Texture analyzer and neural network-based brittleness determination and grading prediction system and method for brittle tilapia mossambica

Abstract

The invention discloses a texture analyzer and neural network-based brittleness measurement and grading prediction system and method for crisp tilapia mossambica. The system comprises a data acquisition module, a model training module and a grading prediction module, the data acquisition module is used for sampling the crisp tilapia mossambica to obtain a to-be-detected sample, and performing brittleness detection and grade scoring on the to-be-detected sample to obtain an evaluation result of the to-be-detected sample; the model training module is used for constructing a grading model based on a BP neural network, preprocessing the evaluation result of the to-be-tested sample, and inputting the preprocessed evaluation result into the grading model for training to obtain a trained grading model; and the grading prediction module is used for randomly extracting a preset number of crisp tilapia samples from to-be-tested samples, and inputting the crisp tilapia samples into the trained grading model to obtain a tilapia brittleness measurement and grading prediction result.

Description

Technical Field

[0001] The invention relates to the technical field of aquatic product quality measurement and evaluation, and in particular to a crispness measurement and grading prediction system and method for crispy tilapia based on a texture analyzer and a neural network. Background Art

[0002] Tilapia is an important high-quality economic white-meat fish in the world, with high production and huge industry. According to the data of the "2023 Fisheries Statistical Yearbook", my country's tilapia production in 2022 was 1.7389 million tons. Due to the advantages of tilapia such as few intramuscular spines, high nutritional value, and delicious and tender meat, it is a key variety for my country's agricultural export and foreign exchange earnings. However, with the development of intensive and high-density tilapia farming, the quality of tilapia meat has declined, resulting in low profits for farmers in tilapia farming, which has reduced the enthusiasm of fishermen to farm tilapia. Crisp meat tilapia is an emerging specialty aquatic product in recent years, showing strong market vitality. Ordinary tilapia has firm meat and crisp taste after eating broad bean feed, and is called crisp meat tilapia. Tilapia can significantly increase sales prices and economic benefits and promote income growth through crisp farming. Compared with ordinary tilapia, crisp meat tilapia farming eliminates the earthy smell and has a broader market prospect. Crispy tilapia is also very suitable for making sauerkraut fish, boiled fish and other dishes that are popular among consumers. The entire industry chain of crispy tilapia will have greater development potential.

[0003] However, the crispness of crispy tilapia still depends on the sensory evaluation results of sensory reviewers. Although the method is simple, it requires a lot of manpower and time. Therefore, it is urgent to establish a time-saving and labor-saving method for crispy tilapia crispness determination and crispness prediction. Summary of the invention

[0004] In view of the above problems, the present invention provides a crispness determination and grading prediction system for crispy tilapia based on a texture analyzer and a neural network, the system comprising:

[0005] Data collection module, model training module and hierarchical prediction module;

[0006] The data acquisition module is used to sample the crispy tilapia to obtain a sample to be tested, perform crispness detection and grade scoring on the sample to be tested, and obtain an evaluation result of the sample to be tested;

[0007] The model training module is used to construct a grade classification model based on the BP neural network, and the evaluation results of the samples to be tested are pre-processed and input into the grade classification model for training to obtain a trained grade classification model;

[0008] The grading prediction module is used to randomly extract a preset number of crispy tilapia samples from the samples to be tested, input the crispy tilapia samples into the trained grading model, and obtain tilapia crispness measurement and grading prediction results.

[0009] Optionally, in the data acquisition module, the process of sampling the crispy tilapia to obtain the sample to be tested specifically includes:

[0010] The dorsal muscle of one side of the tilapia was taken to obtain a first sample to be tested with a length, width and thickness of 2 cm×2 cm×1 cm, and the dorsal muscle of the other side of the tilapia was taken to obtain a second sample to be tested with a length, width and thickness of 1 cm×1 cm×1 cm.

