Intelligent detection and grading system for mutton freshness based on multi-dimensional index and multi-information fusion

Abstract

A kind of mutton freshness multidimensional index multi-information fusion intelligent detection and grading system, including mutton conveying sorting device, visible near infrared spectrometer, data processing device;Visible near infrared spectrometer obtains the original spectral data of mutton conveying sorting device conveying to-be-measured mutton;Data processing device receives the original spectral data of to-be-measured mutton obtained by visible near infrared spectrometer, and the original spectral data is preprocessed, feature wavelength is extracted from the spectral data after pre-processing, the freshness index of to-be-measured mutton is predicted according to the pre-stored mutton freshness index prediction model, to obtain the prediction value of current mutton freshness index, the prediction value of current mutton freshness index and pre-stored mutton freshness grade reference data judge the freshness grade of to-be-measured mutton, and according to the freshness grade of to-be-measured mutton judged by mutton conveying sorting device, the to-be-measured mutton is conveyed to corresponding conveying channel.

Description

TECHNICAL FIELD

[0001] The present application relates to the meat freshness index detection technology field, and particularly relates to a multi-dimensional index multi-information fusion intelligent detection and grading system for mutton freshness. BACKGROUND

[0002] Mutton contains rich nutritional components and is an important source of nutrition in people's lives. In recent years, with the increasing demand for mutton, the quality of mutton is also improving. Mutton freshness is a comprehensive evaluation of the flavor, color, taste and other health standards of mutton. Mutton freshness can be divided into three grades: fresh grade, sub-fresh grade and corruption grade. Through grading, the reliability of the nutritional and safety of mutton can be further comprehensively reflected. Traditional mutton freshness detection methods include sensory evaluation and physicochemical experimental analysis. Although sensory evaluation is simple and intuitive, subjective influence makes it difficult to judge the initial deterioration of mutton. For physicochemical experimental analysis, although it has the characteristics of high accuracy and good reliability, the method is destructive.

[0003] At present, spectral technology has been successfully applied to mutton quality evaluation. Researchers have also invented a multi-index rapid non-destructive detection system for meat products, which discloses a method for determining at least three prediction indexes according to a multi-index prediction model. The prediction indexes can be meat color, tenderness, water holding capacity, pH value, protein content and fat content. However, the above prior art has the following problems: after completing the multi-dimensional index detection, it is unable to comprehensively judge the mutton grade according to the multi-dimensional index and separate mutton with different freshness. SUMMARY

[0004] Therefore, the present application discloses a multi-dimensional index multi-information fusion intelligent detection and grading system for mutton freshness.

[0005] A multi-dimensional index multi-information fusion intelligent detection and grading system for mutton freshness, comprising: a mutton conveying and grading device, a visible near-infrared spectrometer and a data processing device.

[0006] The visible near-infrared spectrometer is electrically connected with the mutton conveying and grading device, the visible near-infrared spectrometer is arranged on the mutton conveying and grading device, and is used for acquiring original spectrum data of the mutton to be measured conveyed by the mutton conveying and grading device; the data processing device receives the original spectrum data of the mutton to be measured acquired by the visible near-infrared spectrometer, and pre-processes the original spectrum data; the data processing device is also used for extracting characteristic wavelengths from the pre-processed spectrum data; the data processing device is also used for predicting the freshness index of the mutton to be measured according to a pre-stored freshness index prediction model of mutton, so as to obtain a predicted value of the current freshness index of mutton; and the data processing device is also used for judging the freshness grade of the mutton to be measured according to the predicted value of the current freshness index of mutton and pre-stored reference data of the freshness grade of mutton, and controlling the mutton conveying and grading device to convey the mutton to be measured to a corresponding conveying channel according to the judged freshness grade of the mutton to be measured.

