Yellowfin sea bream meat quality detection method and system based on hyperspectral imaging technology

Abstract

The present invention provides a method and system for detecting the meat quality of yellowfin sea bream based on hyperspectral imaging technology, which relates to the technical field of yellowfin sea bream meat quality detection. The present invention extracts rich spectral detail information through multi-scale wavelet transform, supplemented by a dynamic weight adjustment mechanism, thereby enhancing the key band recognition capability and the accuracy of spectral feature expression, reducing irrelevant information interference, and realizing high-precision mapping of hyperspectral data and meat quality indicators; taking fat characteristics as an intermediary variable, a nonlinear mapping model is constructed in combination with multidimensional physical meat quality indicators, scientifically revealing the intrinsic connection between fat content and meat quality, and improving the biological rationality of meat quality evaluation; at the same time, a weighted scoring system of multidimensional meat quality indicators is integrated to establish a standardized meat quality grading standard to meet the needs of industrial application and realize rapid, non-destructive, and intelligent detection and grading of yellowfin sea bream meat quality.

Description

Technical Field

[0001] The present invention relates to the technical field of yellowfin sea bream meat quality detection, and in particular to a yellowfin sea bream meat quality detection method and system based on hyperspectral imaging technology. Background Art

[0002] With the rapid development of aquaculture, yellowfin sea bream has attracted widespread attention due to its excellent nutritional value and market demand. However, the quality of its meat varies greatly due to factors such as the breeding environment and feed formulation. Traditional meat quality evaluation methods often rely on sensory evaluation or single physical and chemical indicators. These methods are complex and highly subjective, making it difficult to meet the modern aquaculture industry's demand for fast, accurate, and non-destructive meat quality testing. While existing spectral detection technologies provide non-destructive testing methods, they generally ignore the complex nonlinear relationship between fat characteristics and meat quality in fish. Furthermore, spectral data are often processed using a single scale or fixed weight, which weakens key band information and affects prediction accuracy and model stability.

[0003] In the prior art, publication number CN117092041A discloses a method for rapid detection of live carp muscle quality based on hyperspectral imaging technology, including obtaining hyperspectral data and texture indicators of carp samples; extracting spectral information of the region of interest, calculating and generating the mean of the spectral information of the region of interest, performing variable screening and extraction based on the mean of the spectral information, generating characteristic variables, and generating a sample set based on the characteristic variables and texture indicators; constructing a meat quality index prediction model, and optimizing and screening the meat quality index prediction model based on the sample set to generate the best prediction model; predicting the pixel spectral values ​​of the measured carp samples through the best prediction model to generate texture results of different muscle areas of the measured carp samples. Although it covers a variety of prediction models and spectral preprocessing methods, it does not combine biological fat characteristics as an intermediary variable, and lacks in-depth exploration of the nonlinear intrinsic relationship between fat and meat quality; at the same time, the key bands in the spectral characteristics are insufficiently utilized, and the prediction accuracy and model interpretability are weak, which limits the actual application effect.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for detecting the quality of yellowfin sea bream meat based on hyperspectral imaging technology to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The method for detecting the quality of yellowfin sea bream meat based on hyperspectral imaging technology includes the following specific steps:

[0008] S1: A group feeding experiment was conducted on yellowfin seabream to collect the meat quality indicators and fat characteristics of yellowfin seabream under different feeding conditions, and a mapping model between meat quality indicators and fat characteristics was constructed;

[0009] S2: Using hyperspectral imaging technology to collect spectral data of yellowfin seabream under different feeding conditions, and analyzing the fat characteristics of yellowfin seabream by extracting multi-scale spectral features;

[0010] S3: A dynamic weight adjustment mechanism is constructed based on the fat characteristics of yellowfin sea bream to perform weighted processing on multi-scale spectral features. A prediction model of spectral data and meat quality indicators is constructed based on the weighted spectral features.

[0011] S4: Construct a scoring function for meat quality indicators, calculate the comprehensive meat quality score of yellowfin sea bream based on the predicted meat quality indicators generated by the prediction model, and classify the meat quality grades according to the comprehensive meat quality score.

