Marine food management method based on big data

Abstract

The invention discloses a marine food management method based on big data. The method comprises the steps of obtaining marine food management original data, refining the original data, constructing a marine food detection model, optimizing the detection model and performing marine food double-sided detection management. The invention relates to the technical field of marine food quality management data processing, in particular to a marine food management method based on big data, according to the scheme, a double-sided detection method is innovatively introduced, potential impurities on double sides are effectively captured, and the accuracy of marine food detection is improved; a marine food detection activation function and a value grading model loss function are designed, so that the limitation in processing complex characteristics and high-value food detection is solved, and the detection precision and intelligent management level of the marine food are improved; an algorithm for obtaining detection model parameters is improved by calculating and searching individual difference distances and a collaborative exploration strategy, optimal parameter values are obtained, model hyper-parameters are adjusted, and the accuracy of model output results is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of marine food quality management data processing, and in particular to a marine food management method based on big data. Background Art

[0002] With the rapid development of information technology, data processing and intelligent management technologies have been widely used in all walks of life. In the marine food industry, the instability of food quality is the main problem faced in management. Traditional marine food management methods usually rely on manual monitoring and experience judgment, which makes it difficult to cope with quality fluctuations and changes in production efficiency in large-scale marine food production. It not only increases management risks, but also affects the timeliness and accuracy of decision-making. Therefore, a marine food management method based on big data came into being. This method comprehensively collects various types of data in the production and management of marine food, and combines intelligent analysis technology to provide real-time and accurate decision support. Through this method, every link in the marine food production process can be accurately monitored and efficiently optimized, thereby significantly improving production efficiency, ensuring the stability of product quality, and reducing the risk of producing unqualified products. It enables managers to make scientific quality control decisions based on real-time data, improve the transparency and controllability of the overall production process, ensure the efficient and stable development of the marine food industry, and promote the marine food industry to develop in a more refined and intelligent direction.

[0003] However, traditional marine food management has the technical problem that when the marine food is large, single-sided detection may not be able to fully detect the entire surface of the food. The existing activation function in the marine food detection model is limited and the loss function is insufficient, which leads to low accuracy of marine food detection. The existing detection model has inappropriate built-in parameter settings and the model optimization algorithm has weak ability to obtain the global optimal solution, which leads to inaccurate final model output results. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a marine food management method based on big data. In view of the technical problem that when the marine food is large, single-sided detection may not be able to fully detect the entire surface of the food in traditional marine food management, this solution innovatively introduces a double-sided detection method. By independently detecting the front and back of the marine food, it is ensured that each side of the food is thoroughly inspected, and different potential impurities on the front and back are effectively captured, thereby reducing the risk of missed detection, significantly improving the accuracy and efficiency of marine food detection, and thus improving the quality management level of marine food; in view of the limitations of the activation function and the insufficient loss function in the existing marine food detection model, which leads to low accuracy of marine food detection, this solution innovatively designs a marine food detection activation function and a value grading model loss function, which can more effectively handle the complex feature interactions and nonlinear relationships of multiple impurities in marine food, and improve the control of food quality. The sensitivity of key features is taken into account. By introducing a value grading loss function, foods of different values ​​are treated differently, the detection accuracy of high-value foods is improved, the quality management of high-value foods is ensured, and the limitations of traditional methods in dealing with complex features and high-value food detection are solved, thereby significantly improving the detection accuracy, quality management efficiency and intelligent management level of marine foods; for the existing technical problems applicable to detection models, there are inappropriate built-in parameter settings, and the model optimization algorithm has a weak ability to obtain the global optimal solution, which leads to inaccurate final model output results. This solution can adjust the search intensity at different stages by calculating the search individual difference distance and collaborative exploration strategy, ensuring sufficient global exploration capabilities in the early stage and rapid convergence in the later stage, solving the technical problems of local optimal traps and insufficient exploration and development balance of the optimization algorithm, and can obtain the optimal parameter values ​​and adjust the model hyperparameters, significantly improving the accuracy of the model output results, thereby improving the intelligent level of marine food management.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides a marine food management method based on big data, which comprises the following steps:

[0006] Step S1: Obtaining raw data on marine food management;

[0007] Step S2: Refining the original data;

[0008] Step S3: constructing a marine food detection model;

[0009] Step S4: Detection model optimization;

[0010] Step S5: Double-sided detection management of marine food.