[0011] Optionally, in the data acquisition module, the process of performing brittleness detection and grade scoring on the samples to be tested to obtain evaluation results of the samples to be tested specifically includes: using a texture analyzer equipped with a cylindrical flat-bottom probe with a probe diameter of 6 mm, and using the TPA compression mode to perform hardness measurement on the first sample to be tested; using an expert scoring method to perform brittleness scoring and grading on the second sample to be tested; and obtaining evaluation results of the samples to be tested based on the hardness measurement results and the brittleness scoring and grading results.

[0012] Optionally, in the model training module, the input of the input layer is the evaluation result of the sample to be tested, and the number of neurons is set to 1; the number of neurons in the hidden layer is n; and the output of the output layer is the crispness scoring result.

[0013] Optionally, the calculation method for the number of neurons in the hidden layer is n is:

[0014]

[0015] Among them, n is the number of neurons in the input layer, m is the number of neurons in the output layer, and a is any integer between 1 and 10.

[0016] The present invention also discloses a method for determining the crispness and grading prediction of crispy tilapia based on a texture analyzer and a neural network, the method comprising:

[0017] Step S1, sampling the crispy tilapia to obtain a sample to be tested, performing a crispness test and a grade scoring on the sample to be tested, and obtaining an evaluation result of the sample to be tested;

[0018] Step S2, constructing a grading model based on a BP neural network, pre-processing the evaluation results of the sample to be tested and inputting them into the grading model for training to obtain a trained grading model;

[0019] Step S3, randomly extracting a preset number of crispy tilapia samples from the samples to be tested, inputting the crispy tilapia samples into the trained grading model, and obtaining tilapia crispness determination and grading prediction results.

[0020] Optionally, in step S1, the method of sampling the crispy meat tilapia to obtain the sample to be tested specifically includes: taking the dorsal muscle meat of one side of the tilapia to obtain a first sample to be tested with a length, width and thickness of 2cm×2cm×1cm, and taking the dorsal muscle meat of the other side of the tilapia to obtain a second sample to be tested with a length, width and thickness of 1cm×1cm×1cm.

[0021] Optionally, in step S1, the method for performing brittleness detection and grade scoring on the test sample to obtain the evaluation result of the test sample specifically includes: using a texture analyzer equipped with a cylindrical flat-bottom probe with a probe diameter of 6 mm, and using TPA compression mode to measure the hardness of the first test sample; using an expert scoring method to score and grade the brittleness of the second test sample; and obtaining the evaluation result of the test sample based on the hardness measurement result and the brittleness scoring and grading result.

[0022] Optionally, in step S2, in the grade classification model, the input of the input layer is the evaluation result of the sample to be tested, and the number of neurons is set to 1; the number of neurons in the hidden layer is n; and the output of the output layer is the crispness scoring result.

[0023] Optionally, the calculation method for the number of hidden layer neurons n is:

[0024]

[0025] Among them, n is the number of neurons in the input layer, m is the number of neurons in the output layer, and a is any integer between 1 and 10.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The crispness determination and grading prediction method of crispy tilapia based on texture analyzer and BP neural network in the present invention can realize rapid and scientific quantitative detection of the crispness of crispy tilapia meat, save time and labor costs, and provide technical support for the development of the crispy tilapia industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0029] Figure 1 It is a method step diagram of a method for determining and grading crispness of crispy tilapia based on a texture analyzer and a neural network according to an embodiment of the present invention;

[0030] Figure 2It is a BP neural network training convergence diagram of an embodiment of the present invention;

[0031] Figure 3 This is a comparison diagram of the scoring and fitting of tilapia trained in an embodiment of the present invention;

[0032] Figure 4 is a training scoring error graph of an embodiment of the present invention;

[0033] Figure 5 It is a comparison diagram of tilapia classification training in an embodiment of the present invention;

[0034] Figure 6 is a comparison chart of the scores of tilapia tested in an embodiment of the present invention;

