Aquaculture intelligent control system and method

Abstract

This application relates to the field of intelligent control, specifically disclosing an intelligent aquaculture control system and method. This system uses deep learning-based artificial intelligence technology to extract and encode features from aquatic product growth monitoring videos within a predetermined time period, as well as water quality parameters at multiple predetermined time points within the predetermined time period, to generate a personalized feeding plan. This automated control of aquatic product feeding ensures adequate nutrition for the product while minimizing waste and pollution. This not only improves aquaculture efficiency but also reduces feed costs and environmental impact.

Description

Technical Field

[0001] The present application relates to the field of intelligent control, and in particular to an intelligent aquaculture control system and method. Background Art

[0002] Aquatic products, rich in protein, essential amino acids, vitamins, and minerals, are an essential component of a healthy diet. With global population growth and economic development, the demand for aquatic products—including various marine and freshwater fish such as salmon, cod, carp, and crucian carp—continues to rise, driving the rapid development of the aquaculture industry. Since the turn of the century, this industry has undergone a dramatic transformation, transitioning from traditional small-scale, free-range farming to large-scale, scientifically managed practices. However, feeding methods for aquatic products often still rely heavily on experience and manual labor. These methods include timed and fixed-quantity feeding, as well as adjusting feed amounts according to the season and the growth stage of the aquatic product. This approach relies heavily on the farmer's personal experience and intuition, lacks scientific basis, and is difficult to precisely control feed amounts, which can easily lead to feed waste or insufficient feed.

[0003] Therefore, an optimized aquaculture control solution is desired. Summary of the Invention

[0004] This application is proposed to address the aforementioned technical issues. The embodiments of this application provide an intelligent aquaculture control system and method. These methods utilize deep learning-based artificial intelligence (AI) technologies to extract and encode features from aquatic product growth monitoring videos within a predetermined time period, as well as water quality parameters at multiple predetermined time points within the predetermined time period, to generate a personalized feeding plan. This automated control of aquatic product feeding ensures adequate nutrition for the aquatic product while minimizing waste and pollution. This not only improves aquaculture efficiency but also reduces feed costs and environmental impact.

[0005] According to one aspect of the present application, there is provided an intelligent aquaculture control system comprising:

[0006] An aquatic product growth data acquisition module, configured to acquire aquatic product growth monitoring video within a predetermined time period and water quality environment parameters at a plurality of predetermined time points within the predetermined time period;

[0007] an aquatic product growth feature extraction module, configured to extract and encode features of the aquatic product growth monitoring video within the predetermined time period to obtain a semantic understanding feature vector of aquatic product growth;

[0008] A water quality time series feature extraction module is used to extract and encode the time series features of the water quality environment parameters to obtain a water quality environment feature vector;

[0009] The aquatic product feeding plan generation module is used to comprehensively analyze the aquatic product growth semantic understanding feature vector and the water quality environment feature vector to generate a personalized feeding plan.

[0010] According to another aspect of the present application, there is also provided an intelligent control method for aquaculture, which comprises:

[0011] Acquiring aquatic product growth monitoring video within a predetermined time period and water quality environmental parameters at multiple predetermined time points within the predetermined time period;

[0012] Extracting and encoding features of the aquatic product growth monitoring video within the predetermined time period to obtain a semantic understanding feature vector of aquatic product growth;

[0013] Extracting and encoding the water quality environment parameters in a time series manner to obtain a water quality environment feature vector;

[0014] The aquatic product growth semantic understanding feature vector and the water quality environment feature vector are comprehensively analyzed to generate a personalized feeding plan.

[0015] In summary, the intelligent aquaculture control system and method provided in this application utilizes deep learning-based artificial intelligence technology to extract and encode features from aquatic product growth monitoring videos within a predetermined time period, as well as water quality parameters at multiple predetermined time points within that time period, to generate a personalized feeding plan. This automated control of aquatic product feeding ensures adequate nutrition for the aquatic product while minimizing waste and pollution. This not only improves aquaculture efficiency but also reduces feed costs and environmental impact. 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 used 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. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without inventive effort. In the drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 This is a block diagram of an intelligent aquaculture control system according to an embodiment of the present application.

[0018] Figure 2 This is a block diagram of an aquatic product growth feature extraction module in an aquaculture intelligent control system according to an embodiment of the present application.

[0019] Figure 3This is a block diagram of a semantic understanding unit for aquatic product growth characteristics in an intelligent aquaculture control system according to an embodiment of the present application.

