A fresh sheep meat quality nondestructive detection system and method

Abstract

The application provides a fresh sheep meat quality nondestructive detection system and method, the system comprises: a flexible spectrum impedance bimodal sensor and an operation host computer; the method comprises: impedance signal and spectrum signal acquisition, working mode selection and sheep meat quality grading model identification. The application realizes nondestructive and accurate detection of sheep meat by using a flexible sensor; realizes richness and accuracy of classification data by data sampling through the flexible spectrum impedance bimodal sensor; realizes comprehensive and accurate perception of sheep meat quality in different working environments and scenes through the sheep meat quality grading model; realizes real-time monitoring and visualization of sheep meat quality by connecting the operation host computer through a Bluetooth wireless transmission module.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of food quality detection, in particular to a fresh mutton quality nondestructive detection system and method. BACKGROUND

[0002] Meat has gained an important position in human diet due to its important nutritional value, such as protein, fat and trace elements. With the improvement of people's living quality, fresh meat is deeply welcomed by consumers. However, fresh meat is not easy to store for a long time. In the process of transportation and storage, it is easy to deteriorate and rot under the action of temperature and humidity, oxygen, microorganisms and enzymes, which is a key factor threatening the edible safety and quality of fresh meat.

[0003] According to the investigation and reading of the literature, at present, the perception of fresh meat quality grade mostly relies on artificial sensory evaluation, physical and chemical index detection and rigid sensor nondestructive detection. Artificial sensory evaluation needs to consume a lot of manpower and time, and cannot detect comprehensively with large error; physical and chemical index detection will cause damage to fresh mutton; rigid sensor nondestructive detection needs to be detected on a specific production line, and cannot realize convenient and rapid quality grade detection. SUMMARY

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present application is to provide a fresh mutton quality nondestructive detection system and method, which realizes nondestructive and accurate detection of mutton by using a flexible sensor; realizes the richness and accuracy of classification data by using a flexible spectral impedance bimodal sensor to sample data; realizes comprehensive and accurate perception of mutton quality in different working environments and scenes by using a mutton quality grading model; realizes real-time monitoring and visualization of mutton quality by using a Bluetooth wireless transmission module to connect an operation host computer.

[0005] To achieve the above purpose, the present application provides the following scheme:

[0006] A fresh mutton quality nondestructive detection system, comprising: a flexible spectral impedance bimodal sensor and an operation host computer; the flexible spectral impedance bimodal sensor comprises: a sensor front-end unit and a sensor rear-end unit; the sensor front-end unit comprises: a PMDS flexible substrate and a LIG serpentine electrode; the sensor rear unit comprises: a PI flexible substrate, a flexible Cu circuit, a PDMS protective layer, a master control module, a voltage conversion module, an impedance conversion module, an alligator clip, a spectrum acquisition module, an I2C multi-channel expansion module and a Bluetooth wireless transmission module; the sensor front-end unit and the sensor rear-end unit are connected through the alligator clip;

[0007] The flexible spectrum impedance dual-mode sensor is connected with the operation host computer through the Bluetooth wireless transmission module; the LIG meander electrode is attached to the surface of the PMDS flexible substrate; the flexible Cu circuit is attached to the surface of the PI flexible substrate; the master control module, the voltage conversion module, the impedance conversion module, the spectrum acquisition module, the I2C multi-channel expansion module, the Bluetooth wireless transmission module and the crocodile clamp are fixed on the flexible Cu circuit by means of soldering; the PDMS protective layer is spin-coated on the master control module, the voltage conversion module, the impedance conversion module, the spectrum acquisition module, the I2C multi-channel expansion module, the Bluetooth wireless transmission module and the crocodile clamp; the crocodile clamp is connected with the LIG meander electrode;

[0008] The LIG meander electrode is used for receiving an excitation signal generated by the impedance conversion module, transmitting a transmitting pulse to the surface of the target fresh sheep meat according to the excitation signal, receiving a reflected pulse of the surface of the target fresh sheep meat, and transmitting the reflected pulse to the impedance conversion module; the impedance conversion module is used for generating the excitation signal, performing DFT transformation on the reflected pulse to obtain an impedance signal, and transmitting the impedance signal to the master control module; the spectrum acquisition module is used for irradiating the surface of the target fresh sheep meat, converting a reflected light of the surface of the target fresh sheep meat into an electric signal to obtain a spectrum signal, and transmitting the spectrum signal to the master control module through the I2C multi-channel expansion module; the master control module is used for controlling the impedance conversion module to generate the excitation signal, and transmitting the impedance signal and the spectrum signal to the operation host computer through the Bluetooth wireless transmission module; and the voltage conversion module is used for converting 5V voltage provided by an external lithium battery into 3.3V.

[0009] Preferably, the preparation process of the flexible spectrum impedance dual-mode sensor comprises:

[0010] The pre-designed electrode pattern is introduced into a laser direct writing device;

[0011] The laser direct writing device is used to perform laser-induced graphene on a PI film to obtain the LIG meander electrode;

[0012] A mixed solution of PDMS solution and curing agent is spin-coated on the PI film, and the PI film is placed on a heating table at 60°C for curing for 2h;

[0013] The LIG meander electrode on the cured PI film is transferred to the PMDS flexible substrate to obtain the sensor front-end unit;

[0014] The pre-designed circuit pattern is introduced into the laser direct writing device;

[0015] performing laser etching on the PI / Cu film by using the laser direct writing device to obtain the flexible Cu circuit;

[0016] fixing the master control module, the voltage conversion module, the impedance conversion module, the spectrum acquisition module, the I2C multi-channel expansion module, the Bluetooth wireless transmission module and the crocodile clamp on the surface of the flexible Cu circuit by soldering;

[0017] uniformly applying a mixed solution of PDMS solution and curing agent on the surface of the flexible Cu circuit, and placing the flexible Cu circuit on a heating table at 60°C for curing for 2h to obtain the sensor back-end unit;

[0018] connecting the sensor back-end unit and the sensor front-end unit by using the crocodile clamp to obtain the flexible spectrum and impedance dual-mode sensor.

[0019] Preferably, a non-destructive detection method for fresh sheep meat quality comprises:

[0020] fixing the flexible spectrum and impedance dual-mode sensor on the surface of the target fresh sheep meat;

[0021] acquiring the impedance signal and the spectrum signal based on a fixed time interval by the operation host computer and the Bluetooth wireless transmission module;

[0022] selecting a working mode by an adaptive modal mode; the working mode includes: impedance mode, spectrum mode and fusion mode;

[0023] inputting the impedance signal and the spectrum signal into a sheep meat quality grading model for quality perception to obtain a target quality detection result.

