Aquaculture monitoring method and system based on multi-mode sensing technology

Abstract

The invention discloses an aquaculture monitoring method and system based on a multi-modal sensing technology, and relates to the technical field of aquaculture, a plurality of sensors are integrated according to the layout of an aquaculture farm and the flowing condition of a water body, and a multi-modal sensor array is formed; the multi-mode sensor array is used for regularly collecting water body monitoring data; segmenting the water body monitoring data according to a preset time period, and extracting monitoring sample data; integrating the monitoring sample data by using a data fusion algorithm to form a multi-dimensional monitoring data set; a self-adaptive fusion mechanism is introduced to adjust a fusion strategy according to growth environment characteristics of different aquatic products; and processing the multi-dimensional monitoring data set by adopting a deep learning technology to obtain a monitoring result. The aquatic product and water quality parameters are comprehensively evaluated, the monitoring efficiency and accuracy of the environment and water quality are effectively improved, and the survival rate and quality of the aquatic product are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of aquaculture, and more particularly to an aquaculture monitoring method and system using multimodal sensing technology. Background Art

[0002] Aquaculture is an agricultural activity that has extremely high requirements for the ecological environment. Its success is directly affected by a variety of environmental factors and water quality parameters. These factors include air pressure, oxygen content in the air, dissolved oxygen in water, pH value and ammonia nitrogen content. Each factor is key to the growth and survival of aquatic products. Effective monitoring and management of changes in these factors is essential to improving the survival rate and quality of aquatic products. However, in reality, many aquaculture sites have not yet conducted a comprehensive assessment of aquatic products and water quality data. At the same time, most farm monitoring work still relies on manual operations. Staff usually need to regularly go to the waters for sampling and testing, and then conduct laboratory analysis. This process is not only time-consuming, but also limited by the professional knowledge and experience of technicians. For some minor changes, manual monitoring is often difficult to detect, resulting in low aquaculture efficiency.

[0003] Therefore, how to provide an aquaculture monitoring method and system with multimodal sensing technology to comprehensively evaluate aquatic products and water quality parameters, effectively improve the monitoring efficiency and accuracy of the environment and water quality, and improve the survival rate and quality of aquatic products is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0004] In view of this, the present invention provides an aquaculture monitoring method and system with multimodal sensing technology, which integrates acoustic, optical and environmental parameter monitoring technologies to form a multi-dimensional monitoring system. Through data fusion technology, the data obtained from acoustic, optical and environmental parameter monitoring are integrated and analyzed. Not only the monitoring frequency is increased, but also manual participation is reduced, and the accuracy and timeliness of the data are improved; through intelligent analysis of the environmental conditions of aquaculture, the monitoring efficiency and accuracy of the environment and water quality are effectively improved, the survival rate and quality of aquatic products are improved, and the sustainable development of aquaculture is also guaranteed.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: a method for monitoring aquaculture using multimodal sensing technology, comprising:

[0006] Integrate multiple sensors according to the layout of the aquaculture farm and the flow of water to form a multi-modal sensor array to ensure the comprehensiveness and accuracy of information collection;

[0007] The multimodal sensor array regularly collects water body monitoring data to form a continuous data stream;

[0008] Divide the water body monitoring data according to the preset time period and extract the monitoring sample data;

[0009] The monitoring sample data are integrated by using a data fusion algorithm to form a multi-dimensional monitoring data set; and an adaptive fusion mechanism is introduced to adjust the fusion strategy according to the growth environment characteristics of different aquatic products;

[0010] The multi-dimensional monitoring data set is processed using deep learning technology to obtain monitoring results.

[0011] Preferably, it also includes: based on the monitoring results of the status of aquatic products and the growth environment and historical monitoring data, using a machine learning model to predict the changing trends of water quality parameters and generate a personalized aquaculture plan.