[0035] Figure 7 is an error diagram of the score of tilapia tested in an embodiment of the present invention;

[0036] Figure 8 It is a comparison diagram of tilapia grading tested in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] Embodiment 1

[0039] A crispness determination and grading prediction system for crispy tilapia based on texture analyzer and neural network, such as Figure 1As shown, it specifically includes: a data acquisition module, a model training module and a grading prediction module; the data acquisition module is used to sample the crispy meat tilapia to obtain a sample to be tested, and the crispness detection and grade scoring of the sample to be tested are performed to obtain the evaluation result of the sample to be tested. The fresh crispy meat tilapia is placed in ice water and transported to the laboratory. The live fish is killed by hitting the head. The dorsal muscle of one side is taken and made into 5 samples of 2cm×2cm×1cm (length×width×thickness) to be tested, and placed on ice to measure the hardness value to obtain the first sample to be tested; the dorsal muscle of the other side of the fish is taken and made into 10 samples of 1cm×1cm×1cm (length×width×thickness) to be tested. After steaming in a steamer for 5 minutes, it is cooled to obtain the second sample to be tested, and wait for the sensory assessor to perform sensory evaluation. The first sample to be tested was measured using a texture analyzer equipped with a cylindrical flat-bottom probe with a probe diameter of 6 mm, using the TPA compression mode. The measurement conditions were: speed (before, during and after the test) was 1.5 mm / s; compression distance was 4 mm; trigger value was 5 g; the interval between two times was automatic, and the sample hardness was measured to obtain the sample hardness value. The hardness values ​​of the 5 samples were averaged to obtain the hardness value of each fish. The second sample to be tested was scored and rated using the expert scoring method: a sensory evaluation experiment was conducted in this laboratory. This laboratory has established a sensory evaluation room and a sensory evaluation team. In this experiment, a total of 10 trainers (20-40 years old, 5 males and 5 females) who had undergone sensory training were selected to evaluate the fish meat. All trainers had more than 2 years of experience in food quality and texture evaluation. The trainers scored the sensory crispness of tilapia meat according to Table 1.

[0040] Table 1

[0041] grade Fraction describe Level1 Premium 10～8 The fish meat is chewy, refreshing and has a strong firmness. Level 2 General 8~5 (not including 8) The fish meat is more elastic and firm Level3 Slightly Fragile 5~2 (not including 5 and 2) The fish meat has a chewy texture and a general firmness Level 4 is not fragile <＝2 The fish meat is easy to disperse and has poor firmness

[0042] The hardness of each fish meat, the scoring results and the rating results of each fish were statistically analyzed to obtain the evaluation results of the tested samples, as shown in Table 2.

[0043] Table 2

[0044]

[0045] The model training module is used to construct a grade classification model based on the BP neural network, and the evaluation results of the samples to be tested are preprocessed and input into the grade classification model for training to obtain a trained grade classification model.

[0046] The hierarchical classification model is constructed based on the BP neural network and is divided into three layers, including: input layer, hidden layer and output layer. The input layer inputs the texture and the number of neurons is 1. The empirical formula based on the number of neurons in the hidden layer is as follows:

[0047]

[0048] Among them, n1 is the number of neurons in the input layer, m1 is the number of neurons in the output layer, a ranges from 1 to 10, and the number of neurons in the hidden layer is determined to be 6. The output layer outputs the score, and the number of neurons in the output layer is 1. After the model is built, the evaluation results of the samples to be tested are preprocessed, and the input and output data are normalized using the mapminmax function to convert data of different dimensions to between positive and negative 1;

[0049]

