A method for detecting heterogeneous chicken breast quality based on multi-information fusion

Abstract

The present invention discloses a heterogeneous chicken breast quality detection method based on multi-information fusion, and relates to the field of heterogeneous chicken breast quality detection methods. The present invention discloses a heterogeneous chicken breast quality detection method based on multi-information fusion, which is mainly composed of a detection method and technology, a feature information extraction module and a chicken breast quality grading module; the detection method and technology are composed of three-dimensional synchronous fluorescence technology, Fourier infrared technology and nuclear magnetic resonance technology; the feature parameter extraction module extracts 8 feature parameters of chicken breast, and uses a BP neural network model for modeling and prediction. The chicken breast quality grading module outputs the quality of chicken breast. The present invention determines the quality of chicken breast by collecting spectral information of the top area of ​​chicken breast, thereby realizing non-destructive and rapid detection of chicken breast quality.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural product processing and detection, and in particular to a method for detecting the quality of heterogeneous chicken breast based on three-dimensional fluorescence spectroscopy, near-infrared spectroscopy and nuclear magnetic resonance technology, and relates to the determination of normal chicken breast, cooked chicken breast, noodle chicken breast, white-striped chicken breast and comprehensive heterogeneous chicken breast, and the grading of comprehensive heterogeneous chicken breast. Background Art

[0002] In order to meet the growing demand for chicken and optimize broiler production, breeders use genetics to improve the growth rate of broilers and thus increase meat production. However, artificial selection of genes has also promoted the production of heterogeneous chicken breasts, such as woody breast (WB), spaghetti meat (SM), and whitestriping (WS). Compared with normal chicken breast, woody breast has a yellowish surface color, muscle degeneration, and obvious hardness can usually be detected by tactile evaluation. In addition, the differences in histology and muscle composition of woody breast lead to meat defects and impaired functional properties, such as poor water retention and low sensory attribute acceptance scores. Spaghetti breast was first proposed in 2015. Its typical feature is the detachment of the fiber bundles that make up the pectoral muscle, which manifests as soft and mushy pectoral muscles or split into bundles, similar to noodles. Whitestriping breast is mainly characterized by the appearance of white stripes of varying numbers parallel to the muscle fibers in the pectoral or leg muscles. It is reported that chicken breast meat may have two or more quality problems, which is called comprehensive heterogeneous meat. The appearance of heterogeneous chicken breast meat seriously affects the appearance, nutritional quality and processing characteristics of slaughtered chicken and cut products, causing serious economic losses to the broiler industry.

[0003] At present, the detection and grading methods of heterogeneous chicken breast mostly rely on the artificial visual-tactile joint classification detection method, but the artificial visual-tactile joint classification detection method requires a high level of experience in the inspectors and is not suitable for popularization. In addition, the existing detection methods can only detect a specific quality defect problem and cannot detect multiple types of quality defects at the same time. For comprehensive heterogeneous chicken breast, no corresponding grading method has been established. Therefore, it is necessary to conduct in-depth research on the detection method of heterogeneous chicken breast quality and the grading method of comprehensive heterogeneous meat, in order to develop an objective, accurate and efficient heterogeneous chicken breast quality detection method to achieve rapid detection of heterogeneous chicken breast quality defects. Summary of the invention

[0004] The purpose of the present invention is to design a heterogeneous chicken breast quality detection method based on three-dimensional fluorescence spectroscopy, near-infrared spectroscopy and nuclear magnetic resonance technology to determine the specific quality defects in heterogeneous chicken breast, and divide the comprehensive heterogeneous meat into primary, secondary and tertiary comprehensive heterogeneous meat, so as to meet the use requirements of the poultry processing production line that needs to estimate the quality category of chicken breast.

[0005] To solve the above problems, the present invention provides the following technical solutions:

[0006] A heterogeneous chicken breast quality detection method based on multi-information fusion adopts three-dimensional synchronous fluorescence technology, Fourier transform infrared technology and nuclear magnetic resonance technology to jointly detect chicken breast, and obtains three-dimensional synchronous fluorescence spectrum, Fourier transform infrared original spectrum, nuclear magnetic resonance multi-component relaxation characteristic spectrum and nuclear magnetic resonance image respectively, and then uses characteristic parameter extraction module to extract characteristic parameters from the above spectrum, uses characteristic parameters to construct chicken breast quality classification database, establishes chicken breast quality grading model according to the chicken breast quality classification database, and uses the chicken breast quality grading model to automatically grade chicken breast.