[0011] Furthermore, in step S1, the obtaining of marine food management original data is specifically to obtain the marine food management original data by collecting it from a large marine food processing factory management system; the marine food management original data includes marine food information data and marine food production line detection data; the marine food production line detection data includes marine food front detection data, marine food back detection data and production line environment detection data.

[0012] Furthermore, in step S2, the refined raw data is used to optimize the raw data for marine food management, specifically, data cleaning optimization, data standardization optimization, data enhancement optimization and data feature selection optimization are performed on the raw data for marine food management to obtain the optimized data for marine food management, and the training data and the data to be tested are divided; the data cleaning optimization is used to eliminate invalid and inaccurate data, specifically, data missing value processing, data outlier processing and data duplicate value deletion processing are performed on the data; the data standardization optimization is specifically to standardize the data using the maximum and minimum normalization method; the data enhancement optimization is used to generate diversified data, specifically, to generate diversified data samples by generating adversarial networks, and the data feature selection optimization is specifically to perform feature selection on the data based on a tree model to screen out features related to the qualification of marine food products.

[0013] Further, in step S3, the marine food detection model is constructed to establish a marine food detection model, analyze the quality of marine food, and obtain a trained marine food detection model; specifically, the following steps are included:

[0014] Step S31: extracting the local input feature map of marine food, specifically including the following steps:

[0015] Step S311: perform a convolution operation; the formula used is expressed as follows:

[0016] ;

[0017] In the formula, represents the feature map of layer l, Respectively represent the row and column coordinates of the position in the feature map, represents the ReLU activation function, represents the batch normalization operation, n and m represent the height and width of the convolution kernel, Indicates the convolution kernel in The feature map of the layer, Indicates The weight matrix of the convolutional layer, Representative The bias parameters of the convolutional layer;

[0018] Step S312: Perform a pooling operation to reduce the output dimension of the feature; the formula used is as follows:

[0019] ;

[0020] In the formula, Indicates the position of the feature map of layer l after the pooling operation , represents the area of ​​the pooling window, Represents the weight of each position in the pooling window;

[0021] Step S32: Gather key impurity features to focus on areas that are critical to quality judgment. Specifically, the feature map is weighted through a spatial attention mechanism. The formula used is as follows:

[0022] ;

[0023] ;

[0024] In the formula, Represents the feature map location The attention weight, represents the Tanh activation function, represents the weighted feature map, represents element-wise multiplication, Represents the coordinates of other locations in the feature map;

[0025] Step S33: Design a marine food detection activation function to handle complex feature interactions in marine food double-sided detection; the formula used is as follows:

[0026] ;

[0027] In the formula, represents the marine food detection activation function, represents the input data of the activation function, Represents the scaling parameter that controls the activation function in the negative region. represents the Sigmoid activation function, represents the exponential function;

[0028] Step S34: Obtain the marine food detection output result, the formula used is as follows:

[0029] ;

[0030] In the formula, represents the cth marine food detection output result, represents the output weight matrix, Represents the output bias parameter, Represents the Dropout operation;

[0031] Step S35: Design a value classification model loss function, which is used to design a weighted loss function according to the value of different marine foods. The formula used is as follows:

[0032] ;

[0033] ;

[0034] ;

[0035] In the formula, represents the loss function value of common marine food, represents the loss function value of high-value marine food, Represents the total loss function value of the detection model, represents the actual result of the cth marine food test, Indicates whether marine food c is a high-value marine food. If =1, then marine food c is a high-value marine food, represents the loss weight of conventional marine food, represents the loss weight of high-value food, and C represents the total amount of marine food;

[0036] Step S36: Conduct model training, specifically, use the training data as the training data of the model, calculate the loss through the value grading model loss function, use the back propagation algorithm to optimize the weight of the model, and gradually reduce the error. After multiple rounds of iterations, the model continuously adjusts the weight parameters and optimizes the loss function to minimize the difference between the model prediction result and the true value. When the error reaches a preset threshold, the training process stops, thereby obtaining a trained marine food detection model.