[0012] Preferably, the logic of constructing the mapping model in step S1 is:

[0013] S101: Yellowfin sea bream fry are divided into several groups, fed periodically with feeds of different formulas, and used as samples for testing after they grow up;

[0014] S102: Collect meat quality indicators of yellowfin seabream samples, including hardness, elasticity, cohesion, adhesiveness, chewiness, and resilience, as well as fat characteristics of yellowfin seabream samples, including liver-to-body ratio, organ-to-body ratio, and liver fat content;

[0015] S103: constructing a fat feature vector based on the fat characteristics of the yellowfin sea bream sample, and using the RBF function to establish a mapping model between the meat quality index and the fat characteristics.

[0016] Preferably, the mapping expression of the mapping model between the meat quality index and the fat characteristic is:

[0017] ;

[0018] In the formula Indicates the first The predicted value of meat quality index, Indicates the first The meat quality index corresponds to The weight coefficients of the basis functions, subscript 、 Represent the meat quality index and the index of the basis function in the mapping model, represents the total number of basis functions in the mapping model, represents the fat feature vector, 、 、 Both represent the model parameters to be trained in the mapping model.

[0019] Preferably, in step S2, the logic of extracting multi-scale spectral features to analyze the fat characteristics of yellowfin sea bream is:

[0020] S201: Collect spectral data of yellowfin seabream samples;

[0021] S202: Perform multi-scale wavelet transform on the spectral data to extract multi-scale spectral features. The calculation method is:

[0022] ;

[0023] In the formula Indicates the wavelength is The spectral data at the scale level and pan position The wavelet coefficients of Indicates the wavelength is The spectral data, Indicates the wavelength is The spectral data at the scale level and pan position The wavelet basis function of 、 Represent the index of scale layer and translation position respectively, represents the wavelength, 、 Represent the maximum and minimum values ​​of wavelength respectively;

[0024] S203: Extracting wavelet coefficients of the spectral data at different wavelength scales through multi-scale analysis to enhance the recognition capability of fat features.

[0025] Preferably, in step S3, the logic of constructing a prediction model of spectral data and meat quality indicators based on weighted spectral features is:

[0026] S301: Build a dynamic weight adjustment mechanism to introduce fat characteristic information into the spectral data weighting process to improve the key band identification capability;

[0027] S302: Iteratively calculating the dynamic weight of the spectral data according to the gradient update algorithm to obtain a weighted spectral feature;

[0028] S303: Based on the weighted spectral features, a prediction model of spectral data and meat quality indicators is established using RBF function.

[0029] Preferably, the dynamic weight is calculated as follows when iterating using the gradient update algorithm:

[0030] ;

[0031] In the formula Indicates the wavelength is The spectral data of The dynamic weight at the iteration, represents the learning rate, represents the error loss function;

[0032] The expression of the prediction model of the spectral data and meat quality index is:

[0033] ;

[0034] In the formula Indicates the prediction model The predicted value of meat quality index, Indicates the prediction model The meat quality index corresponds to The weight coefficients of the basis functions, subscript represents the index of the basis function in the prediction model, represents the total number of basis functions in the prediction model, represents the weighted spectral features, 、 、 Both represent the model parameters to be trained in the prediction model.

[0035] Preferably, the expression of the weighted spectral feature is:

[0036] ;

[0037] In the formula Indicates the wavelength is The dynamic weight after the spectral data iteration is completed, subscript The index representing the wavelength, Indicates the maximum value of the wavelength index.

[0038] Preferably, the error loss function is constructed based on the square error between the predicted value of the meat quality index and the measured value of the meat quality index based on the mapping model and the prediction model, and its expression is:

[0039] ;

[0040] In the formula Indicates the measured value of meat quality index, 、 Represent the weight coefficients of the mapping model and the prediction model respectively, both of which are greater than 0, and .