[0007] Preferably, after the three-dimensional synchronous fluorescence spectrum is exported, its characteristic parameters are extracted: the excitation wavelength F1 is the X-axis value of the three-dimensional synchronous fluorescence spectrum, the wavelength difference F2 is the Y-axis value of the three-dimensional synchronous fluorescence spectrum, and the fluorescence intensity F3 is the Z-axis value of the three-dimensional synchronous fluorescence spectrum;

[0008] After the original Fourier infrared spectrum is exported, the characteristic parameters are extracted, the X-axis value is the wave number F4, and the Y-axis value is the absorbance F5;

[0009] The free-induced exponential decay curve obtained by nuclear magnetic resonance analysis was inverted to obtain a multi-component relaxation characteristic spectrum. The data of relaxation time 0-10s was taken as the research object, and the peak area of ​​the peak between t1 and t1+0.05s was calculated to finally obtain the relaxation time F6, that is, (2t1+0.05) / 2, and the peak area F7; the nuclear magnetic resonance imaging image was analyzed using the image HSV color mode to obtain the image brightness F8.

[0010] Preferably, the method comprises the following specific steps:

[0011] Step 1, professionals who are engaged in the field of chicken processing and have experience in evaluating heterogeneous chicken breast meat, according to the quality of the chicken breast meat, determine it as normal chicken breast meat, firewood chicken breast meat, noodle chicken breast meat, white striped chicken breast meat, comprehensive heterogeneous level 1, comprehensive heterogeneous level 2 and comprehensive heterogeneous level 3, and record them with values ​​0, 1, 2, 3, 4, 5 and 6 in sequence;

[0012] Step 2, chicken breast detection method and technology; take a sample of appropriate size at the top of the chicken breast perpendicular to the muscle lines, analyze it using a molecular fluorescence spectrometer, and collect characteristic parameters F1, F2 and F3 of the chicken breast sample; then, cut the chicken breast in this area and perform nuclear magnetic resonance analysis to collect characteristic parameters F6, F7 and F8; then place it in a vacuum freeze dryer for freeze drying for 24 hours, analyze it using a Fourier transform infrared spectrometer, and collect characteristic parameters F4 and F5;

[0013] Step 3, establishing a chicken breast quality classification database; testing the chicken breasts that have been graded by professionals in step 1, extracting characteristic parameters from the original spectral data, the method is the same as step 2; fusing the collected characteristic parameters F1-F8 to construct a chicken breast quality classification database; the data output by three-dimensional synchronous fluorescence, Fourier transform infrared and nuclear magnetic resonance technology are all collected from the same area of ​​the chicken breast sample;

[0014] Step 4, establishing a chicken breast quality grading model; using each record information in the chicken breast quality classification database as a training set, using 8 feature parameters F1-F8 of each record information as input, using the chicken breast quality as output, and establishing a chicken breast quality BP neural network grading model according to the BP neural network modeling method;

[0015] Step 5, grading the quality problems of chicken breast meat; placing the chicken breast meat that has not been graded by professionals in a three-dimensional fluorescence spectrometer, a Fourier transform infrared spectrometer and a nuclear magnetic resonance spectrometer for analysis, extracting the characteristic parameters in the original data and sending them to the characteristic information extraction module, and determining the chicken breast meat quality problems and the comprehensive heterogeneous meat grade according to the chicken breast meat quality grading model established in step 4.

[0016] Preferably, the excitation wavelength F1 is in the range of 190-400 nm, the wavelength difference F2 is in the range of 25-160 nm, and the wave number F4 is in the range of 400-4000 cm -1 , the relaxation time F6 ranges from 0 to 10 s.

[0017] Preferably, the chicken breast quality grading model in step 4 is a single hidden layer BP neural network model, whose input variables are F1~F8, the number of output variables is 1, that is, chicken breast quality, the number of hidden layers is 1, the number of hidden layer neurons is 10, and the activation functions of the hidden layer and the output layer both use the sigmoid function.