[0020] Figure 4 This is a flow chart of the intelligent control method for aquaculture according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] Below, the exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings, clearly and completely describing the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0022] Figure 1 FIG is a block diagram of an intelligent aquaculture control system according to an embodiment of the present application. Figure 1 As shown, the intelligent aquaculture control system 100 according to the embodiment of the present application includes: an aquatic product growth data acquisition module 110, which is used to obtain aquatic product growth monitoring video within a predetermined time period and water quality environment parameters at multiple predetermined time points within the predetermined time period; an aquatic product growth feature extraction module 120, which is used to extract and encode the aquatic product growth monitoring video within the predetermined time period to obtain an aquatic product growth semantic understanding feature vector; a water quality time series feature extraction module 130, which is used to extract and encode the time series features of the water quality environment parameters to obtain a water quality environment feature vector; and an aquatic product feeding plan generation module 140, which is used to comprehensively analyze the aquatic product growth semantic understanding feature vector and the water quality environment feature vector to generate a personalized feeding plan.

[0023] In the aforementioned intelligent aquaculture control system 100, the aquatic product growth data acquisition module 110 is used to acquire aquatic product growth monitoring video within a predetermined time period, as well as water quality parameters at multiple predetermined time points within the predetermined time period. As discussed in the aforementioned background technology, aquatic product feeding methods often rely primarily on experience and manual operation. Specific methods include timed and quantitative feeding, as well as adjusting the feed amount based on the season and the growth stage of the aquatic product. This method relies heavily on the farmer's personal experience and intuition, lacks scientific basis, and is difficult to accurately control, which can easily lead to feed waste or insufficient supply. Therefore, an optimized aquaculture control solution is desired.

[0024] To address these technical challenges, we have proposed an intelligent aquaculture control solution. This solution uses deep learning-based artificial intelligence (AI) to extract and encode features from aquatic product growth monitoring videos and water quality parameters at multiple predetermined time points within a predetermined time period to generate personalized feeding plans. This automated control of aquatic product feeding ensures adequate nutrition while minimizing waste and pollution. This not only improves aquaculture profitability but also reduces feed costs and environmental impact.

[0025] More specifically, in the technical solution of this application, high-definition cameras are installed in the aquaculture area to capture videos of aquatic product growth. Various water quality monitoring sensors are installed in the aquaculture waters to monitor water quality parameters such as water temperature, dissolved oxygen, pH, and ammonia nitrogen content in real time. The cameras and sensors collect real-time data on aquatic product growth and water quality. The collected video and water quality data are then transmitted to a central processing system. Deep learning algorithms are used to analyze the video data to extract growth characteristics of aquatic products, such as size and activity. The water quality monitoring data is then processed to extract key environmental parameter features. A deep learning model is used to learn and analyze the extracted features to identify the growth stage, health status, and feeding requirements of aquatic products. Based on the analysis results, a personalized feeding plan is then developed, including feed amount, feeding time, and feeding frequency. The feeding plan is adjusted to adapt to environmental changes, taking into account water quality and environmental parameters. This intelligent control system enables precise control and management of the aquaculture process, improving aquaculture efficiency, reducing costs, and minimizing environmental impact. This approach also helps improve the quality and market competitiveness of aquatic products.

[0026] Specifically, first, a video monitoring video of aquatic product growth over a predetermined time period is obtained, along with water quality parameters at multiple predetermined time points within the predetermined time period. Video monitoring allows for intuitive observation of aquatic product activity, growth rate, and health. Water quality parameters are key indicators for assessing whether the aquaculture environment is suitable for aquatic product growth. Aquatic product growth and feeding behavior are closely related to environmental parameters, and monitoring these parameters helps formulate appropriate feeding plans. A comprehensive analysis combining video monitoring and water quality monitoring data yields more comprehensive aquaculture management information.

[0027] In the above-mentioned intelligent aquaculture control system 100, the aquatic product growth feature extraction module 120 is used to extract and encode features of the aquatic product growth monitoring video within the predetermined time period to obtain a feature vector for semantic understanding of aquatic product growth.

[0028] Figure 2 FIG. 1 is a block diagram of a module for extracting aquatic product growth characteristics in an intelligent aquaculture control system according to an embodiment of the present application. Figure 2As shown, the aquatic product growth feature extraction module 120 includes: an aquatic product growth monitoring key frame extraction unit 121, used to extract multiple aquatic product growth monitoring key frames from the aquatic product growth monitoring video; an aquatic product monitoring feature extraction unit 122, used to perform feature extraction on the multiple aquatic product growth monitoring key frames respectively to obtain multiple aquatic product growth monitoring feature matrices; an aquatic product feature embedding coding unit 123, used to embed code the multiple aquatic product growth monitoring feature matrices to obtain multiple aquatic product growth monitoring feature vectors; and an aquatic product growth feature semantic understanding unit 124, used to perform semantic understanding coding on the multiple aquatic product growth monitoring feature vectors to obtain aquatic product growth semantic understanding feature vectors.