[0041] Preferably, the expression of the scoring function in step S4 is:

[0042] ;

[0043] In the formula Indicates the comprehensive meat quality score, Indicates the The weight coefficients of meat quality indicators must meet the following requirements:

[0044] ;

[0045] when When the meat of yellowfin sea bream was considered to be of substandard quality;

[0046] when When the meat quality of yellowfin sea bream is considered to be qualified;

[0047] when When it comes to yellowfin sea bream, it is considered that the meat quality is good;

[0048] when When it comes to yellowfin sea bream, it is believed that the meat quality is excellent;

[0049] in 、 、 Both represent the preset scoring thresholds, and .

[0050] A yellowfin sea bream meat quality detection system based on hyperspectral imaging technology, wherein the detection system is used to perform the above-mentioned detection method, specifically comprising:

[0051] A sample detection module, wherein the sample detection module is used to detect the meat quality index and fat characteristics of the yellowfin sea bream sample;

[0052] A hyperspectral acquisition module, which is used to detect spectral data of yellowfin sea bream samples and extract multi-scale spectral features;

[0053] A data analysis module, the data analysis module is used to construct a mapping model between meat quality indicators and fat characteristics and a prediction model between spectral data and meat quality indicators;

[0054] The meat quality scoring module is used to calculate the comprehensive meat quality score of the yellowfin sea bream and classify the meat quality grades according to the comprehensive meat quality score.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] The present invention extracts rich spectral detail information through multi-scale wavelet transform, supplemented by a dynamic weight adjustment mechanism, to enhance the key band recognition ability and the accuracy of spectral feature expression, reduce the interference of irrelevant information, and achieve high-precision mapping of hyperspectral data and meat quality indicators; taking fat characteristics as the intermediary variable, combining multidimensional physical meat quality indicators to construct a nonlinear mapping model, scientifically revealing the intrinsic connection between fat content and meat quality, and improving the biological rationality of meat quality evaluation; at the same time, a weighted scoring system of comprehensive multidimensional meat quality indicators is established to establish a standardized meat quality grading standard to meet the needs of industrial applications and realize rapid, non-destructive and intelligent detection and grading of yellowfin sea bream meat quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 Schematic diagram of the overall method flow of the present invention;

[0058] Figure 2 It is a schematic diagram of the module structure of the present invention. DETAILED DESCRIPTION

[0059] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0060] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0061] Example:

[0062] See also Figure 1~Figure 2 , the present invention provides a technical solution:

[0063] The method for detecting the quality of yellowfin sea bream meat based on hyperspectral imaging technology includes the following specific steps:

[0064] S1: A group feeding experiment was conducted on yellowfin sea bream. The meat quality indicators and fat characteristics of yellowfin sea bream under different feeding conditions were collected, and a mapping model between meat quality indicators and fat characteristics was constructed.

[0065] The logic for constructing the mapping model in step S1 is:

[0066] S101: Yellowfin sea bream fry are divided into several groups, fed periodically with feeds of different formulas, and used as samples for testing after they grow up;

[0067] S102: Collect meat quality indicators of yellowfin seabream samples, including hardness, elasticity, cohesion, adhesiveness, chewiness, and resilience, as well as fat characteristics of yellowfin seabream samples, including liver-to-body ratio, organ-to-body ratio, and liver fat content;