[0018] Preferably, the chicken breast quality grading rule in step 4 is:

[0019] The BP neural network model outputs a chicken breast quality value less than 1.6, and the chicken breast is normal;

[0020] The BP neural network model outputs a chicken breast quality value between 1.6 and 2.6 (including 2.6), and the chicken breast is firewood-cooked chicken breast;

[0021] The BP neural network model outputs a chicken breast quality value between 2.6 and 3.6 (including 2.6), and the chicken breast is noodle chicken breast;

[0022] The BP neural network model outputs a chicken breast quality value between 3.6 and 4.6 (including 4.6), and the chicken breast is white-striped chicken breast;

[0023] The BP neural network model outputs a chicken breast quality value between 4.6 and 5.6 (including 5.6), and the chicken breast is a comprehensive heterogeneous first-class chicken breast;

[0024] The BP neural network model outputs a chicken breast quality value between 5.6 and 7.6 (including 7.6), and the chicken breast is a comprehensive heterogeneous secondary chicken breast;

[0025] The BP neural network model outputs a chicken breast quality value between 7.6 and 10 (including 10), and the chicken breast is a comprehensive heterogeneous third-grade chicken breast.

[0026] The advantages of the present invention are:

[0027] The present invention realizes the automation of chicken breast meat quality category and grade in poultry meat processing, reduces the requirements for the specialization of inspectors, and reduces the costs of manual employment and training. At the same time, it solves the simultaneous detection of multiple quality defect types, standardizes the evaluation criteria for poultry meat quality category and grade, improves the objectivity and accuracy of chicken breast meat quality category and grade detection, and realizes high efficiency of detection in the poultry meat processing process.

[0028] The three detection methods used in the present invention, three-dimensional synchronous fluorescence technology, Fourier infrared technology and nuclear magnetic resonance technology, are the best combination obtained after screening and combining a large number of technologies, and all have the characteristics of being fast and non-destructive. The combination of the three technologies can accurately obtain the spectral information of the chicken breast sample to obtain a large number of characteristic parameters.

[0029] The area measured by the three detection methods of the present invention is located in the top area of ​​the chicken breast, which is representative of the quality defect problem of heterogeneous chicken breast, and can collect characteristic parameters that can distinguish the quality category and grade of chicken breast. In addition, the areas measured by the three detection methods are the same, which ensures the rationality of the data and improves the accuracy and correctness of the classification and grading of the quality of chicken breast.

[0030] The chicken breast quality grading model established in the present invention can directly input the obtained data into the feature information extraction module when the measurement areas of the three detection methods are consistent, and can be applied to chicken breast quality grading detection of different scales without any modification.

[0031] The prediction model of the present invention can be adjusted according to different application scenarios. When it is applied to poultry meat production and processing in the future, the collected original spectra can be preprocessed, such as standard normal transformation, multivariate scattering correction, standard normal transformation, etc., or the characteristic parameters can be reduced in dimension to meet different needs in different scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The composition diagram of the system of the present invention;

[0033] Figure 2 System flow chart of the present invention; DETAILED DESCRIPTION

[0034] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0035] Figure 1 The system composition diagram of the present invention includes detection methods and technologies, feature information extraction module and chicken breast quality grading module. The detection methods and technologies include detection technologies composed of three-dimensional synchronous fluorescence technology, Fourier infrared technology and nuclear magnetic resonance technology, as well as the acquired data information.

[0036] The instrument parameters of the three-dimensional synchronous fluorescence spectroscopy technology are: the scanning mode is set to synchronous mode; the data mode is fluorescence; the starting wavelength of the excitation wavelength is 200nm, the ending wavelength is 750nm, the wavelength increment is 5nm, and the excitation slit is 5nm; the emission wavelength and the excitation wavelength are separated by 25 to 160nm, and the synchronous scanning is performed every 3nm; the scanning speed is medium. After the three-dimensional synchronous fluorescence spectrum data is exported, its characteristic parameters are extracted: the excitation wavelength F1 is the X-axis value of the three-dimensional synchronous fluorescence spectrum, the wavelength difference F2 is the Y-axis value of the three-dimensional synchronous fluorescence spectrum, and the fluorescence intensity F3 is the Z-axis value of the three-dimensional synchronous fluorescence spectrum.

[0037] The instrument parameters for Fourier infrared analysis are: measurement range 400~4000cm -1 , scan signal accumulation 16 times, resolution 4cm -1 , the interference of carbon dioxide and water vapor is deducted in real time during the scanning process. After the original spectrum is exported, the characteristic parameters are extracted, the X-axis value is the wave number F4, and the Y-axis value is the absorbance F5.