[0012] Preferably, the sensor includes an acoustic sensor, an optical sensor and an environmental sensor;

[0013] Monitoring the acoustic wave signal in the water body by means of the acoustic sensor;

[0014] monitoring visual data within the water body by means of the optical sensor;

[0015] The water quality parameters are monitored by the environmental sensors.

[0016] Preferably, the deep learning technology combines a gated recurrent unit network and a convolutional neural network to construct a water quality monitoring model based on a dynamic convolutional neural network algorithm;

[0017] The monitoring sample data is input into a water quality monitoring model based on a dynamic convolutional neural network algorithm, and the monitoring results are output.

[0018] Preferably, a water quality monitoring model based on a dynamic convolutional neural network algorithm is constructed, comprising: the water quality monitoring model comprises an input layer, three tributaries and an output layer;

[0019] The first branch is the branch for learning the deep features of the acoustic signal, which is composed of convolution layer 1, downsampling layer 1, convolution layer 2, downsampling layer 2, fully connected layer 1, gated recurrent unit layer 1, gated recurrent unit layer 2 and fully connected layer 2 in sequence;

[0020] The second branch is the branch for learning deep features of visual data, which is composed of convolution layer 1, downsampling layer 1, convolution layer 2, downsampling layer 2, convolution layer 3, downsampling layer 3, fully connected layer 1, gated recurrent unit layer 1, gated recurrent unit layer 2 and fully connected layer 2 in sequence;

[0021] The third tributary is the tributary for learning the deep features of water quality parameters, which is composed of convolutional layer 1, downsampling layer 1, convolutional layer 2, downsampling layer 2, fully connected layer 1, gated recurrent unit layer 1, gated recurrent unit layer 2 and fully connected layer 2 in sequence.

[0022] Preferably, a comprehensive data preprocessing framework is established to perform denoising, standardization and feature extraction steps on the monitoring sample data, and an adaptive noise identification technology is used to dynamically adjust the preprocessing strategy according to the specific type and quality of the monitoring sample data.

[0023] Preferably, the processing steps of the comprehensive data preprocessing framework include:

[0024] Adaptive noise identification and denoising: Based on the data characteristics of multimodal sensor arrays, a deep learning model including an autoencoder based on a convolutional neural network is used to train the model to identify and remove noise from the data and perform normalization processing;

[0025] Specific data standardization: The dimensions and distribution of data from different modalities vary greatly, and standardization steps are required to ensure data consistency and comparability; making the processed data more suitable for subsequent feature extraction and analysis;

[0026] Deep learning feature extraction: Combining the environmental characteristics of aquatic product growth and aquaculture needs, we designed and trained a deep learning model that included a deep convolutional neural network combined with a recurrent neural network to extract features for improving aquaculture planning from denoised data.

[0027] Preferably, introducing an adaptive fusion mechanism to adjust the fusion strategy according to the growth environment characteristics of different aquatic products includes:

[0028] Identify the sensitivity of different aquatic products to environmental parameters and analyze the impact of each parameter on the growth and health of aquatic products through machine learning algorithms;

[0029] According to the analysis results, the weight of each input data in the database fusion process is automatically adjusted.

[0030] Preferably, the data fusion algorithm adopts a weighted average method to assign weights to data of different dimensions to reflect the impact of different environmental parameters on aquatic products.

[0031] Preferably, an aquaculture monitoring system using multimodal sensing technology comprises:

[0032] Multimodal sensor arrays for regular collection of water monitoring data;

[0033] The sample data extraction module is used to segment the water body monitoring data according to the preset time period and extract the monitoring sample data;

[0034] A data processing and fusion module is used to integrate the monitoring sample data using a data fusion algorithm to form a multi-dimensional monitoring data set, and introduce an adaptive fusion mechanism to adjust the fusion strategy according to the growth environment characteristics of different aquatic products;

[0035] The monitoring module is used to process the multi-dimensional monitoring data set using deep learning technology to obtain monitoring results.