[0029] More specifically, multiple key frames for aquatic product growth monitoring are extracted from aquatic product growth monitoring videos. Video files typically contain large amounts of data, resulting in high processing and storage costs. Extracting key frames significantly reduces this data volume, facilitating subsequent processing and analysis. Key frames are often the most representative images in a video, concisely reflecting the main content of the entire video. Key frames serve as the basis for extracting growth characteristics of aquatic products, such as changes in body shape and activity, facilitating more in-depth analysis. In real-time monitoring systems, key frames can quickly provide the current status of aquatic products, facilitating timely problem detection. Automated systems can use key frames for image processing and pattern recognition, improving analysis efficiency and accuracy. Furthermore, compared to processing the entire video, analyzing key frames significantly reduces computing resource consumption. Key frames also occupy less storage space, facilitating long-term storage and comparative analysis of historical data. Key frames provide the information necessary for rapid decision-making, facilitating the development of timely aquaculture management strategies. Using key frames can reduce misjudgments caused by poor video quality or environmental factors. Methods for extracting keyframes for aquatic product growth monitoring typically include content-based methods (such as motion detection and scene change) and time-based methods (such as fixed-time interval extraction). During the extraction process, it is necessary to ensure that the keyframes truly reflect the growth status of the aquatic product, providing a reliable basis for subsequent analysis and decision-making.

[0030] Specifically, in the embodiment of the present application, the aquatic product growth monitoring key frame extraction unit 121 is used to extract multiple aquatic product growth monitoring key frames from the aquatic product growth monitoring video at a predetermined sampling frequency.

[0031] Specifically, in the embodiment of the present application, the aquatic product monitoring feature extraction unit 122 is used to: pass the multiple aquatic product growth monitoring key frames through an aquatic product growth feature extractor based on a convolutional neural network model to obtain multiple aquatic product growth monitoring feature matrices.

[0032] More specifically, the multiple aquatic product growth monitoring key frames are passed through an aquatic product growth feature extractor based on a convolutional neural network model to obtain multiple aquatic product growth monitoring feature matrices. Convolutional neural networks (CNNs) are capable of automatically learning useful features from images. By using CNN models, rich feature information crucial for aquatic product growth analysis can be extracted from key frames, providing strong data support for intelligent control systems. This is of great significance for achieving precise aquaculture management and improving aquaculture efficiency.

[0033] Specifically, in an embodiment of the present application, the aquatic product monitoring feature extraction unit 122 is used to: use each layer of the aquatic product growth feature extractor based on the convolutional neural network model to perform the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; pooling processing on the convolution feature map along the channel dimension to obtain a pooled feature map; and, nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the aquatic product growth feature extractor based on the convolutional neural network model is the aquatic product growth monitoring feature matrix, and the input of the first layer of the aquatic product growth feature extractor based on the convolutional neural network model is the aquatic product growth monitoring key frame.

[0034] Specifically, in the embodiment of the present application, the aquatic product feature embedding coding unit 123 is used to: pass the multiple aquatic product growth monitoring feature matrices through a linear embedding layer to obtain multiple aquatic product growth monitoring feature vectors.

[0035] More specifically, the multiple aquatic product growth monitoring feature matrices are passed through a linear embedding layer to obtain multiple aquatic product growth monitoring feature vectors. A linear embedding layer is a neural network layer that uses a linear transformation to map input data (feature matrix) to a low-dimensional space (feature vector). Linear embedding layers are commonly used for dimensionality reduction, converting high-dimensional feature matrices into low-dimensional feature vectors to reduce computational complexity. The embedding layer can integrate key information from multiple feature matrices into a unified feature vector for ease of subsequent processing. Through the embedding layer, features from different sources or at different scales can be standardized to give them a unified representation. A key feature of the linear embedding layer is that its weight matrix is ​​learnable, meaning that during training, the model automatically adjusts based on the data weight matrix to find the optimal linear mapping. By using a linear embedding layer, the high-dimensional feature matrix extracted by the convolutional neural network can be effectively converted into a low-dimensional feature vector, providing input for subsequent deep learning tasks (such as semantic understanding and feature fusion). This approach is very common when processing high-dimensional data such as images and text.

[0036] Specifically, in an embodiment of the present application, the aquatic product feature embedding coding unit 123 is used to: use the linear embedding layer to linearly project each aquatic product growth monitoring feature matrix in the multiple aquatic product growth monitoring feature matrices with a learnable embedding matrix to obtain the multiple aquatic product growth monitoring feature vectors.

