Intelligent aquaculture management system and method based on Internet of Things

Abstract

The invention relates to the technical field of the Internet of Things, and more specifically discloses an intelligent aquaculture management system and method based on the Internet of Things, which can provide a fine management scheme, optimize aquaculture conditions and improve yield and efficiency for farmers by monitoring aquaculture environment parameters in real time through a sensor and combining data analysis and artificial intelligence. The risk is reduced, and sustainable development of the aquaculture industry is promoted.

Description

Technical Field

[0001] The present application relates to the field of Internet of Things technology, and more specifically, to an intelligent aquaculture management system and method based on the Internet of Things. Background Art

[0002] Currently, most aquaculture farms still use traditional artificial breeding methods. The disadvantage of traditional artificial breeding is that it relies too much on the personal experience of farmers, and the aquaculture management process is extensive and cannot achieve refined breeding management.

[0003] Therefore, a smart aquaculture management system and method based on the Internet of Things is desired. This can achieve real-time monitoring of the aquaculture water environment, data analysis, and intelligent decision-making support, helping farmers achieve refined management, improve aquaculture efficiency and yield, while reducing aquaculture risks and providing support for the sustainable development of the aquaculture industry. Summary of the Invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a smart aquaculture management system and method based on the Internet of Things, which can achieve refined management of aquaculture, improve aquaculture efficiency and output, and reduce aquaculture risks.

[0005] Accordingly, according to one aspect of the present application, there is provided a smart aquaculture management system based on the Internet of Things, comprising:

[0006] Aquaculture water data collection module, used to extract dissolved oxygen value, pH value, conductivity value and water temperature value of aquaculture waters at multiple predetermined time points including the current time point;

[0007] an aquaculture water area data processing module, configured to extract a dissolved oxygen feature vector, a pH value feature vector, a conductivity feature vector, and a water temperature feature vector from the dissolved oxygen values, pH values, conductivity values, and water temperature values ​​at the plurality of predetermined time points, respectively;

[0008] an aquaculture water data fusion module, configured to construct an aquaculture water parameter feature vector between the dissolved oxygen feature vector, the pH value feature vector, the conductivity feature vector, and the water temperature feature vector, and perform manifold regression-guided kernel space mapping processing on the aquaculture water parameter feature vector to obtain an optimized aquaculture water parameter feature vector;

[0009] The aquaculture waters result analysis module is used to pass the optimized aquaculture waters parameter feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the aquaculture waters environment is suitable.

[0010] According to another aspect of the present application, there is also provided a smart aquaculture management method based on the Internet of Things, which includes:

[0011] Extracting dissolved oxygen values, pH values, conductivity values, and water temperature values ​​in the aquaculture waters at multiple predetermined time points, including the current time point;

[0012] extracting a dissolved oxygen feature vector, a pH value feature vector, a conductivity feature vector, and a water temperature feature vector from the dissolved oxygen values, pH values, conductivity values, and water temperature values ​​at the plurality of predetermined time points, respectively;

[0013] constructing an aquaculture water parameter feature vector between the dissolved oxygen feature vector, the pH value feature vector, the conductivity feature vector, and the water temperature feature vector, and performing a manifold regression-guided kernel space mapping process on the aquaculture water parameter feature vector to obtain an optimized aquaculture water parameter feature vector;

[0014] The optimized aquaculture water parameter feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the aquaculture water environment is suitable.

[0015] Compared with the existing technology, the present application provides an intelligent aquaculture management system and method based on the Internet of Things, which uses sensors to monitor aquaculture environment parameters in real time and combines data analysis and artificial intelligence to provide farmers with refined management plans, optimize aquaculture conditions, increase production and efficiency, reduce risks, and promote the sustainable development of the aquaculture industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 This is a block diagram of an IoT-based smart aquaculture management system according to an embodiment of the present application.

[0018] Figure 2 This is a block diagram of the aquaculture water data processing module in the IoT-based smart aquaculture management system according to an embodiment of the present application.

[0019] Figure 3 This is a block diagram of a dissolved oxygen feature extraction unit in an IoT-based smart aquaculture management system according to an embodiment of the present application.

[0020] Figure 4 This is a flow chart of the smart aquaculture management method based on the Internet of Things according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] Various exemplary embodiments, features, and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0022] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0023] In addition, numerous specific details are provided in the following detailed description to better illustrate the present application. Those skilled in the art will appreciate that the present application can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art are not described in detail in order to highlight the main purpose of the present application.

[0024] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. Throughout the description of this application, "plurality" means two or more, unless otherwise specifically defined.

[0025] Figure 1 The figure shows a block diagram of a smart aquaculture management system based on the Internet of Things according to an embodiment of the present application. Figure 1 As shown, according to an embodiment of the present application, the smart aquaculture management system 100 based on the Internet of Things includes: an aquaculture water data acquisition module 110, which is used to extract the dissolved oxygen value, pH value, conductivity value and water temperature value of the aquaculture water at multiple predetermined time points including the current time point; an aquaculture water data processing module 120, which is used to extract a dissolved oxygen feature vector, a pH value feature vector, a conductivity feature vector and a water temperature feature vector from the dissolved oxygen value, pH value, conductivity value and water temperature value at the multiple predetermined time points respectively; an aquaculture water data fusion module 130, which is used to construct an aquaculture water parameter feature vector between the dissolved oxygen feature vector, the pH value feature vector, the conductivity feature vector and the water temperature feature vector, and perform manifold regression-guided kernel space mapping processing on the aquaculture water parameter feature vector to obtain an optimized aquaculture water parameter feature vector; and an aquaculture water result analysis module 140, which is used to pass the optimized aquaculture water parameter feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the aquaculture water environment is suitable.