[0024] Preferably, a method for constructing a sheep meat quality grading model comprises:

[0025] cutting the sheep meat purchased at the same time into the same shape, randomly grouping and storing in the same environment to obtain a plurality of groups of test samples;

[0026] acquiring data of each group of test samples at a fixed time interval by using the flexible spectrum and impedance dual-mode sensor to obtain a test spectrum and impedance raw data set, and saving the test spectrum and impedance raw data set to a preset knowledge base module;

[0027] determining the TVB-N value, PH value, hardness, water loss rate, color difference and protein content of the acquired data of each group in the test spectrum and impedance raw data set to obtain a quality characterization data set and a sheep meat quality classification data set of the sheep meat;

[0028] construct a mutton spectral impedance grading network according to the mutton quality characterization dataset and the mutton quality classification dataset; the mutton spectral impedance grading network comprises: an impedance information grading network layer, a spectral information grading network layer, and a spectral impedance dual-modal information fusion grading network layer; the spectral impedance dual-modal information fusion grading network layer comprises: a data preprocessing layer, a spatial feature extraction layer, a time series feature extraction layer, a feature fusion layer, a quality prediction layer, a regulation mechanism layer, and an error fusion regulation layer; the impedance information grading network layer is used for mutton quality monitoring on the impedance signals collected under the impedance modality; the spectral information grading network layer is used for mutton quality monitoring on the spectral signals collected under the impedance modality; the spectral impedance dual-modal information fusion grading network layer is used for mutton quality monitoring on the impedance signals and the spectral signals collected under the fusion modality; the data preprocessing layer is used for preprocessing the impedance signals and the spectral signals; the spatial feature extraction layer is used for spatial feature extraction on the impedance signals and the spectral signals after preprocessing; the time series feature extraction layer is used for time series feature extraction on the impedance signals and the spectral signals after spatial feature extraction; the feature fusion layer is used for weighted fusion on the spatial feature extraction data and the time series feature extraction data of the impedance signals and the spectral signals; the quality prediction layer is used for mutton quality prediction according to the fusion result features of the feature fusion layer; the regulation mechanism layer is used for feedback adjustment on the prediction abnormal results of the quality prediction layer; and the error fusion regulation layer is used for feedback adjustment on the fusion errors of the feature fusion layer.

[0029] The mutton spectral impedance grading network is optimized to obtain the mutton quality grading model.

[0030] Preferably, the calculation formulas of the impedance information grading network layer and the spectral information grading network layer are respectively:

[0031] and ;

[0032] wherein, ; ; ; ; ; ; ; ; ; ; is the first final mutton grade of the impedance information grading network layer; is the second final mutton grade of the spectral information grading network layer; is the n-th data in the first final mutton grade; for the n-th data item within the second final mutton grade; for a set of impedance quality grade thresholds for mutton; , and are respectively impedance thresholds for under the mutton fresh grade, under the mutton sub-fresh grade and under the mutton putrid deterioration grade; for a set of spectral quality grade thresholds for mutton; , and are respectively spectral thresholds for under the mutton fresh grade, under the mutton sub-fresh grade and under the mutton putrid deterioration grade; denotes a mapping; represents any one of , and ; represents any one of , and ; and are respectively an upper limit and a lower limit of the impedance threshold under the mutton fresh grade; and are respectively an upper limit and a lower limit of the impedance threshold under the mutton sub-fresh grade; and are respectively an upper limit and a lower limit of the impedance threshold under the mutton putrid deterioration grade; and are respectively an upper limit and a lower limit of the spectral threshold under the mutton fresh grade; and are respectively an upper limit and a lower limit of the spectral threshold under the mutton sub-fresh grade; and are respectively an upper limit and a lower limit of the spectral threshold under the mutton putrid deterioration grade.

[0033] Preferably, the calculation formula of the spectral impedance bimodal information fusion grading network layer comprises:

[0034] ,

[0035] ,

[0036] ,

[0037] ,

[0038] ,

[0039] ,

[0040] 、

[0041] 、

[0042] ;

[0043] wherein, and are impedance data features extracted by spatial feature extraction and spectral data features extracted by spatial feature extraction, respectively; and are impedance data features extracted by temporal feature extraction and spectral data features extracted by temporal feature extraction, respectively; and are impedance fusion weight and spectral fusion weight, respectively; is fusion feature; is prediction result; is spatial feature extraction function; is temporal feature extraction function; is preprocessed impedance data; is preprocessed spectral data; , are impedance spatial features and spectral spatial features, respectively; is score calculation function; , are impedance temporal features and spectral temporal features, respectively; is window size; is score weight; is prediction weight; is prediction bias; is prediction function.

[0044] Preferably, further comprising: a regulation mechanism layer; the calculation formula of the regulation mechanism layer comprises:

[0045] 、

[0046] 、

[0047] 、

[0048] 、

[0049] ;

[0050] wherein, is feedback signal; is feedback function; 、 respectively are impedance first optimization fusion weight and spectrum first optimization fusion weight; is the real quality and the predicted quality after the regulation; , respectively are real quality and predicted quality; , respectively are the i th value in the real quality and the predicted quality; is the learning rate.

[0051] Preferably, it further comprises: an error fusion regulation layer; the calculation formula of the error fusion regulation layer comprises:

[0052] ,

[0053] ,

[0054] ,

[0055] ,

[0056] ;

[0057] wherein, is the fusion error value; is the regulation factor; , respectively are impedance second optimization fusion weight and spectrum second optimization fusion weight; is the fusion feature after the error fusion regulation layer; is the regulation sensitivity coefficient.

[0058] Preferably, the calculation formula of the mutton quality grading model is:

[0059] ;

[0060] is the final loss rate; is the loss function; is the fusion error weight coefficient; is the feedback weight coefficient.

