Fish meat quality detection method and system
By determining the key indicators of fish quality and establishing a detection model, the problem of difficult evaluating the quality of brittle tilapia is solved, and the rapid and accurate detection of fish quality is achieved.
Patent Information
- Application Number
- CN202411611684.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-12
- Publication Date
- 2025-03-07
AI Technical Summary
The existing technology lacks effective references, making it difficult to accurately comprehensively evaluate the quality of brittle tilapia, affecting the consumption experience.
By determining the key indicators that affect fish meat quality, obtain fish meat samples to obtain target map data, and establish multiple indicator detection models based on these data and key indicators to achieve rapid non-destructive testing of fish meat quality.
It improves the accuracy and accuracy of fish quality inspection, realizes rapid non-destructive testing of fish quality, and solves the problem of inconsistent quality.
Smart Images
Figure CN119579963A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to a quality detection technology, and in particular to a fish quality detection method and system, belonging to the technical field of quality detection. Background Art
[0002] Crisp tilapia is a specialty freshwater aquaculture product that has developed rapidly in recent years. It has the advantages of crispy and elastic meat and long cooking without becoming mushy, and is widely welcomed by consumers. However, due to the limitations of crisp tilapia aquaculture technology, the crispness and quality of crisp tilapia aquacultured in the same batch are currently different, affecting consumer experience.
[0003] Since there is no fixed meat quality evaluation index system and lack of effective reference, it is difficult to accurately and comprehensively evaluate the quality of brittle tilapia. Therefore, establishing a rapid and unified quantitative detection method is a problem that needs to be solved. Summary of the invention
[0004] In view of this, the present application provides a fish quality detection method and system to solve or alleviate the technical problems existing in the prior art and at least provide a beneficial option.
[0005] The technical solution of the embodiment of the present application is implemented as follows: A method for detecting fish quality comprises the following steps:
[0006] S100: Identify key indicators that affect fish quality.
[0007] S200: Obtain a first fish meat sample to obtain target atlas data according to the first fish meat sample, and establish multiple indicator detection models based on the target atlas data and the key indicators.
[0008] S300: Obtain fish meat to be tested, and obtain the fish meat quality of the fish meat to be tested according to various indicator detection models and the fish meat to be tested.
[0009] Further preferably: the S100 includes:
[0010] A second fish meat sample is obtained, and fish meat quality index data at multiple brittleness times are measured based on the second fish meat sample, and significance analysis and correlation analysis are performed according to the measurement results.
[0011] Obtain key indicators affecting fish quality based on the analysis results.
[0012] Further preferably, the step of acquiring target atlas data in S200 includes:
[0013] A first fish meat sample is obtained, and information of the first fish meat sample is collected through a hyperspectral camera to obtain initial atlas data.
[0014] Initial spectrum data and initial image data are obtained respectively according to the initial spectrum data.
[0015] Further preferably, the step of acquiring target atlas data in S200 further includes:
[0016] The initial spectral data is preprocessed to obtain target spectral data.
[0017] Perform principal component analysis on the initial image data to obtain target image data.
[0018] The target spectrum data and the target image data are combined to obtain target spectrum data.
[0019] Further preferably, the step of establishing multiple indicator detection models in S200 includes:
[0020] Characteristic wavelengths and texture features are extracted according to the target atlas data, so as to establish a plurality of indicator detection models based on the characteristic wavelengths, texture features and the key indicators.
[0021] Further preferably, the establishing of multiple indicator detection models based on the characteristic wavelength, texture feature and the key indicator comprises:
[0022] The characteristic wavelength and texture features are integrated, and an initial detection model of a support vector machine, a feedforward neural network and a subspace discriminant model is established based on the key indicators to obtain an initial detection model with the best detection accuracy as the indicator detection model of the current key indicators.
[0023] Further preferably, the preprocessing at least includes convolution smoothing, multivariate scatter correction, standard normal variate transformation, baseline correction and normalization.
[0024] Further preferably, it also includes: at least using a regression coefficient method, a continuous projection algorithm and a competitive adaptive reweighting algorithm to extract characteristic wavelengths.
[0025] Further preferably, it also includes: at least using gray level co-occurrence matrix to extract texture features.
[0026] Based on the same concept, the present application also provides a fish quality detection system, comprising:
[0027] Determination module: used to determine the key indicators affecting fish quality.