[0036] Through the above technical solutions, it can be known that compared with the prior art, the present invention discloses an aquaculture monitoring method and system with multimodal sensing technology, including: integrating multiple sensors according to the layout of the aquaculture farm and the flow of water bodies to form a multimodal sensor array; the multimodal sensor array regularly collects water monitoring data; divides the water monitoring data according to the preset time period and extracts the monitoring sample data; uses the data fusion algorithm to integrate the monitoring sample data to form a multidimensional monitoring data set; and introduces an adaptive fusion mechanism to adjust the fusion strategy according to the growth environment characteristics of different aquatic products; uses deep learning technology to process the multidimensional monitoring data set to obtain monitoring results. The present invention integrates acoustic, optical and environmental parameter monitoring technologies to form a multidimensional monitoring system. Through data fusion technology, the data obtained by acoustic, optical and environmental parameter monitoring are integrated and analyzed. Not only the monitoring frequency is increased, but also the manual participation is reduced, and the accuracy and timeliness of the data are improved; through intelligent analysis of the environmental conditions of aquaculture, the monitoring efficiency and accuracy of the environment and water quality are effectively improved, the survival rate and quality of aquatic products are improved, and the sustainable development of aquaculture is also guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0038] Figure 1 A schematic flow chart of an aquaculture monitoring method using a multimodal sensing technology provided by the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] The embodiment of the present invention discloses an aquaculture monitoring method and system using multimodal sensing technology. According to the layout of the aquaculture farm and the flow of water bodies, multiple sensors are integrated to form a multimodal sensor array; the multimodal sensor array regularly collects water monitoring data; the water monitoring data is segmented according to a preset time period, and monitoring sample data is extracted; the monitoring sample data is integrated using a data fusion algorithm to form a multidimensional monitoring data set; and an adaptive fusion mechanism is introduced to adjust the fusion strategy according to the growth environment characteristics of different aquatic products; and deep learning technology is used to process the multidimensional monitoring data set to obtain monitoring results. The embodiment of the present invention comprehensively evaluates aquatic products and water quality parameters, effectively improves the monitoring efficiency and accuracy of the environment and water quality, and improves the survival rate and quality of aquatic products.

[0041] In a specific embodiment of the present invention, a method for monitoring aquaculture using multimodal sensing technology, such as Figure 1 As shown, including:

[0042] Integrate multiple sensors according to the layout of the aquaculture farm and the flow of water to form a multi-modal sensor array to ensure the comprehensiveness and accuracy of information collection;

[0043] The multimodal sensor array collects water body monitoring data regularly (e.g., every minute) to form a continuous data stream;

[0044] The water body monitoring data is divided according to the preset time period, and the monitoring sample data is extracted; the preset time period can be set according to the actual breeding needs and the convenience of data processing, such as every 5 minutes or every 10 minutes as a time period; by reasonably selecting the time period length, it can reduce data redundancy while ensuring that the sample data can accurately reflect the overall characteristics and change trends of the breeding environment during the time period. When extracting sample data, ensure that the sample data is representative and timely;

[0045] The monitoring sample data are integrated by using a data fusion algorithm to form a multi-dimensional monitoring data set; and an adaptive fusion mechanism is introduced to adjust the fusion strategy according to the growth environment characteristics of different aquatic products;

[0046] In the data fusion process, the data collected by different types of sensors are fused at the time, space and semantic levels. The embodiment of the present invention aligns the aquatic product feeding sound wave information obtained by the acoustic sensor with the dissolved oxygen content data measured by the environmental sensor on the time axis, combines the aquatic product behavior images captured by the optical sensor, and comprehensively analyzes the internal connection and mutual influence between these data, thereby forming a comprehensive and multi-dimensional monitoring data set.

[0047] Since different types of aquatic products have different sensitivities to environmental parameters, for fish with high temperature and oxygen requirements, the weight of these monitoring data is increased when fusing the data. Through this adaptive approach, the fused data can better highlight the factors that have a greater impact on the growth of specific aquatic products, so as to formulate more optimized management plans.