[0061] The present application discloses the following technical effects:

[0062] ​The application provides a fresh mutton quality nondestructive detection system and method, which solves the defect of damage caused by conventional detection technology, realizes nondestructive and accurate detection of mutton by adopting a flexible sensor; solves the problem of data source acquisition and the problem of accuracy of different signal collection in different environments by data sampling of the flexible spectral impedance bimodal sensor, realizes enrichment and accuracy of classified data; solves the problem of insufficient expression of mutton quality by a single mode by a mutton quality grading model algorithm, realizes comprehensive and accurate perception of mutton quality; solves the problem of real-time monitoring of mutton quality by connecting an operation host computer by using a Bluetooth wireless transmission module, realizes real-time monitoring and visualization of mutton quality. BRIEF DESCRIPTION OF DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments will be briefly introduced. Obviously, the drawings in the following description only constitute some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor under the premise of the drawings.

[0064] Figure 1 The structure schematic diagram of the fresh mutton quality nondestructive detection system provided by the embodiment of the present application is shown in the figure.

[0065] Figure 2 The schematic diagram of the LIG snake-shaped electrode provided by the embodiment of the present application is shown in the figure.

[0066] Figure 3 The schematic diagram of the flexible Cu circuit provided by the embodiment of the present application is shown in the figure.

[0067] Figure 4 The schematic diagram of the system module structure provided by the embodiment of the present application is shown in the figure.

[0068] Figure 5 The schematic diagram of the fresh mutton quality nondestructive detection process provided by the embodiment of the present application is shown in the figure.

[0069] Figure 6 The schematic diagram of the construction process of the mutton quality grading model provided by the embodiment of the present application is shown in the figure.

[0070] Figure 7 The mutton quality prediction grading flowchart provided by the embodiment of the present application is shown in the figure.

[0071] Explanation of reference signs:

[0072] 1-flexible spectral impedance bimodal sensor, 2-operation host computer. DETAILED DESCRIPTION

[0073] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of the present application.

[0074] The present application aims to provide a fresh sheep meat quality nondestructive detection system and method, which realizes nondestructive and accurate detection of sheep meat by using a flexible sensor, realizes richness and accuracy of classification data by data sampling of the flexible spectral impedance bimodal sensor, realizes comprehensive and accurate perception of sheep meat quality in different working environments and scenarios by a sheep meat quality grading model, and realizes real-time monitoring and visualization of sheep meat quality by connecting an operation host computer by using a Bluetooth wireless transmission module.

[0075] To make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0076] Figure 1 The structure schematic diagram of the fresh sheep meat quality nondestructive detection system provided by the present application is shown in Figure 1 The present application provides a fresh sheep meat quality nondestructive detection system, which comprises a flexible spectral impedance bimodal sensor 1 and an operation host computer 2. The flexible spectral impedance bimodal sensor 1 comprises a sensor front-end unit and a sensor back-end unit. The sensor front-end unit comprises a PMDS flexible substrate and an LIG serpentine electrode. The sensor back-end unit comprises a PI flexible substrate, a flexible Cu circuit, a PDMS protective layer, a main control module, a voltage conversion module, an impedance conversion module, an alligator clip, a spectrum acquisition module, an I2C multi-channel expansion module and a Bluetooth wireless transmission module. The sensor front-end unit and the sensor back-end unit are connected by the alligator clip.

[0077] The flexible spectral impedance bimodal sensor 1 is connected with the operation host computer 2 by the Bluetooth wireless transmission module. The LIG serpentine electrode is attached to the surface of the PMDS flexible substrate (for reference Figure 2 ). The flexible Cu circuit is attached to the surface of the PI flexible substrate (for reference Figure 3 ). The main control module, the voltage conversion module, the impedance conversion module, the spectrum acquisition module, the I2C multi-channel expansion module, the Bluetooth wireless transmission module and the alligator clip are fixed on the flexible Cu circuit by soldering. The PDMS protective layer is spin-coated on the main control module, the voltage conversion module, the impedance conversion module, the spectrum acquisition module, the I2C multi-channel expansion module, the Bluetooth wireless transmission module and the alligator clip. The alligator clip is connected with the LIG serpentine electrode.

[0078] ReferenceFigure 4 The LIG meander electrode is used for receiving an excitation signal generated by the impedance conversion module, and transmitting a transmission pulse to the surface of the target fresh mutton according to the excitation signal, and receiving a reflection pulse of the surface of the target fresh mutton, and transmitting the reflection pulse to the impedance conversion module; the impedance conversion module is used for generating the excitation signal, and performing DFT transformation on the reflection pulse to obtain an impedance signal, and transmitting the impedance signal to the master control module; the spectrum acquisition module is used for irradiating the surface of the target fresh mutton, and converting a reflected light of the surface of the target fresh mutton into an electric signal to obtain a spectrum signal, and transmitting the spectrum signal to the master control module through the I2C multi-channel expansion module; the master control module is used for controlling the impedance conversion module to generate the excitation signal, and transmitting the impedance signal and the spectrum signal to the operation host computer 2 through the Bluetooth wireless transmission module; and the voltage conversion module is used for converting 5V voltage provided by an external lithium battery into 3.3V.

[0079] Preferably, the operation host computer 2 comprises a display function, a process control function, an alarm function and a knowledge base.

[0080] Further, the preparation process of the flexible spectrum impedance dual-mode sensor comprises:

[0081] The pre-designed electrode pattern is introduced into a laser direct writing device;

[0082] The laser direct writing device is used to perform laser-induced graphene on the PI film to obtain the LIG meander electrode;

[0083] The mixed solution of the PDMS solution and the curing agent is spin-coated on the PI film, and the PI film is placed on a heating table at 60 DEG C for curing for 2h;

[0084] The LIG meander electrode on the cured PI film is transferred to the PMDS flexible substrate to obtain a sensor front-end unit;

[0085] The pre-designed circuit pattern is introduced into a laser direct writing device;

[0086] The laser direct writing device is used to perform laser etching on the PI / Cu film to obtain a flexible Cu circuit;

[0087] The master control module, the voltage conversion module, the impedance conversion module, the spectrum acquisition module, the I2C multi-channel expansion module, the Bluetooth wireless transmission module and the crocodile clip are fixed on the surface of the flexible Cu circuit by soldering;

[0088] The mixed solution of the PDMS solution and the curing agent is evenly applied on the surface of the flexible Cu circuit, and the flexible Cu circuit is placed on a heating table at 60 DEG C for curing for 2h to obtain a sensor back-end unit;

[0089] The sensor rear end unit is connected with the sensor front end unit by using the crocodile clip, so as to obtain the flexible spectrum impedance dual-mode sensor 1.