[0026] In an embodiment of the present application, the aquaculture water data acquisition module 110 is used to extract the dissolved oxygen value, pH value, conductivity value, and water temperature value of the aquaculture water area at multiple predetermined time points, including the current time point. It should be understood that these parameters are important indicators for assessing water quality and the health of the aquaculture environment. By monitoring the values ​​of these parameters at different time points, changes in water quality can be monitored in a timely manner, helping farmers adjust aquaculture conditions to ensure that the water quality meets the needs of aquaculture organisms. Changes in parameters such as dissolved oxygen, pH value, conductivity, and water temperature may affect the growth and health of aquaculture organisms. By regularly monitoring these parameters, the occurrence of diseases can be prevented and the survival rate of aquaculture organisms can be improved. Maintaining suitable water quality conditions is crucial for increasing the growth rate of aquaculture organisms, increasing production, and improving aquaculture efficiency. By monitoring and adjusting these parameters, the aquaculture environment can be optimized and production efficiency can be improved.

[0027] In an embodiment of the present application, the aquaculture water data processing module 120 is used to extract dissolved oxygen feature vectors, pH value feature vectors, conductivity feature vectors and water temperature feature vectors from the dissolved oxygen values, pH values, conductivity values ​​and water temperature values ​​at the multiple predetermined time points, respectively. It should be understood that the original dissolved oxygen value, pH value, conductivity value and water temperature value data may contain a large amount of noise and redundant information. Through feature extraction, these data can be converted into more representative and interpretable feature vectors, which helps the model to better understand and process the data. Extracting feature vectors can effectively reduce the dimension of the data, reduce the impact of redundant information, and improve data processing efficiency and model training speed. The feature vector contains richer information, which helps to reveal the patterns and laws hidden in the data, enabling the model to better perform classification, prediction or other related tasks. By extracting the feature vectors of dissolved oxygen, pH value, conductivity value and water temperature value, more effective input data can be provided for subsequent models, thereby improving the performance and accuracy of the model.

[0028] Specifically, in one embodiment of the present application, Figure 2 The figure shows a block diagram of the aquaculture water data processing module in the smart aquaculture management system based on the Internet of Things according to an embodiment of the present application. Figure 2 As shown, in the above-mentioned smart aquaculture management system 100 based on the Internet of Things, the aquaculture water data processing module 120 includes: a dissolved oxygen feature extraction unit 121, used to extract the dissolved oxygen feature vector from the dissolved oxygen values ​​at the multiple predetermined time points; a pH value feature extraction unit 122, used to extract the pH value feature vector from the pH values ​​at the multiple predetermined time points; a conductivity feature extraction unit 123, used to extract the conductivity feature vector from the conductivity values ​​at the multiple predetermined time points; and a water temperature feature extraction unit 124, used to extract the water temperature feature vector from the water temperature values ​​at the multiple predetermined time points.

[0029] Accordingly, in a specific example of the present application, the dissolved oxygen feature extraction unit 121 is used to extract the ambient temperature feature vector from the ambient temperature values ​​at the multiple predetermined time points. It should be understood that ambient temperature is an important environmental parameter that has an important impact on many organisms and systems. By extracting the feature vector of the ambient temperature, it can help to identify the changing patterns, trends and periodicity of temperature, and thus better understand and respond to environmental changes. Extracting the feature vector of the ambient temperature helps to conduct a more in-depth analysis and mining of temperature data, revealing the potential patterns and information in the temperature data, and providing support for subsequent data processing and decision-making. Data analysis based on the ambient temperature feature vector can provide more effective features for establishing a prediction model, thereby realizing the prediction of future ambient temperature changes and providing a reference basis for decision-making and planning. By extracting the ambient temperature feature vector, the ambient temperature changes can be better monitored and managed, which helps to optimize the environmental control strategy and improve production efficiency and quality.

[0030] further, Figure 3 The figure shows a block diagram of a dissolved oxygen feature extraction unit in an IoT-based smart aquaculture management system according to an embodiment of the present application. Figure 3 As shown, in the aquaculture water data processing module 120 of the above-mentioned smart aquaculture management system 100 based on the Internet of Things, the dissolved oxygen feature extraction unit 121 includes: a dissolved oxygen arrangement vector subunit 1211, which is used to arrange the dissolved oxygen values ​​of the multiple predetermined time points into a dissolved oxygen input vector according to the time dimension; and a dissolved oxygen multi-scale extraction subunit 1212, which is used to pass the dissolved oxygen input vector through a dissolved oxygen value extractor based on a multi-scale neighborhood feature extraction module to obtain the dissolved oxygen feature vector.

[0031] Specifically, the dissolved oxygen arrangement vector subunit 1211 is used to arrange the dissolved oxygen values ​​at the multiple predetermined time points according to the time dimension into a dissolved oxygen input vector. It should be understood that time is an important dimension and has a key impact on the change of dissolved oxygen values. Arranging the dissolved oxygen values ​​according to the time dimension can retain the sequential information in the time series and help capture the dynamic characteristics of dissolved oxygen changes over time. Arranging the dissolved oxygen values ​​according to the time dimension to form an input vector can apply sequence modeling methods, such as recurrent neural networks (RNN) or long short-term memory networks (LSTM), to better explore patterns and regularities in time series data. Based on the dissolved oxygen input vector arranged according to the time dimension, time series prediction and analysis can be performed, such as predicting the future trend of dissolved oxygen value changes or detecting outliers, so as to better understand the state and change patterns of the water body. The dissolved oxygen values ​​in the time series data can be used as a feature sequence and arranged in the time dimension to form an input vector, which helps to extract richer and more meaningful features and provide more effective input for subsequent data processing and analysis.

[0032] Specifically, the dissolved oxygen multi-scale extraction subunit 1212 is configured to pass the dissolved oxygen input vector through a dissolved oxygen value extractor based on a multi-scale neighborhood feature extraction module to obtain the dissolved oxygen feature vector. It should be understood that the multi-scale neighborhood feature extraction module can effectively capture characteristic information of dissolved oxygen values ​​at different scales and levels. This helps extract richer and more comprehensive features, including local and global features, to better describe the complexity and diversity of dissolved oxygen data. Multi-scale neighborhood feature extraction enables feature extraction and characterization of dissolved oxygen data at different spatial scales, improving the expressiveness and discriminability of features, making the extracted features more representative and discriminative. The multi-scale neighborhood feature extraction module improves the characterization capabilities of dissolved oxygen data, enabling better model training and performance. This helps improve the quality and effectiveness of dissolved oxygen feature vectors, thereby enhancing subsequent data analysis and application. In complex environments, dissolved oxygen values ​​may be affected by a variety of factors, including local environmental changes and global water body characteristics. The multi-scale neighborhood feature extraction module can better meet the needs of dissolved oxygen data feature extraction in complex environments.