[0007] In the above-mentioned intelligent detection and grading system of the multi-dimensional index and multi-information fusion of mutton freshness, the visible near-infrared spectrometer is arranged on the mutton conveying and grading device, and is used for acquiring original spectrum data of the mutton to be measured conveyed by the mutton conveying and grading device; the data processing device receives the original spectrum data of the mutton to be measured acquired by the visible near-infrared spectrometer, and pre-processes the original spectrum data; the data processing device is also used for extracting characteristic wavelengths from the pre-processed spectrum data; the data processing device is also used for predicting the freshness index of the mutton to be measured according to a pre-stored freshness index prediction model of mutton, so as to obtain a predicted value of the current freshness index of mutton; the data processing device is also used for judging the freshness grade of the mutton to be measured according to the predicted value of the current freshness index of mutton and pre-stored reference data of the freshness grade of mutton, and controlling the mutton conveying and grading device to convey the mutton to be measured to a corresponding conveying channel according to the judged freshness grade of the mutton to be measured, thereby realizing comprehensive evaluation and comprehensive grading of the multi-dimensional freshness index of mutton quality. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 It is a functional module schematic diagram of the intelligent detection and grading system of the multi-dimensional index and multi-information fusion of mutton freshness.

[0009] Figure 2 It is a modeling flowchart of the freshness index prediction model of the intelligent detection and grading system of the multi-dimensional index and multi-information fusion of mutton freshness.

[0010] Figure 3 It is a structure schematic diagram of the mutton conveying and grading device. Figure 1 It is a structure schematic diagram of the mutton conveying and grading device.

[0011] Figure 4 It is a structure schematic diagram of the mutton conveying and grading device. Figure 1 It is a structure schematic diagram of the mutton conveying and grading device.

[0012] Figure 5The logic flow chart of the freshness grade determination unit.

[0013] In the figure: the intelligent detection and grading system of mutton freshness multi-dimensional index multi-information fusion 10, the mutton conveying and grading device 20, the support 21, the detection channel 22, the fresh grade mutton conveying channel 23, the sub-fresh grade mutton conveying channel 24, the corruption grade mutton conveying channel 25, the first grading baffle 26, the second grading baffle 27, the third grading baffle 28, the visible near-infrared spectrometer 30, the data processing device 40, the spectral data preprocessing unit 41, the characteristic wavelength extraction unit 42, the freshness index prediction unit 43, the freshness grade determination unit 44, the input and output unit 45, and the prediction model storage unit 46. DETAILED DESCRIPTION

[0014] To make the objectives, technical solutions, and advantages of the present application clearer, the technical solutions of the present application will be described below in connection with specific embodiments of the present application and corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0015] As shown in Figure 1 The present application provides an intelligent detection and grading system of mutton freshness multi-dimensional index multi-information fusion 10, which comprises a mutton conveying and grading device 20, a visible near-infrared spectrometer 30, and a data processing device 40. The mutton conveying and grading device 20, the visible near-infrared spectrometer 30, and the data processing device 40 are electrically connected.

[0016] The visible near-infrared spectrometer 30 is arranged on the mutton conveying and grading device 20. The visible near-infrared spectrometer 30 is used to acquire the original spectral data of the mutton to be measured conveyed by the mutton conveying and grading device 20. The collection wavelength band of the visible near-infrared spectrometer is 400-2500 nm. The visible near-infrared spectrometer 30 mainly comprises a light source, a sample chamber, a grating, a photoelectric detector, an amplifier, an analog-to-digital converter, and the like. The light source emits light, which is irradiated on the mutton to be measured through the sample chamber. The sample reflects or transmits part of the light, which is decomposed by the grating. The photoelectric detector receives the light signal and converts it into an electric signal. The amplifier amplifies the electric signal. The analog-to-digital converter converts the signal into a digital signal and transmits it to the data processing device 40.