[0048] The multi-dimensional monitoring data set is processed using deep learning technology to obtain monitoring results.

[0049] Through deep learning technology, multi-dimensional monitoring data sets are deeply analyzed to identify the behavior patterns, growth stages and health status of aquatic products. The embodiment of the present invention determines whether the fish is in a state of stress or disease by analyzing its swimming trajectory, speed and interaction with the surrounding environment.

[0050] Specifically, it also includes: based on the monitoring results of the status of aquatic products and the growth environment as well as historical monitoring data, using machine learning models to predict the changing trends of water quality parameters and generate personalized aquaculture plans. By analyzing historical data and real-time monitoring data, it predicts changes in water temperature and oxygen content fluctuations, so as to make early adjustments.

[0051] Establish a feedback mechanism to continuously optimize the layout of sensors and data collection frequency based on monitoring data and breeding results to improve the effectiveness and accuracy of monitoring.

[0052] Specifically, the sensors include acoustic sensors, optical sensors and environmental sensors;

[0053] The acoustic sensor is used to monitor the acoustic wave signals in the water body, such as the feeding status of aquatic products;

[0054] The optical sensor is used to monitor visual data in the water body, such as real-time monitoring of the growth status of aquatic products;

[0055] The environmental sensors are used to monitor water quality parameters, such as water temperature, pH value, dissolved oxygen, ammonia nitrogen and other key environmental parameters, to provide real-time data support for aquaculture management.

[0056] Wherein, the monitoring sample data includes: sound wave signals, visual data and water quality parameter data.

[0057] Specifically, the deep learning technology combines the gated recurrent unit network and the convolutional neural network to construct a water quality monitoring model based on a dynamic convolutional neural network algorithm;

[0058] The monitoring sample data is input into a water quality monitoring model based on a dynamic convolutional neural network algorithm, and the monitoring results are output.

[0059] Specifically, a water quality monitoring model based on a dynamic convolutional neural network algorithm is constructed, including: the water quality monitoring model includes an input layer, three tributaries and an output layer;

[0060] The first branch is the branch for learning the deep features of the acoustic signal, which is composed of convolution layer 1, downsampling layer 1, convolution layer 2, downsampling layer 2, fully connected layer 1, gated recurrent unit layer 1, gated recurrent unit layer 2 and fully connected layer 2 in sequence;

[0061] The second branch is the branch for learning deep features of visual data, which is composed of convolution layer 1, downsampling layer 1, convolution layer 2, downsampling layer 2, convolution layer 3, downsampling layer 3, fully connected layer 1, gated recurrent unit layer 1, gated recurrent unit layer 2 and fully connected layer 2 in sequence;

[0062] The third tributary is the tributary for learning the deep features of water quality parameters, which is composed of convolutional layer 1, downsampling layer 1, convolutional layer 2, downsampling layer 2, fully connected layer 1, gated recurrent unit layer 1, gated recurrent unit layer 2 and fully connected layer 2 in sequence.

[0063] The convolution kernels in the convolution layer are all connected to the input. Each convolution kernel extracts features at different positions of the input data through a sliding window and outputs a feature. The features of the convolution layer will continue to propagate as the input of the next layer.

[0064] The input of the downsampling layer is the output of the convolution layer, and the features at different positions are fused, and the feature information is reduced in dimension and normalized before output; the normalization process refers to linearly transforming the output of the downsampling layer and mapping the data to a closed interval [0,1];

[0065] The gated recurrent unit layer uses a gated recurrent neural network with self-recurrent links to extract the temporal relationship in the features;

[0066] The fully connected layer uses a fully connected operation to linearly transform the output of the downsampling layer from one feature space to another feature space; the output layer is a 2-output softmax output layer, and the softmax function is used in the output layer to obtain the probability distribution of the category.