[0033] Accordingly, the dissolved oxygen multi-scale extraction subunit includes: a first-scale dissolved oxygen encoding secondary subunit, configured to use the first convolution layer of the dissolved oxygen value extractor based on the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the dissolved oxygen input vector using a one-dimensional convolution kernel having a first scale to obtain a first-scale dissolved oxygen feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel having a first length; a second-scale dissolved oxygen encoding secondary subunit, configured to use the second convolution layer of the dissolved oxygen value extractor based on the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the dissolved oxygen input vector using a one-dimensional convolution kernel having a second scale to obtain a second-scale dissolved oxygen feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel having a second length, and the first length is different from the second length; and a multi-scale dissolved oxygen cascade secondary subunit, configured to cascade the first-scale dissolved oxygen feature vector and the second-scale dissolved oxygen feature vector to obtain the dissolved oxygen feature vector.

[0034] Accordingly, in a specific example of the present application, the pH value feature extraction unit 122 is configured to extract the pH value feature vector from the pH values ​​at the multiple predetermined time points. It should be understood that pH values ​​may exhibit certain regularities and trends over time. Therefore, extracting pH values ​​from multiple time points can capture the dynamic changes in pH values ​​over time, forming time series features. These time series features can be used to analyze the periodicity, trends, and abnormal changes in pH values. Extracting the pH values ​​at multiple time points into a single feature vector reduces the dimensionality of the data, making it easier to process and analyze. This feature extraction method integrates information from multiple time points into a single vector, reducing data redundancy. By extracting the pH value feature vector, machine learning or deep learning models can be applied for pattern recognition and prediction. These models can use the extracted pH value feature vectors to predict future pH value trends or detect anomalies, helping to monitor the health of water bodies. Extracting pH value feature vectors can help understand the relationship between pH and other factors. These feature vectors can be used as input to explore the correlation between pH and factors such as dissolved oxygen, temperature, and water quality, facilitating in-depth analysis of the characteristics and patterns of change in water bodies.

[0035] Specifically, the pH value feature extraction unit includes: a pH value arrangement vector subunit, which is used to arrange the pH values ​​of the multiple predetermined time points into a pH input vector according to the time dimension; and a pH value multi-scale extraction subunit, which is used to pass the pH input vector through a pH value extractor based on a multi-scale neighborhood feature extraction module to obtain the pH value feature vector.

[0036] Accordingly, the pH value arrangement vector subunit is configured to arrange the pH values ​​at the multiple predetermined time points according to the time dimension into a pH input vector. It should be understood that arranging the pH values ​​according to the time dimension preserves temporal order information, ensuring that the model can capture the changing trends of the pH values ​​over time. This helps the model better understand the evolution of the pH values, thereby improving the model's understanding and prediction capabilities of the data. Arranging the pH values ​​according to the time dimension to form a sequence input is suitable for sequence models (such as recurrent neural networks (RNNs) or long short-term memory networks (LSTMs)). These models can effectively process time series data and capture temporal correlations within the sequence, thereby better predicting future pH value changes. Arranging the pH values ​​into a time series allows for better time series feature extraction. The model can learn time-related features from the continuous pH value sequence, including trends, periodicity, and sudden changes, thereby better capturing the dynamic characteristics of the data. Arranging the pH values ​​according to the time dimension allows for better model training. During training, the model can better learn the temporal changes of the pH values, improving the modeling capabilities and prediction accuracy of time series data. Arranging pH values ​​as input vectors along the time dimension can make the data structure more neat and standardized, facilitating data processing and model training. This helps improve the interpretability of the data and the interpretability of the model.

[0037] Furthermore, the pH multi-scale extraction subunit is configured to pass the pH input vector through a pH extractor based on a multi-scale neighborhood feature extraction module to obtain the pH feature vector. It should be understood that the pH extractor based on the multi-scale neighborhood feature extraction module can extract pH features at different scales, including local details and global information. This helps capture the diversity and complexity of pH data, improving the richness and representation of features. Multi-scale neighborhood feature extraction can better capture the spatial and temporal correlations in pH data, enhancing the expressiveness of features. This helps the model better understand the inherent structure and patterns of pH data, improving the discriminative and generalization capabilities of features. Multi-scale neighborhood feature extraction can help suppress noise interference in the data and improve the stability and robustness of features. By extracting features at different scales, interference factors in the data can be better filtered out, improving the quality and reliability of features. Utilizing the pH extractor based on the multi-scale neighborhood feature extraction module can improve model performance and generalization. By more comprehensively extracting features from pH data, the model's understanding and prediction capabilities can be enhanced, thereby improving the model's performance in pH analysis and prediction tasks. Multi-scale neighborhood feature extraction can integrate information at different scales to obtain more comprehensive and accurate feature representations. This helps improve the diversity and richness of features, making feature vectors more representative and discriminative.

[0038] Accordingly, in a specific example of the present application, the conductivity feature extraction unit 123 is configured to extract the conductivity feature vector from the conductivity values ​​at the multiple predetermined time points. It should be understood that by extracting the conductivity feature vector, the conductivity values ​​at the multiple time points can be represented as a fixed-dimensional feature vector, thereby reducing the dimensionality of the data. This helps simplify data representation and processing, reducing the complexity of model training. The conductivity feature vector can help extract key characteristic information, including trends, periodicity, and outliers in the conductivity values. These features help the model better understand the patterns and regularities in the conductivity data, thereby improving data characterization and predictive performance. The conductivity values ​​extracted as feature vectors can be used as model input for training and prediction. The feature vectors can better represent the characteristics of the conductivity data, enabling the model to more effectively learn the patterns and regularities in the data. The conductivity feature vectors can be used in time series analysis to help understand the temporal trends of the conductivity data. By extracting the feature vectors, the dynamic characteristics of the conductivity data can be better captured, thereby improving the accuracy and reliability of time series analysis. Extracting conductivity feature vectors is part of feature engineering, which helps transform and process raw data to make it more suitable for model training and prediction. By extracting key conductivity features, you can improve model performance and effectiveness.