[0017] The data processing device 40 receives the raw spectral data of the mutton to be measured acquired by the visible near-infrared spectrometer 30, and performs spectral pretreatment on the raw spectral data to obtain pretreated spectral data, wherein the purpose of spectral pretreatment is to improve the signal-to-noise ratio of the spectral data and reduce the variability between spectra, so as to more accurately interpret the spectral data and extract effective information therefrom, for example, spectral pretreatment includes signal denoising, background correction, spectral alignment, sample drift correction, etc. The data processing device 40 is also used to extract characteristic wavelengths from the pretreated spectral data, wherein the characteristic wavelength refers to a wavelength within a certain wavelength range at which the absorption or reflection of light by the mutton sample changes significantly. In spectral analysis, characteristic wavelengths are usually related to the physical and chemical properties of the mutton sample. The biochemical components in fresh mutton are different from those in aged mutton, and the absorption or reflection of these components at different wavelengths is also different. By analyzing the spectral data at different wavelengths, the characteristic wavelengths related to the freshness indicator of mutton can be found, and the relationship between the characteristic wavelengths and the freshness can be established. In this way, the freshness of mutton can be quickly and accurately determined by detecting the spectral data at the characteristic wavelengths. The data processing device 40 predicts the freshness indicator of the mutton to be measured according to a pre-stored mutton freshness indicator prediction model to obtain a predicted value of the current mutton freshness indicator, wherein the mutton freshness indicator prediction model is composed of spectral data corresponding to characteristic wavelengths related to the physical and chemical properties of the sample mutton. The data processing device 40 is also used to determine the freshness grade of the mutton to be measured according to the predicted value of the current mutton freshness indicator and pre-stored mutton freshness grade reference data, and to control the mutton conveying and grading device 20 to convey the mutton to be measured to the corresponding conveying channel according to the determined freshness grade of the mutton to be measured. The current mutton freshness indicator includes the total number of colonies TVC value, volatile base nitrogen TVB-N and meat color L* and a* value of the mutton. The freshness grade of the mutton is determined according to the TVC value, TVB-N content and meat color L* and a* value of the mutton. The unit of TVC is CFU / g, and the unit of TVB-N is mg / (100g). The specific grading threshold may vary due to factors such as production environment and processing method, so the grading threshold can be adjusted according to actual conditions in specific applications. The freshness grade is divided into fresh grade, sub-fresh grade and spoilage grade.

[0018] The data processing device 40 is a single-chip microcomputer or microcomputer running a computer application program for completing the multi-dimensional freshness index comprehensive evaluation and comprehensive grading of mutton quality. After running the computer application program, the data processing device 40 generates the following functional units: a spectral data preprocessing unit 41, a characteristic wavelength extraction unit 42, a freshness index prediction unit 43, a freshness grade determination unit 44, and an input / output unit 45. The spectral data preprocessing unit 41 is used to receive the raw spectral data of the mutton to be tested obtained by the visible near-infrared spectrometer and preprocess the raw spectral data. For example, the spectral data collected by the visible near-infrared spectrometer is read using PySpectra, and the standard normal variable (SNV) preprocessing algorithm is selected. In Python, the SNV transformation is implemented using the pandas library. The characteristic wavelength extraction unit 42 is used to extract characteristic wavelengths from the preprocessed spectral data. For example, in Python, the LogisticRegressionCV class in the sklearn library is used to implement the competitive adaptive reweighted sampling (CARS) algorithm to extract characteristic wavelengths. The freshness index prediction unit 43 predicts the freshness index of the mutton to be tested according to the pre-stored mutton freshness index prediction model to obtain the predicted value of the current mutton freshness index. The freshness grade determination unit 44 is used to determine the freshness grade of the mutton to be tested according to the predicted value of the current mutton freshness index and the pre-stored mutton freshness grade reference data, and controls the mutton conveying and grading device 20 to convey the mutton to be tested to the corresponding conveying channel according to the determined freshness grade of the mutton to be tested. The input / output unit 45 is used to output the determined current mutton freshness index and freshness grade to the display screen for display. The input / output unit 45 is also used to respond to the modification operation of the operator on the keyboard or touch screen connected to the data processing device 40 to modify the threshold value of the freshness grade.

[0019] Further, the data processing device 40 also generates a prediction model storage unit 46 for storing the mutton freshness index prediction model, for example, using the pickle module in Python to save the mutton freshness index prediction model in the form of a binary file to the disk, in order to improve the practicability of the system, the mutton freshness index prediction model can be updated according to the needs in the later use, wherein the mutton freshness index prediction model involves the indexes including the total number of colonies TVC, volatile base nitrogen TVB-N and meat color L* and a* values. Please refer to Table 1, wherein the pre-stored mutton freshness grade reference data includes grade data and the TVC value, TVB-N content and meat color L* and a* value of mutton corresponding to the grade data, the grade data includes fresh level, sub-fresh level and corruption level.