[0038] The instrument parameters of the NMR analysis are: the sequence is set to Q-CPMG, the spectrum width SW is 100, the center frequency is 416796.321Hz, the test temperature is set to 32℃, the pulse width p1 is 11μs, the delay time d1 is 1s, and the number of scans ns is 8. The free induced exponential decay curve is inverted to obtain a multi-component relaxation characteristic spectrum. The data of relaxation time 0-10s is used as the research object. The peak area calculation of the peak between t1 and t1+0.05s finally obtains the relaxation time F6, that is, (2t1+0.05) / 2, and the peak area F7. The image HSV color mode is used to analyze the NMR imaging image to obtain the image brightness F8.

[0039] The feature information extraction module can be used to extract 8 characteristic variables of chicken breast, including excitation wavelength F1, wavelength difference F2, fluorescence intensity F3, wave number F4, absorbance F5, relaxation time F6, peak area F7, and image brightness F8, and the BP neural network model is used for modeling and prediction.

[0040] The characteristic parameters obtained from the detection technology and method module are modeled and predicted using a single hidden layer BP neural network model. The number of input variables is 8 (excitation wavelength F1, wavelength difference F2, fluorescence intensity F3, wave number F4, absorbance F5, relaxation time F6, peak area F7, image brightness F8), the number of output variables is 1 (chicken breast quality), the number of hidden layers is 1, the number of hidden layer neurons is 10, and the activation functions of the hidden layer and the output layer are all sigmoid functions. The excitation wavelength F1 ranges from 190 to 400 nm, the wavelength difference F2 ranges from 25 to 160 nm, and the wave number F4 ranges from 400 to 4000 cm -1 , the relaxation time F6 ranges from 0 to 10s.

[0041] The chicken breast quality grading module is used to classify chicken breast into normal chicken breast, cooked chicken breast, noodle chicken breast, white striped chicken breast, comprehensive heterogeneous grade one, comprehensive heterogeneous grade two and comprehensive heterogeneous grade three chicken breast.

[0042] Figure 2This is a flow chart of the system of the present invention, and the process includes manual grading, calibration of detection area, detection technology determination, acquisition of characteristic parameters, establishment of chicken breast quality classification database, establishment of chicken breast quality classification model and output of chicken breast quality. Among them, after manual grading, the chicken breast is divided into normal chicken breast, firewood chicken breast, noodle chicken breast, white striped chicken breast, comprehensive heterogeneous first level, comprehensive heterogeneous second level and comprehensive heterogeneous third level chicken breast. The detection area refers to the area of ​​10mm×5mm×8mm located at the top of the chicken breast in the direction of the vertical muscle texture. The detection technology consists of three-dimensional synchronous fluorescence technology, Fourier infrared technology and nuclear magnetic resonance technology. After the characteristic information of the obtained data is extracted, a chicken breast quality classification database and classification model are established, and the chicken breast quality is finally output.

[0043] A method for detecting the quality of heterogeneous chicken breast meat using multi-information fusion comprises the following steps:

[0044] (1) Manual grading of chicken breast quality: Professionals engaged in the chicken processing field and with experience in evaluating heterogeneous chicken breasts will classify the chicken breasts into normal chicken breasts, cooked chicken breasts, noodle chicken breasts, white striped chicken breasts, comprehensive heterogeneous grade 1, comprehensive heterogeneous grade 2, and comprehensive heterogeneous grade 3 based on their quality, and record them with values ​​of 0, 1, 2, 3, 4, 5, and 6.

[0045] (2) Calibration of the test area: Take a sample of 10 mm × 5 mm × 8 mm in size from the top of the chicken breast perpendicular to the muscle lines.

[0046] (3) Detection methods and technologies for chicken breast meat; G8800A molecular fluorescence spectrometer was used for analysis, and a three-dimensional synchronous fluorescence spectrum was derived; low-field nuclear magnetic resonance analysis was then performed, and the free-induced exponential decay curve was inverted to obtain a multi-component relaxation characteristic spectrum. After the sample was freeze-dried in a vacuum freeze dryer for 24 hours, it was analyzed using an IRAffinity-IS Fourier transform infrared spectrometer to obtain the original spectrum and collect characteristic parameters F4 and F5. Based on the three-dimensional synchronous fluorescence spectrum, multi-component relaxation characteristic spectrum, nuclear magnetic resonance image, and Fourier transform infrared original spectrum data, 8 characteristic parameters of chicken breast meat quality were extracted, and the characteristic descriptions are as follows:

[0047] Excitation wavelength F1: X-axis value of the three-dimensional synchronous fluorescence spectrum.

[0048] Wavelength difference F2: Y-axis value of the three-dimensional synchronous fluorescence spectrum.

[0049] Fluorescence intensity F3: Z-axis value of the three-dimensional synchronous fluorescence spectrum.