[0037] More specifically, the multiple aquatic product growth monitoring feature vectors are passed through a transformer-based aquatic product growth semantic feature extractor to obtain aquatic product growth semantic understanding feature vectors. The aquatic product growth semantic feature extractor is a transformer-based context encoder model. Transformer models, particularly context encoders, can capture contextual information in feature vectors, which is crucial for understanding complex patterns in aquatic product growth. Transformer models can effectively capture long-range dependencies, which is very useful for analyzing long-term trends and patterns in aquatic product growth. A context encoder model is a deep learning model that utilizes a transformer architecture to encode contextual information in input data. In natural language processing, a context encoder generally refers to a model that encodes the contextual information of a word or phrase into its representation. In the context of aquatic product growth monitoring, a context encoder model can take one of the following forms: 1. BERT (Bidirectional Encoder Representations from Transformers): A bidirectional context encoder that considers both preceding and following context information. 2. GPT (Generative Pre-trained Transformer): A unidirectional autoregressive language model that generates the next item based on the previous context. 3. XLM (Cross-lingual Language Model): A cross-lingual transformer model capable of handling multiple languages. These models utilize an attention mechanism to capture complex relationships within input sequences and generate feature representations rich in contextual information. In aquatic growth monitoring, context encoders can help the model understand complex patterns and trends within feature vectors, thereby providing more accurate feature vectors for semantic understanding of growth and supporting the development of more scientific farming strategies.

[0038] Figure 3 FIG. 1 is a block diagram of a semantic understanding unit for aquatic product growth characteristics in an intelligent aquaculture control system according to an embodiment of the present application. Figure 3As shown, the aquatic product growth feature semantic understanding unit 124 is used for: a query vector construction subunit 11, which is used to arrange the multiple aquatic product growth monitoring feature vectors into an input vector; a vector conversion subunit 12, which is used to convert the input vector into a query vector and a key vector respectively through a learnable embedding matrix; a self-attention subunit 13, which is used to calculate the product between the query vector and the transposed vector of the key vector to obtain a self-attention association matrix; a normalization subunit 14, which is used to normalize the self-attention association matrix to obtain a standardized self-attention association matrix; an attention calculation subunit 15, which is used to input the standardized self-attention association matrix into a Softmax activation function for activation to obtain a self-attention feature matrix; and an attention application subunit 16, which is used to multiply the self-attention feature matrix with each aquatic product growth monitoring feature vector in the multiple aquatic product growth monitoring feature vectors to obtain the multiple aquatic product growth semantic context feature vectors; a cascade subunit 17, which is used to cascade the multiple aquatic product growth semantic context feature vectors to obtain an aquatic product growth semantic understanding feature vector.

[0039] In the above-mentioned intelligent aquaculture control system 100, the water quality time series feature extraction module 130 is used to extract and encode the time series features of the water quality environment parameters to obtain a water quality environment feature vector.

[0040] Specifically, in the embodiment of the present application, the water quality time series feature extraction module 130 is used to: pass the water quality environment parameter through a water quality environment time series encoder to obtain a water quality environment feature vector.

[0041] More specifically, the water quality environment parameters are passed through a water quality environment time series encoder to obtain a water quality environment feature vector. The water quality environment time series encoder is a time series encoder comprising a one-dimensional convolutional layer and a fully connected layer. The water quality parameters change over time, forming time series data. The time series encoder can analyze this time-varying data and extract useful features. A time series encoder is a neural network structure used to process time series data, which can convert time series data into a feature vector of fixed length. A time series encoder typically includes the following key components: 1. One-dimensional convolutional layer: used to extract local features in the time dimension. One-dimensional convolution can capture local patterns and trends in time series data. 2. Pooling layer: usually used in combination with a one-dimensional convolutional layer to reduce the dimension of the features, extract key information, and increase invariance to time changes. 3. Recurrent layer: such as a long short-term memory network (LSTM) or a gated recurrent unit (GRU), used to process long-term dependencies in time series data. 4. Fully Connected Layer (Dense Layer): At the end of the network, it is used to convert the extracted features into the final feature vector. 5. Normalization and Activation Function: Normalization and activation functions, such as batch normalization and ReLU activation functions, are used in different parts of the network to improve model performance and stability. By using a time series encoder, key features reflecting water quality changes can be extracted from time series data of water quality environmental parameters, providing accurate water quality information for intelligent control systems, thereby achieving more effective water quality management and decision support.

[0042] Specifically, in an embodiment of the present application, the water quality environment temporal encoder is a temporal encoder model using a one-dimensional convolutional layer and a fully connected layer.