[0020] Table 1 Mutton freshness grade reference data

[0021]

[0022] Please also refer to Figure 2The freshness index prediction model of mutton in the embodiment is established according to the following method: first, the spectral data of sample mutton is collected for subsequent analysis and modeling; further, the spectral data is preprocessed, the collected spectral data is preprocessed, the algorithm is consistent with the preprocessing algorithm in the preprocessing module, so as to subsequent data processing and analysis; further, feature extraction, CARS algorithm is used to extract useful features from spectral data, the algorithm is consistent with the feature wavelength extraction algorithm in the feature wavelength extraction module, which is used for subsequent model construction and prediction; further, the model is established, the partial least squares regression (PLSR) modeling algorithm is selected, and the spectral model well fitted with the sample data is established; further, model verification and optimization, the prediction performance of the established model is verified using the test data set, and optimization is performed to improve the prediction accuracy and stability of the model; wherein, in the establishment of the freshness index prediction model, the corresponding chemical values need to be determined, including: the total number of colonies TVC, volatile base nitrogen TVB-N and meat color L* and a* value of mutton; wherein, the total number of colonies TVC determination method is: the specific method refers to GB 4789.2-2022 “National food safety standard food microbiological examination determination of total number of colonies”, and is used as the reference value of quantitative analysis, in order to ensure accuracy, at least 6 times determination is made for each sample, and the average value is taken as the TVC value of the sample; wherein, the determination method of TVB-N is: the specific method refers to GB 5009.228-2016 “Determination of volatile base nitrogen in food”, and is used as the reference value of quantitative analysis, in order to ensure accuracy, at least 6 times determination is made for each sample, and the average value is taken as the TVB-N content of the sample; wherein, the determination method of meat color L* and a* value is: using colorimeter to determine, and is used as the reference value of quantitative analysis, in order to ensure accuracy, at least 6 times determination is made for each sample, and the average value is taken as the L* and a* value of the sample.

[0023] In the embodiment, please refer to Figure 5, the freshness grade judging unit 44 receives the current mutton freshness index, and performs the following steps: first, initialize the defined variables, X, F are natural numbers, let X=F=0; further, according to the TVC value to determine whether the mutton is fresh level, when the TVC value is less than or equal to 5*10^6, X=X+1; further, according to the TVB-N content to determine whether the mutton is fresh level, when the TVB-N content is less than 15, X=X+1; further, according to the L* value to determine whether the mutton is fresh level, when the L* value is greater than 40.0, X=X+1; further, according to the a* value to determine whether the mutton is fresh level, when the a* value is greater than 19.3, X=X+1; further, according to the TVC value to determine whether the mutton is putrid level, when the TVC value is greater than 7*10^6, F=F+1; according to the TVB-N content to determine whether the mutton is putrid level, when the TVB-N content is greater than 25, F=F+1; further, according to the L* value to determine whether the mutton is putrid level, when the L* value is less than 38.9, F=F+1; further, according to the a* value to determine whether the mutton is putrid level, when the a* value is less than 18.9, F=F+1; further, judge the fresh level variable X size, when the fresh level variable is greater than or equal to 3, then the measured mutton is fresh level; further, judge the putrid level variable F size, when the putrid level variable is greater than 1, the measured mutton is putrid level; further, determine the fresh level variable is less than 3, the putrid level variable is not greater than 1, the measured mutton is sub fresh level.

[0024] Further, please see Figure 3 and Figure 4The mutton conveying and grading device 20 comprises a support 21, a detection channel 22, a fresh grade mutton conveying channel 23, a sub-fresh grade mutton conveying channel 24, a decay grade mutton conveying channel 25, a first grading baffle 26, a second grading baffle 27, and a third grading baffle 28. The detection channel 22, the fresh grade mutton conveying channel 23, the sub-fresh grade mutton conveying channel 24, the decay grade mutton conveying channel 25, the first grading baffle 26, the second grading baffle 27, and the third grading baffle 28 are arranged on the support 21. The near-infrared spectrometer 30 is arranged on the detection channel 22. The fresh grade mutton conveying channel 23, the sub-fresh grade mutton conveying channel 24, and the decay grade mutton conveying channel 25 are arranged side by side. The fresh grade mutton conveying channel 23 is connected with the detection channel 22 to receive the mutton to be detected output from the detection channel 22. The first grading baffle 26, the second grading baffle 27, and the third grading baffle 28 are arranged at intervals on one side of the fresh grade mutton conveying channel 23, the sub-fresh grade mutton conveying channel 24, and the decay grade mutton conveying channel 25. The first grading baffle 26, the second grading baffle 27, and the third grading baffle 28 are connected with a motor. The freshness grade determination unit 44 controls the motor connected with the first grading baffle 26, the second grading baffle 27, and the third grading baffle 28 to work according to the determined freshness grade of the mutton to be detected, so as to convey the fresh grade mutton to the fresh grade mutton conveying channel 23, the sub-fresh grade mutton to the sub-fresh grade mutton conveying channel 24, and the decay grade mutton to the decay grade mutton conveying channel 25.