[0028] Establishment module: used to obtain a first fish meat sample, to obtain target atlas data according to the first fish meat sample, and to establish multiple indicator detection models based on the target atlas data and the key indicators.
[0029] Detection module: used for obtaining the fish meat to be detected, and obtaining the fish meat quality of the fish meat to be detected according to various indicator detection models and the fish meat to be detected.
[0030] The embodiment of the present application adopts the above technical solution, which has the following advantages:
[0031] The use of atlas data to detect fish meat quality improves the accuracy and precision of model detection and realizes rapid and non-destructive detection of fish meat quality.
[0032] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present application will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0034] Figure 1 This is a flow chart of the fish quality detection method described in this application;
[0035] Figure 2 This is a framework diagram of the fish quality detection system described in this application. DETAILED DESCRIPTION
[0036] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present application. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.
[0037] The embodiments of the present application are described in detail below with reference to the accompanying drawings.
[0038] like Figure 1 As shown, the embodiment of the present application provides a method for detecting fish quality, comprising the following steps:
[0039] S100: Identify key indicators that affect fish quality.
[0040] In this example, specifically: S100 includes:
[0041] A second fish meat sample is obtained, and fish meat quality index data at multiple brittleness times are measured based on the second fish meat sample, and significance analysis and correlation analysis are performed according to the measurement results.
[0042] Obtain key indicators affecting fish quality based on the analysis results.
[0043] In some embodiments, the present application is applied to the quality detection of tilapia; crispy tilapia at different brittle stages are collected, and hyperspectral data are collected in the state of fish fillets, and spectral and image data of 400-1000nm and 900-1700nm are collected respectively. After the data is collected, the data is corrected using ENVI+IDL 4.8 software, and a region of interest (ROI) is divided from the corrected data to extract the information required for analysis.
[0044] Among them, crispy tilapia cultured for 0 days (not brittle), 20 days, 40 days, 60 days and 90 days were selected as experimental objects. Among them, the samples that were brittle for 60 days were finished products that could be shipped out of the pond, while the samples that were 90 days were products that were put on the market after a normal brittle cycle.
[0045] Furthermore, the crispy tilapia was processed immediately after the atlas data was collected. The samples were cut, and the texture analyzer TPA mode was used to obtain the comprehensive results of the texture characteristics, and the hardness, cohesion, elasticity, adhesiveness, chewiness and recovery of the samples were obtained. In terms of chemical composition detection, the moisture content of tilapia at different brittle stages was obtained by freeze drying after homogenization, and the protein, fat and collagen content of the freeze-dried samples were further determined.
[0046] After obtaining the physical and chemical data of the samples, the sample data were statistically analyzed. First, the significance of each physical and chemical index was analyzed with the crisping time as the group, and the indexes with phased changes during the crisping period were obtained. Then, the physical and chemical indexes of all samples were taken as separate variables, and the correlation analysis was used to explore the correlation of different variables. Finally, principal component analysis combined with factor analysis was used to further confirm the key indicators affecting the quality of crispy tilapia.
[0047] Finally, key indicators such as moisture content, fat content, hardness value, cohesion, chewiness and recovery were determined.
[0048] S200: Obtain a first fish meat sample to obtain target atlas data according to the first fish meat sample, and establish multiple indicator detection models based on the target atlas data and the key indicators.
[0049] In this example, specifically: the step of obtaining target atlas data in S200 includes:
[0050] A first fish meat sample is obtained, and information of the first fish meat sample is collected through a hyperspectral camera to obtain initial atlas data.
[0051] Initial spectrum data and initial image data are obtained respectively according to the initial spectrum data.
[0052] In this example, specifically: the step of acquiring target atlas data in S200 further includes:
[0053] The initial spectral data are preprocessed to obtain target spectral data.
[0054] Wherein, the preprocessing at least includes convolution smoothing, multivariate scatter correction, standard normal variable transformation, baseline correction and normalization.
[0055] Specifically: Convolution smoothing (SG smoothing) is a weighted filtering of spectral data, but its weighted weight is obtained by least square fitting of a given high-order polynomial; compared with other smoothing methods, SG smoothing can more effectively retain the signal change information. The SG smoothing operation was performed in The Unscaramber X 10.4 software, where the parameters Polynomial order was set to 2 and Smoothing points were set to 4.
