Spectral prediction model for milk fat content in camel milk

Through mid-infrared spectroscopy technology and partial least squares regression model, combined with characteristic bands and pretreatment methods, the accuracy problem of milk fat content detection in camel milk is solved, and the rapid and accurate detection of milk fat content in camel milk is achieved, which is suitable for camel performance determination and quality inspection.

CN119000593APending Publication Date: 2024-11-22HUAZHONG AGRI UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410803543.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-20
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The existing technology lacks a fast batch detection model for the cream content in camel milk with high accuracy, which makes it difficult to achieve rapid, efficient and accurate measurement of camel milk production performance and quality detection.

Method used

Mid-infrared spectroscopy technology is adopted to collect MIR data of characteristic bands and use second-order differential and partial least squares regression models to predict the cream content, select characteristic bands containing some water absorption areas, and combine optimal preprocessing and algorithm combination to improve the The accuracy of the model.

Benefits of technology

It realizes fast, accurate and low-cost testing of the cream content in camel milk, with the testing time only 10-15 seconds, improving the detection efficiency and suitable for camel performance measurement and quality testing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure BDA0004904047420000041
    Figure BDA0004904047420000041
  • Figure BDA0004904047420000051
    Figure BDA0004904047420000051
  • Figure BDA0004904047420000061
    Figure BDA0004904047420000061
Patent Text Reader

Abstract

The invention belongs to the field of camel performance determination and camel milk quality detection, and discloses a spectrum prediction model for the content of milk fat in camel milk. The applicant uses a method of manual selection and multiple traversal, and breaks through the screening characteristics of a common algorithm. And finally, selecting a characteristic wave band for modeling, particularly an absorption area containing part of water, and proving that the accuracy of the model can be improved by increasing part of the water absorption wave band. The applicant selects the optimal preprocessing and algorithm combination of the model at the same time, the optimal parameter is determined, and the accuracy of the model is improved. The method realizes rapid, accurate and low-cost detection of the milk fat content in the raw material camel milk, realizes rapid batch detection, only needs 10-15 seconds for determination of each sample, improves the detection efficiency, has strong practicability, and can be widely applied to camel performance determination and camel milk quality detection.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of camel performance measurement and camel milk quality detection, and particularly relates to a spectral prediction model for milk fat content in camel milk. Background Art

[0002] Camels are divided into two types according to the number of humps: dromedary camels and Bactrian camels. They are adapted to extreme environments, and their physiological structure and behavioral habits enable them to survive in deserts and arid areas. Camel milk produced by Bactrian camels is called "desert white gold" and has high nutritional and medicinal value, including playing a role in the treatment of diabetes, hepatitis, and aiding digestion. Camel milk is white in color, slightly salty, and has a sweet aftertaste.

[0003] Fat is a triglyceride composed of glycerol and fatty acids, among which fatty acids are divided into saturated fatty acids, monounsaturated fatty acids and polyunsaturated fatty acids. The main function of fat is to provide energy for the human body. Each gram of fat can provide 9 kilocalories of energy, which is more than twice that of carbohydrates and proteins. In addition, fat also has the functions of protecting organs, maintaining body temperature, participating in hormone synthesis and cell formation in the body, and can serve as a carrier of fat-soluble vitamins to promote the absorption of vitamins.

[0004] The fat content of camel milk varies greatly between dromedary and Bactrian camels. The fat content of dromedary camel milk is similar to that of cow milk, at 3.61±1.01%, while the fat content of Bactrian camel milk is higher, at 5.77±1.07%. Compared with the milk of species such as cows and buffaloes, camel milk fat globules have a smaller diameter and are more susceptible to lipolytic enzymes, making them easier to digest in the human gastrointestinal system[9]. Compared with cow milk and human milk, camel milk contains only a small amount of short-chain fatty acids, but has a higher concentration of long-chain saturated fatty acids

[10] . The proportion of unsaturated fatty acids in camel milk is also significantly higher than that of cow milk, goat milk and human milk, especially the content of linolenic acid, which is an important essential fatty acid for the human body and can play an important role in regulating blood sugar and preventing fatty liver.