[0068] Specifically, hardness reflects the meat's ability to resist deformation, affecting its firmness; elasticity reflects the meat's ability to return to its original shape and is closely related to its tenderness; cohesion indicates the binding force between meat molecules and affects the overall structural stability of the meat; adhesiveness measures the adhesion of the meat to the mouth when bitten, determining the stickiness of the mouthfeel; chewiness, a combination of hardness, elasticity, and cohesion, reflects the ease of chewing; and resilience reflects the meat's ability to recover after being subjected to force, affecting the residual taste after eating. Together, these indicators constitute the multidimensional sensory and physical manifestations of meat quality, which can comprehensively evaluate the quality of yellowfin sea bream meat and are positively correlated with meat quality. Furthermore, these indicators can be obtained through actual physical measurement methods. The yellowfin sea bream samples are cut into small pieces of equal size and regular shape, and measured using a texture analyzer under constant temperature conditions: a two-stage compression test can be performed. After the first compression is completed, the indenter is lifted, and the sample elastically recovers after the indenter is removed. The ratio of the height of the sample after the first compression to the height recovered during rebound is used as the resilience, the ratio of the maximum rebound force to the maximum compression force is used as the elasticity, and the maximum compression force in the two compressions is used as the hardness value; cohesion indicates the stability of the intermolecular bonding force inside the meat, which can be defined as the ratio of the reaction force generated by the sample on the indenter during the second compression to the first compression force, which is used to reflect the integrity and coherence of the meat structure; and For example, taking adhesion as an example, the record is made when the indenter leaves the sample before the second compression, and the instrument records the negative force area generated when the indenter is pulled up. The negative force (pulling force) generated during the removal of the indenter changes with time. The integral represents the negative force area, and the negative force area is used as the parameter to quantify adhesion; for example, taking chewiness as an example, hardness, elasticity and cohesion can be quantified by simply weighting them; the same is true for the other parameters. The mechanical parameters used to quantify these parameters can be actually measured. They can be defined and quantified according to general physical meanings, or defined according to the professional knowledge of the testers. After quantification, they can be normalized to eliminate dimensional differences and directly input into the model. Therefore, the specific definition method and normalization method will not be repeated here.

[0069] Liver-to-body ratio and organ-to-body ratio are important biological indicators that reflect the distribution of accumulated fat in fish bodies, and can reflect the metabolic status and storage of fat in fish bodies. The fat content in the liver directly reflects the amount of accumulated fat, and the fat content is closely related to the flavor, tenderness and texture of the meat. As the main physiological basis for the formation of meat quality, fat characteristics can effectively reflect the mechanism of the influence of feed on meat quality. Therefore, by establishing a mapping model between fat characteristics and meat quality indicators, and using fat characteristics to predict meat quality indicators such as muscle elasticity, adhesion and chewiness, the purpose of indirectly predicting meat quality through spectral data can be achieved.

[0070] S103: constructing a fat feature vector based on the fat characteristics of the yellowfin sea bream sample, and using the RBF function to establish a mapping model between the meat quality index and the fat characteristics.

[0071] The RBF function is used because it has strong nonlinear fitting ability and good generalization performance. It can adapt to the complex variation of the influence of fat characteristics on different meat quality indicators, reduce the risk of overfitting, and improve the prediction accuracy of unknown samples.

[0072] The mapping expression of the mapping model between meat quality index and fat characteristics is:

[0073] ;

[0074] In the formula Indicates the first The predicted value of meat quality index, Indicates the first The meat quality index corresponds to The weight coefficients of the basis functions, subscript 、 Represent the meat quality index and the index of the basis function in the mapping model, represents the total number of basis functions in the mapping model, represents the fat feature vector, 、 、 Both represent the model parameters to be trained in the mapping model, where the fat feature vector can be expressed as:

[0075] ;

[0076] in 、 、 represent liver-to-body ratio, organ-to-body ratio, and liver fat content, respectively.

[0077] It is understandable that, unlike nutritional richness, which can directly measure the content of various nutrients in meat, meat quality testing is more subjective and more difficult to quantify. Therefore, in traditional meat quality evaluation, classification is usually based on the professional experience of the testers, and the human subjective factors in the classification results account for a large proportion. In the present invention, by adopting multiple groups of feed for group feeding, a variety of fat characteristics are obtained, combined with the accurately measured multidimensional meat quality indicators, and using the radial basis function (RBF) nonlinear mapping model, it is possible to capture the inherent connection between complex, nonlinear fat characteristics and meat quality, breaking through the limitations of simple linear or empirical models in the prior art. In addition, the fat characteristics serve as an intermediary variable between feed conditions and meat quality, making the model more consistent with biological mechanisms, improving the model's explanatory power and the scientific nature of its application.