[0056] Multivariate scatter correction (MSC) is one of the commonly used algorithms for hyperspectral data preprocessing. MSC can effectively eliminate the spectral differences caused by different scattering levels. This method corrects the baseline shift and offset of spectral data through ideal spectra, thereby enhancing the correlation between spectra and data. MSC is operated in The Unscaramber X 10.4 software, in which Full MSC is selected to establish a complete MSC model.
[0057] The standard normal variate transform (SNV) is mainly used to eliminate the effects of solid particles, optical path changes and surface scattering on near-infrared spectra. The function of SNV is similar to that of MSC. The SNV operation was performed in TheUnscaramber X 10.4 software with default settings.
[0058] Furthermore, in reality, since environmental factors cannot be completely controlled and the equipment is easily affected by humidity and temperature, the spectral data components are very messy. In addition to some characteristic variables, the background baseline and noise in the spectrum will affect the data analysis. Therefore, before building the model, it is necessary to filter out the background baseline from the high-dimensional, non-zero variable data, and finally select the characteristic wavelength corresponding to the predicted object based on the spectral data. The baseline correction operation is performed in The UnscaramberX 10.4 software, and the method selects Baseline offset.
[0059] Normalization is a dimensionless processing method that converts the absolute value of the physical system into a relative value relationship. It is an effective way to simplify calculations and reduce the value of a quantity. The Normalizer operation maps the data to the range of 0 to 1 or -1 to 1 for processing, which is more convenient and fast, and converts dimensional expressions into dimensionless expressions, so that indicators of different units or magnitudes can be compared and weighted. Normalizing the data can eliminate the impact of the data dimension and make the data indicators comparable. The result of normalization is to limit the processed data to a certain range, thereby eliminating the adverse effects caused by singular sample data. The normalization operation was performed in The Unscaramber X 10.4 software, and the method selected was Area normalization.
[0060] Perform principal component analysis on the initial image data to obtain target image data.
[0061] The target spectrum data and the target image data are combined to obtain target spectrum data.
[0062] In this example, specifically: the establishment of multiple indicator detection models in S200 includes:
[0063] Characteristic wavelengths and texture features are extracted according to the target atlas data, so as to establish a plurality of indicator detection models based on the characteristic wavelengths, texture features and the key indicators.
[0064] Among them, at least a regression coefficient method, a continuous projection algorithm and a competitive adaptive reweighting algorithm are used to extract characteristic wavelengths.
[0065] Specifically: The regression coefficient is a parameter that indicates the influence of the independent variable on the dependent variable in the regression equation. The larger the regression coefficient, the greater the influence of the independent variable on the dependent variable. A positive regression coefficient indicates that the dependent variable increases as the independent variable increases, while a negative regression coefficient indicates the opposite. Combined with the PLSR model to extract features, principal component analysis and 5-fold cross validation were performed.
[0066] The Successive Projections Algorithm (SPA) is a forward variable selection algorithm that minimizes the collinearity of the vector space. Its advantage is that it can extract several characteristic wavelengths of the entire band, eliminate redundant information in the original spectral matrix, and can be used to screen spectral characteristic wavelengths. The number of variables extracted in the parameter setting is controlled to be a maximum of 30 and a minimum of 2, and a 5-fold cross validation is performed.
[0067] The competitive adaptive reweighting algorithm (CARS) selects the wavelength points with large absolute values of regression coefficients in the PLS model through the adaptive reweighting sampling (ARS) technology, removes the points with small weights, and uses interactive validation to select the subset with the lowest RMSE value, which can effectively find the optimal variable combination. Combined with the PLSR model to extract features, perform principal component analysis and 5-fold cross validation, and iterate 100 times.
[0068] And, at least a gray level co-occurrence matrix is used to extract texture features.
[0069] Among them, the principal component grayscale image (i.e., the target image data) is obtained by performing principal component analysis on the initial image data; the gray level co-occurrence matrix (GLCM) values are extracted from the principal component grayscale image, which are four values of contrast, correlation, energy, and inverse difference, and the image texture feature values are obtained according to the principal component image.
[0070] In this example, specifically: the multiple indicator detection models are established based on the characteristic wavelength, texture features and key indicators, including:
[0071] The characteristic wavelength and texture features are integrated, and an initial detection model of a support vector machine, a feedforward neural network and a subspace discriminant model is established based on the key indicators to obtain an initial detection model with the best detection accuracy as the indicator detection model of the current key indicators.
