minutes. More preferably, in step (1), the nitrocellulose membrane is immersed in the sensitive material for 10 minutes.

[0032] Preferably, the drying time of the nitrocellulose membrane in step (1) is 40-80 min. More preferably, the drying time of the nitrocellulose membrane in step (1) is 60 min.

[0033] Preferably, the thickness of the optical adhesive layer in step (2) is 50-200 mm. More preferably, the thickness of the optical adhesive layer in step (2) is 100-150 mm. Further preferably, the thickness of the optical adhesive layer in step (2) is 125 mm.

[0034] Preferably, the diameter of the circular indicator in step (2) is 4-10 mm. More preferably, the diameter of the circular indicator in step (2) is 6 mm.

[0035] Preferably, the top PVC layer of the colorimetric array sensor is torn off before use.

[0036] A third aspect of the present invention provides the use of the above-mentioned colorimetric array sensor in detecting different levels of freshness of salmon meat.

[0037] A fourth aspect of the present invention provides a salmon product that can display different levels of freshness, the product comprising salmon meat, the above-mentioned colorimetric array sensor and outer packaging.

[0038] A fifth aspect of the present invention provides a method for determining different levels of freshness of the above-mentioned salmon product, the method comprising:

[0039] Method A:

[0040] Observe the color changes of the reaction points of the colorimetric array sensor in the tested salmon product. If the reaction points of bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline and bromothymol blue are blue, brown, green, yellow, dark purple, green, red and yellow respectively, it means that the salmon is spoiled; or

[0041] Method B:

[0042] (1) Using a camera to take a picture of the colorimetric array sensor in the salmon product to be tested, the original image obtained by taking the picture is preprocessed, and Photoshop is used to obtain the three color features Rm, Gm and Bm of each reaction point, and the Euclidean distance ED value is calculated;

[0043] (2) judging the freshness of salmon based on the change of ED value: when the ED value is between 100 and 200, the salmon is in the fresh stage; when the ED value is between 200 and 400, the salmon is in the semi-fresh stage; and when the ED value exceeds 400, the salmon is in the spoilage stage; or

[0044] Method C:

[0045] (1) Using a camera to take a picture of the colorimetric array sensor in the salmon product to be tested, the original image obtained by taking the picture is preprocessed, and Photoshop is used to obtain 18 color features Rm, Gm, Bm, RGBn, Rm / RGBn, Gm / RGBn, Bm / RGBn, ΔE, Hm, Sm, Vm, L*m, a*m, b*m, Cm, Mm, Km and Ym of each reaction point;

[0046] (2) Substituting the measured color characteristics b1, H1, H4, b4, ΔE1, H3, R7, G4, K1, and b7 values ​​into the partial least squares regression model of color characteristics and TVC values ​​to calculate the TVC value of the salmon product to be tested;

[0047] (3) Determine the freshness of the salmon according to the TVC value calculated in step (2): when the TVC value is less than 5 lg (CFU / g), the salmon is in the fresh stage; when the TVC value is between 5 and 5.69 lg (CFU / g), the salmon is in the sub-fresh stage; when the TVC value is greater than 5.69 lg (CFU / g), the salmon is in the spoilage stage.

[0048] Preferably, in step (1) of method B, a Sony zv-1 camera with a resolution of 300 dpi is used to take photos, and the camera parameters are: shutter speed 1 / 100, aperture 2.8, iso100.

[0049] Preferably, the meanings of the three color characteristics in step (1) of method B are as shown in Table 1 of the present invention.

[0050] Preferably, the Euclidean distance ED value in step (1) of method B is expressed as a color difference value (ΔE), and its calculation formula is as follows: Among them, R a , G a and B a are the pixel values ​​of the sensitive material solution in the red, green and blue channels, respectively, b , G b and B b It is the pixel value of the nitrocellulose membrane in the same channel after immersion.

[0051] Preferably, the meanings of the 18 color characteristics in step (1) of method C are as shown in Table 1 of the present invention.

[0052] Preferably, the partial least squares regression model formula in step (2) of method C is as follows: TVC value = 4.722 + 0.36*b1-0.113*H1 + 0.171*H4-0.5*b4 + 0.826*ΔE1 + 0.249*H3 + 0.157*R7 + 0.764*G4 + 0.322*K1-0.948*b7.

[0053] The technical effects produced by the present invention are:

[0054] 1. The present invention screened a variety of paper-based materials and unexpectedly found that nitrocellulose membrane has a good gas-sensitive dye carrying capacity, and by adding a PVC backing to the nitrocellulose membrane, the problem of curling and becoming brittle after drying and being unable to undergo subsequent processing was effectively solved.

[0055] 2. The present invention selects 8 indicators as sensitive materials and uses a solution impregnation method to prepare a colorimetric sensor based on a nitrocellulose membrane. The sensor has a rapid response in detecting the freshness of salmon meat at different levels and can provide accurate results in a short time. The detection speed is much better than that of traditional detection methods. In addition, the array-type colorimetric sensor of the present invention can make a comprehensive judgment on the freshness through multiple color change points, which significantly improves the accuracy and reliability of the detection results compared with other single indicator films.