[0037] Further, in step S4, the detection model optimization is specifically to obtain the optimal hyperparameter combination of the trained marine food detection model by improving the optimization algorithm to obtain the optimized marine food detection model, including the following steps:

[0038] Step S41: Initialize the search individuals, specifically randomly generate initial population individuals in the solution space; the formula used is as follows:

[0039] ;

[0040] In the formula, represents the initial position of the i-th population individual, and represent the lower and upper limits of the population position, respectively. Represents a random number in the range [0,1];

[0041] Step S42: Calculate the individual fitness value, specifically calculate the individual fitness value f in the population i ; The performance of the trained marine food detection model established based on the individual position is used as the individual fitness value, and the individuals are sorted from best to worst according to the fitness value to obtain the individual position with the highest global fitness value ;

[0042] Step S43: Calculate the difference distance of the searched individuals, specifically, obtain the distance between the current individual and the individual with the worst fitness value, and the formula used is as follows:

[0043] ;

[0044] In the formula, represents the difference distance of the i-th individual, t represents the number of iterations, represents a random number in the range [0,1]. Indicates the corresponding position of the worst fitness in the current iteration, represents the position of the i-th individual in the t-th generation population, Indicates the maximum number of iterations;

[0045] Step S44: updating the positions of the individuals in the population, specifically including the following steps:

[0046] Step S441: collaborative exploration updates individual positions, specifically updating positions according to collaborative exploration strategies, using the following formula:

[0047] ;

[0048] In the formula, represents the position of the i-th individual in the t+1th generation, Represents a random number in the range [0,1]; represents the average value of the positions of three randomly selected individuals in the population, represents the average value of the positions of two randomly selected individuals in the population;

[0049] Step S442: Randomly update individual positions, specifically update positions according to a variation diversification strategy, and the formula used is as follows:

[0050] ;

[0051] In the formula, represents a random number in the range [0,1]. represents the position of a randomly selected individual in the population;

[0052] Step S443: Update the individual position according to the iteration value, specifically update the position according to the dynamic behavior simulation strategy, and the formula used is as follows:

[0053] ;

[0054] In the formula, is the gravitational acceleration constant, Represents a random number in the range [0,1];

[0055] Step S45: The worst search individual position randomly jumps, and the formula used is as follows:

[0056] ;

[0057] In the formula, represents a random number in the range [0,0.45], represents a random number in the range [0.45,1], represents the difference between the positions of two randomly selected individuals in the population, represents a binary random variable;

[0058] Step S46: Individuals with low fitness are mutated, specifically, the fitness value of each individual is recalculated, the individuals are sorted from best to worst according to the fitness value, and 5%N individuals with low fitness are selected for individual mutation. The formula used is as follows:

[0059] ;

[0060] In the formula, represents individuals with low fitness. represents the optimal fitness of the current population, represents the worst fitness of the current population, and N represents the total number of search individuals;

[0061] Step S47: Obtain the optimal individual position, specifically when the individual fitness value f i When the fitness threshold is higher than the maximum number of iterations, the search is terminated and the individual global optimal position is obtained, and the optimal hyperparameter combination of the marine food detection model is obtained;

[0062] Step S48: obtaining an optimized marine food detection model, specifically adjusting the hyperparameters of the trained marine food detection model according to the optimal hyperparameter combination of the marine food detection model to obtain the optimized marine food detection model.

[0063] Further, in step S5, the marine food double-sided detection management is used for intelligent management of marine food, specifically realizing intelligent management of marine food through the results of the marine food double-sided detection; comprising the following steps:

[0064] Step S51: marine food front detection: specifically, the marine food information data and the marine food front detection data in the data to be detected are used as input data of the optimized marine food detection model, and the optimized marine food detection model is used to detect the front of the marine food to obtain the output result of the marine food front detection;

[0065] Step S52: marine food back side detection: specifically, using the marine food information data and the marine food back side detection data in the data to be detected as input data of the optimized marine food detection model, and using the optimized marine food detection model to detect the back side of the marine food to obtain the output result of the marine food back side detection;

[0066] Step S53: Intelligent management of marine food. Based on the output results of the front side detection of marine food and the output results of the back side detection of marine food, the safety of marine food is comprehensively monitored and managed. If both output results are qualified, the marine food is sent to the freezing and refrigeration system. If the detection result of any side is unqualified, the unqualified marine food is discarded to ensure that unqualified products do not enter the next production link, thereby realizing efficient and accurate intelligent management of marine food.