[0050] The 'trainlm' function with fast convergence speed, high accuracy and insensitivity to initial weights is selected as the training function. The activation function from the input layer to the hidden layer neurons is the logarithmic S function 'logsig', and the activation function from the hidden layer to the output is the tangent s-type function 'tansig'. The maximum number of iterations, expected error, learning rate, and maximum number of losses are as follows: set the maximum number of iterations to 1000; set the learning rate to 0.01; set the learning target to 0.001; and set the maximum number of losses to 50. Training convergence is as follows Figure 2 As shown in Figure 2, the BP neural network training converges quickly, and the expected error is reached after the 14th iteration. The training fitting effect is shown in Figure 2. Figure 3 As shown in the figure, the score value of the training fitting simulation is very close to the actual score value. In the figure, the circle represents the BP neural network training simulation value, and the * represents the expected value. Figure 4 is the error graph, from Figure 4 It can be seen that the error is relatively small. Figure 5 This is a comparison chart of tilapia grading for training. As you can see from the chart, all are correct.

[0051] The grading prediction module is used to randomly extract a preset number of crispy tilapia samples from the samples to be tested, and input the crispy tilapia samples into the trained grading model to obtain the tilapia crispness determination and grading prediction results. After the model is trained, 10 samples are randomly selected for testing. The test results are shown in Table 3. The maximum score error is 0.1240, and the grade error is 0. Because the grade error is an interval evaluation, it has a certain fault tolerance.

[0052] Table 3

[0053]

[0054] From the Tilapia scoring error graph ( Figure 6 ) and the test tilapia score error diagram show that ( Figure 7 ), the artificial neural network simulation and actual value of the score are very close, and the BP neural network simulation is more accurate. From the test tilapia grading comparison chart ( Figure 8) It can be seen that the accuracy of BP neural network in assessing the grade of tilapia is 100%. The model is further evaluated from the goodness of fit R2, root mean square error RMSE, mean deviation MBE, mean absolute error MAE, and mean square error MSE. Goodness of fit refers to the degree of fit of the regression line to the observed value. The statistic that measures the goodness of fit is the coefficient of determination (also known as the coefficient of determination) R 2 . R 2 The maximum value is 1. 2 The closer the value of is to 1, the better the regression line fits the observed value; on the contrary, R 2 The smaller the value of is, the worse the regression line fits the observed value. 2 The calculation formula is as follows:

[0055]

[0056] In the formula, x i Represents the predicted value, y i Indicates the actual value, Represents the mean of the actual value. The root mean square error (RMSE) is the square root of the ratio of the sum of squares of the deviations between the predicted value and the actual value to the number of monitoring times. It is used to measure the deviation between the predicted value and the measured value. The formula is as follows:

[0057]

[0058] The mean deviation MBE is the arithmetic mean of the absolute values ​​of the deviations between the values ​​in the series and their arithmetic mean. The mean deviation is a measure of the degree of deviation of the values ​​in the series from their mean. The formula is as follows:

[0059]

[0060] The mean absolute error (MAE) is the average of the absolute values ​​of the deviations of all individual observations from the arithmetic mean. The mean absolute error can avoid the problem of errors canceling each other out, and thus can accurately reflect the size of the actual prediction error. The formula is as follows:

[0061]

[0062] The mean square error (MSE) is the square root of the ratio of the sum of squares of the deviations between the predicted value and the actual value to the number of monitoring times. It is used to measure the deviation between the predicted value and the measured value. The formula is as follows:

[0063]

[0064] The test result evaluation parameters are shown in Table 4:

[0065] Table 4

[0066] name <![CDATA[R 2 ]]> RMSE MBE MAE MSE Numeric 0.99 0.039 0.013 0.011 0.0015

[0067] Embodiment 2

[0068] The present invention provides a method for determining the crispness and grading prediction of crispy tilapia based on a texture analyzer and a neural network. Figure 1 As shown, the method comprises: step S1, sampling the crispy meat tilapia to obtain a sample to be tested, and performing a crispness test and a grade score on the sample to be tested to obtain an evaluation result of the sample to be tested. The fresh crispy meat tilapia is placed in ice water and transported to the laboratory, and the live fish is killed by hitting the head, and the dorsal muscle of one side of the fish is taken, and it is made into 5 samples of 2cm×2cm×1cm (length×width×thickness) to be tested, and placed on ice to determine the hardness value to obtain the first sample to be tested; the dorsal muscle of the other side of the fish is taken, and it is made into 10 samples of 1cm×1cm×1cm (length×width×thickness) to be tested, and steamed in a steamer for 5min, and then cooled to obtain a second sample to be tested, and wait for the sensory assessor to perform sensory evaluation.