[0056] 3. The present invention uses a colorimetric array sensor to perform real-time nondestructive detection of different levels of freshness of refrigerated salmon meat, and combines HCA, PLS-DA, LDA and RF analysis methods to identify the color characteristics of the sensor, screens out key color features through VIP value and average decrease precision, and further predicts the changes of TVB-N and TVC through the PLSR model. This not only provides a new technical means for the quality control of salmon meat, but also provides a useful reference for the freshness detection of other aquatic products. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 are the TVB-N values ​​(a) and TVC values ​​(b) of salmon meat during storage at 4°C;

[0058] Figure 2 The application of colorimetric array in salmon samples during storage at 4°C (a), color change (b) and Euclidean distance change (c) over time;

[0059] Figure 3 It is the HCA analysis of the color characteristics of the colorimetric array during salmon cold storage;

[0060] Figure 4 It is the 2d score graph of PLS-DA analysis of colorimetric array (a) and the score graph of the top 15 VIP values ​​(b);

[0061] Figure 5 It is the LDA analysis chart of the quality attribute index of salmon during storage at 4°C;

[0062] Figure 6 The random forest analysis graph (a) and mean decreasing precision graph (b) of the colorimetric array are shown;

[0063] Figure 7 This is the fitting diagram of the regression model of TVB-N value (a) and TVC value (b) based on PLSR;

[0064] Figure 8 The ED value changes of the colorimetric array stored at different temperatures (a) and the detected ED value after 30 days of storage (b);

[0065] Fig. 9 It is the appearance result of the screening sample of paper-based material;

[0066] Fig.10 is the ED value of the screening sample of paper-based materials;

[0067] Fig.11 It is a graph showing the carrying capacity of nitrocellulose membrane for different gas-sensitive dyes;

[0068] Fig.12 These are the test results of colorimetric arrays constructed with different gas-sensitive materials in salmon samples of different freshness;

[0069] Fig.13 It is the result of one-way ANOVA of 162 color characteristics in the gas-sensitive material screening test. DETAILED DESCRIPTION

[0070] The present invention is further described below in conjunction with examples, but the embodiments of the present invention are not limited thereto. The experimental methods used in the following examples are conventional methods unless otherwise specified.

[0071] Experimental Example 1: Detection of salmon freshness based on nitrocellulose membrane colorimetric array sensor

[0072] 1. Materials and instruments

[0073] Experimental samples: Whole Norwegian salmon (species: Atlantic salmon, weight: 6.5kg±0.2kg, country of origin: Norway) were purchased from Beijing Haiyun Baichuan Food Co., Ltd. The samples were killed on the same day, airlifted to Beijing Capital International Airport within 24 hours, and then transported to the laboratory on ice.

[0074] Experimental reagents: bromophenol blue (indicator grade), chlorophenol red (indicator grade), bromocresol green (indicator grade), hydroxynaphthol blue (indicator grade), bromothymol blue (indicator grade), parosaniline (indicator grade), 3,3,5,5-tetraiodophenolsulfonphthalein (indicator grade, Shanghai Aladdin Biochemical Technology Co., Ltd.), thymol blue (analytical grade, Shanghai MacLean Biochemical Technology Co., Ltd.), 95% ethanol (analytical grade, Sinopharm Chemical Reagent Co., Ltd.), plate count agar medium (Beijing Luqiao Technology Co., Ltd.), LG optical glue (Huizhou Langu Electronic Technology Co., Ltd.), nitrocellulose membrane with backing (CN140, Shanghai Jieyi Biotechnology Co., Ltd.)

[0075] Instruments and equipment: automatic Kjeldahl nitrogen analyzer (Kjeltec 8400, FOSS, Switzerland), texture analyzer (CTXTM, Brookfield, USA), colorimeter (3nh-NR110, Trinity, Shenzhen, China), hole puncher (6mm, Deli, China); electronic balance (accuracy 0.00001g, ME403E, METTLER TOLEDO, Switzerland), laboratory pH meter (S20K, METTLER TOLEDO, Switzerland), Sony camera (ZV-1, Sony (China) Co., Ltd.

[0076] 2. Test methods

[0077] 2.1. Determination of TVB-N value

[0078] The TVB-N values ​​of salmon meat samples stored for different times under refrigerated conditions were determined by the automatic Kjeldahl nitrogen determination method in accordance with GB 5009.228-2016 "National Food Safety Standard Determination of Volatile Basic Nitrogen in Food". TVB-N was calculated by the volume of the consumed hydrochloric acid standard titration solution in mg / 100g. Three parallels were measured for each sample, and the average value was taken for result analysis.

[0079] 2.2 Determination of TVC value

[0080] The total colony count is determined according to the method in GB 4789.2-2022 "National Food Safety Standard - Determination of Total Colony Count for Food Microbiological Examination" to determine the total colony count of the sample during storage, and the count results of total live bacteria are expressed in lg CFU / g.