[0039] Specifically, the conductivity feature extraction unit includes: a conductivity vector arrangement subunit, configured to arrange the conductivity values ​​at the plurality of predetermined time points into a conductivity input vector according to the time dimension; and a conductivity multi-scale extraction subunit, configured to pass the conductivity input vector through a conductivity value extractor based on a multi-scale neighborhood feature extraction module to obtain the conductivity feature vector.

[0040] Furthermore, the conductivity vector arrangement subunit is configured to arrange the conductivity values ​​at the multiple predetermined time points according to the time dimension into a conductivity input vector. It should be understood that arranging the conductivity values ​​according to the time dimension preserves the characteristics of a time series, enabling the model to better capture trends and patterns in data over time. The sequential nature of time in data is crucial for analyzing and predicting conductivity values. Arranging the conductivity values ​​according to the time dimension preserves the information order of the data, ensuring that the model can correctly understand the correlations and temporal order between data points during training and prediction. This helps the model better understand the dynamic characteristics of the data. Many models require input formats for time series data, often arranging data according to the time dimension. Arranging the conductivity values ​​according to the time dimension into an input vector satisfies these input format requirements and enables the model to effectively process time series data. Arranging the conductivity values ​​according to the time dimension facilitates feature extraction and pattern recognition. Arranging the conductivity values ​​according to the time dimension allows the model to better capture features and patterns in the data, improving its understanding and prediction capabilities. Arranging the conductivity values ​​according to the time dimension simplifies data processing, making the data easier for the model to process and analyze. The arrangement of the time dimension helps to perform effective input and output operations on data and improve the efficiency of data processing.

[0041] Furthermore, the conductivity multi-scale extraction subunit is configured to pass the conductivity input vector through a conductivity value extractor based on a multi-scale neighborhood feature extraction module to obtain the conductivity feature vector. It should be understood that the conductivity value extractor based on the multi-scale neighborhood feature extraction module can help extract feature information from neighborhoods of different scales. This can more comprehensively capture local and global features of the data, improving the richness and representational power of features. Multi-scale neighborhood feature extraction can make the conductivity feature vector more representative. Feature extraction at different scales can capture information at different levels, allowing the feature vector to better describe the complexity and diversity of the data. By introducing the multi-scale neighborhood feature extraction module, the performance of the conductivity value extractor can be improved, enabling it to more effectively learn and extract features from the data. This helps enhance the model's understanding and predictive capabilities of conductivity data. Multi-scale feature extraction can help suppress noise and interference in the data, making the feature vector more robust. By combining features at different scales, the impact of noise on feature extraction can be reduced, improving the accuracy and stability of features. Introducing a multi-scale neighborhood feature extraction module can enhance the generalization ability of the model. Multi-scale feature extraction can make the model more adaptable to different data distributions and scenarios, thereby improving the model's performance in different situations.

[0042] Accordingly, in a specific example of the present application, the water temperature feature extraction unit 124 is used to extract the water temperature feature vector from the water temperature values ​​at the multiple predetermined time points. It should be understood that the water temperature values ​​have certain trends and patterns as they change over time, and extracting the water temperature feature vector can retain the characteristics of the time series data. This helps to capture the changes in water temperature over time and provides a basis for subsequent analysis and prediction. By extracting the water temperature feature vector, the water temperature values ​​at multiple time points can be converted into a feature vector with lower dimensions, simplifying the data representation. This helps to reduce the complexity of the data and improve the efficiency of data processing. Extracting the water temperature feature vector can help identify important features and patterns in the water temperature data. This helps to discover regularities in the water temperature data and provide a reliable feature representation for subsequent analysis and prediction. The extraction of the water temperature feature vector is part of feature engineering, and the performance of the model can be improved by selecting an appropriate feature extraction method. Appropriate feature representation can help the model better understand the data and improve prediction accuracy. Many machine learning models need to represent data in the form of feature vectors for processing. Extracting the water temperature feature vector can meet the input requirements of the model, allowing the model to effectively process water temperature data.

[0043] Specifically, the water temperature feature extraction unit includes: a water temperature arrangement vector subunit, which is used to arrange the water temperature values ​​of the multiple predetermined time points into a water temperature input vector according to the time dimension; a water temperature multi-scale extraction subunit, which is used to pass the water temperature input vector through a water temperature value extractor based on a multi-scale neighborhood feature extraction module to obtain the water temperature feature vector.

[0044] Furthermore, the water temperature arrangement vector subunit is configured to arrange the water temperature values ​​at the multiple predetermined time points according to the time dimension into a water temperature input vector. It should be understood that water temperature data exhibits certain sequential characteristics and trends over time, and arranging according to the time dimension can preserve this time series information. This helps capture the patterns of data change over time, providing important temporal information for subsequent analysis and prediction. Arranging water temperature values ​​according to the time dimension can make the data structure clearer and more organized. This helps simplify data representation, making the data easier to process and analyze. Arranging water temperature values ​​according to the time dimension facilitates feature extraction and engineering. Arranging water temperature values ​​according to the time dimension can help identify temporal patterns and trends in the data, providing a foundation for feature extraction. Many machine learning models require data to be represented as input vectors in a specific format for processing. Arranging water temperature values ​​according to the time dimension can meet the model's input requirements, enabling the model to effectively process time series data. Arranging water temperature values ​​according to the time dimension facilitates data analysis and visualization. Arranging water temperature values ​​according to the time dimension can help researchers better understand time-related characteristics in the data and discover regularities and trends within the data.

[0045] Furthermore, the water temperature multi-scale extraction subunit is configured to pass the water temperature input vector through a water temperature value extractor based on a multi-scale neighborhood feature extraction module to obtain the water temperature feature vector. It should be understood that the multi-scale neighborhood feature extraction module can help the system capture water temperature data features at different scales. This helps understand water temperature data from different perspectives, including local details and overall trends, and improves the comprehensiveness and accuracy of feature extraction. Multi-scale neighborhood feature extraction can better represent complex features and patterns in water temperature data. This helps improve the representational capabilities of feature vectors, making the extracted features richer and more meaningful. Multi-scale neighborhood feature extraction can map the high-dimensional feature space of the original water temperature data to a lower-dimensional feature space. This helps reduce data complexity and the dimensionality of feature vectors while retaining important information. Multi-scale neighborhood feature extraction can improve the generalization ability of the model. This approach enables the model to better adapt to water temperature data of different scales and complexities, improving the model's applicability and robustness. The water temperature feature vectors obtained through multi-scale neighborhood feature extraction can improve the data representation capabilities, thereby helping to improve the predictive performance of subsequent models. Richer and more accurate feature representations can help models better understand the data and make accurate predictions.