[0090] Reference Figure 5 A nondestructive detection method for fresh sheep meat quality, comprising:

[0091] The flexible spectrum impedance dual-mode sensor 1 is fixed on the surface of the target fresh sheep meat;

[0092] The impedance signal and the spectrum signal are collected based on the fixed time interval by operating the upper computer 2 and the Bluetooth wireless transmission module;

[0093] The working mode is selected by the adaptive modal mode; the working mode includes: impedance mode, spectrum mode and fusion mode;

[0094] The impedance signal and the spectrum signal are input into the sheep meat quality grading model for quality perception, so as to obtain the target quality detection result.

[0095] Specifically, a method for constructing a sheep meat quality grading model, comprising:

[0096] The sheep meat purchased at the same time is cut into the same shape, randomly grouped and stored separately under the same environmental conditions, so as to obtain a plurality of groups of test samples;

[0097] The flexible spectrum impedance dual-mode sensor 1 is used to collect data of each group of test samples at a fixed time interval, so as to obtain a test spectrum impedance original data set, and the test spectrum impedance original data set is saved to a preset knowledge base module;

[0098] The collected data of each group in the test spectrum impedance original data set are determined for TVB-N value, PH value, hardness, water loss rate, color difference and protein content, so as to obtain a quality characterization data set of the sheep meat and a sheep meat quality classification data set;

[0099] The sheep meat spectral impedance grading network is constructed according to the sheep meat quality characterization dataset and the sheep meat quality classification dataset; the sheep meat spectral impedance grading network comprises an impedance information grading network layer, a spectral information grading network layer and a spectral impedance dual-modal information fusion grading network layer; the spectral impedance dual-modal information fusion grading network layer comprises a data preprocessing layer, a spatial feature extraction layer, a time sequence feature extraction layer, a feature fusion layer, a quality prediction layer, a regulation mechanism layer and an error fusion regulation layer; the impedance information grading network layer is used for monitoring the quality of sheep meat by using the impedance signals collected in the impedance mode; the spectral information grading network layer is used for monitoring the quality of sheep meat by using the spectral signals collected in the impedance mode; the spectral impedance dual-modal information fusion grading network layer is used for monitoring the quality of sheep meat by using the impedance signals and the spectral signals collected in the fusion mode; the data preprocessing layer is used for preprocessing the impedance signals and the spectral signals; the spatial feature extraction layer is used for extracting spatial features of the preprocessed impedance signals and the spectral signals; the time sequence feature extraction layer is used for extracting time sequence features of the spatial feature extracted impedance signals and the spectral signals; the feature fusion layer is used for weighting and fusing the spatial feature extraction data and the time sequence feature extraction data of the impedance signals and the spectral signals; the quality prediction layer is used for predicting the quality of sheep meat according to the fusion result features of the feature fusion layer; the regulation mechanism layer is used for feeding back and adjusting the abnormal prediction result of the quality prediction layer; and the error fusion regulation layer is used for feeding back and adjusting the fusion error of the feature fusion layer.

[0100] The sheep meat spectral impedance grading network is optimized to obtain a sheep meat quality grading model.

[0101] Preferably, the calculation formulas of the impedance information grading network layer and the spectral information grading network layer are respectively:

[0102] and ;

[0103] wherein, ; ; ; ; ; ; ; ; ; ; is the first final sheep meat grade of the impedance information grading network layer; is the second final sheep meat grade of the spectral information grading network layer; is the n th data in the first final sheep meat grade; is the n th data in the second final sheep meat grade; is a set of sheep meat impedance quality grade thresholds; , and respectively are impedance threshold values under the fresh grade of mutton, under the sub-fresh grade of mutton and under the putrefactive grade of mutton; is a set of spectral quality grade threshold values of mutton; , and respectively are spectral threshold values under the fresh grade of mutton, under the sub-fresh grade of mutton and under the putrefactive grade of mutton; represents a mapping; represents any one of , and ; represents any one of , and ; and respectively are upper and lower limits of the impedance threshold values under the fresh grade of mutton; and respectively are upper and lower limits of the impedance threshold values under the sub-fresh grade of mutton; and respectively are upper and lower limits of the impedance threshold values under the putrefactive grade of mutton; and respectively are upper and lower limits of the spectral threshold values under the fresh grade of mutton; and respectively are upper and lower limits of the spectral threshold values under the sub-fresh grade of mutton; and respectively are upper and lower limits of the spectral threshold values under the putrefactive grade of mutton.

[0104] Further, the calculation formula of the spectral impedance bimodal information fusion grading network layer comprises:

[0105] ,

[0106] ,

[0107] ,

[0108] ,

[0109] ,

[0110] ,

[0111] ,

[0112] 、

[0113] ;

[0114] wherein, and are impedance data features extracted by spatial feature extraction and spectral data features extracted by spatial feature extraction, respectively; and are impedance data features extracted by temporal feature extraction and spectral data features extracted by temporal feature extraction, respectively; and are impedance fusion weight and spectral fusion weight, respectively; is fusion feature; is prediction result; is spatial feature extraction function; is temporal feature extraction function; is preprocessed impedance data; is preprocessed spectral data; 、 are impedance spatial feature and spectral spatial feature, respectively; is score calculation function; 、 are impedance temporal feature and spectral temporal feature, respectively; is window size; is score weight; is prediction weight; is prediction bias; is prediction function.

[0115] Specifically, further comprising: a regulation mechanism layer; the calculation formula of the regulation mechanism layer comprises:

[0116] 、

[0117] 、

[0118] 、

[0119] 、

[0120] ;

[0121] wherein, is feedback signal; is feedback function; 、 are impedance first optimization fusion weight and spectral first optimization fusion weight, respectively; is the regulation mechanism layer after regulation; ; , respectively are the real quality and the predicted quality; , respectively are the i-th value within the real quality and the predicted quality; is a learning rate.

[0122] Preferably, further comprising: an error fusion regulation layer; a calculation formula of the error fusion regulation layer comprises:

[0123] ,

[0124] ,

[0125] ,

[0126] ,

[0127] ;

[0128] wherein, is a fusion error value; is a regulation factor; , respectively are the impedance second optimized fusion weight and the spectrum second optimized fusion weight; is the fusion feature after the error fusion regulation layer; is a regulation sensitivity coefficient.

[0129] Further, a calculation formula of the mutton quality grading model is:

[0130] ;

[0131] is a final loss rate; is a loss function; is a fusion error weight coefficient; is a feedback weight coefficient.