[0081] 2.3. Preparation of colorimetric sensor array

[0082] A nitrocellulose membrane with a backing was selected as the solid phase carrier of the indicator, and 8 indicators including bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline, and bromothymol blue were selected as sensitive materials to construct a colorimetric array, and the sensitive materials were fixed by the immersion adsorption method. After the cut nitrocellulose membranes were immersed in different sensitive materials (200ppm) for 10 minutes, the nitrocellulose membranes were taken out and dried naturally at room temperature for 60 minutes. The dried nitrocellulose membranes were then attached to 125 mm LG optical adhesives, and finally cut into circular indicators with a diameter of 6 mm using a cutter, and finally pasted one by one onto the PP plate to form a 3 × 3 colorimetric array sensor (bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline, and bromothymol blue correspond to reaction points 1-4 and 6-9, respectively, and point 5 is a blank point. For specific positions, see Figure 2 b (lower left corner schematic diagram), and use this array to distinguish the freshness of salmon. The prepared colorimetric array sensor consists of a PVC layer, a sensitive material layer, an optical adhesive layer, and a PP layer from top to bottom. The top PVC layer needs to be torn off before use.

[0083] 2.4. Determination of salmon freshness using colorimetric array sensor

[0084] The salmon samples were washed with 4°C precooled physiological saline, peeled, de-bone, and cut into fish pieces, trying to ensure that each piece had the same shape, size, and weight. About 100g of fish pieces were weighed and packed into food vacuum bags. The colorimetric array sensor prepared above was placed in the center of the fish piece surface in the vacuum bag. At the same time, the air was exhausted to simulate the sales environment. The packaged samples were placed in a 4°C constant temperature refrigerator. The first day was set within 24 hours after the salmon died. Samples were taken every 1 day to test the TVB-N value and TVC value, and the test was continued for 12 days. At the same time, a camera was used to take pictures every 4 hours, and the time for each sample to be taken out of the refrigerator and photographed did not exceed 2 minutes.

[0085] 2.5 Image color feature extraction

[0086] To ensure the consistency of shooting conditions, all colorimetric array images were taken in a specific dark room with the same ambient light source and the same resolution of 300dpi using a Sony zv-1 camera to obtain the initial image. The camera parameters were: shutter speed 1 / 100, aperture 2.8, iso100. At the same time, the image was preprocessed by L value normalization. According to the blank in the center, the L value was used as the normalization object to adjust the brightness of the image to the same degree. After image preprocessing, Photoshop was used to explore the color features of the color array in RGB, HSV, L*a*b* and CMKY color spaces. Table 1 shows the 18 color features extracted from each indicator, and finally 144 (8 indicators × 18 color features) features were obtained. The specific color feature is named as the color feature plus the reaction point position, such as R1 represents the R color feature of the first reaction point. The color difference value (ΔE) is expressed by Euclidean distance (ED), and the calculation formula is as follows:

[0087]

[0088] Among them, R a , G a and B a are the pixel values ​​of the sensitive material solution in the red, green and blue channels, respectively, b , G b and B b It is the pixel value of the paper-based material on the same channel.

[0089] Table 1 List of color features for freshness prediction

[0090]

[0091] 2.6 Data Analysis Methods

[0092] All data are the average of 3 independent repeated experiments, expressed as mean ± standard deviation. IBM SPSSStatistics 26 software was used for difference significance analysis, p < 0.05 represented significant differences, and Excel 2021 was used to organize the data. Origin2021 pro was used for HCA and LDA analysis, MetaboAnalyst 6.0 (http: / / www.metaboanalyst.ca) was used for PLS-DA and random forest visualization analysis, SIMCA software was used to establish a PLS-DA model to verify the data test set, and SPSSPRO (https: / / www.spsspro.com / ) was used to build a random forest prediction model and a partial least squares regression (PLSR) prediction model. Color feature extraction was performed using Photoshop, selecting an 8-bit color channel, an average sampling size of 11×11 pixels, and 5 extractions for each reaction point.

[0093] 3. Test results

[0094] 3.1. Determination of TVB-N and TVC values ​​and sample division

[0095] The TVB-N value reflects the process of producing alkaline compounds such as ammonia and trimethylamine during the decomposition of proteins by enzymes and microorganisms. These compounds can cause changes in the color, flavor, total colony count and sensory characteristics of salmon meat. Therefore, TVB-N is widely used as a traditional indicator to measure freshness. GB 2733-2015 "National Food Safety Standard for Fresh and Frozen Animal Aquatic Products" requires that the TVB-N content of fresh marine fish should be controlled below 30mg / 100g. Figure 1 Figure a shows the changing trend of TVB-N value of salmon during storage at 4°C. As the storage time increases, the TVB-N value of the refrigerated salmon sample reaches 31.34 mg / 100 g on the 11th day, exceeding the standard upper limit of 30 mg / 100 g, indicating that the salmon has become spoiled and cannot be eaten. Although the TVB-N value only exceeds 30 mg / 100 g on the 11th day, refrigerated salmon cannot be eaten raw if stored for more than a week. Judging the freshness of salmon based solely on the TVB-N value has limitations, and a comprehensive evaluation of multiple indicators is required.