[0069] The first sample to be tested was measured using a texture analyzer equipped with a cylindrical flat-bottom probe with a probe diameter of 6 mm, using the TPA compression mode. The measurement conditions were: speed (before, during and after the test) was 1.5 mm / s; compression distance was 4 mm; trigger value was 5 g; the interval between two times was automatic, and the sample hardness was measured to obtain the sample hardness value. The hardness values ​​of the 5 samples were averaged to obtain the hardness value of each fish. The second sample to be tested was scored and rated using the expert scoring method: a sensory evaluation experiment was conducted in this laboratory. This laboratory has established a sensory evaluation room and a sensory evaluation team. In this experiment, a total of 10 trainers (20-40 years old, 5 males and 5 females) who had undergone sensory training were selected to evaluate the fish meat. All trainers had more than 2 years of experience in food quality and texture evaluation. The trainers scored the sensory crispness of tilapia meat according to Table 5.

[0070] Table 5

[0071] grade Fraction describe Level1 Premium 10～8 The fish meat is chewy, refreshing and has a strong firmness. Level 2 General 8~5 (not including 8) The fish meat is more elastic and firm Level3 Slightly Fragile 5~2 (not including 5 and 2) The fish meat has a chewy texture and a general firmness Level 4 is not fragile <＝2 The fish meat is easy to disperse and has poor firmness

[0072] The hardness of each fish meat, the scoring results and the rating results of each fish were statistically analyzed to obtain the evaluation results of the tested samples, as shown in Table 6.

[0073] Table 6

[0074]

[0075] Step S2, constructing a classification model based on the BP neural network. In this embodiment, the threshold range of the classification is 0-10g. The evaluation results of the samples to be tested are preprocessed and input into the classification model for training to obtain a trained classification model.

[0076] The hierarchical classification model is constructed based on the BP neural network and is divided into three layers, including: input layer, hidden layer and output layer. The input layer inputs the texture and the number of neurons is 1. The empirical formula based on the number of neurons in the hidden layer is as follows:

[0077]

[0078] Among them, n1 is the number of neurons in the input layer, m1 is the number of neurons in the output layer, a ranges from 1 to 10, and the number of neurons in the hidden layer is determined to be 6. The output layer outputs the score, and the number of neurons in the output layer is 1. After the model is built, the evaluation results of the samples to be tested are preprocessed, and the input and output data are normalized using the mapminmax function to convert data of different dimensions to between positive and negative 1;

[0079]

[0080] The 'trainlm' function with fast convergence speed, high accuracy and insensitivity to initial weights is selected as the training function. The activation function from the input layer to the hidden layer neurons is the logarithmic S function 'logsig', and the activation function from the hidden layer to the output is the tangent s-type function 'tansig'. The maximum number of iterations, expected error, learning rate, and maximum number of losses are as follows: set the maximum number of iterations to 1000; set the learning rate to 0.01; set the learning target to 0.001; and set the maximum number of losses to 50. Training convergence is as follows Figure 2 As shown in Figure 2, the BP neural network training converges quickly, and the expected error is reached after the 14th iteration. The training fitting effect is shown in Figure 2. Figure 3 As shown in the figure, the score value of the training fitting simulation is very close to the actual score value. In the figure, the circle represents the BP neural network training simulation value, and the * represents the expected value. Figure 4 is the error graph, from Figure 4 It can be seen that the error is relatively small. Figure 5 This is a comparison chart of tilapia grading during training. It can be seen from the chart that all the grading is correct.