[0067] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0068] (1) In view of the technical problem that single-sided inspection may not be able to fully detect the entire surface of marine food when the marine food is large, this solution innovatively introduces a double-sided inspection method. By independently inspecting the front and back of the marine food, it ensures that each side of the food is thoroughly inspected, effectively captures different potential impurities on the front and back, reduces the risk of missed inspections, and significantly improves the accuracy and efficiency of marine food inspections, thereby improving the quality management level of marine food.

[0069] (2) In response to the technical problems of the limitations of activation functions and insufficient loss functions in existing models applicable to marine food detection, which lead to low accuracy in marine food detection, this scheme innovatively designs marine food detection activation functions and value grading model loss functions, which can more effectively handle the complex feature interactions and nonlinear relationships of various impurities in marine foods, improve the sensitivity to key features of food quality, and introduce value grading loss functions to differentiate foods of different values, thereby improving the detection accuracy of high-value foods and ensuring the quality management of high-value foods. It solves the limitations of traditional methods in handling complex features and high-value food detection, thereby significantly improving the detection accuracy, quality management efficiency and intelligent management level of marine foods.

[0070] (3) In view of the technical problems that the existing detection models have inappropriate built-in parameter settings and the model optimization algorithm has a weak ability to obtain the global optimal solution, which leads to inaccurate final model output results, this solution can adjust the search intensity at different stages by calculating the search individual difference distance and collaborative exploration strategy, ensuring sufficient global exploration capabilities in the early stage and rapid convergence in the later stage. It solves the technical problems of local optimal traps and insufficient exploration and development balance in the optimization algorithm, can obtain the optimal parameter values ​​and adjust the model hyperparameters, significantly improving the accuracy of the model output results, and thus improving the level of intelligent marine food management. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A schematic diagram of a process of a marine food management method based on big data provided by the present invention;

[0072] Figure 2 is a schematic diagram of the process of step S3;

[0073] Figure 3 is a schematic flow chart of step S4;

[0074] Figure 4 is a schematic flow chart of step S5;

[0075] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0076] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0077] In the description of the present invention, it is necessary to understand that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred system or element must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as a limitation on the present invention.

[0078] Example 1, see Figure 1 The technical solution adopted by the present invention is as follows: The present invention provides a marine food management method based on big data, which comprises the following steps:

[0079] Step S1: obtaining the original marine food management data, specifically, collecting the original marine food management data from the management system of a large marine food processing factory;

[0080] Step S2: refining the original data for optimizing the original marine food management data, specifically, performing data cleaning optimization, data standardization optimization, data enhancement optimization and data feature selection optimization on the original marine food management data to obtain the optimized marine food management data;

[0081] Step S3: constructing a marine food detection model, which is used to establish a marine food detection model and analyze the quality of marine food. Specifically, the marine food local input feature map is extracted and combined with the spatial attention mechanism to obtain key impurity features, and the marine food detection activation function is designed to generate output results. The training is performed according to the value grading model loss function, and finally the trained marine food detection model is obtained;

[0082] Step S4: Optimizing the detection model, specifically obtaining the optimal hyperparameter combination of the trained marine food detection model by improving the optimization algorithm, and obtaining the optimized marine food detection model;

[0083] Step S5: Double-sided detection management of marine food, which is used for intelligent management of marine food, specifically through the results of double-sided detection of marine food, to achieve intelligent management of marine food.

[0084] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the marine food management original data is obtained by collecting the marine food management original data from the management system of a large marine food processing factory; the marine food management original data includes marine food information data and marine food production line detection data; the marine food information data includes marine food product type, marine food size, marine food weight and marine food fishing area information; the marine food production line detection data includes marine food front detection data, marine food back detection data and production line environment detection data; the marine food front detection data and the marine food back detection data both include impurity type, impurity position, impurity size, impurity volume and impurity density; the production line environment detection data includes production line temperature data and production line humidity data.

[0085] Example 3, see Figure 1This embodiment is based on the above embodiment. In step S2, the refined original data is used to optimize the original marine food management data, specifically, data cleaning optimization, data standardization optimization, data enhancement optimization and data feature selection optimization are performed on the original marine food management data to obtain the optimized marine food management data, and the training data and the data to be tested are divided; the data cleaning optimization is used to eliminate invalid and inaccurate data, specifically, missing value processing, data outlier processing and data duplicate value deletion processing are performed on the data; the data standardization optimization is specifically to standardize the data using the maximum and minimum normalization method; the data enhancement optimization is used to generate diversified data, specifically to generate diversified data samples through a generative adversarial network, and the data feature selection optimization is specifically to perform feature selection based on a tree model on the data to screen out features related to qualified marine food products.