[0096] During transportation, storage, cutting and sales, salmon is susceptible to microbial contamination. The metabolic activity of spoilage microorganisms is the main cause of odor and texture changes. GB10136-2015 "Aquatic Products of Animal Origin" stipulates that the microbial limit for ready-to-eat raw aquatic products of animal origin is 4.69lg (CFU / g) for the total colony count and 5lg (CFU / g) for the maximum safety limit. Figure 1 As shown in b, the initial TVC value of the refrigerated salmon sample was 2.74lg (CFU / g). On the 7th day, the TVC value rose to 5.17lg (CFU / g), exceeding the maximum safety limit of ready-to-eat raw animal aquatic products in the standard. At this time, the salmon is no longer suitable for immediate consumption. Therefore, the 0-6th day can be regarded as the fresh interval. According to the provisions of the International Food Microbiological Specification Committee on Microbiological Inspection and Food Safety Control, the acceptable level of total colony count of fish is 5.69lg (CFU / g). The TVC value on the 10th day has exceeded the upper limit of the international standard. Therefore, considering the above standards, 0 to 6 days are the fresh period, 7 to 9 days are the sub-fresh period, and after 10 days, it enters the corruption period.

[0097] Based on the above classification of freshness levels, the Kennard-Stone (KS) method was used to analyze the 115 sample data with the TVC value as a reference. The KS method is a strict validation sample selection technique that can select representative samples between the training set and the validation set. Finally, the 115 samples were divided into a training set of 76 samples and a validation set of 32 samples.

[0098] 3.2. Colorimetric array response analysis

[0099] When the colorimetric array is used to measure the freshness of salmon at different levels, each array point produces different responses, showing color changes visible to the naked eye ( Figure 2 a), As the storage time increases, the color of the reaction points of the colorimetric array changes, and the color gradually accumulates ( Figure 2 b), indicating that its color change is closely related to the quality change process of salmon. When the storage time reaches 10 days, the salmon sample is completely corrupted, and multiple reaction points in the colorimetric array, such as bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline and bromothymol blue, appear blue, brown, green, yellow, dark purple, green, red and yellow, respectively. These color changes provide an intuitive basis for judging the freshness of salmon. Further analysis of the ED values ​​of the colorimetric array at different storage times ( Figure 2c), it can be found that the ED value increases significantly with the increase of storage time (p<0.05). According to the change of ED value, the freshness of salmon can be distinguished: when the ED value is between 100 and 200, the salmon can be regarded as fresh; when the ED value is between 200 and 400, the salmon is in the sub-fresh stage; and when the ED value exceeds 400, it indicates that the salmon has entered the corruption stage. It should be noted that the ED value reflects the integration of multidimensional information, and the color characteristics of a single reaction point cannot be characterized by its value alone. The color changes of many reaction points are not simple linear changes, and their change process is greatly affected by the shooting environment. Related studies have shown that the change of ED value can also be affected by environmental factors. Therefore, it is recommended to use ED value as one of the reference indicators of color characteristics. In order to more accurately evaluate the freshness of salmon, it is also necessary to combine multivariate statistical methods to establish a colorimetric array recognition model, and explore the relationship between TVB-N value, TVC value and the color characteristics of each reaction point in the colorimetric array through a regression model, so as to verify the effectiveness of the colorimetric array in salmon freshness recognition.

[0100] 3.3 HCA analysis

[0101] Hierarchical Cluster Analysis (HCA) is a method that can group the most similar variables into one category. It calculates the distance between samples, gradually aggregates samples with higher similarity, and finally merges all samples into one. This test case uses the HCA method, takes storage time as the clustering basis, and performs cluster analysis on the color characteristics of the colorimetric array of salmon at different storage times (4°C), with the number of clusters set to 4. The cluster analysis results are shown in the figure below. Figure 3 As shown. The analysis showed that the parallel samples with the same storage time were basically clustered in the same category, which verified the good stability of the colorimetric array in the detection. According to the results of HCA, the first category was blank samples; the second category was samples from the 1st to the 6th day, indicating the fresh stage; the third category was samples from the 7th to the 9th day, indicating the sub-fresh stage; the fourth category was samples from the 10th to the 12th day, indicating the corruption stage. Among the 108 samples, the classification of fresh and sub-fresh samples was completely accurate, and 3 samples of the corruption samples were misclassified as sub-fresh, and the classification accuracy of HCA was 97.2%. These results show that the application of the colorimetric sensor prepared in this experimental example in the freshness grading of salmon is feasible and has high accuracy.