[0078] S2: Based on hyperspectral imaging technology, spectral data of yellowfin seabream under different feeding conditions were collected, and the fat characteristics of yellowfin seabream were analyzed by extracting multi-scale spectral features.

[0079] In step S2, the logic for extracting multi-scale spectral features to analyze the fat characteristics of yellowfin sea bream is:

[0080] S201: Collect spectral data of yellowfin seabream samples;

[0081] S202: Perform multi-scale wavelet transform on the spectral data to extract multi-scale spectral features. The calculation method is:

[0082] ;

[0083] In the formula Indicates the wavelength is The spectral data at the scale level and pan position The wavelet coefficients of Indicates the wavelength is The spectral data, Indicates the wavelength is The spectral data at the scale level and pan position The wavelet basis function of 、 Represent the index of scale layer and translation position respectively, represents the wavelength, 、 Represent the maximum and minimum wavelengths respectively.

[0084] The multi-scale wavelet transform is used here because traditional hyperspectral analysis focuses on overall band or single-scale features, making it difficult to fully capture detailed variations at different scales (frequency components) within the spectrum. As a time-frequency localization analysis tool, the wavelet transform can simultaneously capture spectral details (high-frequency features) and trends (low-frequency features) at different scales, greatly enhancing the richness and layering of spectral features. Furthermore, the wavelet transform inherently suppresses noise, separating useful signals from noise components. This helps stabilize spectral data analysis and subsequent model training, thereby improving model robustness.

[0085] S203: Extracting wavelet coefficients of the spectral data at different wavelength scales through multi-scale analysis to enhance the recognition capability of fat features.

[0086] Specifically, multiscale spectral features refer to the collection of features derived from analyzing the original spectral signal at different scales (i.e., different resolutions or frequency levels). These features simultaneously reflect both the macroscopic trends (low-frequency components) and microscopic details (high-frequency components) of the spectral signal, helping to fully capture the biochemical information hidden in the spectrum. Wavelet coefficients are the specific mathematical expression of multiscale spectral features, reflecting the detailed representation of the spectral signal at different wavelength scales and are a direct result of multiscale analysis.

[0087] Because changes in fat content are usually reflected in subtle wavelength changes in spectral data, multi-scale analysis can effectively mine these wavelet coefficients, avoiding the problem of insufficient recognition ability of single-band or single-scale methods for complex biochemical components, thereby improving the sensitivity and recognition accuracy of fat characteristics.

[0088] In this step, multi-scale extraction can capture the macroscopic reflection of fat in the spectrum and deeply identify subtle fluctuations, significantly improving the information content and characterization capabilities of spectral data, laying a solid foundation for subsequent dynamic weight adjustment and meat quality prediction.

[0089] S3: A dynamic weight adjustment mechanism is constructed based on the fat characteristics of yellowfin sea bream, multi-scale spectral features are weighted, and a prediction model of spectral data and meat quality indicators is constructed based on the weighted spectral features.

[0090] In step S3, the logic for constructing a prediction model of spectral data and meat quality indicators based on weighted spectral features is as follows:

[0091] S301: Build a dynamic weight adjustment mechanism to introduce fat characteristic information into the spectral data weighting process to improve the key band identification capability;

[0092] S302: Iteratively calculating the dynamic weight of the spectral data according to the gradient update algorithm to obtain a weighted spectral feature;

[0093] S303: Based on the weighted spectral features, a prediction model of spectral data and meat quality indicators is established using RBF function.

[0094] The dynamic weight is calculated when iterating using the gradient update algorithm:

[0095] ;

[0096] In the formula Indicates the wavelength is The spectral data of The dynamic weight at the iteration, represents the learning rate, represents the error loss function.

[0097] The number of dynamic weight iterations can be set to stop after a specific number of iterations, or it can be set to stop after the loss function converges (that is, the decrease in the loss function is lower than a preset threshold for several consecutive iterations (such as 5 to 10 consecutive times). It can also be set to stop when the absolute value of the gradient of the loss function with respect to the weight is lower than a threshold (this indicates that the weight update amplitude is extremely small)... The specific method can be set based on the professional experience of the tester.