[0086] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S3, the marine food detection model is constructed to establish a marine food detection model, analyze the quality of marine food, and obtain a trained marine food detection model; specifically, the following steps are included:

[0087] Step S31: extracting the local input feature map of marine food, specifically including the following steps:

[0088] Step S311: perform a convolution operation; the formula used is expressed as follows:

[0089] ;

[0090] In the formula, represents the feature map of layer l, Respectively represent the row and column coordinates of the position in the feature map, represents the ReLU activation function, represents the batch normalization operation, n and m represent the height and width of the convolution kernel, Indicates the convolution kernel in The feature map of the layer, Indicates The weight matrix of the convolutional layer, Representative The bias parameters of the convolutional layer;

[0091] Step S312: Perform a pooling operation to reduce the output dimension of the feature; the formula used is as follows:

[0092] ;

[0093] In the formula, Indicates the position of the feature map of layer l after the pooling operation , represents the area of ​​the pooling window, Represents the weight of each position in the pooling window;

[0094] Step S32: Gather key impurity features to focus on areas that are critical to quality judgment. Specifically, the feature map is weighted through a spatial attention mechanism. The formula used is as follows:

[0095] ;

[0096] ;

[0097] In the formula, Represents the feature map location The attention weight, represents the Tanh activation function, represents the weighted feature map, represents element-wise multiplication, Represents the coordinates of other locations in the feature map;

[0098] Step S33: Design a marine food detection activation function to handle complex feature interactions in marine food double-sided detection; the formula used is as follows:

[0099] ;

[0100] In the formula, represents the marine food detection activation function, represents the input data of the activation function, Represents the scaling parameter that controls the activation function in the negative region. represents the Sigmoid activation function, represents the exponential function;

[0101] Step S34: Obtain the marine food detection output result, the formula used is as follows:

[0102] ;

[0103] In the formula, represents the cth marine food detection output result, represents the output weight matrix, Represents the output bias parameter, Represents the Dropout operation;

[0104] Step S35: Design a value classification model loss function, which is used to design a weighted loss function according to the value of different marine foods. The formula used is as follows:

[0105] ;

[0106] ;

[0107] ;

[0108] In the formula, represents the loss function value of common marine food, represents the loss function value of high-value marine food, Represents the total loss function value of the detection model, represents the actual result of the cth marine food test, Indicates whether marine food c is a high-value marine food. If =1, then marine food c is a high-value marine food, represents the loss weight of conventional marine food, represents the loss weight of high-value food, and C represents the total amount of marine food;

[0109] Step S36: Conduct model training, specifically, use the training data as the training data of the model, calculate the loss through the value grading model loss function, use the back propagation algorithm to optimize the weight of the model, and gradually reduce the error. After multiple rounds of iterations, the model continuously adjusts the weight parameters and optimizes the loss function to minimize the difference between the model prediction result and the true value. When the error reaches a preset threshold, the training process stops, thereby obtaining a trained marine food detection model.

[0110] By performing the above operations, in order to solve the technical problems of limitations of activation functions and insufficient loss functions in existing models applicable to marine food detection, which lead to low accuracy of marine food detection, this scheme innovatively designs marine food detection activation functions and value grading model loss functions, which can more effectively handle the complex feature interactions and nonlinear relationships of various impurities in marine foods, improve the sensitivity to key characteristics of food quality, and introduce value grading loss functions to differentiate foods of different values, improve the detection accuracy of high-value foods, ensure the quality management of high-value foods, and solve the limitations of traditional methods in handling complex features and high-value food detection, thereby significantly improving the detection accuracy, quality management efficiency and intelligent management level of marine foods.