[0043] Specifically, in the embodiment of the present application, the water quality time series feature extraction module 130 is configured to: a vector construction unit for arranging the water quality environment parameters at the multiple time points into an input vector; a fully connected encoding unit for using the fully connected layer of the water quality environment time series encoder to fully connect the input vector using the following fully connected encoding formula to extract high-dimensional implicit features of the eigenvalues ​​at each position in the input vector, wherein the fully connected encoding formula is:

[0044]

[0045] Where X is the input vector, Y is the output vector, W is the weight matrix, and B is the bias vector. represents matrix multiplication; and a one-dimensional convolutional encoding unit, configured to perform one-dimensional convolution encoding on the input vector using the one-dimensional convolution layer of the water quality environment temporal encoder using the following one-dimensional convolution formula to extract high-dimensional implicit correlation features of the correlation between the eigenvalues ​​at each position in the input vector, wherein the one-dimensional convolution formula is:

[0046]

[0047] Wherein, a is the width of the first convolution kernel in the x direction, F(a) is the first convolution kernel parameter vector, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, X represents the input vector, and Cov(X) represents one-dimensional convolution encoding of the input vector.

[0048] In the above-mentioned intelligent aquaculture control system 100, the aquatic product feeding plan generation module 140 is used to comprehensively analyze the aquatic product growth semantic understanding feature vector and the water quality environment feature vector to generate a personalized feeding plan.

[0049] Specifically, in an embodiment of the present application, the aquatic product feeding plan generation module 140 includes: an aquatic product feeding feature fusion unit, used to fuse the water quality environment feature vector and the aquatic product growth semantic understanding feature vector to obtain an aquatic product feeding feature matrix; an optimization unit, used to perform a coherent attention response based on deep anchoring features on the aquatic product feeding feature matrix to obtain an optimized aquatic product feeding feature matrix; and an aquatic product feeding plan generation unit, used to pass the optimized aquatic product feeding feature matrix through a generator to generate a personalized feeding plan.

[0050] More specifically, the water quality and environmental feature vectors are fused with the aquatic product growth semantic understanding feature vectors to generate an aquatic product feeding feature matrix. Water quality and growth characteristics may contain complementary information, and fusion can fully utilize this information to improve the model's predictive power and accuracy. The fusion of these two feature vectors can be achieved through a variety of methods, such as feature concatenation, weighted averaging, and feature interaction. The specific method chosen depends on the characteristics of the data and the specific needs of aquaculture management. The fused feature matrix serves as the basis for generating personalized feeding plans within the intelligent control system.

[0051] Specifically, in the embodiment of the present application, the aquatic product feeding feature fusion unit is used to fuse the water quality environment feature vector and the aquatic product growth semantic understanding feature vector using the following aquatic product feeding feature fusion formula to obtain an aquatic product feeding feature matrix; wherein the aquatic product feeding feature fusion formula is:

[0052]

[0053] in, represents vector multiplication, M represents the aquatic product feeding feature matrix, F1 represents the water quality environment feature vector, F2 represents the aquatic product growth semantic understanding feature vector, Represents the transpose of the aquatic product growth semantic understanding feature vector.

[0054] In particular, in the technical solution of this application, the aquatic product feeding feature matrix is ​​a key information carrier that combines water quality parameters and aquatic product growth characteristics, and is used to generate personalized feeding plans. Improving the feature significance of this feature matrix and avoiding feature redundancy are crucial to ensuring the scientific and effective feeding plans. Specifically, significant features can more accurately reflect key information about the water quality environment and the growth status of aquatic products, thereby generating more precise feeding plans. Significant features help the system adapt to the growth needs of different aquatic products and changes in the water quality environment, improving the applicability of feeding plans. Significant features help predict the growth trends and nutritional requirements of aquatic products, thereby formulating more scientific feeding strategies. Furthermore, removing redundant features can reduce computational complexity and improve the speed and efficiency of feeding plan generation. Redundant features can lead to model overfitting, affecting its performance in practical applications. Avoiding feature redundancy helps improve the model's generalization ability. Furthermore, significant features help generate more personalized feeding plans that meet the specific needs of different aquatic products. Precise features can avoid unnecessary feed waste and improve feed utilization efficiency. Significant features help formulate feeding plans that better promote the healthy growth of aquatic products. Based on this, a coherent attention response based on deep anchoring features is performed on the aquatic product feeding feature matrix to obtain an optimized aquatic product feeding feature matrix.

[0055] Specifically, in an embodiment of the present application, the optimization unit includes: performing a coherent attention response based on a deep anchor feature on the aquatic product feeding feature matrix to obtain an optimized aquatic product feeding feature matrix, including: a feature flattening subunit, used to perform feature flattening on the aquatic product feeding feature matrix to obtain an aquatic product feeding feature vector; a coherent attention response subunit, used to perform a coherent attention response based on a deep anchor feature on the aquatic product feeding feature vector to obtain an optimized aquatic product feeding feature vector; and a reverse feature aggregation subunit, used to perform reverse feature aggregation on the optimized aquatic product feeding feature vector in accordance with the feature flattening method to obtain the optimized aquatic product feeding feature matrix.