[0102] 3.4 PLS-DA analysis

[0103] Partial Least Squares Discriminant Analysis (PLS-DA) is a supervised multivariate statistical analysis method, which is often used to deal with classification and discrimination problems. As a pattern recognition tool, PLS-DA can effectively classify and identify the freshness level of salmon according to the color characteristics of the colorimetric array, and help to find the key variables that affect the differences between groups. Use SIMCA14.1 software to import the training set data and validation set data. The format of the data is that the first column is the primary ID, the second column is the Class ID, and all subsequent columns are color feature columns. In the Home column, click on Model type to select the PLS-DA model, click Autofit to fit, and after fitting, set the prediction data set to the prediction data under the Predict column, and click Classification list to view the classification results of the prediction data.

[0104] The test results show that the cumulative contribution rate of the first principal component and the second principal component is 66.4% ( Figure 4 a), indicating that the PLS-DA model can well explain most of the variables. In the PLS-DA model, the accuracy (Accuracy), the determination coefficient (R 2 ) and prediction ability (Q2) are important indicators for evaluating the effectiveness of the model. The higher the value, the better the model classification effect. The model was verified by the ten-fold cross-validation method. When four components were selected, the accuracy of the model was 0.99231 and the determination coefficient (R 2 ) is 0.98391, and the prediction ability (Q2) is 0.9678. The above results show that the four components are the best choice, which can achieve good accuracy, R 2 Provides a good balance between Q1 and Q2.

[0105] VIP (Variable Importance in the Projection) score is used to measure the importance of variables in the model, which reflects the contribution of each variable to the model fit and classification ability. Usually, variables with a VIP score greater than 1 are considered to have a significant impact on the model. According to the analysis results, variables such as H4, H1, ΔE1, a4, M4, ΔE4, b1, G4, ΔE7, M1, V1, R7, L1, RGBn1, and V3 play an important role in the model ( Figure 4b), these color variables are mainly concentrated in reaction points 1, 4 and 7 (corresponding to bromophenol blue, thymol blue and 3,3,5,5-tetraiodophenolsulfonphthalein indicators, respectively), indicating that these three reaction points are key indicators for distinguishing the freshness level of salmon. The validation set was judged by SIMCA, and the results are shown in Table 2. In the training set, most of the fresh, sub-fresh and corrupt samples were correctly classified, with an average discrimination accuracy of 96.37%, and the average discrimination accuracy of the test set was 94.59%. Despite this, the model still has a small number of misjudgments between fresh and sub-fresh, and sub-fresh and corrupt. In order to further improve the classification performance of the model, especially to reduce these misjudgments, the color variables can be optimized and screened by VIP scores, and a more accurate classification model can be built based on this to ensure a higher accuracy and stronger generalization ability.

[0106] 3.5 LDA analysis

[0107] Linear Discriminant Analysis (LDA), also known as Fisher linear discriminant, is a statistical method that determines the attribution of unknown category samples by establishing a linear combination function based on the properties of known category samples. This test example uses the VIP score graph of the LDA model and selects the first 15 important color features (H4, H1, ΔE1, a4, M4, ΔE4, b1, G4, ΔE7, M1, V1, R7, L1, RGBn1, V3) for Fisher discriminant analysis. Use OriginPro 2021 to import the training set data and validation set data. The data format is that the first column is the sample name column and the second column is the classification column. Select the training set data, click Multivariate Analysis in Statistics, and then click Discriminant Analysis. Select the training sample grouping as the second column, check the prediction test data, and then add the location of the prediction data. View the classification results of the prediction data in Test Result.

[0108] from Figure 5 It can be seen that the contribution rates of LDA discriminant function 1 and discriminant function 2 are 90.87% and 9.13% respectively, and the cumulative contribution rate is 100%. In addition, the degree of separation between the groups is large, indicating that the constructed LDA model can effectively distinguish fresh, sub-fresh and corrupt samples.

[0109] The results of Wilk's Lambda test showed that the differences between the groups were statistically significant (P<0.05). The freshness of salmon meat during storage at 4°C was classified by the discriminant function, and the results are shown in Table 2. In the training set, fresh, sub-fresh and corrupt samples were correctly classified, with an average discrimination accuracy of 100%; the average discrimination accuracy of cross-validation was also 100%. In addition, all 32 samples in the test set were correctly classified, with an error rate of 0%. These results show that the LDA model established in this test case has excellent predictive ability and stability. Therefore, the salmon freshness discrimination model based on LDA can effectively and accurately judge the freshness of salmon.

[0110] 3.6 RF Analysis

[0111] Random Forest (RF) is a powerful ensemble supervised learning method. Its core idea is to build multiple decision tree models and increase the diversity of the model by randomly selecting features and samples, thereby reducing the variance of the model and improving the generalization performance. Finally, RF obtains more accurate and robust prediction results by voting or averaging the prediction results of all decision trees. In a random forest, the training data of each decision tree is generated by sampling with replacement, which means that there may be some unused data in the training data set of each tree. This part of data is called out-of-bag samples. The out-of-bag error rate (OOB) is the model error rate calculated on these out-of-bag samples, which is used to evaluate the performance of the model. Mean Decrease Accuracy is used as an indicator to measure the importance of features and is used to evaluate the contribution of each variable to the model.