[0111] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S4, the detection model optimization is specifically to obtain the optimal hyperparameter combination of the trained marine food detection model by improving the optimization algorithm to obtain the optimized marine food detection model, including the following steps:

[0112] Step S41: Initialize the search individuals, specifically randomly generate initial population individuals in the solution space; the formula used is as follows:

[0113] ;

[0114] In the formula, represents the initial position of the i-th population individual, and represent the lower and upper limits of the population position, respectively. Represents a random number in the range [0,1];

[0115] Step S42: Calculate the individual fitness value, specifically calculate the individual fitness value f in the population i ; The performance of the trained marine food detection model established based on the individual position is used as the individual fitness value, and the individuals are sorted from best to worst according to the fitness value to obtain the individual position with the highest global fitness value ;

[0116] Step S43: Calculate the difference distance of the searched individuals, specifically, obtain the distance between the current individual and the individual with the worst fitness value, and the formula used is as follows:

[0117] ;

[0118] In the formula, represents the difference distance of the i-th individual, t represents the number of iterations, represents a random number in the range [0,1]. Indicates the corresponding position of the worst fitness in the current iteration, represents the position of the i-th individual in the t-th generation population, Indicates the maximum number of iterations;

[0119] Step S44: updating the positions of the individuals in the population, specifically including the following steps:

[0120] Step S441: collaborative exploration updates individual positions, specifically updating positions according to collaborative exploration strategies, using the following formula:

[0121] ;

[0122] In the formula, represents the position of the i-th individual in the t+1th generation, Represents a random number in the range [0,1]; represents the average value of the positions of three randomly selected individuals in the population, represents the average value of the positions of two randomly selected individuals in the population;

[0123] Step S442: Randomly update individual positions, specifically update positions according to a variation diversification strategy, and the formula used is as follows:

[0124] ;

[0125] In the formula, represents a random number in the range [0,1]. represents the position of a randomly selected individual in the population;

[0126] Step S443: Update the individual position according to the iteration value, specifically update the position according to the dynamic behavior simulation strategy, and the formula used is as follows:

[0127] ;

[0128] In the formula, is the gravitational acceleration constant, Represents a random number in the range [0,1];

[0129] Step S45: The worst search individual position randomly jumps, and the formula used is as follows:

[0130] ;

[0131] In the formula, represents a random number in the range [0,0.45], represents a random number in the range [0.45,1], represents the difference between the positions of two randomly selected individuals in the population, represents a binary random variable;

[0132] Step S46: Individuals with low fitness are mutated, specifically, the fitness value of each individual is recalculated, the individuals are sorted from best to worst according to the fitness value, and 5%N individuals with low fitness are selected for individual mutation. The formula used is as follows:

[0133] ;

[0134] In the formula, represents individuals with low fitness. represents the optimal fitness of the current population, represents the worst fitness of the current population, and N represents the total number of search individuals;

[0135] Step S47: Obtain the optimal individual position, specifically when the individual fitness value f i When the fitness threshold is higher than the maximum number of iterations, the search is terminated and the individual global optimal position is obtained, and the optimal hyperparameter combination of the marine food detection model is obtained;

[0136] Step S48: obtaining an optimized marine food detection model, specifically adjusting the hyperparameters of the trained marine food detection model according to the optimal hyperparameter combination of the marine food detection model to obtain the optimized marine food detection model.

[0137] By performing the above operations, the existing detection models have inappropriate built-in parameter settings, and the model optimization algorithm has a weak ability to obtain the global optimal solution, which leads to inaccurate final model output results. Technical problems, this solution can adjust the search intensity at different stages by calculating the search individual difference distance and collaborative exploration strategy, ensuring sufficient global exploration capabilities in the early stage and rapid convergence in the later stage, solving the technical problems of local optimal traps and insufficient exploration and development balance in the optimization algorithm, and can obtain the optimal parameter values ​​and adjust the model hyperparameters, significantly improving the accuracy of the model output results, thereby improving the level of intelligent marine food management.

[0138] Example 6, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S5, the marine food double-sided detection management is used to intelligently manage marine food. Specifically, the marine food double-sided detection strategy is used to realize the intelligent management of marine food. The embodiment includes the following steps:

[0139] Step S51: marine food front detection: specifically, the marine food information data and the marine food front detection data in the data to be detected are used as input data of the optimized marine food detection model, and the optimized marine food detection model is used to detect the front of the marine food to obtain the output result of the marine food front detection;

[0140] Step S52: marine food back side detection: specifically, using the marine food information data and the marine food back side detection data in the data to be detected as input data of the optimized marine food detection model, and using the optimized marine food detection model to detect the back side of the marine food to obtain the output result of the marine food back side detection;

[0141] Step S53: Intelligent management of marine food. Based on the output results of the front side detection of marine food and the output results of the back side detection of marine food, the safety of marine food is comprehensively monitored and managed. If both output results are qualified, the marine food is sent to the freezing and refrigeration system. If the detection result of any side is unqualified, the unqualified marine food is discarded to ensure that unqualified products do not enter the next production link, thereby realizing efficient and accurate intelligent management of marine food.