[0005] Mid-infrared spectroscopy (MIR) technology can achieve rapid and non-destructive batch testing of samples, including qualitative and quantitative testing. Specific chemical bonds or functional groups in molecules have different characteristic absorption peaks for mid-infrared light, which can be used to identify the structure of organic components and predict the properties and content of material components, thereby obtaining information such as the physiological state of organisms. In recent years, the research and application of MIR detection technology in the nutritional and physiological indicators of dairy cows have gradually deepened and improved. Models of substances such as milk fat, protein, lactose, and total solids in dairy milk have been established and applied to actual production. However, there is currently no highly accurate mid-infrared spectroscopy rapid batch detection model specifically for camel milk.

[0006] At present, there are many studies on the use of mid-infrared spectroscopy models to predict the material components in dairy milk at home and abroad, but there are very few reports on camel milk, which are almost absent in China. Since camels and dairy cows are two different species, their milk quality also has its own characteristics, so the dairy milk detection model is not necessarily suitable for camel milk, and the characteristic band of the spectrum is also different. Therefore, it is urgent to establish a milk fat content detection method suitable for camel milk with my country's independent property rights, thereby realizing fast, efficient, accurate and non-invasive determination of the important performance index of milk fat content in camel milk, laying the foundation for camel milk production performance determination and genetic improvement, and providing reference for dairy processing industry and consumers. The purpose of the present invention is to solve the existing problems, establish a special camel milk milk fat content rapid detection technology, and provide technical guarantee for the development of China's dairy industry. Summary of the invention

[0007] The purpose of the present invention is to provide a mid-infrared rapid batch detection method for milk fat content in camel milk. The method is simple, rapid and has a high accuracy rate compared with the true value.

[0008] Another object of the present invention is to provide an application of a mid-infrared rapid batch detection method for milk fat content in camel milk. In order to achieve the above object, the present invention adopts the following technical measures:

[0009] The mid-infrared rapid batch detection method for milk fat content in camel milk comprises the following steps:

[0010] 1. The characteristic band in the infrared spectrum of camel milk samples is 933.636cm -1 —1180.548cm -1 , 1473.756cm -1 —1639.65cm -1 , 1685.946cm -1 —1909.71cm -1 , 2006.16cm -1 —2430.54cm -1 , 2469.12cm -1 —2851.062cm -1 , 2854.92cm -1 —3020.814cm -1 , 3391.182cm -1 —3622.662cm -1 ,3634.236cm -1 —3788.556cm -1 ,4074.048cm -1 —4193.646cm -1 ,4413.552cm -1—4660.464cm -1 , 4783.92cm -1 —4807.068cm -1 MIR data in;

[0011] 2. Substitute the MIR data obtained by the measurement into the second-order difference + second-order difference + partial least squares regression model to output the prediction result of the milk fat content in camel milk;

[0012] The principal components of the partial least squares regression are 7.

[0013] In the method described above, preferably, a gap of two wave points is allowed before and after each section of the characteristic wave band.

[0014] The protection scope of the present invention also includes:

[0015] Application of mid-infrared rapid batch detection method for milk protein in camel milk in camel milk quality testing.

[0016] In the above-mentioned application, preferably, the milk quality detection is milk fat content detection.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] 1. Using the mid-infrared spectrum of camel milk, a special prediction model for the fat content of camel milk is established to break the current situation that the determination of camel milk production performance relies on the mid-infrared spectrum model of dairy cow milk.

[0019] 2. In terms of the selection of characteristic bands, the method of manual selection + multiple traversal was used to break the common algorithm screening features. Finally, the characteristic bands used for modeling were selected, especially the absorption area containing some water, and it was proved that adding some water absorption bands can improve the accuracy of the model.

[0020] 3. The optimal preprocessing and algorithm combination of the model was selected, the optimal parameters were determined, and the accuracy of the model was improved.

[0021] 4. It has achieved rapid, accurate and low-cost detection of the milk fat content in raw camel milk, and realized rapid batch detection. The measurement time for each sample is only 10-15 seconds, which improves the detection efficiency. It has strong practicality and can be widely used in camel performance measurement and camel milk quality detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The infrared spectrum (a) and average spectrum (b) of untreated camel milk sample.