[0112] Use SPSS PRO software to import the training set data. The data format is that the first column is the sample name column and the second column is the classification column. Select random forest classification in machine learning classification, put the second classification column into the categorical Y variable, and put the remaining color features into quantitative X. Select 0.7 for the training ratio, select model prediction in output result 5, import the validation set data, and view the classification results of the prediction set data in the prediction results.

[0113] In this test case, the classification of blank, fresh, sub-fresh and corrupt samples was correct, and the OOB error rate was 0.0( Figure 6 a), indicating that the random forest contrast color array has a very high recognition accuracy and can effectively judge the freshness of salmon. Through the mean descent precision evaluation, features such as b1, H1, H4, b4, ΔE1, H3, R7, G4, K1, b7, M1, L1, K4, S4, and C1 are considered to be important variables affecting model prediction ( Figure 6b) From the above test results, it can be seen that the key color features are mainly concentrated in reaction points 1, 4 and 7, which is consistent with the PLS-DA analysis results. The results of using RF to discriminate the validation set are shown in Table 2. In the training set, fresh, sub-fresh and corrupt samples were correctly classified, and the average discrimination accuracy reached 100%. In addition, all 32 samples in the test set were also correctly classified, with an error rate of 0%. This further verifies the feasibility of the colorimetric array in distinguishing the freshness level of salmon.

[0114] Table 2 Prediction results of different models for salmon freshness

[0115]

[0116]

[0117] 3.7. Construction of prediction model

[0118] Partial Least Squares Regression (PLSR) combines the advantages of principal component regression and multivariate linear regression and is an effective linear data decomposition method. PLSR is widely used to construct prediction models by converting the most effective features into the first few principal components and associating these transformed data with salmon quality indicators. According to the average decline accuracy of RF, variables such as b1, H1, H4, b4, ΔE1, H3, R7, G4, K1, and b7 were selected for modeling to construct a prediction model for TVB-N value and TVC value, and the prediction effect of the model was evaluated by calculating the relative error of the prediction. The model fitting effect is shown in Figure 2. Figure 7 The relative error rate of TVB-N value is 7.36%, including 5.26% for fresh, 16.57% for second-fresh, and 3.59% for corruption; the relative error rate of TVC value is 7.61%, including 9.57% for fresh, 6.05% for second-fresh, and 5.82% for corruption. The regression model constructed is TVB-N value = 20.151+2.39*b1+0.539*H1+3.082*H4-2.33*b4+5.025*ΔE1+4.273*H3+2.764*R7+3.168*G4-0.229*K1-4.101*b7; TVC value = 4.722+0.36*b1-0.113*H1+0.171*H4-0.5*b4+0.826*ΔE1+0.249*H3+0.157*R7+0.764*G4+0.322*K1-0.948*b7.

[0119] 3.8 Sensor stability analysis

[0120] Colorimetric array sensors are easily affected by environmental factors during storage, which may lead to a decrease in their performance. Therefore, it is particularly important to analyze their stability. The colorimetric array sensor prepared in this test example is wrapped with a layer of PVC film on the outside, which can effectively isolate the air and significantly reduce the interference of the external environment on the sensor performance. In order to evaluate the impact of different temperatures on the stability of the colorimetric array, this test example stored the colorimetric array in a constant temperature and humidity chamber at 4°C, 25°C and 37°C and a relative humidity of 50%. By calculating the total response Euclidean distance (ED) and combining it with actual sample testing, the stability and effectiveness of the colorimetric array were evaluated.

[0121] The test results showed that after 30 days of storage, the total ED of the colorimetric array was around 40 ( Figure 8 a), the color change is small. When the colorimetric array after storage at different temperature conditions for 30 days is applied to salmon detection, the ED value changes ( Figure 8 b) No significant difference was found compared with the above colorimetric array response analysis results. The above test results show that the colorimetric array prepared in this test example has excellent storage performance, can effectively resist the influence of environmental changes without the need for additional storage environment, and will not affect its subsequent use effect.

[0122] In summary, this test case divides the freshness of refrigerated salmon into three levels according to multiple national standards: 0-6 days as the fresh range, 7-9 days as the sub-fresh range, and 10 days and above as the spoilage range. Through the response analysis of the colorimetric array, the fish meat of different freshness levels can be judged by the naked eye. When monitoring refrigerated salmon samples, the accuracy of HCA cluster analysis was 97.2%, and in PLS-DA analysis, the accuracy was 0.99231, and R 2 is 0.98391, and Q2 is 0.9678, all of which indicate that the model has a good classification effect. From the comparison of the model classification effects, it can be seen that the LDA model constructed based on the VIP score has an accuracy of 100% in both the validation set and the prediction set, the OOB error rate of the RF analysis is 0.0, and the accuracy of the validation set is 100%, indicating that the color characteristics of the colorimetric array have good recognition and the best classification effect. By comparing the prediction model results of PLSR with those of traditional TVB-N values ​​and TVC values, it is shown that the colorimetric array sensor is similar to these traditional indicators in performance. At the same time, the colorimetric array sensor prepared by the present invention has excellent storage performance and can effectively resist the influence of environmental changes.