[0046] In this embodiment of the present application, the aquaculture waters data fusion module 130 is configured to construct an aquaculture waters parameter feature vector from the dissolved oxygen feature vector, the pH feature vector, the conductivity feature vector, and the water temperature feature vector, and perform manifold regression-guided kernel space mapping on the aquaculture waters parameter feature vector to obtain an optimized aquaculture waters parameter feature vector. It should be understood that dissolved oxygen, pH, conductivity, and water temperature are very important parameters in aquaculture waters, reflecting the fundamental characteristics of water quality. Fusion of the feature vectors of these parameters can provide more comprehensive and integrated water quality information, facilitating a comprehensive assessment of the environmental conditions of aquaculture waters. These parameters often have certain correlations and influences. For example, water temperature can affect dissolved oxygen content, and pH can also be affected by conductivity. Therefore, fusing them together can better understand their interactions and influences. Fusion of feature vectors of different parameters can provide richer and more diverse feature information. This helps improve the representational capabilities of the feature vectors, making the feature matrix more representative and informative, and better describing the characteristics of aquaculture waters. Fusion of multiple feature vectors into a single feature matrix can reduce the dimensionality of the data, simplifying data processing and analysis. This helps reduce data processing complexity and improve computational efficiency. After fusing the eigenvectors of multiple parameters into a feature matrix, various machine learning and data analysis algorithms can be more conveniently applied for modeling and prediction.

[0047] Accordingly, in one embodiment of the present application, the aquaculture water data fusion module includes: a parameter arrangement unit, used to arrange the dissolved oxygen characteristic vector, the pH value characteristic vector, the conductivity characteristic vector and the water temperature characteristic vector according to the parameter dimension to obtain an aquaculture water parameter input matrix; a spatial attention unit, used to pass the aquaculture water parameter input matrix through the first convolutional neural network model of spatial attention to obtain the aquaculture water parameter characteristic vector; a vector optimization unit, used to perform manifold regression-guided kernel space mapping processing on the aquaculture water parameter characteristic vector to obtain an optimized aquaculture water parameter characteristic vector.

[0048] Accordingly, in a specific example of the present application, the parameter arrangement unit is configured to arrange the dissolved oxygen feature vector, the pH value feature vector, the conductivity feature vector, and the water temperature feature vector according to parameter dimensions to obtain an aquaculture water parameter input matrix. It should be understood that arranging feature vectors according to parameter dimensions can maintain data structure consistency, ensuring that the positions of features of different parameters in the matrix correspond to their parameter positions in actual aquaculture waters. This helps maintain the overall structure and accuracy of the data. Arranging feature vectors according to parameter dimensions allows features of the same type to be arranged consecutively in the matrix, facilitating data processing and analysis. For example, features of specific parameters can be easily extracted for further processing or analysis. Arranging feature vectors of different parameters into a matrix format according to parameter dimensions can better meet the input requirements of various machine learning models. Models generally expect input data to have a certain structure and dimension, and arranging according to parameter dimensions can meet these model input requirements. Arranging feature vectors according to parameter dimensions can improve data utilization efficiency. In practical applications, data of specific parameters needs to be frequently accessed and processed, and arranging according to parameter dimensions can make these operations more efficient. Arranging feature vectors according to parameter dimensions can make data visualization more intuitive and easier to understand. The structure of the feature matrix is ​​clear, which can easily show the relationship and change trend between different parameters.

[0049] Accordingly, in a specific example of the present application, the spatial attention unit is used to pass the aquaculture water parameter input matrix through the first convolutional neural network model of spatial attention to obtain the aquaculture water parameter feature vector. It should be understood that convolutional neural networks (CNNs) perform well in processing spatial structured data and can effectively extract spatial features in the data. Through the CNN model, the features in the aquaculture water parameter input matrix can be learned, including the correlation and importance between different parameters. The introduction of the spatial attention mechanism can enable the model to pay more attention to the spatial relationship between different parameters when learning features, and improve the model's ability to model complex interactions between parameters. This helps to more accurately capture the correlation and importance between aquaculture water parameters. Through the CNN model, the high-dimensional aquaculture water parameter input matrix can be converted into a more representative and efficient feature matrix. This helps to reduce data dimensionality, reduce computational complexity, and extract the most important feature information. The CNN model has good generalization capabilities in processing images and spatial data, and can be applied to data with spatial structure such as aquaculture water parameters. By training the CNN model, the model can better understand and learn the complex relationships between aquaculture water parameters. The aquaculture water parameter feature vectors obtained through the CNN model can better reflect the correlation and importance between parameters, thereby improving the accuracy and effectiveness of subsequent models in water environment condition prediction or other tasks.

[0050] Here, when different types of water quality parameters (such as dissolved oxygen, pH, conductivity, and water temperature) are converted into their respective feature vectors, although these feature vectors can reflect the trends and characteristics of their respective parameters over time, the complex interactions and intrinsic correlations that may exist between them may not be fully captured and utilized. For example, the relationship between dissolved oxygen levels and water temperature may be affected by other factors (such as seasonal changes, biological activity, etc.), and simple linear or nonlinear combinations may not be able to fully reveal these complex interactions. In addition, when further processed by convolutional neural networks, although this method is good at capturing local spatial patterns, it has limited ability to capture overall structural information across multiple parameter dimensions, especially when long-range dependencies or global contextual information are involved. Traditional methods may not be able to effectively identify and utilize this information. Therefore, in order to optimize this process, the aquaculture water parameter feature vectors are subjected to manifold regression-guided kernel space mapping processing to obtain optimized aquaculture water parameter feature vectors.