[0081] Step S3, randomly extracting a preset number of crispy tilapia samples from the samples to be tested, inputting the crispy tilapia samples into the trained grading model, and obtaining tilapia crispness determination and grading prediction results.

[0082] After training the model, 10 samples are randomly selected for testing. The test results are shown in Table 7. The maximum score error is 0.1240 and the grade error is 0. Because the grade error is an interval evaluation, it has a certain fault tolerance.

[0083] Table 7

[0084]

[0085] From the Tilapia scoring error graph ( Figure 6 ) and the test tilapia score error diagram show that ( Figure 7 ), the artificial neural network simulation and actual value of the score are very close, and the BP neural network simulation is more accurate. From the test tilapia grading comparison chart ( Figure 8 ) It can be seen that the accuracy of BP neural network in assessing the grade of tilapia is 100%. The model is further evaluated from the goodness of fit R2, root mean square error RMSE, mean deviation MBE, mean absolute error MAE, and mean square error MSE. Goodness of fit refers to the degree of fit of the regression line to the observed value. The statistic that measures the goodness of fit is the coefficient of determination (also known as the coefficient of determination) R 2 . R 2 The maximum value is 1. 2 The closer the value of is to 1, the better the regression line fits the observed value; on the contrary, R 2 The smaller the value of is, the worse the regression line fits the observed value. 2 The calculation formula is as follows:

[0086]

[0087] In the formula, x i Represents the predicted value, y i Indicates the actual value, Represents the mean of the actual value. The root mean square error (RMSE) is the square root of the ratio of the sum of squares of the deviations between the predicted value and the actual value to the number of monitoring times. It is used to measure the deviation between the predicted value and the measured value. The formula is as follows:

[0088]

[0089] The mean deviation MBE is the arithmetic mean of the absolute values ​​of the deviations between the values ​​in the series and their arithmetic mean. The mean deviation is a measure of the degree of deviation of the values ​​in the series from their mean. The formula is as follows:

[0090]

[0091] The mean absolute error (MAE) is the average of the absolute values ​​of the deviations of all individual observations from the arithmetic mean. The mean absolute error can avoid the problem of errors canceling each other out, and thus can accurately reflect the size of the actual prediction error. The formula is as follows:

[0092]

[0093] The mean square error (MSE) is the square root of the ratio of the sum of squares of the deviations between the predicted value and the actual value to the number of monitoring times. It is used to measure the deviation between the predicted value and the measured value. The formula is as follows:

[0094]

[0095] The test result evaluation parameters are shown in Table 8:

[0096] Table 8

[0097] name <![CDATA[R 2 ]]> RMSE MBE MAE MSE Numeric 0.99 0.039 0.013 0.011 0.0015

[0098] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A crispness determination and grading prediction system for crispy tilapia based on texture analyzer and neural network, characterized in that: The system includes: a data acquisition module, a model training module and a classification prediction module; The data acquisition module is used to sample the crispy tilapia to obtain a sample to be tested, perform crispness detection and grade scoring on the sample to be tested, and obtain an evaluation result of the sample to be tested; The model training module is used to construct a grade classification model based on the BP neural network, and the evaluation results of the samples to be tested are pre-processed and input into the grade classification model for training to obtain a trained grade classification model; The grading prediction module is used to randomly extract a preset number of crispy tilapia samples from the samples to be tested, input the crispy tilapia samples into the trained grading model, and obtain tilapia crispness measurement and grading prediction results.

2. The crispness determination and grading prediction system of crispy tilapia based on texture analyzer and neural network according to claim 1, characterized in that: In the data acquisition module, the process of sampling the crispy tilapia to obtain the sample to be tested specifically includes: The dorsal muscle of one side of the tilapia was taken to obtain a first sample to be tested with a length, width and thickness of 2 cm×2 cm×1 cm, and the dorsal muscle of the other side of the tilapia was taken to obtain a second sample to be tested with a length, width and thickness of 1 cm×1 cm×1 cm.