[0023] Figure 2 This is the mid-infrared spectrum of the camel milk sample after D2 pretreatment.

[0024] Figure 3 The overall spectrum of the eleven selected characteristic bands (a) and the enlarged view of each characteristic band (b).

[0025] Figure 4 This is the correlation and fitted straight line graph of the true value and predicted value of the camel milk data of the model. DETAILED DESCRIPTION

[0026] The technical solutions described in the present invention, unless otherwise specified, are all conventional solutions in the art; the reagents or materials described, unless otherwise specified, are all from commercial channels.

[0027] 1. Experimental Materials

[0028] The experimental materials were healthy and disease-free Xinjiang Junggar Bactrian camels from five large-scale camel farms in Urumqi, Xinjiang, my country. A milk sample was collected from each camel, and a total of 161 milk samples were collected. The sampling work was carried out by experienced camel breeders. First, the milk room was wiped with a sterilized towel, and then the udder was disinfected with an iodine-glycerol mixed solution. After squeezing out the first three handfuls of milk, the milk sample was collected. 100-150ml of each milk sample was collected (packed in 2-3 DHI sampling tubes), numbered in sequence, and bronopol preservative was immediately added to each sampling bottle, and slowly shaken to fully dissolve it. A part of the milk sample was stored at 4°C for the determination of its conventional composition (dairy cow prediction model) and extraction of mid-infrared spectra (MIRS) by mid-infrared spectrometer the next day. Spectral data were collected twice for all milk samples. A part of the samples was frozen and stored for the determination of milk fat content in camel milk.

[0029] 2. Mid-infrared spectrum measurement and collection

[0030] The sample was poured into a cylindrical sample tube with a diameter of 3.5 cm and a height of 9 cm, and then placed in a water bath at 42°C for 15-20 min. The sample was scanned using the MilkoScan TM The 7RM milk composition detector extends the solid fiber optic probe into the liquid, mixes the sample and then scans it.

[0031] 3. Detection method of the true (reference) value of milk fat content in camel milk

[0032] 3.1 Instruments, equipment and reagents

[0033] Analytical balance (Mettler-Toledo Instrument Co., Ltd. (AL204)): sensitivity 0.0001gg; electric blast drying oven (Shanghai Jinghong Testing Equipment Co., Ltd. (DHG-9140A)); liposuction bottle (100ml centrifuge tube with cork); constant temperature water bath (Huanghua Feisifu).

[0034] Ammonia (NH3·H2O) (Aladdin, mass fraction is 25%); Ethanol (C2H5OH) (Sinopharm, volume fraction is 95%); Anhydrous ether (C4H 10 O)(Sinopharm); Petroleum ether(C n H 2n+2 ) (Chinese medicine, boiling range is 30℃~6

[0035] 0℃); hydrochloric acid (HCl) (Chinese medicine).

[0036] 3.2 Experimental methods

[0037] Determination of milk fat by alkaline hydrolysis

[0038] (1) Solution preparation: 1. Mixed solvent: Mix ether and petroleum ether in equal volumes and prepare before use. 2. Hydrochloric acid solution (6 mol / L): Measure 50 mL of hydrochloric acid and slowly pour it into 40 mL of water, make up to 100 mL, and mix well.

[0039] (2) Alkaline hydrolysis of the sample: Weigh 10 g (accurate to 0.0001 g) of the thoroughly mixed sample into a liposuction bottle. Add 2.0 mL of ammonia water and immediately place the liposuction bottle in a 65°C ± 5°C water bath after thorough mixing. Heat for 15 to 20 minutes, taking it out and shaking it from time to time. After taking it out, cool it to room temperature. Let it stand for 30 seconds.