Claims

1. A smart detection and grading system for mutton freshness that integrates multiple dimensions and information, characterized in that, The system includes a mutton conveying and grading device, a visible-near-infrared spectrometer, and a data processing device. The mutton conveying and grading device, the visible-near-infrared spectrometer, and the data processing device are electrically connected. The visible-near-infrared spectrometer is mounted on the mutton conveying and grading device to acquire the raw spectral data of the mutton to be tested conveyed by the device. The data processing device receives the raw spectral data of the mutton to be tested acquired by the visible-near-infrared spectrometer and preprocesses the raw spectral data. The data processing device is also used to extract characteristic wavelengths from the preprocessed spectral data. Furthermore, the data processing device uses pre-stored... The mutton freshness index prediction model predicts the freshness index of the mutton to be tested, so as to obtain the predicted value of the current mutton freshness index. The current mutton freshness index includes the total bacterial count (TVC) value, volatile basic nitrogen (TVB-N) content, and meat color (L* and a* values). The data processing device is also used to determine the freshness grade of the mutton to be tested based on the predicted value of the current mutton freshness index and the pre-stored mutton freshness grade benchmark data, and to control the mutton conveying and grading device to transport the mutton to be tested to the corresponding conveying channel according to the determined freshness grade of the mutton to be tested. The pre-stored benchmark data for mutton freshness grades includes grade data and the corresponding total bacterial count (TVC), volatile basic nitrogen (TVB-N) content, and meat color (L* and a* values). The grade data includes fresh, semi-fresh, and spoiled grades. The data processing device includes a spectral data preprocessing unit, a characteristic wavelength extraction unit, a freshness index prediction unit, and a freshness grade determination unit. The spectral data preprocessing unit receives the raw spectral data of the mutton to be tested from the visible and near-infrared spectrometer and preprocesses the raw spectral data. The feature wavelength extraction unit is used to extract feature wavelengths from the preprocessed spectral data; the freshness index prediction unit predicts the freshness index of the mutton to be tested according to the pre-stored mutton freshness index prediction model to obtain the predicted value of the current mutton freshness index. The indicators involved in the mutton freshness index prediction model include total bacterial count (TVC), volatile basic nitrogen (TVB-N) content, and meat color (L* and a* values); the freshness grade determination unit is used to determine the freshness grade of the mutton to be tested according to the predicted value of the current mutton freshness index and the pre-stored mutton freshness grade benchmark data, and controls the mutton conveying and grading device to convey the mutton to be tested to the corresponding conveying channel according to the determined freshness grade of the mutton to be tested. The freshness grade determination unit receives the current freshness index of the mutton and performs the following steps: First, initialize the defined variables, X and F are natural numbers, and set X=F=0; Next, determine whether the mutton is fresh based on the total bacterial count (TVC) value. When the TVC value is less than or equal to 5*10^6, X=X+1; Next, determine whether the mutton is fresh based on the volatile basic nitrogen (TVB-N) content. When the TVB-N content is less than 15, X=X+1; Next, determine whether the mutton is fresh based on the L* value. When the L* value is greater than 40.0, X=X+1; Next, determine whether the mutton is fresh based on the a* value. When the a* value is greater than 19.3, X=X+1; Finally, determine whether the mutton is spoiled based on the total bacterial count (TVC) value. The classification process is as follows: First, if the total bacterial count (TVC) is greater than 7 * 10^6, F = F + 1. Second, if the volatile basic nitrogen (TVB-N) content is greater than 25, F = F + 1. Third, if the L* value is less than 38.9, F = F + 1. Fourth, if the a* value is less than 18.9, F = F + 1. Fifth, if the freshness grade variable X is greater than or equal to 3, the tested mutton is fresh. Sixth, if the spoilage grade variable F is greater than 1, the tested mutton is spoiled. Finally, if the freshness grade variable is less than 3 and the spoilage grade variable is not greater than 1, the tested mutton is considered slightly fresh.