[0050] Wave number F4: X-axis value of the original Fourier infrared spectrum.

[0051] Absorbance F5: Y-axis value of the original Fourier infrared spectrum

[0052] Relaxation time F6: the middle value of relaxation time t1 to t1+0.05 in the multi-component relaxation characteristic spectrum.

[0053] Peak area F7: Peak area between t1 and t1+0.05s in the multi-component relaxation characteristic spectrum

[0054] Image brightness F8: Image brightness of MRI images in HSV color mode.

[0055] (4) Establish a chicken breast quality classification database; test the chicken breasts graded by professionals in (1) and extract characteristic parameters from the original spectral data in the same way as in step (2). Perform data fusion on the eight characteristic parameters collected to construct a chicken breast quality classification database. The data output by three-dimensional synchronous fluorescence, Fourier transform infrared and nuclear magnetic resonance techniques are all collected from the same area of ​​the chicken breast sample.

[0056] (5) Establishing a chicken breast quality classification model; taking each record in the chicken breast quality classification database as a training set, taking the eight characteristic parameters of each record as input, and taking the chicken breast quality as output, and according to the BP neural network modeling method, establishing a chicken breast quality BP neural network grading model.

[0057] (6) Classifying the quality problems of chicken breast meat; placing the chicken breast meat that has not been graded by professionals in a three-dimensional fluorescence spectrometer, a Fourier transform infrared spectrometer and a nuclear magnetic resonance spectrometer for analysis, extracting characteristic parameters from the original data and sending them to a characteristic information extraction module, and determining the quality problems of the chicken breast meat and the grade of the comprehensive heterogeneous meat according to the chicken breast meat quality grading model established in step (5).

[0058] The judgment rule for chicken breast quality grading in step (5) is:

[0059] The BP neural network model outputs a chicken breast quality value less than 1.6, and the chicken breast is normal;

[0060] The BP neural network model outputs a chicken breast quality value between 1.6 and 2.6 (including 2.6), and the chicken breast is firewood-cooked chicken breast;

[0061] The BP neural network model outputs a chicken breast quality value between 2.6 and 3.6 (including 2.6), and the chicken breast is noodle chicken breast;

[0062] The BP neural network model outputs a chicken breast quality value between 3.6 and 4.6 (including 4.6), and the chicken breast is white-striped chicken breast;

[0063] The BP neural network model outputs a chicken breast quality value between 4.6 and 5.6 (including 5.6), and the chicken breast is a comprehensive heterogeneous first-class chicken breast;

[0064] The BP neural network model outputs a chicken breast quality value between 5.6 and 7.6 (including 7.6), and the chicken breast is a comprehensive heterogeneous secondary chicken breast;

[0065] The BP neural network model outputs a chicken breast quality value between 7.6 and 10 (including 10), and the chicken breast is a comprehensive heterogeneous third-grade chicken breast.

[0066] It is known from common technical knowledge that the present invention can be implemented by other embodiments that do not deviate from its spirit or essential features. Therefore, the above disclosed embodiments are only illustrative in all respects and are not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are included in the present invention.