[0040] (2) Extraction: 1. Add 10 mL of ethanol and mix gently but thoroughly to avoid the liquid getting too close to the bottleneck. 2. Add 25 mL of ether, plug the bottle, keep the liposuction bottle in a horizontal position, clamp the extended part of the ball upward on the shaker, and shake at about 100 times / min for 1 minute. Manual shaking can also be used. However, care should be taken to avoid the formation of a persistent emulsion. After the liposuction bottle has cooled, carefully open the stopper and rinse the stopper and bottleneck with a small amount of mixed solvent to allow the flushing liquid to flow into the liposuction bottle. 3. Add 25 mL of petroleum ether, plug the re-wetted stopper, and gently shake for 30 seconds as described in 2. 4. Place the stoppered liposuction bottle in a centrifuge and centrifuge at 500 r / min~600 r / min for 5 minutes. Otherwise, let the liposuction bottle stand for at least 30 minutes until the upper liquid is clear and clearly separated from the water phase. 5. Carefully open the stopper and rinse the stopper and the inner wall of the bottleneck with a small amount of mixed solvent to allow the flushing liquid to flow into the liposuction bottle. If the interface between the two phases is lower than the junction of the ball and the bottle body, slowly add water along the edge of the bottle wall to make the liquid level higher than the junction of the ball and the bottle body for easy pouring.

[0041] (3) Weighing: Combine all the extracts and remove the solvent in the fat collection bottle by distillation or by evaporating it to dryness in a boiling water bath. Rinse the inside of the bottleneck with a small amount of mixed solvent before distillation. Dry the fat collection bottle in an oven at 100℃±5℃ for 1 hour, take it out and cool it in a desiccator for 0.5 hour before weighing. Repeat the above steps until constant weight is reached (until the difference between the two weighings does not exceed 2 mg).

[0042] (4) Calculation:

[0043]

[0044] Where:

[0045] X——the fat content in the sample, in grams per hundred grams (g / 100g);

[0046] m1——the mass of the fat collection bottle and fat after constant weight, in grams (g);

[0047] m2——the mass of the fat collection bottle, in grams (g);

[0048] m3——the mass of the fat collection bottle and the extract after constant weight in the blank test, in grams (g);

[0049] m4——the mass of the fat collection bottle in the blank test, in grams (g);

[0050] m——the mass of the sample, in grams (g);

[0051] 100——Conversion factor.

[0052] The calculation result is expressed to one decimal place.

[0053] (5) Repeat the test twice for each sample and take the average value.

[0054] 4. Selection of effective samples

[0055] Among the 161 samples, invalid data caused by sample deterioration, loss, abnormal mid-infrared spectrum measurement, abnormal sample reference value measurement, etc. were eliminated. A total of 9 abnormal samples were eliminated, and 152 samples were selected for model establishment and optimization.

[0056] Figure 1 (a) shows the infrared spectra of 152 untreated camel milk samples. Figure 1 (b) is the average spectrum of 152 samples.

[0057] Embodiment 1:

[0058] Algorithm selection for the prediction model of milk fat in camel milk:

[0059] The purpose of this application is to establish a quantitative determination model for milk fat in camel milk, so the modeling algorithm used is a regression algorithm. There are many types of regression algorithms. This embodiment mainly uses the partial least squares regression (PLSR) algorithm to establish the model for the following reasons:

[0060] The PLSR algorithm is one of the most effective algorithms in multi-feature samples. Spectral features often have a strong linear correlation with the detected substance, so the algorithm that can better handle linear problems and occupy less computing resources becomes the first choice, among which PLSR is one of the most commonly used classical algorithms and the standard construction algorithm for the model used in FT-MIRS instruments. PLSR can better solve the problem of multicollinearity, associate the main components extracted from the features and target values, and use multivariate linear regression for model training to quickly obtain the optimal results, so this algorithm is selected in this embodiment.

[0061] Embodiment 2:

[0062] Preliminary screening of the mid-infrared spectroscopy model algorithm and its optimal preprocessing:

[0063] Effective feature screening is a basic operation for processing spectral data, the purpose is to eliminate noise and lay a good foundation for extracting features. Effective feature screening mainly includes feature extraction, feature preprocessing and feature dimension reduction. This embodiment mainly uses 12 methods such as None (no preprocessing), MMS (normalization), SS (standardization), CT (mean centering), SNV (standard normal variable transformation), MA (moving average smoothing), SG (convolution smoothing), MSC (multiple scattering correction), D1 (first-order derivative), D2 (second-order derivative), DT (trend correction), WAVE (wavelet transform) to perform feature preprocessing on spectral data.