[0051] Specifically, the vector optimization unit is used to: first, establish a superpixel manifold correlation matrix of the aquaculture water parameter feature vector, which is expressed as follows:

[0052]

[0053] Wherein, V represents the characteristic vector of the aquaculture water parameters, v i and v j Respectively represent the eigenvalues ​​of the i-th and j-th positions of the aquaculture water parameter eigenvector, d(v i ,v j ) represents the calculation of Euclidean distance, D i,j Represents the eigenvalue of the (i, j) position of the superpixel manifold incidence matrix.

[0054] Specifically, the discrete multi-dimensional parameter representation is transformed into a manifold spatial representation with an explicit topological structure. By constructing a superpixel manifold association matrix, it is possible not only to quantify the synergistic interaction patterns of parameters such as dissolved oxygen, pH, conductivity, and water temperature in their spatiotemporal evolution, but also to capture the dynamic coupling mechanisms of different sensor data within local regions. This manifold learning approach reveals the underlying topological associations between multidimensional parameters, transforming the traditional discrete feature point mapping into a superpixel manifold association matrix with semantic associations. This overcomes the limitations of traditional feature engineering for independent parameter analysis and provides physically meaningful constraints for subsequent kernel space mapping.

[0055] Secondly, deformable convolution is used to perform kernel domain representation learning on the superpixel manifold association matrix to obtain the kernel domain response dynamic matrix of the aquaculture water parameters, which is expressed as follows:

[0056] M=Conv(D)

[0057] Wherein, D represents the superpixel manifold association matrix, Conv represents the deformable convolution layer, and M represents the aquaculture water parameter kernel domain response dynamic matrix.

[0058] That is, through the dynamic receptive field adjustment mechanism of deformable convolution, the geometric deformation correlation patterns between superpixel units are adaptively captured within the kernel space. Specifically, through offset learning of the convolution kernel parameterization, the generated aquaculture water parameter kernel domain response dynamic matrix can encode the non-rigid deformation patterns implicit in the aquaculture parameter correlation manifold, providing an interpretable representation of the cross-parameter dynamic interaction mechanism for subsequent decision-making models, thereby improving the predictive robustness of "sudden multi-parameter coupling anomalies."

[0059] Then, the intrinsic mode decomposition is performed on the superpixel manifold correlation matrix to obtain a set of multi-scale intrinsic component encoding vectors of the aquaculture water parameters, which is expressed as follows:

[0060]

[0061] Where T represents the transpose of the vector, Λ represents the diagonal matrix, λ1 and λ mThey represent the first and mth eigenvalues ​​of the diagonal matrix respectively, U represents the set of multi-scale eigencomponent encoding vectors of aquaculture water parameters, x1, x2, x m Represent the multi-scale eigencomponent encoding vectors of the first, second and mth aquaculture water parameters respectively.

[0062] In other words, through intrinsic mode decomposition, the complex superpixel manifold correlation matrix can be decoupled into multi-scale eigencomponent encoding vectors of aquaculture water parameters that reflect the dynamics of different scales. Essentially, this establishes an orthogonal modal space for the parameter correlation structure, allowing the cross-scale interaction mechanisms originally mixed in the global correlation manifold to be explicitly expressed. Specifically, through multi-scale projection of orthogonal basis vectors, the hierarchical physical interaction mechanisms implicit in the aquaculture environment parameter system are extracted, providing an interpretable modal feature basis for subsequent abnormal pattern recognition.

[0063] Next, the multi-scale intrinsic component encoding vector of the aquaculture water parameter is input into the sequence model based on the self-attention mechanism to obtain a set of calibrated intrinsic component encoding vectors of the aquaculture water parameter, which is expressed as follows:

[0064] Y=Transformer{[x1,x2,…,x m ]}=[y1,y2,…,y m ]

[0065] Among them, Transformer represents a sequence model based on the self-attention mechanism, Y represents the set of encoding vectors of the intrinsic components of the aquaculture water parameter calibration, y1, y2, y m They represent the first, second and mth aquaculture water parameter calibration eigencomponent coding vectors respectively.

[0066] Specifically, a cross-modal dynamic weight allocation network is established through a sequence model based on the self-attention mechanism. The self-attention mechanism is used to conditionally model the nonlinear coupling relationship of multi-scale intrinsic components (e.g., automatically enhancing the self-attention coefficient of the corresponding high-frequency mode when the pH value fluctuates abnormally), thereby breaking through the bottleneck of fixed weight fusion in representing complex environmental interactions. Specifically, the self-attention mechanism can adaptively suppress noise-dominated modes and enhance physical mechanism-related modes. This allows the set of calibrated aquaculture water parameter calibration intrinsic component encoding vectors to dynamically focus on the dominant action chain that is most likely to affect aquaculture risks while retaining multi-scale physical laws, ultimately achieving an improvement in the multi-level linkage monitoring capability of "hidden ecological chain reactions."

[0067] Then, the set of the aquaculture water parameter calibration intrinsic component encoding vectors and the aquaculture water parameter core domain response dynamic matrix are tensor-contracted to obtain the set of aquaculture water parameter intrinsic component core domain self-attention encoding vectors, which can be expressed as follows:

[0068]

[0069] in, represents matrix multiplication, S represents the characteristic scale of the dynamic matrix of the kernel domain response of the aquaculture water parameters, and y i represents the encoding vector of the intrinsic component of the calibration of the i-th aquaculture water parameter, L represents the length of the encoding vector of the intrinsic component of the calibration of the aquaculture water parameter, z i Represents the kernel self-attention encoding vector of the intrinsic component of the i-th aquaculture water parameter.

[0070] That is, by establishing a cross-modal implicit channel through high-order tensor operations, the intrinsic component encoding vector of the aquaculture water parameter calibration can dynamically perceive the local mutation signals implicit in the kernel domain, thereby reconstructing a physically interpretable composite feature space. Specifically, through the implicit feature selection mechanism in the tensor dimensionality reduction process, the macro-ecological laws and micro-dynamic responses are integrated into a unified aquaculture water parameter intrinsic component self-attention encoding vector. This allows the feature representation to retain the causal main chain of multi-scale parameter evolution while embedding nonlinear perturbation traces of local emergencies, thereby improving the robustness of the discrimination of complex working conditions.