[0098] The expression of the prediction model of spectral data and meat quality indicators is:

[0099] ;

[0100] In the formula Indicates the prediction model The predicted value of meat quality index, Indicates the prediction model The meat quality index corresponds to The weight coefficients of the basis functions, subscript represents the index of the basis function in the prediction model, represents the total number of basis functions in the prediction model, represents the weighted spectral features, 、 、 Both represent the model parameters to be trained in the prediction model.

[0101] It can be understood that both the prediction model and the mapping model are constructed based on the RBF function. In essence, the spectral data of the yellowfin sea bream is collected through hyperspectral imaging, and the spectral features are extracted using multi-scale wavelet transform. Combined with the dynamic weight adjustment mechanism, the fat characteristics in the fish body are identified and quantified. This step completes the mapping from spectral signals to fat characteristics; the fat characteristics are then used to predict meat quality indicators such as muscle elasticity, adhesion and chewiness, thereby achieving the purpose of indirectly predicting meat quality through spectral data.

[0102] The expression of weighted spectral characteristics is:

[0103] ;

[0104] In the formula Indicates the wavelength is The dynamic weight after the spectral data iteration is completed, subscript The index representing the wavelength, Indicates the maximum value of the wavelength index.

[0105] Traditional spectral analysis often assigns equal weights to all bands, diluting some key bands and limiting their ability to fully exploit their discriminative power. The dynamic weight adjustment mechanism incorporates fat signature information and uses a gradient update algorithm to automatically optimize the weights of each band, highlighting important spectral bands and suppressing irrelevant or interfering bands, thereby improving the effectiveness of signal representation and discriminative power. Furthermore, the spectral features extracted by the multi-scale wavelet transform contain information at different frequencies and levels of detail. Combined with dynamic weighting, this mechanism can more comprehensively and accurately reflect the distribution and changing characteristics of fat in yellowfin sea bream.

[0106] The error loss function is constructed based on the square error between the predicted value of the meat quality index and the measured value of the meat quality index using the mapping model and the prediction model. Its expression is:

[0107] ;

[0108] In the formula Indicates the measured value of meat quality index, 、 Represent the weight coefficients of the mapping model and the prediction model respectively, both of which are greater than 0, and .

[0109] In this step, the spectral feature weights were optimized through gradient iteration, combined with multi-scale wavelet features and RBF nonlinear mapping models, to form a scientific, reasonable and efficient spectral data and meat quality index prediction algorithm. This not only reduces the interference of useless information and improves the accuracy and stability of meat quality prediction, but also combines the design of fat characteristics as intermediary variables to make the interpretation of spectral data in the hyperspectral spectrum more biologically meaningful, solves the problem of the "black box" model, improves the credibility of the prediction results, and can achieve fast, non-destructive and accurate intelligent grading and quality control of yellowfin sea bream meat.

[0110] S4: Construct a scoring function for meat quality indicators, calculate the comprehensive meat quality score of yellowfin sea bream based on the predicted meat quality indicators generated by the prediction model, and classify the meat quality grades according to the comprehensive meat quality score.

[0111] The expression of the scoring function in step S4 is:

[0112] ;

[0113] In the formula Indicates the comprehensive meat quality score, Indicates the The weight coefficients of meat quality indicators must meet the following requirements:

[0114] ;

[0115] when When the meat of yellowfin sea bream was considered to be of substandard quality;

[0116] when When the meat quality of yellowfin sea bream is considered to be qualified;

[0117] when When it comes to yellowfin sea bream, it is considered that the meat quality is good;

[0118] when When it comes to yellowfin sea bream, it is believed that the meat quality is excellent;

[0119] in 、 、 Both represent the preset scoring thresholds, and The scoring threshold here can be initially determined based on expert experience and subsequently adjusted based on measured data, for example:

[0120] We collected different yellowfin seabream and calculated their comprehensive meat quality scores. Experts then manually graded the meat quality of each yellowfin seabream into four categories: unqualified, qualified, good, and excellent.