2. The intelligent detection and grading system for mutton freshness based on multi-dimensional indicators and multi-information fusion as described in claim 1, characterized in that: The data processing device also includes a prediction model storage unit, which is used to store prediction models for mutton freshness indicators.

3. The intelligent detection and grading system for mutton freshness based on multi-dimensional indicators and multi-information fusion as described in claim 2, characterized in that: The mutton freshness index prediction model was established as follows: First, sample spectral data was collected for subsequent analysis and modeling. Next, spectral data preprocessing was performed, using the same algorithm as the preprocessing module, to facilitate subsequent data processing and analysis. Then, feature extraction was performed, using the CARS algorithm to extract useful features from the spectral data, consistent with the feature wavelength extraction algorithm in the feature wavelength extraction module, for subsequent model building and prediction. Next, model building was conducted, selecting partial least squares regression to establish a spectral model that fits well with the sample data. Finally, model validation and optimization were performed, using a test dataset to validate the predictive performance of the established model and optimizing it to improve its prediction accuracy and stability. In establishing the freshness index prediction model, relevant chemical values ​​needed to be measured, including: total bacterial count (TVC) value, volatile basic nitrogen (TVB-N) content, and meat color L* and a* values. The method for determining the total bacterial count (TVC) value is as follows: Refer to GB [GB standard]. The determination of total bacterial count (TVC) in food is based on GB 5009.228-2016, "Determination of Volatile Basic Nitrogen in Food," and is used as a reference value for quantitative analysis. To ensure accuracy, each sample should be measured at least 6 times, and the average value should be taken as the TVC value of the sample. The determination of volatile basic nitrogen (TVB-N) content is based on GB 5009.228-2016, "Determination of Volatile Basic Nitrogen in Food," and is used as a reference value for quantitative analysis. To ensure accuracy, each sample should be measured at least 6 times, and the average value should be taken as the TVB-N content of the sample. The determination of meat color L* and a* values ​​is based on colorimetry and is used as a reference value for quantitative analysis. To ensure accuracy, each sample should be measured at least 6 times, and the average value should be taken as the meat color L* and a* values ​​of the sample.

4. The intelligent detection and grading system for mutton freshness based on multi-dimensional index and multi-information fusion as described in claim 1, characterized in that: The mutton conveying and grading device includes a support frame, a detection channel, a fresh-grade mutton conveying channel, a slightly fresh-grade mutton conveying channel, a spoiled-grade mutton conveying channel, a first-grade baffle, a second-grade baffle, and a third-grade baffle. The detection channel, the fresh-grade mutton conveying channel, the slightly fresh-grade mutton conveying channel, the spoiled-grade mutton conveying channel, the first-grade baffle, the second-grade baffle, and the third-grade baffle are mounted on the support frame. The visible and near-infrared spectrometer is installed in the detection channel. The fresh-grade mutton conveying channel, the slightly fresh-grade mutton conveying channel, and the spoiled-grade mutton conveying channel are arranged side-by-side, and the fresh-grade mutton conveying channel is connected to the detection channel to receive signals from the detection channel. The mutton to be tested is arranged with three grading baffles (number one, number two, and number three) spaced apart on one side of the fresh mutton conveyor channel, the sub-fresh mutton conveyor channel, and the spoiled mutton conveyor channel. Each of the three grading baffles is connected to a motor. The freshness grade determination unit controls the corresponding motor connected to the three grading baffles to work according to the determined freshness grade of the mutton to be tested, so as to send the fresh mutton to the fresh mutton conveyor channel, push the sub-fresh mutton to the sub-fresh mutton conveyor channel, and push the spoiled mutton to the spoiled mutton conveyor channel.

Images