3. The crispness determination and classification prediction system of crispy tilapia based on texture analyzer and neural network according to claim 2, characterized in that: In the data acquisition module, the process of performing brittleness detection and grade scoring on the sample to be tested to obtain the evaluation result of the sample to be tested specifically includes: Using a texture analyzer equipped with a cylindrical flat-bottom probe with a probe diameter of 6 mm, the hardness of the first sample to be tested is measured in TPA compression mode; Using an expert scoring method to score and grade the brittleness of the second sample to be tested; The evaluation results of the tested samples are obtained based on the hardness measurement results and the brittleness scoring and rating results.

4. The crispness determination and grading prediction system of crispy tilapia based on texture analyzer and neural network according to claim 3, characterized in that: In the model training module, the input of the input layer is the evaluation result of the sample to be tested, and the number of neurons is set to 1; the number of neurons in the hidden layer is n; and the output of the output layer is the crispness scoring result.

5. The crispness determination and grading prediction system of crispy tilapia based on texture analyzer and neural network according to claim 4, characterized in that: The calculation method for the number of hidden layer neurons n is: Among them, n is the number of neurons in the input layer, m is the number of neurons in the output layer, and a is any integer between 1 and 10.

6. A method for determining the crispness and grading prediction of crispy tilapia based on a texture analyzer and a neural network, the method using the grading prediction system according to any one of claims 1 to 5, characterized in that: The method comprises: Step S1, sampling the crispy tilapia to obtain a sample to be tested, performing a crispness test and a grade scoring on the sample to be tested, and obtaining an evaluation result of the sample to be tested; Step S2, constructing a grading model based on a BP neural network, pre-processing the evaluation results of the sample to be tested and inputting them into the grading model for training to obtain a trained grading model; Step S3, randomly extracting a preset number of crispy tilapia samples from the samples to be tested, inputting the crispy tilapia samples into the trained grading model, and obtaining tilapia crispness determination and grading prediction results.

7. The method for determining the crispness and grading prediction of crispy tilapia based on a texture analyzer and a neural network according to claim 6, characterized in that: In step S1, the method of sampling the crispy tilapia to obtain the sample to be tested specifically includes: The dorsal muscle of one side of the tilapia was taken to obtain a first sample to be tested with a length, width and thickness of 2 cm×2 cm×1 cm, and the dorsal muscle of the other side of the tilapia was taken to obtain a second sample to be tested with a length, width and thickness of 1 cm×1 cm×1 cm.

8. The method for determining the crispness and grading prediction of crispy tilapia based on a texture analyzer and a neural network according to claim 7, characterized in that: In step S1, the method of performing brittleness detection and grade scoring on the sample to be tested to obtain the evaluation result of the sample to be tested specifically includes: Using a texture analyzer equipped with a cylindrical flat-bottom probe with a probe diameter of 6 mm, the hardness of the first sample to be tested is measured in TPA compression mode; Using an expert scoring method to score and grade the brittleness of the second sample to be tested; The evaluation results of the tested samples are obtained based on the hardness measurement results and the brittleness scoring and rating results.

9. The method for determining the crispness and grading prediction of crispy tilapia based on a texture analyzer and a neural network according to claim 8, characterized in that: In step S2, in the grading model, the input of the input layer is the evaluation result of the sample to be tested, and the number of neurons is set to 1; the number of neurons in the hidden layer is n; and the output of the output layer is the crispness scoring result.

10. The method for determining the crispness and grading prediction of crispy tilapia based on a texture analyzer and a neural network according to claim 9, characterized in that: The calculation method for the number of hidden layer neurons n is: Among them, n is the number of neurons in the input layer, m is the number of neurons in the output layer, and a is any integer between 1 and 10.

Images