[0067] The gated recurrent unit mainly includes two gate calculations and a candidate state variable update equation, as follows:

[0068] Update gate z t :z t =σ(x t U Z +h t-1 W z b z );

[0069] Reset Gate t :rt =σ(x t U r +h t-1 W r b r );

[0070] Update equation

[0071] The output of the gated recurrent unit is the current state variable h t :

[0072] Among them, x t is the input at the current moment, h t-1 is the hidden state output of the previous moment, U z , W z , U r , W r , U h , W h are the weight matrices of the network, b z 、b r and b h are the bias vectors of the network, σ is the sigmoid function, and its range is (0,1). Represents element-wise multiplication.

[0073] Specifically, a comprehensive data preprocessing framework is established to perform denoising, standardization and feature extraction steps on the monitoring sample data, and an adaptive noise recognition technology is used to dynamically adjust the preprocessing strategy according to the specific type and quality of the monitoring sample data to improve processing efficiency and data quality.

[0074] Specifically, the processing steps of the comprehensive data preprocessing framework include:

[0075] Adaptive noise identification and denoising: Based on the data characteristics of multimodal sensor arrays, a deep learning model including an autoencoder based on a convolutional neural network is used to train the model to identify and remove noise from the data and perform normalization processing;

[0076] Specific data standardization: The dimensions and distribution of data from different modalities vary greatly, and standardization steps are required to ensure data consistency and comparability; making the processed data more suitable for subsequent feature extraction and analysis;

[0077] Deep learning feature extraction: Combining the environmental characteristics of aquatic product growth and aquaculture needs, we designed and trained a deep learning model that included a deep convolutional neural network combined with a recurrent neural network to extract features for improving aquaculture planning from denoised data.

[0078] Specifically, the introduction of an adaptive fusion mechanism to adjust the fusion strategy according to the growth environment characteristics of different aquatic products includes:

[0079] Identify the sensitivity of different aquatic products to environmental parameters (such as temperature, pH, dissolved oxygen, etc.), and analyze the impact of each parameter on the growth and health of aquatic products through machine learning algorithms;

[0080] According to the analysis results, the weights of each input data in the database fusion process are automatically adjusted to ensure that the fused data set can better reflect the key environmental factors affecting the growth of specific aquatic products.

[0081] Specifically, the data fusion algorithm adopts a weighted average method to assign weights to data of different dimensions to reflect the impact of different environmental parameters on aquatic products.

[0082] In a specific embodiment of the present invention, an aquaculture monitoring system using multimodal sensing technology comprises:

[0083] Multimodal sensor arrays for regular collection of water monitoring data;

[0084] The sample data extraction module is used to segment the water body monitoring data according to the preset time period and extract the monitoring sample data;

[0085] A data processing and fusion module is used to integrate the monitoring sample data using a data fusion algorithm to form a multi-dimensional monitoring data set, and introduce an adaptive fusion mechanism to adjust the fusion strategy according to the growth environment characteristics of different aquatic products;

[0086] The monitoring module is used to process the multi-dimensional monitoring data set using deep learning technology to obtain monitoring results.

[0087] It also includes: a prediction module, which uses machine learning models to predict changing trends in water quality parameters and generate personalized aquaculture plans based on the monitoring results of the status of aquatic products and the growth environment as well as historical monitoring data.

[0088] Specifically, aquaculture farms have various layouts, including open ponds, closed circulating water systems, cage farming, etc. Different layouts have different requirements for the installation location of the sensor.

[0089] Open pond aquaculture: In open ponds, water flow is mainly affected by natural wind and water flow. Sensors should be installed in the center and edge areas of the pond to monitor water quality changes at different locations. At the same time, considering the wind direction and water flow direction, sensors should try to avoid blind spots to ensure the comprehensiveness of monitoring data.

[0090] Closed circulating water system: A closed circulating water system has a highly controllable water flow. In this system, sensors should be installed at the water inlet, outlet, and main water flow channels to monitor changes in water quality in real time. In addition, since closed systems are prone to local water quality differences, sensors should be evenly arranged in different areas of the system.