[0123] It can be seen that the present invention provides a freshness monitoring method based on a colorimetric array sensor, which not only provides a new technical means for the quality control of salmon, but also provides a useful reference for the freshness detection of other aquatic products. The non-destructive, real-time detection method of the present invention has a wide range of application prospects and can meet consumers' high requirements for food safety and freshness.

[0124] Test Example 2: Paper-based Screening Test of Colorimetric Array Sensor

[0125] 1. Test methods

[0126] Using a common gas-sensitive dye (methyl red), a screening experiment of various paper-based materials was conducted. The 12 paper-based materials screened included wood pulp cellulose paper, cotton cellulose paper, non-wood pulp cellulose paper (cellulose mixture), quantitative filter paper (cotton cellulose), slow qualitative filter paper (cotton cellulose), medium-speed qualitative filter paper (cotton cellulose), fast qualitative filter paper (cotton cellulose), nitrocellulose membrane (no backing), A4 paper (mixed plant cellulose), glass cellulose paper, polyester fiber paper and special absorbent paper. They were immersed in a 200 ppm methyl red solution for 10 minutes, and the images before and after the reaction were taken under the same environment. In Photoshop, the colors of 5 points were selected for sampling, and the average sampling area was 31×31 pixels. The Euclidean distance (ED) between the color of the methyl red solution and the methyl red solution was calculated to measure the carrying capacity of the methyl red. The ED value is a standard for measuring the difference between two colors, and its calculation formula is as follows:

[0127]

[0128] Among them, R a , G a and B a are the pixel values ​​of the methyl red solution in the red, green and blue channels, respectively, R b , G b and B b It is the pixel value of the paper-based material on the same channel.

[0129] 2. Test results

[0130] Preliminary observations showed that the nitrocellulose membrane had the best gas sensitivity and its carrying capacity for methyl red was significantly better than other paper-based materials. However, the impregnated paper showed obvious deformation problems. The nitrocellulose membrane shrank after being impregnated with methyl red, and the paper became very fragile. Even with slight pressure, the paper was easily broken (see Fig. 9Sample 8). This brittleness and deformation phenomenon caused great inconvenience in the subsequent preparation of colorimetric arrays, limiting its further application. After repeated experiments, it was found that adding PVC backing to the nitrocellulose membrane can effectively solve the problem of paper curling and brittleness. The nitrocellulose membrane with PVC backing maintains the stability and toughness of the paper while carrying the gas-sensitive dye, avoiding the breakage caused by excessive fragility (see Fig. 9 This improvement significantly improves the performance of the paper-based material and provides a more ideal paper-based choice for subsequent experiments.

[0131] The larger the calculated ED value, the more significant the color change, indicating that the paper-based material has a poorer carrying capacity for methyl red solution; conversely, the smaller the ED value, the smaller the color change and the stronger the carrying capacity. After testing 12 different paper-based materials, it was found that the nitrocellulose membrane with PVC backing had the smallest ED value among all the tested samples (see Fig.10 This indicates that the color difference between it and the methyl red solution is the smallest, indicating that it can better carry the gas-sensitive dye and maintain a relatively stable color response, thus having the best gas-sensitive material carrying capacity.

[0132] Test Example 3: Screening test of gas-sensitive materials for colorimetric array sensors

[0133] 1. Test methods

[0134] This test example investigates the carrying capacity of nitrocellulose membrane for other common gas-sensitive dyes except methyl red. Several common dyes including bromophenol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, bromothyme blue, beetroot red, chlorophenol red, phenolphthalein, hydroxynaphthol blue, bromocresol green, cresol red, etc. were selected and prepared into a 20ppm solution, and the nitrocellulose membrane was immersed in it to further evaluate its carrying performance. The Euclidean distance (ED value) of these dyes was calculated to evaluate the response of nitrocellulose membrane paper to different gas-sensitive materials.

[0135] 2. Test results

[0136] Fig.11 a is a visual display of the carrying capacity of nitrocellulose membrane for different dyes. Fig.11The data in b show that, except for beet infrared, the ED values ​​of other gas-sensitive dyes generally remain around 50. This shows that nitrocellulose membrane performs well in the carrying capacity of most common gas-sensitive dyes and can effectively maintain color stability and consistency of reaction. Therefore, nitrocellulose membrane is an ideal material choice when making colorimetric arrays, and its excellent dye carrying capacity provides reliable support for colorimetric measurement and color reaction. Therefore, nitrocellulose membrane with PVC backing was finally selected as the main paper-based material. This choice not only ensures the stability and repeatability of the experimental results, but also provides more reliable paper-based support for the practical application of gas-sensitive dyes.