[0132] Preferably, a calculation formula of the adaptive modal mode adjustment comprises:

[0133]

[0134]

[0135] If / >T, the impedance or spectrum signal is disturbed by the environment; wherein, is an environmental disturbance score, and T is an environmental disturbance factor, For newly collected impedance and spectral data; when both impedance and spectral signals are not disturbed, select the fusion mode; when the spectral signal is disturbed, select the impedance mode; when the impedance signal is disturbed, select the spectral mode; is the average value of impedance data; is the mean of the spectral data; is the standard deviation of the impedance data; is the standard deviation of the spectral data.

[0136] refer to Figure 6 ,The construction process of the mutton quality grading model includes:

[0137] Fresh mutton purchased at the same time was cut into the same shape, randomly grouped, and stored separately under the same environmental conditions to obtain several groups of test samples;

[0138] Flexible impedance sensors and flexible spectral sensors are used to collect data from each group of test samples at fixed time intervals to obtain the original test spectral impedance data set, which is then saved to the knowledge base module. The specific data set is , where Z is impedance data and S is spectral data; 、 Represented as impedance data and spectral data collected at different times;

[0139] For each group of test samples, TVB-N value, pH value, hardness, water loss rate, color difference, and protein content were measured at fixed intervals to obtain the mutton quality characterization dataset and mutton quality classification dataset, which were saved in the knowledge base module. The specific datasets are:

[0140]

[0141]

[0142] in, 、 、 、 、 、 、 They are the specific values ​​of mutton's TVB-N value, pH value, hardness, water loss rate, color difference, and protein content.

[0143] Constructing a mutton spectral impedance grading network; the mutton spectral impedance grading network includes: an impedance information grading network layer, a spectral information grading network layer, and a spectral impedance dual-modal information fusion grading network layer;

[0144] The mutton spectral impedance grading network was optimized to obtain the mutton quality grading model.

[0145] Preferably, the data preprocessing layer removes noise from the spectral data through smoothing processing and processes the spectral data and impedance data to the same dimension. The specific calculation formula of the data preprocessing layer is:

[0146]

[0147] wherein is the spectral noise elimination data, is the window size, , are the preprocessed data of the spectrum and impedance respectively, .

[0148] Further, for the impedance modal, the impedance information hierarchical network layer specifically includes:

[0149] According to the formula , , the impedance data is divided into three hierarchical regions according to , wherein , represent the impedance signals of fresh sheep meat, sub-fresh sheep meat and rotten and deteriorated sheep meat respectively.

[0150] According to the formula , the average values of the impedance signals of the same type of sheep meat are taken and compared with the average values of the impedance signals of another type of sheep meat in the same category. The threshold coefficient is set according to the average values of the impedance signals of the two types of sheep meat. Wherein, is the final sheep meat grade, is the threshold parameter of each grade, is the impedance threshold coefficient.

[0151] Specifically, for the spectral modal, the spectral information hierarchical network layer specifically includes:

[0152] According to the formula , , the impedance data is divided into three hierarchical regions according to , wherein , represent the spectral signals of fresh sheep meat, sub-fresh sheep meat and rotten and deteriorated sheep meat respectively.

[0153] According to the formula , , , , , The impedance signals of the same type of mutton are averaged, and the average value of the impedance signals of another type of mutton in the same category is compared. The threshold coefficient is set according to the average values of the impedance signals of the two types of mutton. Among them, is the final mutton grade, is the threshold parameter of each grade, is the impedance threshold coefficient.

[0154] Preferably, the spectral impedance bimodal information fusion grading network layer of the fusion mode specifically includes:

[0155] According to the formula , The spatial features are extracted from the preprocessed impedance and spectral data; wherein and are the impedance data features extracted by spatial features and the spectral data features extracted by spatial features, respectively, is a spatial feature extraction function.

[0156] According to the formula , The time series features are extracted from the preprocessed impedance and spectral data; wherein and are the impedance data features extracted by spatial features and the spectral data features extracted by time series features, respectively, is a time series feature extraction function.

[0157] According to the formula , , , The fusion features are obtained by weighted fusion of the data after spatial and time series feature extraction; wherein , are the fusion weights, is the calculation score of the time series features, is the calculation score weight, is the fusion feature, and exp is the exponential function.

[0158] According to the formula The quality is predicted according to the fusion features, wherein is the prediction result, is the prediction function, is the prediction weight, is the prediction bias. According to the formula , , , , A feedback mechanism is introduced into the network structure, is a feedback signal, is a feedback function, is a learning rate, and are dynamic feature adjustment fusion weights respectively, is an updated fusion feature.

[0159] According to the formula , , the network is fused with the error definition and corrected, is the fusion error, is the regulation factor, is the regulation sensitivity coefficient, and is the fusion weight adjusted according to the error fusion, is the fusion feature adjusted by the fusion error.

[0160] According to the mutton optimization algorithm , the network structure is optimized, wherein is the final loss rate, is the loss function, is the feedback weight coefficient, is the fusion error weight coefficient.

[0161] Specifically, the thickness of the PDMS flexible substrate of the sensor front-end unit is 1mm to 3mm, and specifically, it can be 2mm. The thickness of the LIG serpentine electrode 12 is 0.05mm. The size of the sensor front-end unit is 50mm*50mm.