[0091] Cage aquaculture: The water exchange in cage aquaculture mainly depends on external water flow. Sensors should be installed at the inlet and outlet of the cage and in the center of the cage to monitor the changes in water quality when water flows through the cage. At the same time, considering the difference in water quality inside and outside the cage, sensors should also be installed at appropriate locations outside the cage.

[0092] The impact of water flow conditions on sensor installation Water flow conditions are another important factor affecting the sensor installation location. Reasonable use of water flow characteristics can improve the accuracy of monitoring data.

[0093] Water flow speed: In areas with faster water flow, the sensor should be installed in the center of the water flow channel to ensure that the water quality changes in the water flow can be accurately captured. In areas with slower water flow, the sensors should be appropriately densely arranged to monitor local water quality changes.

[0094] Water flow direction: Sensors should be installed along the direction of water flow as much as possible to monitor the dynamic changes of water quality at different locations. At the same time, sensors should also be installed at the water flow turning or bifurcation points to capture the impact of water flow changes on water quality.

[0095] Water stratification: In deep-water aquaculture areas, water is prone to stratification of parameters such as temperature and salinity. In this case, a multi-level sensor layout should be used to monitor water quality parameters at different depths to fully understand changes in the water environment.

[0096] Sensors and their installation locations

[0097] Dissolved oxygen sensor: Dissolved oxygen is a critical parameter in aquaculture. The sensor should be installed in areas with high activity of aquaculture organisms and areas with weak water flow to ensure that the lack of dissolved oxygen can be detected in time.

[0098] pH sensor: pH value has a great influence on the growth of cultured organisms. Sensors should be evenly distributed throughout the culture area to monitor pH changes at different locations.

[0099] Temperature Sensor: Temperature has a direct impact on the growth rate and health of cultured organisms. Sensors should be installed at different depths in the culture area to monitor vertical changes in water temperature.

[0100] Ammonia nitrogen sensor: Ammonia nitrogen is an important indicator for measuring the degree of water pollution. The sensor should be installed in areas where the excrement of farmed organisms is concentrated and where the water flow is slow to monitor the changes in ammonia nitrogen in real time.

[0101] In aquaculture farms, the key to ensuring the accuracy and reliability of monitoring data is to properly select the installation location of sensors. It is necessary to comprehensively consider the layout of the farm, the flow of water bodies, and the monitoring needs of different sensors. Only through scientific and reasonable layout and installation can the role of sensors in aquaculture environmental monitoring be truly brought into play, providing strong protection for the healthy growth of farmed organisms.

[0102] The embodiment of the present invention integrates and analyzes the data obtained from acoustic, optical and environmental parameter monitoring through data fusion technology, thereby improving the accuracy and reliability of monitoring.

[0103] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0104] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring aquaculture using multimodal sensing technology, characterized in that: include: Integrate multiple sensors according to the layout of the aquaculture farm and the flow of water to form a multimodal sensor array; The multimodal sensor array regularly collects water monitoring data; Divide the water body monitoring data according to the preset time period and extract the monitoring sample data; Using a data fusion algorithm, the monitoring sample data is integrated to form a multi-dimensional monitoring data set; And introduce adaptive fusion mechanism to adjust fusion strategy according to the growth environment characteristics of different aquatic products; The multi-dimensional monitoring data set is processed using deep learning technology to obtain monitoring results.

2. The aquaculture monitoring method using multimodal sensing technology according to claim 1 is characterized in that: Also includes: Based on the monitoring results of the status of aquatic products and the growth environment as well as historical monitoring data, machine learning models are used to predict the changing trends of water quality parameters and generate personalized aquaculture plans.