[0137] Bromophenol blue, chlorophenol red, bromocresol green, hydroxynaphthol blue, thyme blue, bromothyme blue, parosaniline, 2,4-dinitrophenylhydrazine and 3,3,5,5-tetraiodophenolsulfonphthalein were selected as gas-sensitive materials for making colorimetric array reaction points, and screening experiments were carried out. Fig.12 It can be seen that the colorimetric array has a good response to salmon of different freshness, forming different fingerprints. The 18-dimensional color features of each point were extracted through Photoshop to obtain 162 color features for subsequent analysis (see Fig.13 ). From the one-way ANOVA, it can be seen that most of the color features have undergone significant changes (p < 0.05), among which the insignificant color features are concentrated in the 7th reaction point (2,4-dinitrophenylhydrazine). Therefore, 2,4-dinitrophenylhydrazine was excluded in the subsequent selection of gas-sensitive materials for the reaction points of the colorimetric array.

[0138] Although specific embodiments of the present invention have been described, it will be appreciated by those skilled in the art that various changes and modifications may be made to the present invention without departing from the scope or spirit of the present invention. Therefore, the present invention is intended to cover all such changes and modifications that fall within the scope of the appended claims and their equivalents.

Claims

1. A colorimetric array sensor for determining different levels of freshness of salmon meat, characterized in that: The colorimetric array sensor comprises a sensitive material layer consisting of a solid phase carrier and a sensitive material, wherein the solid phase carrier is a nitrocellulose membrane, and the sensitive material is selected from one or more of bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline and bromothymol blue indicators.

2. The colorimetric array sensor according to claim 1, characterized in that: The colorimetric array sensor further comprises an optical adhesive layer, which is located below the sensitive material layer.

3. The colorimetric array sensor according to claim 1, characterized in that: The colorimetric array sensor further comprises a PVC layer, which is on the upper layer of the sensitive material layer and serves as a backing layer of the solid phase carrier.

4. The colorimetric array sensor according to claim 1, characterized in that: The colorimetric array sensor further comprises a polypropylene board layer, and the polypropylene board layer is located under the optical adhesive layer.

5. The method for preparing the colorimetric array sensor according to any one of claims 1 to 4, characterized in that: The method comprises the following steps: (1) soaking the nitrocellulose membrane with PVC backing into different sensitive materials respectively, taking out the nitrocellulose membrane and drying it at room temperature to prepare a sensitive material layer with a PVC layer; (2) attaching the sensitive material layer with the PVC layer obtained in step (1) to the optical adhesive layer respectively, and then cutting it into circular indicators; (3) pasting the circular indicators obtained in step (2) onto a polypropylene plate in sequence to obtain a colorimetric array sensor; The sensitive material is selected from one or more of bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline and bromothymol blue indicators.

6. The preparation method according to claim 5, characterized in that: The concentration of the sensitive material in step (1) is 100-300 ppm.

7. The preparation method according to claim 5, characterized in that: The thickness of the optical adhesive layer in step (2) is 50-200 mm.

8. Use of the colorimetric array sensor according to any one of claims 1 to 4 in detecting different levels of freshness of salmon meat.

9. A salmon product capable of displaying different levels of freshness, characterized in that: The product comprises salmon meat, the colorimetric array sensor according to any one of claims 1 to 4, and an outer package.

10. A method for determining different levels of freshness of salmon products according to claim 9, characterized in that: The method comprises: Method A: Observe the color changes of the reaction points of the colorimetric array sensor in the tested salmon product. If the reaction points of bromophenol blue, chlorophenol red, bromocresol green, thymol blue, hydroxynaphthol blue, 3,3,5,5-tetraiodophenolsulfonphthalein, rosaniline and bromothymol blue are blue, brown, green, yellow, dark purple, green, red and yellow respectively, it means that the salmon is spoiled; or Method B: (1) Using a camera to take a picture of the colorimetric array sensor in the salmon product to be tested, the original image obtained by taking the picture is preprocessed, and Photoshop is used to obtain the three color features Rm, Gm and Bm of each reaction point, and the Euclidean distance ED value is calculated; (2) judging the freshness of salmon based on the change of ED value: when the ED value is between 100 and 200, the salmon is in the fresh stage; when the ED value is between 200 and 400, the salmon is in the semi-fresh stage; and when the ED value exceeds 400, the salmon is in the spoilage stage; or Method C: (1) Using a camera to take a picture of the colorimetric array sensor in the salmon product to be tested, the original image obtained by taking the picture is preprocessed, and Photoshop is used to obtain 18 color features Rm, Gm, Bm, RGBn, Rm / RGBn, Gm / RGBn, Bm / RGBn, ΔE, Hm, Sm, Vm, L*m, a*m, b*m, Cm, Mm, Km and Ym of each reaction point; (2) Substituting the measured color characteristics b1, H1, H4, b4, ΔE1, H3, R7, G4, K1, and b7 values ​​into the partial least squares regression model of color characteristics and TVC values ​​to calculate the TVC value of the salmon product to be tested; (3) Determine the freshness of the salmon according to the TVC value calculated in step (2): when the TVC value is less than 5 lg (CFU / g), the salmon is in the fresh stage; when the TVC value is between 5 and 5.69 lg (CFU / g), the salmon is in the sub-fresh stage; when the TVC value is greater than 5.69 lg (CFU / g), the salmon is in the spoilage stage.

Images