[0056] Specifically, in an embodiment of the present application, the consistency attention response subunit includes: determining the mean of the aquatic product feeding feature vector; accumulating and summing the aquatic product feeding feature vectors to obtain a total value of the aquatic product feeding feature vectors; subtracting the mean of the aquatic product feeding feature vector from the total value of the aquatic product feeding feature vector, dividing the result by a predetermined hyperparameter, and then dividing the result by the length of the aquatic product feeding feature vector to obtain a first quotient value; calculating an exponential function value with a natural constant as the base for the first quotient value to obtain a first natural exponential function value; and multiplying the first natural exponential function value by the aquatic product feeding feature vector by a position point to obtain an optimized aquatic product feeding feature vector.

[0057] More specifically, in an embodiment of the present application, the consistency attention response subunit is configured to perform a consistency attention response based on the depth anchoring feature on the aquatic product feeding feature vector using the following consistency attention response formula to obtain an optimized aquatic product feeding feature vector, wherein the consistency attention response formula is:

[0058]

[0059] Among them, V1 represents the aquatic product feeding feature vector, v i represents the eigenvalue of the i-th position of the aquatic product feeding eigenvector, represents the mean of the aquatic product feeding feature vector, L represents the length of the aquatic product feeding feature vector, ε represents a predetermined hyperparameter, ⊙ represents the position point multiplication, V1 ′ Represents the optimized aquatic product feeding feature vector.

[0060] Specifically, in order to improve the feature significance of the aquatic product feeding feature matrix and avoid feature redundancy, in the technical solution of the present application, a coherent attention response based on a depth anchor feature is performed on the aquatic product feeding feature matrix, which uses the global mean of the aquatic product feeding feature matrix as its dynamically formed depth anchor feature, and applies a coherent attention response mechanism based on the depth anchor feature to the original feature distribution, so as to perform position scattering response depth correlation correction in the class divergence space with depth difference, so that the original feature distribution has a highly significant depth correlation distribution characteristic, so as to improve the feature significance of the aquatic product feeding feature matrix and avoid feature redundancy.

[0061] The optimized aquatic product feeding feature matrix is ​​then passed through a generator to generate a personalized feeding plan. Aquatic products in each aquaculture pond may have different growth states and needs, and the generator can generate personalized feeding plans based on these differences. The generator automatically creates feeding plans, reducing the need for manual decision-making and improving efficiency. By precisely controlling feeding amounts and frequency, the healthy growth of aquatic products can be promoted and the risk of disease can be reduced. Specifically, some examples of personalized feeding plans are as follows: 1. Growth stage adjustment: If the system detects that fish in a particular aquaculture pond are experiencing a period of rapid growth, the personalized feeding plan may increase the number of daily feedings and the amount of each feeding to meet the fish's increased nutritional needs. 2. Health status monitoring: If surveillance video and water quality parameter analysis indicate that certain fish are experiencing health issues, such as loss of appetite, the personalized feeding plan may reduce the feeding amount and adjust the feed formula to increase ingredients that aid recovery. 3. Environmental adaptation: During hot seasons, dissolved oxygen levels may decrease. The personalized feeding plan may adjust feeding times to avoid periods of lowest dissolved oxygen levels to reduce stress on the fish due to hypoxia. 4. Behavioral pattern learning: By learning the behavioral patterns of aquatic products, the system finds that the fish in a certain breeding pond are most active in feeding in the early morning. The personalized feeding plan may arrange for larger-scale feeding during this time period.

[0062] Specifically, in an embodiment of the present application, the model of the generator can be a generative adversarial network model or a variational autoencoder model, which is trained on the optimized aquatic product feeding feature matrix to map the optimized aquatic product feeding feature matrix to a representation of a personalized feeding plan.

[0063] In summary, the intelligent aquaculture control system according to the embodiments of this application has been described. It uses artificial intelligence technologies based on deep learning to extract and encode features from aquatic product growth monitoring videos within a predetermined time period, as well as water quality parameters at multiple predetermined time points within the predetermined time period, to generate a personalized feeding plan. This automated control of aquatic product feeding ensures that the aquatic product receives adequate nutrition while avoiding waste and pollution. This not only improves aquaculture efficiency but also reduces feed costs and environmental impact.