[0142] By performing the above operations, in order to address the technical problem in traditional marine food management that when the marine food is large, single-sided detection may not be able to fully detect the entire surface of the food, this solution innovatively introduces a double-sided detection method. By independently detecting the front and back of the marine food, it ensures that each side of the food is thoroughly inspected, effectively captures different potential impurities on the front and back, reduces the risk of missed detection, and significantly improves the accuracy and efficiency of marine food detection, thereby improving the quality management level of marine food.

[0143] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0144] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0145] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A marine food management method based on big data, characterized by: The method comprises the following steps: Step S1: obtaining the original marine food management data, specifically obtaining the original marine food management data by collecting; Step S2: refining the original data for optimizing the original marine food management data, specifically refining the original marine food management data to obtain the marine food management optimization data, and dividing the training data and the data to be tested; Step S3: constructing a marine food detection model, which is used to establish a marine food detection model and analyze the quality of marine food, specifically by extracting a local input feature map of marine food and combining it with a spatial attention mechanism to obtain key impurity features, designing a marine food detection activation function, generating output results, and training according to a value grading model loss function to finally obtain a trained marine food detection model, including the following steps: step S31: extracting a local input feature map of marine food, step S32: aggregating key impurity features, step S33: designing a marine food detection activation function, step S34: obtaining marine food detection output results, step S35: designing a value grading model loss function, and step S36: performing model training; Step S4: Detection model optimization, by calculating and searching individual difference distances and improving the optimization algorithm through collaborative exploration strategy to obtain the optimal parameter combination of the marine food detection model, and obtain the optimized marine food detection model; Step S5: Double-sided detection of marine food is used for intelligent management of marine food, specifically, intelligent management of marine food is achieved based on the results of double-sided detection of marine food.

2. The marine food management method based on big data according to claim 1, characterized in that: In step S3, the construction of the marine food detection model specifically includes the following steps: Step S31: extracting a local input feature map of marine food, specifically performing convolution operations and pooling operations to obtain a feature map; Step S32: Gather key impurity features to focus on areas that are critical to quality judgment. Specifically, the feature map is weighted through a spatial attention mechanism. The formula used is as follows: ； ； In the formula, Represents the feature map location The attention weight, represents the Tanh activation function, represents the weighted feature map, represents element-wise multiplication, Represents the coordinates of other locations in the feature map, R represents the area of ​​the pooling window, Indicates the position of the feature map of layer l after the pooling operation ; Step S33: designing a marine food detection activation function for processing complex feature interactions in marine food double-sided detection; The formula used is as follows: ； In the formula, represents the marine food detection activation function, represents the input data of the activation function, Represents the scaling parameter that controls the activation function in the negative region. represents the Sigmoid activation function, represents the exponential function; Step S34: Obtain the marine food detection output result, the formula used is as follows: ； In the formula, represents the cth marine food detection output result, represents the output weight matrix, Represents the output bias parameter, Represents the Dropout operation; Step S35: Design a value classification model loss function, which is used to design a weighted loss function according to the value of different marine foods. The formula used is as follows: ； ； ； In the formula, represents the loss function value of common marine food, represents the loss function value of high-value marine food, Represents the total loss function value of the detection model, represents the actual result of the cth marine food test, Indicates whether marine food c is a high-value marine food. If =1, then marine food c is a high-value marine food, represents the loss weight of conventional marine food, represents the loss weight of high-value food, and C represents the total amount of marine food; Step S36: Conduct model training, specifically, use the training data as the training data of the model, calculate the loss through the value grading model loss function, use the back propagation algorithm to optimize the weight of the model, and gradually reduce the error. After multiple rounds of iterations, the model continuously adjusts the weight parameters and optimizes the loss function to minimize the difference between the model prediction result and the true value. When the error reaches a preset threshold, the training process stops, thereby obtaining a trained marine food detection model.