[0064] In this embodiment, each sample corresponds to a piece of MIR spectrum data. The full spectrum band is substituted into the model to compare and analyze the accuracy of the model to determine the initial optimal preprocessing of the model. The results are as follows:

[0065] The comparison results are as follows:

[0066]

[0067]

[0068] After comparison, it is found that D2 preprocessing has better effect, the training set result is higher, and the test set effect has reached the level of preliminary application, without obvious overfitting. The training set effect of other preprocessing algorithms (except D1) is slightly worse overall, and the overall effect of D1 is slightly worse than that of D2. Figure 2 This is the infrared spectrum of the camel milk sample after D2 pretreatment.

[0069] Embodiment 3:

[0070] Establishment of a method for detecting milk fat content in camel milk using mid-infrared spectroscopy:

[0071] 1. Division of modeling data set

[0072]

[0073] In the modeling data set division in this embodiment, 75% is the training set and 25% is the test set. The ratio of the training set to the test set is 3:1. During the model training process, 10-fold cross validation is performed, so the training set is also called the cross validation set.

[0074] 2. Manual selection process and determination of modeling feature bands

[0075] There are many methods for selecting feature bands, mainly including algorithmic feature selection and manual feature selection. The principle of algorithmic feature selection is mainly derived from the correlation between each wave point and the reference value. The advantages are fast speed and high efficiency, but the disadvantages are that it ignores the collinearity problem between adjacent wave points and the idea is relatively simple; the advantage of manual feature selection is that the role of the band (that is, adjacent wave points) can be strengthened during the selection process, and at the same time, the original information state of the spectrum can be retained more in the process of improving the model. It has stronger inclusiveness and generalization ability and accurate band selection. The disadvantages are slow selection speed and low efficiency.

[0076] This embodiment selects the characteristic band by manual selection, and the selection steps are as follows:

[0077] (1) Determine the initial optimal pretreatment. From Example 2, it can be seen that the initial pretreatment effect of D2 is better.

[0078] (2) Determine the optimal second pretreatment. During the modeling process, sometimes only one pretreatment will limit the generalization ability and optimization potential of the model. Integrated pretreatment has been proven to produce better results in the model building process. In the modeling process, this embodiment found that a second pretreatment can better optimize the model. After comparison, it was found that the second pretreatment using D2 further improved the effect on the training set, and the test set effect was significantly improved, with greater optimization potential, so D2 was finally selected as the best second pretreatment method (the results are shown in the following table).

[0079]

[0080] The manual selection process of modeling feature bands is as follows:

[0081] (1) The whole band is segmented into 100 wave points per segment, and finally divided into 11 segments. According to the preliminary experiment, it is found that retaining some spectral regions related to water absorption can improve the accuracy of the model. Therefore, in this step, the spectral regions related to water absorption are not removed.

[0082] (2) Taking 50-80 wave points as a group, the PLSR algorithm is used to first increase or decrease a group of wave points at both ends of the critical point of the first band to find the optimal effect, and then perform similar operations on the second band based on this. Finally, the first traversal is completed after all 11 bands have completed a round of operations.

[0083] (3) After the first traversal is completed, perform a second, third, or more manual traversals until all wave points no longer change, which is the optimal characteristic band.

[0084] Finally, after fifteen rounds of screening, the optimal results were obtained, as shown in the table:

[0085]

[0086] The final selected characteristic band result is: 933.636cm -1 —1180.548cm -1 , 1473.756cm -1 —1639.65cm -1 , 1685.946cm -1 —1909.71cm -1 , 2006.16cm -1 —2430.54cm -1 , 2469.12cm -1 —2851.062cm -1 , 2854.92cm -1 —3020.814cm -1 , 3391.182cm -1 —3622.662cm -1 ,3634.236cm -1 —3788.556cm -1 ,4074.048cm -1 —4193.646cm -1 ,4413.552cm -1 —4660.464cm -1 , 4783.92cm -1 —4807.068cm -1, and a gap of two wave points is allowed before and after each segment. The results show that after adding the first segment of water absorption and part of the second segment of water absorption area to the model, the model can achieve the optimal effect, indicating that the characteristic band of milk fat in camel milk contains part of the water absorption area. Figure 3 The overall spectrum of the eleven selected characteristic bands (a) and the enlarged view of each characteristic band (b).