3. The aquaculture monitoring method using multimodal sensing technology according to claim 1 is characterized in that: The sensors include acoustic sensors, optical sensors and environmental sensors; Monitoring the acoustic wave signal in the water body by means of the acoustic sensor; monitoring visual data within the water body by means of the optical sensor; The water quality parameters are monitored by the environmental sensors.

4. The aquaculture monitoring method using multimodal sensing technology according to claim 1 is characterized in that: The deep learning technology combines the gated recurrent unit network and the convolutional neural network to construct a water quality monitoring model based on the dynamic convolutional neural network algorithm; The monitoring sample data is input into a water quality monitoring model based on a dynamic convolutional neural network algorithm, and the monitoring results are output.

5. The aquaculture monitoring method using multimodal sensing technology according to claim 4 is characterized in that: Constructing a water quality monitoring model based on a dynamic convolutional neural network algorithm, including: the water quality monitoring model includes an input layer, three tributaries and an output layer; The first branch is the branch for learning the deep features of the acoustic signal, which is composed of convolution layer 1, downsampling layer 1, convolution layer 2, downsampling layer 2, fully connected layer 1, gated recurrent unit layer 1, gated recurrent unit layer 2 and fully connected layer 2 in sequence; The second branch is the branch for learning deep features of visual data, which is composed of convolution layer 1, downsampling layer 1, convolution layer 2, downsampling layer 2, convolution layer 3, downsampling layer 3, fully connected layer 1, gated recurrent unit layer 1, gated recurrent unit layer 2 and fully connected layer 2 in sequence; The third tributary is the tributary for learning the deep features of water quality parameters, which is composed of convolutional layer 1, downsampling layer 1, convolutional layer 2, downsampling layer 2, fully connected layer 1, gated recurrent unit layer 1, gated recurrent unit layer 2 and fully connected layer 2 in sequence.

6. The aquaculture monitoring method using multimodal sensing technology according to claim 1, characterized in that: A comprehensive data preprocessing framework is established to perform denoising, standardization and feature extraction steps on the monitoring sample data, and an adaptive noise identification technology is used to dynamically adjust the preprocessing strategy according to the specific type and quality of the monitoring sample data.

7. The aquaculture monitoring method using multimodal sensing technology according to claim 6 is characterized in that: The processing steps of the comprehensive data preprocessing framework include: Adaptive noise identification and denoising: Based on the data characteristics of multimodal sensor arrays, a deep learning model including an autoencoder based on a convolutional neural network is used to train the model to identify and remove noise from the data and perform normalization processing; Deep learning feature extraction: Combining the environmental characteristics of aquatic product growth and aquaculture needs, we designed and trained a deep learning model that included a deep convolutional neural network combined with a recurrent neural network to extract features for improving aquaculture planning from denoised data.

8. The aquaculture monitoring method using multimodal sensing technology according to claim 1, characterized in that: The introduction of adaptive fusion mechanism adjusts the fusion strategy according to the growth environment characteristics of different aquatic products, including: Identify the sensitivity of different aquatic products to environmental parameters and analyze the impact of each parameter on the growth and health of aquatic products through machine learning algorithms; According to the analysis results, the weight of each input data in the database fusion process is automatically adjusted.

9. The aquaculture monitoring method using multimodal sensing technology according to claim 1, characterized in that: The data fusion algorithm adopts a weighted average method to assign weights to data of different dimensions to reflect the impact of different environmental parameters on aquatic products.

10. An aquaculture monitoring system using multimodal sensing technology, characterized in that: include: Multimodal sensor arrays for regular collection of water monitoring data; The sample data extraction module is used to segment the water body monitoring data according to the preset time period and extract the monitoring sample data; A data processing and fusion module is used to integrate the monitoring sample data using a data fusion algorithm to form a multi-dimensional monitoring data set, and introduce an adaptive fusion mechanism to adjust the fusion strategy according to the growth environment characteristics of different aquatic products; The monitoring module is used to process the multi-dimensional monitoring data set using deep learning technology to obtain monitoring results.

Images