[0071] Finally, the set of kernel self-attention encoding vectors of the aquaculture water parameter intrinsic components is cascaded and fused to obtain the optimized aquaculture water parameter feature vector, which is expressed as follows:

[0072] V′=Concat{z1,z2,…,z m}

[0073] Among them, Concat represents the cascade function, z1, z2, z m They represent the first, second and mth aquaculture water parameter intrinsic component kernel self-attention encoding vectors respectively, and V' represents the optimized aquaculture water parameter feature vector.

[0074] That is, through cascade fusion, the independent information of the kernel self-attention encoding vector of the intrinsic components of all aquaculture water parameters can be retained, and a more compact comprehensive representation can be sought to facilitate the following legal processing.

[0075] In this embodiment of the present application, the aquaculture waters result analysis module 140 is configured to pass the optimized aquaculture waters parameter feature vector through a classifier to obtain a classification result. The classification result indicates whether the aquaculture waters environment is suitable. It should be understood that the classifier can effectively convert the optimized aquaculture waters parameter feature vector into a specific classification result, thereby determining the suitability of the aquaculture waters environment. The classification result can help farmers understand whether the current environmental conditions meet specific standards or requirements, allowing them to take appropriate measures or adjustments. The classification results obtained by the classifier can provide farmers with rapid decision-making support. When the classification result indicates that the aquaculture waters environment is unsuitable, farmers can take timely action to adjust their aquaculture strategies or environmental management to ensure the smooth progress of aquaculture activities. Using a classifier to classify the optimized aquaculture waters parameter feature vectors can achieve automated environmental monitoring. Automatic classification to determine the suitability of the aquaculture waters environment can reduce manual monitoring costs, improve monitoring efficiency, and promptly detect environmental anomalies. By learning patterns and regularities in the feature matrix, the accuracy of predicting the suitability of the aquaculture waters environment can be improved. Continuously monitoring changes in environmental suitability can help farmers adjust their aquaculture strategies in a timely manner, ensuring the sustainable development of the aquaculture industry.

[0076] Accordingly, in one embodiment of the present application, the aquaculture water result analysis module includes: a water parameter fully-connected encoding unit, used to use the fully-connected layer of the classifier to fully-connected encode the optimized aquaculture water parameter feature vector to obtain a water parameter fully-connected encoded feature vector; a water parameter probability acquisition unit, used to pass the water parameter fully-connected encoded feature vector through the Softmax classification function of the classifier to obtain a first probability belonging to the suitable aquaculture water environment and a second probability belonging to the unsuitable aquaculture water environment; a water parameter classification result determination unit, used to determine the classification result based on a comparison between the first probability and the second probability.

[0077] In summary, the IoT-based smart aquaculture management system and method described in the embodiments of the present application monitors aquaculture environment parameters in real time through sensors and combines data analysis with artificial intelligence to provide farmers with refined management plans, optimize aquaculture conditions, increase production and efficiency, reduce risks, and promote the sustainable development of the aquaculture industry.

[0078] As described above, the IoT-based smart aquaculture management system 100 according to the embodiments of the present application can be implemented in various terminal devices, such as servers of the IoT-based smart aquaculture management system. In one example, the IoT-based smart aquaculture management system 100 can be integrated into the terminal device as a software module and / or hardware module. For example, the IoT-based smart aquaculture management system 100 can be a software module in the operating system of the terminal device, or an application developed specifically for the terminal device; of course, the IoT-based smart aquaculture management system 100 can also be one of the many hardware modules of the terminal device.

[0079] Alternatively, in another example, the IoT-based smart aquaculture management system 100 and the terminal device may also be separate devices, and the IoT-based smart aquaculture management system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0080] Figure 4 Flowchart of the smart aquaculture management method based on the Internet of Things according to the embodiment of the present application. Figure 4 As shown, the smart aquaculture management method based on the Internet of Things according to the embodiment of the present application includes the following steps: S110, extracting the dissolved oxygen value, pH value, conductivity value and water temperature value of the aquaculture water area at multiple predetermined time points including the current time point; S120, extracting the dissolved oxygen feature vector, pH value feature vector, conductivity feature vector and water temperature feature vector from the dissolved oxygen value, pH value, conductivity value and water temperature value at the multiple predetermined time points respectively; S130, constructing an aquaculture water area parameter feature vector between the dissolved oxygen feature vector, the pH value feature vector, the conductivity feature vector and the water temperature feature vector, and performing manifold regression-guided kernel space mapping processing on the aquaculture water area parameter feature vector to obtain an optimized aquaculture water area parameter feature vector; S140, passing the optimized aquaculture water area parameter feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the aquaculture water area environment is suitable.

[0081] Here, those skilled in the art will understand that the specific operations of each step in the above-mentioned smart aquaculture management method based on the Internet of Things have been referred to above. Figures 1 to 3 The invention has been introduced in detail in the description of the smart aquaculture management system based on the Internet of Things, and therefore, its repeated description will be omitted.

[0082] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0083] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0084] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0085] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0086] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0087] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0088] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A smart aquaculture management system based on the Internet of Things, characterized by: include: Aquaculture water data collection module, used to extract dissolved oxygen value, pH value, conductivity value and water temperature value of aquaculture waters at multiple predetermined time points including the current time point; an aquaculture water area data processing module, configured to extract a dissolved oxygen feature vector, a pH value feature vector, a conductivity feature vector, and a water temperature feature vector from the dissolved oxygen values, pH values, conductivity values, and water temperature values ​​at the plurality of predetermined time points, respectively; an aquaculture water data fusion module, configured to construct an aquaculture water parameter feature vector between the dissolved oxygen feature vector, the pH value feature vector, the conductivity feature vector, and the water temperature feature vector, and perform manifold regression-guided kernel space mapping processing on the aquaculture water parameter feature vector to obtain an optimized aquaculture water parameter feature vector; The aquaculture waters result analysis module is used to pass the optimized aquaculture waters parameter feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the aquaculture waters environment is suitable.