Claims

1. A method for detecting the quality of heterogeneous chicken breast based on multi-information fusion, characterized in that: The chicken breast is tested by using three-dimensional synchronous fluorescence technology, Fourier infrared technology and nuclear magnetic resonance technology, and three-dimensional synchronous fluorescence spectrum, Fourier infrared original spectrum, nuclear magnetic resonance multi-component relaxation characteristic spectrum and nuclear magnetic resonance image are obtained respectively. The characteristic parameter extraction module is then used to extract characteristic parameters from the above spectrum, and a chicken breast quality classification database is constructed by using the characteristic parameters. A chicken breast quality grading model is established according to the chicken breast quality classification database, and the chicken breast is automatically graded by using the chicken breast quality grading model. After the three-dimensional synchronous fluorescence spectrum is exported, its characteristic parameters are extracted: the excitation wavelength F1 is the X-axis value of the three-dimensional synchronous fluorescence spectrum, the wavelength difference F2 is the Y-axis value of the three-dimensional synchronous fluorescence spectrum, and the fluorescence intensity F3 is the Z-axis value of the three-dimensional synchronous fluorescence spectrum; after the original Fourier infrared spectrum is exported, the characteristic parameters are extracted, the X-axis value is the wave number F4, and the Y-axis value is the absorbance F5; the free induced exponential decay curve obtained by nuclear magnetic resonance analysis is inverted to obtain a multi-component relaxation characteristic spectrum, and the data of relaxation time 0-10 s is used as the research object, and the peak area of ​​the peak between t1 and t1+0.05 s is calculated to finally obtain the relaxation time F6, that is, (2t1+0.05) / 2, and the peak area F7; the nuclear magnetic resonance imaging image is analyzed using the image HSV color mode to obtain the image brightness F8; The method comprises the following specific steps: Step 1, professionals who are engaged in the field of chicken processing and have experience in evaluating heterogeneous chicken breast meat, according to the quality of the chicken breast meat, determine it as normal chicken breast meat, firewood chicken breast meat, noodle chicken breast meat, white striped chicken breast meat, comprehensive heterogeneous level 1, comprehensive heterogeneous level 2 and comprehensive heterogeneous level 3, and record them with values ​​0, 1, 2, 3, 4, 5 and 6 in sequence; Step 2, chicken breast detection method and technology; take a sample of appropriate size at the top of the chicken breast perpendicular to the muscle lines, analyze it using a molecular fluorescence spectrometer, and collect characteristic parameters F1, F2 and F3 of the chicken breast sample; then, perform nuclear magnetic resonance analysis on the chicken breast sample, and collect characteristic parameters F6, F7 and F8; then, freeze-dry it in a vacuum freeze dryer for 24 hours, analyze it using a Fourier transform infrared spectrometer, and collect characteristic parameters F4 and F5; Step 3, establishing a chicken breast quality classification database; testing the chicken breasts that have been graded by professionals in step 1, extracting characteristic parameters from the original spectral data, the method is the same as step 2; fusing the collected characteristic parameters F1-F8 to construct a chicken breast quality classification database; the data output by three-dimensional synchronous fluorescence, Fourier transform infrared and nuclear magnetic resonance technology are all collected from the same area of ​​the chicken breast sample; Step 4, establishing a chicken breast quality grading model; using each record information in the chicken breast quality classification database as a training set, using 8 feature parameters F1-F8 of each record information as input, using the chicken breast quality as output, and establishing a chicken breast quality BP neural network grading model according to the BP neural network modeling method; Step 5, grading the quality problems of chicken breast meat; placing the chicken breast meat that has not been graded by professionals in a three-dimensional fluorescence spectrometer, a Fourier transform infrared spectrometer and a nuclear magnetic resonance spectrometer for analysis, extracting the characteristic parameters in the original data and sending them to the characteristic information extraction module, and determining the chicken breast meat quality problems and the comprehensive heterogeneous meat grade according to the chicken breast meat quality grading model established in step 4.

2. The method for detecting heterogeneous chicken breast quality based on multi-information fusion according to claim 1, characterized in that: The excitation wavelength F1 ranges from 190 to 400 nm, the wavelength difference F2 ranges from 25 to 160 nm, and the wave number F4 ranges from 400 to 4000 cm -1 , the relaxation time F6 ranges from 0 to 10 s.

3. The method for detecting heterogeneous chicken breast quality based on multi-information fusion according to claim 2, characterized in that: The chicken breast quality grading model in step 4 is a single hidden layer BP neural network model, whose input variables are F1~F8, the number of output variables is 1, i.e. chicken breast quality, the number of hidden layers is 1, the number of hidden layer neurons is 10, and the activation functions of the hidden layer and the output layer both use the sigmoid function.

4. The method for detecting heterogeneous chicken breast quality based on multi-information fusion according to claim 3, characterized in that: The judgment rule for chicken breast quality grading in step 4 is: The BP neural network model outputs a chicken breast quality value less than 1.6, and the chicken breast is normal; The BP neural network model outputs a chicken breast quality value between 1.6 and 2.6, indicating that the chicken breast is firewood-cooked chicken breast; The BP neural network model outputs a chicken breast quality value between 2.6 and 3.6, and the chicken breast is noodle chicken breast; The BP neural network model outputs a chicken breast quality value between 3.6 and 4.6, and the chicken breast is white-striped chicken breast; The BP neural network model outputs a chicken breast quality value between 4.6 and 5.6, and the chicken breast is a comprehensive heterogeneous first-class chicken breast; The BP neural network model outputs a chicken breast quality value between 5.6 and 7.6, and the chicken breast is a comprehensive heterogeneous second-level chicken breast; The BP neural network model outputs a chicken breast quality value between 7.6 and 10, indicating that the chicken breast is a comprehensive heterogeneous third-grade chicken breast.

Images