[0087] 3. Screening and determination of model parameters

[0088] The model parameters include the parameters of the preprocessing method and the parameters of the algorithm. The D2 preprocessing method involved in this model has no parameters; the main parameters are the parameters of the partial least squares regression algorithm: the principal component (n_component). The comparison of parameter selection results is as follows (partial):

[0089]

[0090] According to the comparison results, the main component (n_component) is finally selected as 7.

[0091] After comparative analysis, the best regression model for camel milk fat content is: D2 (second-order difference) + D2 (second-order difference) + PLSR (n_component = 7) model. The determination coefficients of the training set and the test set are 0.9960 and 0.9923 respectively; the root mean square errors are 0.0868 and 0.1410 respectively; and the RPDs are 15.8249 and 11.4195 respectively. Figure 4 The correlation and fitted straight line graph of the true value and predicted value of the camel milk training set and test set data for the model.

[0092] Example 4

[0093] Application of mid-infrared spectroscopy rapid batch detection method for milk fat content in camel milk:

[0094] The established optimal regression model for milk fat in camel milk (D2(second-order difference)+D2(second-order difference)+PLSR(n_component=7)) was used to predict 10 randomly selected external camel milk samples (not one of the 152 experimental materials), and the predicted results were compared with the true values.

[0095] 1. The characteristic band of the infrared spectrum of camel milk samples is: 933.636cm -1 —1180.548cm -1 , 1473.756cm -1 —1639.65cm -1 , 1685.946cm -1 —1909.71cm -1 , 2006.16cm-1 —2430.54cm -1 , 2469.12cm -1 —2851.062cm -1 , 2854.92cm -1 —3020.814cm -1 , 3391.182cm -1 —3622.662cm -1 ,3634.236cm -1 —3788.556cm -1 ,4074.048cm -1 —4193.646cm -1 ,4413.552cm -1 —4660.464cm -1 , 4783.92cm -1 —4807.068cm -1 MIR data in;

[0096] At the same time, the alkaline hydrolysis method was used to detect the true value of the milk fat content in the same batch of camel milk.

[0097] 2. Substitute the MIR data obtained by the measurement into the D2 (second-order difference) + D2 (second-order difference) + PLSR (n_component = 7) model constructed in Example 3 to output the prediction result of the milk fat content in camel milk.

[0098] The results predicted by the model are very close to the actual results (as shown in the following table), so the model has high accuracy and can be used to predict the milk fat content in camel milk.

[0099]

Claims

1. A mid-infrared rapid batch detection method for milk fat content in camel milk, comprising the following steps: (1) The characteristic band in the infrared spectrum of camel milk samples is 933.636 cm -1 —1180.548cm -1 , 1473.756cm -1 —1639.65cm -1 , 1685.946cm -1 —1909.71cm -1 , 2006.16cm -1 —2430.54cm -1 , 2469.12cm -1 —2851.062cm -1 , 2854.92cm -1 —3020.814cm -1 , 3391.182cm -1 —3622.662cm -1 ,3634.236cm -1 —3788.556cm -1 ,4074.048cm -1 —4193.646cm -1 ,4413.552cm -1 —4660.464cm -1 , 4783.92cm -1 —4807.068cm -1 MIR data in; (2) Substituting the MIR data obtained by the measurement into the second-order difference + second-order difference + partial least squares regression model, the prediction result of the milk fat content in camel milk can be output; The principal components of the partial least squares regression are 7.

2. The detection method according to claim 1, characterized in that: For the characteristic band, a gap of two wave points is allowed before and after each section.

3. Application of the detection method according to claim 1 in camel milk quality detection.

4. The use according to claim 3, wherein the milk quality detection is milk fat content detection.

Citation Information

Patent Citations

  • Intermediate infrared rapid batch detection method for fat content in buffalo milk and application

    CN114166785A

  • Intermediate infrared spectrum detection method for protein content in buffalo milk and application

    CN114166787A

  • Intermediate infrared rapid batch detection method for content of free methionine in milk

    CN114166790A

  • Intermediate infrared rapid batch detection method for alpha-lactalbumin in milk

    CN114184572A