2. The smart aquaculture management system based on the Internet of Things according to claim 1 is characterized in that: The aquaculture waters data processing module includes: a dissolved oxygen feature extraction unit, configured to extract the dissolved oxygen feature vector from the dissolved oxygen values ​​at the plurality of predetermined time points; a pH value feature extraction unit, configured to extract the pH value feature vector from the pH values ​​at the plurality of predetermined time points; a conductivity feature extraction unit, configured to extract the conductivity feature vector from the conductivity values ​​at the plurality of predetermined time points; The water temperature feature extraction unit is configured to extract the water temperature feature vector from the water temperature values ​​at the plurality of predetermined time points.

3. The smart aquaculture management system based on the Internet of Things according to claim 2 is characterized in that: The dissolved oxygen feature extraction unit comprises: A dissolved oxygen arranging vector quantum unit, configured to arrange the dissolved oxygen values ​​at the plurality of predetermined time points into a dissolved oxygen input vector according to a time dimension; The dissolved oxygen multi-scale extraction subunit is used to pass the dissolved oxygen input vector through a dissolved oxygen value extractor based on a multi-scale neighborhood feature extraction module to obtain the dissolved oxygen feature vector.

4. The smart aquaculture management system based on the Internet of Things according to claim 3 is characterized in that: The dissolved oxygen multi-scale extraction subunit comprises: a first-scale dissolved oxygen encoding secondary subunit, configured to use a first convolutional layer of the dissolved oxygen value extractor based on the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the dissolved oxygen input vector using a one-dimensional convolution kernel having a first scale to obtain a first-scale dissolved oxygen feature vector, wherein the first convolutional layer has a first one-dimensional convolution kernel having a first length; a second-scale dissolved oxygen encoding secondary subunit, configured to use a second convolutional layer of the dissolved oxygen value extractor based on the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the dissolved oxygen input vector using a one-dimensional convolution kernel having a second scale to obtain a second-scale dissolved oxygen feature vector, wherein the second convolutional layer has a second one-dimensional convolution kernel having a second length, and the first length is different from the second length; The multi-scale dissolved oxygen cascade secondary subunit is configured to cascade the first-scale dissolved oxygen feature vector and the second-scale dissolved oxygen feature vector to obtain the dissolved oxygen feature vector.

5. The smart aquaculture management system based on the Internet of Things according to claim 4 is characterized in that: The pH value feature extraction unit comprises: a pH value arranging vector subunit, configured to arrange the pH values ​​of the plurality of predetermined time points into a pH input vector according to a time dimension; The pH value multi-scale extraction subunit is used to pass the pH input vector through a pH value extractor based on a multi-scale neighborhood feature extraction module to obtain the pH value feature vector.

6. The smart aquaculture management system based on the Internet of Things according to claim 5 is characterized in that: The conductivity feature extraction unit includes: a conductivity arranging vector quantum unit, configured to arrange the conductivity values ​​at the plurality of predetermined time points into a conductivity input vector according to a time dimension; The conductivity multi-scale extraction subunit is used to pass the conductivity input vector through the conductivity value extractor based on the multi-scale neighborhood feature extraction module to obtain the conductivity feature vector.

7. The smart aquaculture management system based on the Internet of Things according to claim 6, characterized in that: The water temperature feature extraction unit comprises: A water temperature arrangement vector subunit, configured to arrange the water temperature values ​​at the plurality of predetermined time points into a water temperature input vector according to a time dimension; The water temperature multi-scale extraction subunit is used to pass the water temperature input vector through a water temperature value extractor based on a multi-scale neighborhood feature extraction module to obtain the water temperature feature vector.

8. The smart aquaculture management system based on the Internet of Things according to claim 7 is characterized in that: The aquaculture waters data fusion module includes: a parameter arrangement unit, configured to arrange the dissolved oxygen characteristic vector, the pH value characteristic vector, the conductivity characteristic vector, and the water temperature characteristic vector according to parameter dimensions to obtain an aquaculture water parameter input matrix; A spatial attention unit, configured to input the aquaculture water parameter into a matrix and pass it through a first convolutional neural network model of spatial attention to obtain a feature vector of the aquaculture water parameter; The vector optimization unit is used to perform manifold regression-guided kernel space mapping processing on the aquaculture water parameter characteristic vector to obtain an optimized aquaculture water parameter characteristic vector.

9. The smart aquaculture management system based on the Internet of Things according to claim 8, characterized in that: The vector optimization unit is used to: Establishing a superpixel manifold correlation matrix of the aquaculture water parameter feature vector; Using deformable convolution to perform kernel domain representation learning on the superpixel manifold association matrix to obtain a kernel domain response dynamic matrix of aquaculture water parameters; Performing intrinsic mode decomposition on the superpixel manifold correlation matrix to obtain a set of multi-scale eigencomponent encoding vectors of aquaculture water parameters; Inputting the multi-scale intrinsic component encoding vector of the aquaculture water parameter into a sequence model based on a self-attention mechanism to obtain a set of calibrated intrinsic component encoding vectors of the aquaculture water parameter; Performing tensor contraction on the set of the aquaculture water parameter calibration intrinsic component encoding vectors and the aquaculture water parameter core domain response dynamic matrix to obtain a set of aquaculture water parameter intrinsic component core domain self-attention encoding vectors; The set of kernel domain self-attention encoding vectors of the aquaculture water parameter intrinsic components are cascaded and fused to obtain the optimized aquaculture water parameter feature vector.

10. A smart aquaculture management method based on the Internet of Things, characterized in that: include: Extracting dissolved oxygen values, pH values, conductivity values, and water temperature values ​​in the aquaculture waters at multiple predetermined time points, including the current time point; extracting a dissolved oxygen feature vector, a pH value feature vector, a conductivity feature vector, and a water temperature feature vector from the dissolved oxygen values, pH values, conductivity values, and water temperature values ​​at the plurality of predetermined time points, respectively; constructing an aquaculture water parameter feature vector between the dissolved oxygen feature vector, the pH value feature vector, the conductivity feature vector, and the water temperature feature vector, and performing a manifold regression-guided kernel space mapping process on the aquaculture water parameter feature vector to obtain an optimized aquaculture water parameter feature vector; The optimized aquaculture water parameter feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the aquaculture water environment is suitable.

Images