3. The marine food management method based on big data according to claim 1, characterized in that: In step S4, the detection model optimization specifically includes the following steps: Step S41: Initialize the search individuals, specifically, randomly generate initial population individuals in the solution space; Step S42: Calculate the individual fitness value, specifically calculate the individual fitness value f in the population i ; The performance of the trained marine food detection model established based on the individual position is used as the individual fitness value, and the individuals are sorted from best to worst according to the fitness value to obtain the individual position with the highest global fitness value ; Step S43: Calculate the difference distance of the searched individuals, specifically, obtain the distance between the current individual and the individual with the worst fitness value, and the formula used is as follows: ； In the formula, represents the difference distance of the i-th individual, t represents the number of iterations, represents a random number in the range [0,1]. Indicates the corresponding position of the worst fitness in the current iteration, represents the position of the i-th individual in the t-th generation population, Indicates the maximum number of iterations; Step S44: updating the positions of the individuals in the population, specifically including the following steps: Step S441: collaborative exploration updates individual positions, specifically updating positions according to collaborative exploration strategies, using the following formula: ； In the formula, represents the position of the i-th individual in the t+1th generation, Represents a random number in the range [0,1]; represents the average value of the positions of three randomly selected individuals in the population, represents the average value of the positions of two randomly selected individuals in the population; Step S442: Randomly update individual positions, specifically update positions according to a variation diversification strategy, and the formula used is as follows: ； In the formula, represents a random number in the range [0,1]. represents the position of a randomly selected individual in the population; Step S443: Update the individual position according to the iteration value, specifically update the position according to the dynamic behavior simulation strategy, and the formula used is as follows: ； In the formula, is the gravitational acceleration constant, Represents a random number in the range [0,1]; Step S45: The worst search individual position randomly jumps, and the formula used is as follows: ； In the formula, represents a random number in the range [0,0.45], represents a random number in the range [0.45,1], represents the difference between the positions of two randomly selected individuals in the population, represents a binary random variable; Step S46: Individuals with low fitness are mutated, specifically, the fitness value of each individual is recalculated, the individuals are sorted from best to worst according to the fitness value, and 5%N individuals with low fitness are selected for individual mutation. The formula used is as follows: ； In the formula, represents individuals with low fitness. represents the optimal fitness of the current population, represents the worst fitness of the current population, N represents the total number of search individuals, Represents the current individual fitness value; Step S47: Obtain the optimal individual position, specifically when the individual fitness value f i When the fitness threshold is higher than the maximum number of iterations, the search is terminated and the individual global optimal position is obtained, and the optimal hyperparameter combination of the marine food detection model is obtained; Step S48: obtaining an optimized marine food detection model, specifically adjusting the hyperparameters of the trained marine food detection model according to the optimal hyperparameter combination of the marine food detection model to obtain the optimized marine food detection model.

4. The marine food management method based on big data according to claim 1, characterized in that: In step S5, the marine food double-sided detection management is used for intelligent management of marine food, specifically realizing intelligent management of marine food through the results of the marine food double-sided detection; and comprises the following steps: Step S51: marine food front detection: specifically, the marine food information data and the marine food front detection data in the data to be detected are used as input data of the optimized marine food detection model, and the optimized marine food detection model is used to detect the front of the marine food to obtain the output result of the marine food front detection; Step S52: marine food back side detection: specifically, using the marine food information data and the marine food back side detection data in the data to be detected as input data of the optimized marine food detection model, and using the optimized marine food detection model to detect the back side of the marine food to obtain the output result of the marine food back side detection; Step S53: Intelligent management of marine food. Based on the output results of the front side detection of marine food and the output results of the back side detection of marine food, the safety of marine food is comprehensively monitored and managed. If both output results are qualified, the marine food is sent to the freezing and refrigeration system. If the detection result of any side is unqualified, the unqualified marine food is discarded to ensure that unqualified products do not enter the next production link, thereby realizing efficient and accurate intelligent management of marine food.

5. The marine food management method based on big data according to claim 1, characterized in that: In step S1, the marine food management original data is obtained by collecting the marine food management original data from the marine food processing factory management system; the marine food management original data includes marine food information data and marine food production line detection data; the marine food production line detection data includes marine food front detection data, marine food back detection data and production line environment detection data.

6. The marine food management method based on big data according to claim 1, characterized in that: In step S2, the refined raw data is used to optimize the raw data for marine food management, specifically, data cleaning optimization, data standardization optimization, data enhancement optimization and data feature selection optimization are performed on the raw data for marine food management to obtain the optimized data for marine food management, and the training data and the data to be tested are divided; the data enhancement optimization is used to generate diversified data, specifically, diversified data samples are generated by generating adversarial networks.

Images