[0121] Based on the sensory grading results, the numerical range of the corresponding comprehensive score is calculated, and a score distribution histogram or box plot is drawn. The reasonable interval split points are determined by analyzing the central trend (mean, median) of the scores of different grade groups and the overlap of intervals;

[0122] Using these critical sensory ratings as thresholds ensures clear classification boundaries and a low false positive rate. Furthermore, when determining the optimal split point, auxiliary methods such as the maximum inter-class variance method (Otsu's method), cluster analysis (such as K-means), or receiver operating characteristic (ROC) curve analysis can also be used. Since these methods are mature existing technologies, they will not be discussed in detail here.

[0123] In this step, the multidimensional meat quality indicators predicted based on spectral data in the previous step are converted into a unified comprehensive score, forming a complete "prediction-evaluation-grading" closed loop, which provides a standardized output form for hyperspectral detection results. This not only improves the comprehensiveness and scientific nature of meat quality evaluation, but also facilitates quality management and grading applications in actual production, helping to enhance the practicality and market adaptability of yellowfin sea bream meat quality detection solutions.

[0124] The yellowfin sea bream meat quality detection system based on hyperspectral imaging technology is used to perform the above-mentioned detection method, specifically including:

[0125] A sample detection module is used to detect the meat quality index and fat characteristics of the collected yellowfin sea bream samples;

[0126] Hyperspectral acquisition module, which is used to detect the spectral data of yellowfin sea bream samples and extract multi-scale spectral features;

[0127] Data analysis module, which is used to build a mapping model between meat quality indicators and fat characteristics and a prediction model between spectral data and meat quality indicators;

[0128] The meat quality scoring module is used to calculate the comprehensive meat quality score of yellowfin sea bream and classify the meat quality grades according to the comprehensive meat quality score.

[0129] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0130] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0131] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0132] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for detecting the quality of yellowfin sea bream meat based on hyperspectral imaging technology, characterized in that: The specific steps include: S1: A group feeding experiment was conducted on yellowfin seabream to collect the meat quality indicators and fat characteristics of yellowfin seabream under different feeding conditions, and a mapping model between meat quality indicators and fat characteristics was constructed; The logic for building the mapping model is: S101: Yellowfin sea bream fry are divided into several groups, fed periodically with feeds of different formulas, and used as samples for testing after they grow up; S102: Collect meat quality indicators of yellowfin seabream samples, including hardness, elasticity, cohesion, adhesiveness, chewiness, and resilience, as well as fat characteristics of yellowfin seabream samples, including liver-to-body ratio, organ-to-body ratio, and liver fat content; S103: constructing a fat feature vector based on the fat characteristics of the yellowfin sea bream sample, and using the RBF function to establish a mapping model between the meat quality index and the fat characteristics; S2: Using hyperspectral imaging technology to collect spectral data of yellowfin seabream under different feeding conditions, and analyzing the fat characteristics of yellowfin seabream by extracting multi-scale spectral features; The logic for extracting multi-scale spectral features to analyze the fat characteristics of yellowfin sea bream is as follows: S201: Collect spectral data of yellowfin seabream samples; S202: Perform multi-scale wavelet transform on the spectral data to extract multi-scale spectral features. The calculation method is: In the formula Indicates the wavelength is The spectral data at the scale level and pan position The wavelet coefficients of Indicates the wavelength is The spectral data, Indicates the wavelength is The spectral data at the scale level and pan position The wavelet basis function of 、 Represent the index of scale layer and translation position respectively, represents the wavelength, 、 Represent the maximum and minimum values ​​of wavelength respectively; S203: extracting wavelet coefficients of spectral data at different wavelength scales through multi-scale analysis to enhance the recognition capability of fat features; S3: A dynamic weight adjustment mechanism is constructed based on the fat characteristics of yellowfin sea bream to perform weighted processing on multi-scale spectral features. A prediction model of spectral data and meat quality indicators is constructed based on the weighted spectral features. The logic of constructing a prediction model of spectral data and meat quality indicators based on weighted spectral features is as follows: S301: Build a dynamic weight adjustment mechanism to introduce fat characteristic information into the spectral data weighting process to improve the key band identification capability; S302: Iteratively calculating the dynamic weight of the spectral data according to the gradient update algorithm to obtain a weighted spectral feature; S303: Based on the weighted spectral features and the mapping model between the meat quality index and the fat feature, a prediction model of the spectral data and the meat quality index is established using an RBF function; S4: Construct a scoring function for meat quality indicators, calculate the comprehensive meat quality score of yellowfin sea bream based on the predicted meat quality indicators generated by the prediction model, and classify the meat quality grades according to the comprehensive meat quality score.