[0162] Preferably, the main control module comprises an STM32 microcontroller, a crystal oscillator, a resistor, and a capacitor, which controls the operation of the whole sensor backend unit; the voltage conversion module comprises a voltage conversion chip, a capacitor, and a resistor, which provides corresponding working voltage for each module of the sensor backend unit and stabilizes the voltage; the impedance conversion module comprises an impedance conversion chip, a capacitor, and a resistor, which can stimulate excitation signals of different frequencies to detect the biological dielectric impedance of fresh mutton; the spectrum acquisition module mainly comprises a photoelectric acquisition chip, a flash memory chip, an LED, a capacitor, and a resistor, the flash memory chip stores the underlying program of the photoelectric acquisition chip to ensure its normal work, and the LED is a light source that provides a light signal meeting the working conditions for the photoelectric acquisition chip; the I2C multi-channel expansion module ensures that the photoelectric acquisition chips arranged in a diagonal array can work simultaneously; the Bluetooth wireless transmission module comprises a Bluetooth transmission chip, a resistor, and a capacitor, which is connected with the Bluetooth wireless host computer to start, receive, and save the collected data; the alligator clip is used to connect with the sensor frontend unit to form the sensor frontend unit. The detection principle is as follows: the sensor frontend unit is closely attached to the surface of the fresh mutton, the frequency synthesizer inside the impedance conversion chip generates excitation signals of multiple frequencies, the excitation signals are emitted through the output electrode of the flexible impedance sensing electrode, when the signals pass through the surface of the fresh mutton, they are received by the input electrode of the flexible impedance sensing electrode, and then the signals are transmitted to the STM32 microcontroller through the DFT transformation of the DSP module of the impedance conversion chip, the signals are transmitted to the Bluetooth wireless transmission module through the USART serial port of the STM32 microcontroller, and the Bluetooth wireless transmission module transmits the collected data to the Bluetooth wireless host computer. The flexible spectrum impedance bimodal sensor 1 can detect the biological dielectric impedance in the range of 1 Hz to 100000 Hz. The sensor backend unit is closely attached to the surface of the fresh mutton, and the LED irradiation light source on the surface of the fresh mutton is affected by the internal components of the mutton. The LED irradiates the surface of the mutton, the photoelectric acquisition chip collects the light after the diffuse reflection of the surface of the mutton to generate corresponding photoelectric signal changes, the photoelectric signals are transmitted to the main control module through the I2C multi-channel expansion module, the main control module transmits the data to the Bluetooth wireless transmission module through its internal serial port, and the Bluetooth wireless transmission module transmits the collected data to the operation host computer 2.

[0163] Optionally, the thickness of the PI flexible substrate of the flexible spectrum impedance bimodal sensor 1 is 0.05 mm to 0.35 mm, and specifically, it can be 0.20 mm; the thickness of the flexible Cu circuit is one of 0.035 mm, 0.05 mm, or 0.07 mm, for example, it can be 0.05 mm; the thickness of the PDMS protective layer is just enough to cover the height of the electronic components. The size of the sensor backend unit is 120 mm*60 mm.

[0164] Furthermore, the number of turns, length, width, and overall size of the LIG serpentine electrode were designed using circuit design software. The electronic components were designed to meet the sensor's functional requirements, ensuring they met the circuit system's requirements. The circuit line widths were 0.2mm and 0.5mm.

[0165] Preferably, a first laser etching is performed on a flexible Cu film with a thickness of 0.2 mm and a PI substrate using a laser direct writing device at a power of 5 W and a speed of 100 mm / s, and then a second laser etching is performed at the same power and speed to completely etch the Cu film except for the circuit, so that it loses its conductive ability.

[0166] Furthermore, the laser-etched flexible Cu circuit was cleaned with deionized water. After cleaning, the flexible Cu circuit was placed on a workbench and soldered using solder paste and a welding gun. After soldering, the circuit was cooled and solidified at room temperature for 3 minutes.

[0167] refer to Figure 7 This embodiment also provides a more specific method for predicting and grading mutton quality, which specifically includes:

[0168] 1): Number the fresh mutton to be monitored, and start the flexible spectral impedance dual-modal sensor 1 corresponding to the number by operating the host computer 2;

[0169] 2): The flexible spectral impedance dual-modal sensor 1 system is initialized to determine whether the sensor is working properly; if it is not working properly, the sensor is restarted for adjustment; if it still does not work properly after adjustment, an alarm is issued to remind manual inspection;

[0170] 3): The flexible spectral impedance dual-modal sensor 1 collects spectral signals and impedance signals of fresh mutton, and collects them multiple times within the same period of time to ensure data stability; the spectral signals and impedance signals are compared with the knowledge base to determine whether the collected signals are abnormal data; if they are unavailable, re-collect them; if the re-collected signals are still abnormal data, an alarm is issued and manual inspection is performed;

[0171] Prioritizing, the simultaneous application of the three modalities can more accurately determine the freshness of the current fresh mutton; in single-modality mode, basic quality judgments can be made on fresh mutton, which is suitable for working environments that cannot simultaneously meet the needs of two modalities and where the quality of fresh mutton needs to be quickly screened; in fusion mode, fresh mutton can be accurately measured and the quality indicators and quality status of the current mutton can be predicted, which is suitable for high-end restaurants and research-level mutton preparation.

[0172] The database collectively includes: impedance signals and spectral signals of fresh mutton collected in the experiment

[0173]

[0174] The fresh sheep meat quality indexes and quality grades collected in the test

[0175]

[0176]

[0177] 4) Determine whether the spectrum signal and the impedance signal are disturbed by the environment, when both are disturbed, start the fusion mode; when the impedance signal is disturbed, start the spectrum mode; when the spectrum signal is disturbed, start the impedance mode;

[0178] The environmental disturbance is specifically: when the sensor is in an environment with high external light intensity, the data collected by the spectrum acquisition module will be affected, causing the sensor to work abnormally; when the sensor is disturbed by electromagnetic interference, and is in an environment with poor temperature and humidity, the serpentine sensing electrode will be affected, causing the data to be abnormal;

[0179] The specific formula for signal interference judgment is: , , if / >T, the impedance or spectrum signal is disturbed by the environment

[0180] 5) After selecting the determined mode, the data is imported into the sheep meat quality prediction grading model for data preprocessing

[0181]

[0182] 6) If it is the impedance mode, it is input into the impedance information grading network layer, and the specific formula includes:

[0183] If it is the spectrum mode, it is input into the spectrum information grading network layer, and the specific formula includes:

[0184] If it is the fusion mode, it is input into the spectrum impedance dual-mode information fusion grading network layer, and the specific formula includes:

[0185] ,

[0186] ,

[0187]

[0188]

[0189]

[0190]

[0191]

[0192] Introducing feedback mechanism in the network, the specific formula contains:

[0193]

[0194]

[0195]

[0196]

[0197]

[0198] Introducing error fusion regulatory layer in the network, the specific formula contains:

[0199]

[0200]

[0201]

[0202]

[0203] According to the adaptive monitoring algorithm, the model is optimized, and the specific formula contains:

[0204]

[0205] Preferably, since the chemical substances inside the fresh mutton can be measured in the visible / near-infrared spectrum, and have a relatively stable spectrum. When the LED light irradiates the surface of the mutton, the vibration and rotation energy level of the molecules inside the mutton changes, which causes the absorption peak to change. The visible / near-infrared spectrum area expresses the absorption details of C-O, C=O, O-H and N-H key chemical bonds, which can reflect the internal quality of the mutton.