[0072] Among them, the obtained preprocessed data and the selected characteristic variables are combined with key indicators to establish the prediction model, and the data is divided into training set and test set with a division ratio of 3: 1. The prediction model is established using partial least squares (PLSR), support vector machine (SVM / SVR), subspace discrimination, feedforward neural network and other algorithms, and the r2 and root mean square error (RMSE) of different models are compared to finally obtain the best model.
[0073] S300: Obtain fish meat to be tested, and obtain the fish meat quality of the fish meat to be tested according to various indicator detection models and the fish meat to be tested.
[0074] Based on the same concept, Figure 2 As shown, the present application also provides a fish quality detection system, comprising:
[0075] Determination module: used to determine the key indicators affecting fish quality.
[0076] Establishment module: used to obtain a first fish meat sample, to obtain target atlas data according to the first fish meat sample, and to establish multiple indicator detection models based on the target atlas data and the key indicators.
[0077] Detection module: used for obtaining the fish meat to be detected, and obtaining the fish meat quality of the fish meat to be detected according to various indicator detection models and the fish meat to be detected.
[0078] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. A method for detecting fish quality, characterized in that: The following steps are involved: S100: Determine the key indicators that affect fish quality; S200: Obtaining a first fish meat sample to obtain target atlas data according to the first fish meat sample, and establishing a plurality of indicator detection models based on the target atlas data and the key indicators; S300: Obtain fish meat to be tested, and obtain the fish meat quality of the fish meat to be tested according to various indicator detection models and the fish meat to be tested.
2. The fish meat quality detection method according to claim 1, characterized in that: The S100 includes: Obtaining a second fish meat sample, measuring fish meat quality index data at multiple brittleness times based on the second fish meat sample, and performing significance analysis and correlation analysis according to the measurement results; Obtain key indicators affecting fish quality based on the analysis results.
3. The fish meat quality detection method according to claim 2, characterized in that: The step of acquiring target atlas data in S200 includes: Obtaining a first fish meat sample, and collecting information on the first fish meat sample using a hyperspectral camera to obtain initial atlas data; Initial spectrum data and initial image data are obtained respectively according to the initial spectrum data.
4. The fish meat quality detection method according to claim 3, characterized in that: The step of acquiring target atlas data in S200 further includes: Preprocessing the initial spectral data to obtain target spectral data; Performing principal component analysis on the initial image data to obtain target image data; The target spectrum data and the target image data are combined to obtain target spectrum data.
5. The fish meat quality detection method according to claim 4, characterized in that: The step of establishing multiple indicator detection models in S200 includes: Characteristic wavelengths and texture features are extracted according to the target atlas data, so as to establish a plurality of indicator detection models based on the characteristic wavelengths, texture features and the key indicators.
6. The fish meat quality detection method according to claim 5, characterized in that: The establishing of multiple indicator detection models based on the characteristic wavelength, texture feature and key indicator includes: The characteristic wavelength and texture features are integrated, and an initial detection model of a support vector machine, a feedforward neural network and a subspace discriminant model is established based on the key indicators to obtain an initial detection model with the best detection accuracy as the indicator detection model of the current key indicators.
7. The fish meat quality detection method according to claim 4, characterized in that: The preprocessing includes at least convolution smoothing, multivariate scatter correction, standard normal variate transformation, baseline correction and normalization.
8. The fish quality detection method according to claim 5, characterized in that: Also includes: At least a regression coefficient method, a continuous projection algorithm and a competitive adaptive reweighting algorithm are used to extract characteristic wavelengths.
9. The fish meat quality detection method according to claim 5, characterized in that: Also includes: At least a gray level co-occurrence matrix is used to extract texture features.
10. A system using the fish quality detection method according to any one of claims 1 to 9, characterized in that: include: Determination module: used to determine the key indicators affecting fish quality; Establishment module: used for obtaining a first fish meat sample, obtaining target atlas data according to the first fish meat sample, and establishing a plurality of indicator detection models based on the target atlas data and the key indicators; Detection module: used for obtaining the fish meat to be detected, and obtaining the fish meat quality of the fish meat to be detected according to various indicator detection models and the fish meat to be detected.