2. The method for detecting the quality of yellowfin sea bream meat based on hyperspectral imaging technology according to claim 1, characterized in that: The mapping expression of the mapping model between the meat quality index and the fat characteristic is: In the formula Indicates the first The predicted value of meat quality index, Indicates the first The meat quality index corresponds to The weight coefficients of the basis functions, subscript 、 Represent the meat quality index and the index of the basis function in the mapping model, represents the total number of basis functions in the mapping model, represents the fat feature vector, 、 、 Both represent the model parameters to be trained in the mapping model.

3. The method for detecting the quality of yellowfin sea bream meat based on hyperspectral imaging technology according to claim 1, characterized in that: The dynamic weight is calculated as follows when iterating using the gradient update algorithm: In the formula Indicates the wavelength is The spectral data of The dynamic weight at the iteration, represents the learning rate, represents the error loss function; The expression of the prediction model of the spectral data and meat quality index is: In the formula Indicates the prediction model The predicted value of meat quality index, Indicates the prediction model The meat quality index corresponds to The weight coefficients of the basis functions, subscript represents the index of the basis function in the prediction model, represents the total number of basis functions in the prediction model, represents the weighted spectral features, 、 、 Both represent the model parameters to be trained in the prediction model.

4. The method for detecting the quality of yellowfin sea bream meat based on hyperspectral imaging technology according to claim 3, characterized in that: The expression of the weighted spectral feature is: In the formula Indicates the wavelength is The dynamic weight after the spectral data iteration is completed, subscript represents the index of wavelength, Indicates the maximum value of the wavelength index.

5. The method for detecting the quality of yellowfin sea bream meat based on hyperspectral imaging technology according to claim 3, characterized in that: The error loss function is constructed based on the square error between the predicted value of the meat quality index and the measured value of the meat quality index based on the mapping model and the prediction model, and its expression is: In the formula Indicates the measured value of meat quality index, 、 Represent the weight coefficients of the mapping model and the prediction model respectively, both of which are greater than 0, and .

6. The method for detecting the quality of yellowfin sea bream meat based on hyperspectral imaging technology according to claim 3, characterized in that: The expression of the score function in step S4 is: In the formula Indicates the comprehensive meat quality score, Indicates the The weight coefficients of meat quality indicators must meet the following requirements: when When the meat of yellowfin sea bream was considered to be of substandard quality; when When the meat quality of yellowfin sea bream is considered to be qualified; when When it comes to yellowfin sea bream, it is considered that the meat quality is good; when When it comes to yellowfin sea bream, it is believed that the meat quality is excellent; in 、 、 Both represent the preset scoring thresholds, and .

7. Yellowfin sea bream meat quality detection system based on hyperspectral imaging technology, characterized by: The detection system is used to perform the detection method according to any one of claims 1 to 6, specifically comprising: A sample detection module, wherein the sample detection module is used to detect the meat quality index and fat characteristics of the yellowfin sea bream sample; A hyperspectral acquisition module, which is used to detect spectral data of yellowfin sea bream samples and extract multi-scale spectral features; A data analysis module, the data analysis module is used to construct a mapping model between meat quality indicators and fat characteristics and a prediction model between spectral data and meat quality indicators; The meat quality scoring module is used to calculate the comprehensive meat quality score of the yellowfin sea bream and classify the meat quality grades according to the comprehensive meat quality score.

Images