[0206] Optionally, since the cell structure inside the fresh mutton can be measured and expressed in impedance, and has a relatively stable impedance amplitude and impedance angle. When the sensor front-end unit is close to the surface of the mutton, the impedance amplitude and impedance angle of the internal cells of the mutton change due to the changes in water loss and structure, which can reflect the internal quality of the mutton.

[0207] Specifically, the spectral impedance bimodal information network layer adopts a spatial feature extraction layer as a feature extractor to capture complex long-term dependencies and local patterns. The spatial feature extraction layer rolls over the input sequence through a sliding window, capturing local patterns at different time steps, which helps the model learn the spatial features of the data. The temporal feature extraction layer combines forward and backward information flow to better capture long-term dependencies in sequential data. In addition, local features in the lamb spectral and impedance data are crucial for understanding lamb freshness, and the spatial feature extraction layer plays a key role in extracting these features. Combined with the temporal feature extraction layer, the model can consider historical and predictive information to better understand the context and dependencies in time series data. At the same time, the introduction of attention mechanism enables the model to intelligently select and weight basic features to adaptively adjust the attention to different parts of the data. This approach enables the model to flexibly learn and reconstruct data, especially suitable for dealing with random noise in lamb spectral impedance data, avoiding the need for complex feature engineering. In terms of adaptive anomaly detection, the network structure is dynamically adjusted by adopting a feedback mechanism and error fusion regulation layer.

[0208] 7): Weighted scoring of the output of the spectral impedance bimodal hierarchical network layer, outputting the quality grade of fresh lamb, the specific formula is:

[0209]

[0210]

[0211] wherein, is the total quality score, are the quality scores in 6), the value of is one of (0, 0.5, 0.75, 1); all scoring standards are scored by the knowledge base module.

[0212] The beneficial effects of the present application are as follows:

[0213] The present application realizes non-destructive and accurate detection of lamb by using a flexible sensor; realizes the richness and accuracy of classification data by using a flexible spectral impedance bimodal sensor for data sampling; realizes comprehensive and accurate perception of lamb quality in different working environments and scenarios through a lamb quality grading model; realizes real-time monitoring and visualization of lamb quality by connecting an upper computer through a Bluetooth wireless transmission module.

[0214] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be mutually referred to.

[0215] The principles and implementation manners of the present application are described by using specific examples in the present application, and the above examples are only used to help understand the method of the present application and its core idea; meanwhile, for the general technical personnel in the art, the specific implementation manners and application ranges will be changed according to the idea of the present application. In conclusion, the content of the present specification should not be understood as the limitation of the present application.

Claims

1. A non-destructive detection system for fresh sheep meat quality, characterized in that, The application relates to a flexible spectrum impedance dual-mode sensor and an operation host computer. The flexible spectrum impedance dual-mode sensor comprises a sensor front-end unit and a sensor rear-end unit; the sensor front-end unit comprises a PMDS flexible substrate and an LIG meander electrode; the sensor rear-end unit comprises a PI flexible substrate, a flexible Cu circuit, a PDMS protective layer, a master control module, a voltage conversion module, an impedance conversion module, an alligator clip, a spectrum acquisition module, an I2C multi-channel expansion module and a Bluetooth wireless transmission module; the sensor front-end unit and the sensor rear-end unit are connected through the alligator clip; The flexible spectrum impedance dual-mode sensor is connected with the operation host computer through the Bluetooth wireless transmission module; the LIG meander electrode is attached to the surface of the PMDS flexible substrate; the flexible Cu circuit is attached to the surface of the PI flexible substrate; the master control module, the voltage conversion module, the impedance conversion module, the spectrum acquisition module, the I2C multi-channel expansion module, the Bluetooth wireless transmission module and the alligator clip are fixed on the flexible Cu circuit through tin soldering; the PDMS protective layer is spin-coated on the master control module, the voltage conversion module, the impedance conversion module, the spectrum acquisition module, the I2C multi-channel expansion module, the Bluetooth wireless transmission module and the alligator clip; the alligator clip is connected with the LIG meander electrode; The LIG meander electrode is used for receiving an excitation signal generated by the impedance conversion module, sending an emission pulse to the surface of target fresh sheep meat according to the excitation signal, receiving a reflection pulse of the surface of the target fresh sheep meat and sending the reflection pulse to the impedance conversion module; the impedance conversion module is used for generating the excitation signal and performing DFT transformation on the reflection pulse to obtain an impedance signal and send the impedance signal to the master control module; the spectrum acquisition module is used for irradiating the surface of the target fresh sheep meat, converting a reflection light of the surface of the target fresh sheep meat into an electric signal to obtain a spectrum signal and sending the spectrum signal to the master control module through the I2C multi-channel expansion module; the master control module is used for controlling the impedance conversion module to generate the excitation signal and sending the impedance signal and the spectrum signal to the operation host computer through the Bluetooth wireless transmission module; and the voltage conversion module is used for converting 5V voltage provided by an external lithium battery into 3.3V. The preparation process of the flexible spectrum impedance dual-mode sensor comprises the following steps:

2. The non-destructive testing system for fresh sheep meat quality according to claim 1, characterized in that, introducing a pre-designed electrode pattern into a laser direct writing device; performing laser-induced graphene on a PI film by using the laser direct writing device to obtain the LIG meander electrode; spin-coating a mixed solution of a PDMS solution and a curing agent on the PI film and placing the PI film on a heating table at 60 DEG C for curing for 2 hours; transferring the LIG meander electrode on the cured PI film to the PMDS flexible substrate to obtain the sensor front-end unit; introducing a pre-designed circuit pattern into the laser direct writing device; ​ The laser direct writing equipment is used for laser etching on a PI / Cu film to obtain the flexible Cu circuit; The main control module, the voltage conversion module, the impedance conversion module, the spectrum acquisition module, the I2C multi-channel expansion module, the Bluetooth wireless transmission module and the crocodile clamp are fixed on the surface of the flexible Cu circuit through soldering; A mixed solution of a PDMS solution and a curing agent is uniformly applied on the surface of the flexible Cu circuit, and the flexible Cu circuit is placed on a heating table at 60 DEG C for curing for 2 hours to obtain the sensor rear-end unit; The sensor rear-end unit and the sensor front-end unit are connected by the crocodile clamp to obtain the flexible spectrum and impedance dual-mode sensor.