[0064] As described above, the aquaculture intelligent control system 100 according to the embodiments of the present application can be implemented in various terminal devices, such as an aquaculture intelligent control server. In one example, the aquaculture intelligent control system 100 according to the embodiments of the present application can be integrated into the terminal device as a software module and / or hardware module. For example, the aquaculture intelligent control system 100 can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device. Of course, the aquaculture intelligent control system 100 can also be one of the many hardware modules of the terminal device.

[0065] Alternatively, in another example, the aquaculture intelligent control system 100 and the terminal device may also be separate devices, and the aquaculture intelligent control 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.

[0066] Based on the same inventive concept, an embodiment of the present application also provides an intelligent control method for aquaculture, which can be used to implement the system described in the above embodiment, as described in the following embodiments.

[0067] Figure 4 FIG. 1 is a flow chart of an intelligent control method for aquaculture according to an embodiment of the present application. Figure 4 As shown, the intelligent control method for aquaculture according to the embodiment of the present application includes the steps of: S110, obtaining a monitoring video of aquatic product growth within a predetermined time period and water quality environment parameters at multiple predetermined time points within the predetermined time period; S120, performing feature extraction and encoding on the monitoring video of aquatic product growth within the predetermined time period to obtain a semantic understanding feature vector of aquatic product growth; S130, performing time series feature extraction and encoding on the water quality environment parameters to obtain a water quality environment feature vector; and S140, comprehensively analyzing the semantic understanding feature vector of aquatic product growth and the water quality environment feature vector to generate a personalized feeding plan.

[0068] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. Here, for the aquaculture intelligent control method disclosed in the embodiment, those skilled in the art will understand that the specific operations of each step in the above-mentioned aquaculture intelligent control method have been referred to above. Figures 1 to 3 It has been introduced in detail in the description of the aquaculture intelligent control system, so the description is relatively simple. For relevant details, please refer to the description of the aquaculture intelligent control system part, and therefore, its repeated description will be omitted.

[0069] In summary, the intelligent aquaculture control method according to the embodiments of this application has been described. It uses deep learning-based artificial intelligence technology to extract and encode features from aquatic product growth monitoring videos within a predetermined time period, as well as water quality parameters at multiple predetermined time points within the predetermined time period, to generate a personalized feeding plan. This automated control of aquatic product feeding ensures that the aquatic product receives adequate nutrition while avoiding waste and pollution. This not only improves aquaculture efficiency but also reduces feed costs and environmental impact.

[0070] In summary, after reading this detailed disclosure, those skilled in the art will appreciate that the foregoing detailed disclosure may be presented by way of example only and may not be limiting. Although not explicitly stated herein, those skilled in the art will understand that this application is intended to encompass various reasonable changes, improvements, and modifications to the embodiments. Such changes, improvements, and modifications are intended to be proposed by this application and are within the spirit and scope of the exemplary embodiments of this application.

[0071] In addition, certain terms in this application have been used to describe embodiments of the present application. For example, "one embodiment," "an embodiment," and / or "some embodiments" mean that a particular feature, structure, or characteristic described in conjunction with that embodiment may be included in at least one embodiment of the present application. Therefore, it is emphasized and should be understood that two or more references to "an embodiment," "one embodiment," or "an alternative embodiment" in various parts of this specification do not necessarily refer to the same embodiment. Furthermore, particular features, structures, or characteristics may be appropriately combined in one or more embodiments of the present application.

[0072] Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to encompass non-exclusive inclusion, such that an article or device comprising a list of elements includes not only those elements but also other elements not explicitly listed or inherent to such article or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the article or device comprising the aforementioned elements.

[0073] It should be understood that in the foregoing description of the embodiments of this application, in order to facilitate understanding of a feature and to simplify this application, this application combines various features into a single embodiment, figure, or description thereof. However, this does not mean that the combination of these features is required. When reading this application, it is entirely possible for those skilled in the art to extract some of the features and understand them as separate embodiments. In other words, the embodiments of this application can also be understood as the integration of multiple secondary embodiments. This also applies when the content of each secondary embodiment is less than all the features of a single aforementioned disclosed embodiment.

[0074] Finally, it should be understood that the embodiments of the application disclosed herein are illustrations of the principles of the embodiments of the present application. Other modified embodiments are also within the scope of the present application. Therefore, the embodiments disclosed in this application are merely examples and not limitations. Those skilled in the art can adopt alternative configurations to implement the application in this application based on the embodiments in this application.

[0075] Therefore, the embodiments of the present application are not limited to the precise embodiments described in the application.