3. A method for non-destructive detection of fresh lamb meat quality, characterized in that, The detection method is applied to the fresh sheep meat quality nondestructive detection system of claim 1, and the detection method comprises: The flexible spectrum and impedance dual-mode sensor is fixed on the surface of the target fresh sheep meat; The impedance signal and the spectrum signal are collected based on a fixed time interval through the operation host computer and the Bluetooth wireless transmission module; An adaptive modal mode is used for working mode selection, and the working mode comprises an impedance mode, a spectrum mode and a fusion mode; The impedance signal and the spectrum signal are input into a sheep meat quality grading model for quality perception to obtain a target quality detection result.

4. A method for constructing a mutton quality grading model, characterized in that, The construction method is applied to the fresh sheep meat quality nondestructive detection method of claim 3, and the construction method comprises: Sheep meat purchased at the same time is cut into the same shape, randomly grouped and stored separately under the same environmental condition to obtain a plurality of groups of test samples; The flexible spectrum and impedance dual-mode sensor is used for data collection of each group of test samples at a fixed time interval to obtain a test spectrum and impedance original data set, and the test spectrum and impedance original data set is saved to a preset knowledge base module; TVB-N value, PH value, hardness, water loss rate, color difference and protein content of the collected data of each group in the test spectrum and impedance original data set are determined to obtain a sheep meat quality characterization data set and a sheep meat quality classification data set; According to the sheep meat quality characterization dataset and the sheep meat quality classification dataset, a sheep meat spectrum impedance grading network is constructed; the sheep meat spectrum impedance grading network comprises: an impedance information grading network layer, a spectrum information grading network layer, and a spectrum impedance bimodal information fusion grading network layer; the spectrum impedance bimodal information fusion grading network layer comprises: a data preprocessing layer, a spatial feature extraction layer, a time series feature extraction layer, a feature fusion layer, a quality prediction layer, a regulation mechanism layer, and an error fusion regulation layer; the impedance information grading network layer is used for monitoring the quality of sheep meat by using the impedance signals collected in the impedance mode; the spectrum information grading network layer is used for monitoring the quality of sheep meat by using the spectrum signals collected in the impedance mode; the spectrum impedance bimodal information fusion grading network layer is used for monitoring the quality of sheep meat by using the impedance signals and the spectrum signals collected in the fusion mode; the data preprocessing layer is used for preprocessing the impedance signals and the spectrum signals; the spatial feature extraction layer is used for extracting spatial features of the preprocessed impedance signals and spectrum signals; the time series feature extraction layer is used for extracting time series features of the spatial feature extracted impedance signals and spectrum signals; the feature fusion layer is used for weighted fusion of the spatial feature extraction data and the time series feature extraction data of the impedance signals and the spectrum signals; the quality prediction layer is used for predicting the quality of sheep meat according to the fusion result features of the feature fusion layer; the regulation mechanism layer is used for feedback adjustment of the prediction abnormal result of the quality prediction layer; the error fusion regulation layer is used for feedback adjustment of the fusion error of the feature fusion layer. The sheep meat spectrum impedance grading network is optimized to obtain the sheep meat quality grading model.

5. The method for constructing a lamb quality grading model according to claim 4, characterized in that, The calculation formulas of the impedance information grading network layer and the spectrum information grading network layer are respectively: and ; wherein ; ; ; ; ; ; ; ; ; ; is a first final mutton grade for the impedance information grading network layer; is a second final mutton grade for the spectral information grading network layer; is an nth item of data within the first final mutton grade; is an nth item of data within the second final mutton grade; is a set of mutton impedance quality grade thresholds; , and are impedance thresholds under the mutton fresh grade, under the mutton sub-fresh grade, and under the mutton putrid grade, respectively; is a set of mutton spectral quality grade thresholds; , and are spectral thresholds under the mutton fresh grade, under the mutton sub-fresh grade, and under the mutton putrid grade, respectively; denotes a mapping; represents any one of , and ; represents any one of , and ; and are an upper limit and a lower limit of the impedance thresholds under the mutton fresh grade, respectively; and are an upper limit and a lower limit of the impedance thresholds under the mutton sub-fresh grade, respectively; and are an upper limit and a lower limit of the impedance thresholds under the mutton putrid grade, respectively; and are an upper limit and a lower limit of the spectral thresholds under the mutton fresh grade, respectively; and are an upper limit and a lower limit of the spectral thresholds under the mutton sub-fresh grade, respectively; and are an upper limit and a lower limit of the spectral thresholds under the mutton putrid grade, respectively.

6. The method for constructing a sheep meat quality grading model according to claim 4, characterized in that, The calculation formula of the spectrum impedance bimodal information fusion grading network layer comprises: 、 、 、 、 、 、 、 、 ； wherein, and are impedance data features extracted by spatial feature extraction and spectral data features extracted by spatial feature extraction, respectively; and are impedance data features extracted by temporal feature extraction and spectral data features extracted by temporal feature extraction, respectively; and are impedance fusion weight and spectral fusion weight, respectively; is fusion feature; is prediction result; is spatial feature extraction function; is temporal feature extraction function; is preprocessed impedance data; is preprocessed spectral data; , are impedance spatial feature and spectral spatial feature, respectively; is score calculation function; , are impedance temporal feature and spectral temporal feature, respectively; is window size; is score weight; is prediction weight; is prediction bias; is prediction function.

7. The method according to claim 6, wherein, Further comprising: The regulation mechanism layer; The calculation formula of the regulation mechanism layer Further comprising: 、 、 、 、 ； wherein, is a feedback signal; is a feedback function; , are an impedance first optimized fusion weight and a spectrum first optimized fusion weight, respectively; is the said ; , are a real quality and a predicted quality, respectively; , are an i-th value within the said real quality and the said predicted quality, respectively; is a learning rate.

8. The method according to claim 7, characterized in that, Further comprising: The error fusion regulation layer; The calculation formula of the error fusion regulation layer Further comprising: 、 、 、 、 ； wherein, is a fusion error value; is a regulation factor; , are an impedance second optimized fusion weight and a spectrum second optimized fusion weight, respectively; is the fusion feature after the error fusion regulation layer; is a regulation sensitivity coefficient.

9. The method for constructing a lamb quality grading model according to claim 8, characterized in that, The calculation formula of the sheep meat quality grading model is: ； wherein, is the final loss rate; is the loss function; is the fusion error weight coefficient; is the backpropagation weight coefficient.

Images