Claims

1. An intelligent aquaculture control system, characterized in that: include: An aquatic product growth data acquisition module, configured to acquire aquatic product growth monitoring video within a predetermined time period and water quality environment parameters at a plurality of predetermined time points within the predetermined time period; an aquatic product growth feature extraction module, configured to extract and encode features of the aquatic product growth monitoring video within the predetermined time period to obtain a semantic understanding feature vector of aquatic product growth; A water quality time series feature extraction module is used to extract and encode the time series features of the water quality environment parameters to obtain a water quality environment feature vector; An aquatic product feeding plan generation module, configured to comprehensively analyze the aquatic product growth semantic understanding feature vector and the water quality environment feature vector to generate a personalized feeding plan; Wherein, the aquatic product feeding plan generation module includes: an aquatic product feeding feature fusion unit, configured to fuse the water quality environment feature vector and the aquatic product growth semantic understanding feature vector to obtain an aquatic product feeding feature matrix; an optimization unit, configured to perform a coherent attention response based on deep anchoring features on the aquatic product feeding feature matrix to obtain an optimized aquatic product feeding feature matrix; an aquatic product feeding plan generating unit, configured to pass the optimized aquatic product feeding characteristic matrix through a generator to generate a personalized feeding plan; Wherein, the optimization unit includes: a feature flattening subunit, configured to perform feature flattening on the aquatic product feeding feature matrix to obtain an aquatic product feeding feature vector; a coherence attention response subunit, configured to perform a coherence attention response based on a deep anchor feature on the aquatic product feeding feature vector to obtain an optimized aquatic product feeding feature vector; The reverse feature aggregation subunit is used to perform reverse feature aggregation on the optimized aquatic product feeding feature vector according to the feature flattening method to obtain the optimized aquatic product feeding feature matrix.

2. The aquaculture intelligent control system according to claim 1, characterized in that: The aquatic product growth feature extraction module includes: an aquatic product growth monitoring key frame extraction unit, configured to extract a plurality of aquatic product growth monitoring key frames from the aquatic product growth monitoring video; an aquatic product monitoring feature extraction unit, configured to extract features from the plurality of aquatic product growth monitoring key frames to obtain a plurality of aquatic product growth monitoring feature matrices; an aquatic product feature embedding coding unit, configured to embed and code the plurality of aquatic product growth monitoring feature matrices to obtain a plurality of aquatic product growth monitoring feature vectors; The aquatic product growth feature semantic understanding unit is used to perform semantic understanding encoding on the multiple aquatic product growth monitoring feature vectors to obtain an aquatic product growth semantic understanding feature vector.

3. The aquaculture intelligent control system according to claim 2, characterized in that: The aquatic product monitoring feature extraction unit is used to: The multiple aquatic product growth monitoring key frames are passed through an aquatic product growth feature extractor based on a convolutional neural network model to obtain multiple aquatic product growth monitoring feature matrices.

4. The intelligent aquaculture control system according to claim 3, characterized in that: The aquatic product feature embedding coding unit is used to: The multiple aquatic product growth monitoring feature matrices are passed through a linear embedding layer to obtain multiple aquatic product growth monitoring feature vectors.

5. The intelligent aquaculture control system according to claim 4, characterized in that: The aquatic product growth feature semantic understanding unit is used to: The multiple aquatic product growth monitoring feature vectors are passed through a converter-based aquatic product growth semantic feature extractor to obtain an aquatic product growth semantic understanding feature vector.

6. The aquaculture intelligent control system according to claim 5, characterized in that: The water quality time series feature extraction module is used to: The water quality environment parameters are passed through a water quality environment time series encoder to obtain a water quality environment feature vector.

7. The aquaculture intelligent control system according to claim 6, characterized in that: The coherence attention response subunit includes: Determining the mean of the aquatic product feeding characteristic vector; Accumulating and summing the aquatic product feeding characteristic vectors to obtain a total value of the aquatic product feeding characteristic vectors; subtracting the mean of the aquatic product feeding feature vector from the sum of the aquatic product feeding feature vectors, dividing the sum by a predetermined hyperparameter, and then dividing by the length of the aquatic product feeding feature vector to obtain a first quotient; Calculating an exponential function value with a natural constant as a base on the first quotient to obtain a first natural exponential function value; The first natural exponential function value is multiplied by the aquatic product feeding characteristic vector at each position point to obtain an optimized aquatic product feeding characteristic vector.

8. An intelligent control method for aquaculture, using the intelligent aquaculture control system according to claim 1, characterized in that: include: Acquiring aquatic product growth monitoring video within a predetermined time period and water quality environmental parameters at multiple predetermined time points within the predetermined time period; Extracting and encoding features of the aquatic product growth monitoring video within the predetermined time period to obtain a semantic understanding feature vector of aquatic product growth; Extracting and encoding the water quality environment parameters in a time series manner to obtain a water quality environment feature vector; The aquatic product growth semantic understanding feature vector and the water quality environment feature vector are comprehensively analyzed to generate a personalized feeding plan.

Images