Intelligent buoy water quality on-line monitoring system and method based on multi-modal data fusion

Abstract

The invention discloses an intelligent buoy water quality on-line monitoring system and method based on multi-modal data fusion, belongs to the technical field of water quality on-line monitoring, and aims to solve the problems that pollution diffusion tracks are difficult to draw from spatial dimensions and dynamic adjustment cannot be performed according to water quality change conditions. The system is composed of a plurality of intelligent buoys provided with diversified in-situ sensors, GPS positioning devices, wireless communication devices, autonomous moving propellers and microprocessors, and the system covers a monitored water area in a distributed layout mode and is used for collecting water quality in-situ data and receiving control instructions. According to the method, the intelligent buoy sampling strategy adjustment scheme is generated through multi-modal data fusion in combination with intelligent buoy in-situ data, satellite remote sensing images and meteorological and hydrological data, the position of a pollution source can be determined more accurately, the problem that single-point data of a traditional buoy is difficult to trace is solved, and precious time is won for emergency disposal of an emergent pollution event.

Description

Technical Field

[0001] This invention belongs to the field of online water quality monitoring technology, specifically relating to an intelligent buoy-based online water quality monitoring system and method based on multimodal data fusion. Background Technology

[0002] Water quality monitoring, as an important means of water pollution prevention and control, plays a crucial role in timely understanding water quality conditions, early warning of pollution incidents, and the formulation of effective remediation measures. Intelligent buoy online water quality monitoring is a new type of environmental monitoring method that utilizes modern Internet of Things (IoT), sensors, and data analysis technologies to conduct real-time, continuous, and automated monitoring of the aquatic environment.

[0003] Smart buoys can collect multiple parameters in water bodies in real time and transmit the data to a monitoring center via wireless communication technology, enabling continuous online monitoring of water quality. The smart buoy carrier is a floating platform equipped with multiple water quality sensors, typically featuring solar power, GPS positioning, and wireless transmission capabilities, adaptable to different water bodies such as rivers, lakes, and oceans. The multi-parameter sensors are used to detect key indicators in real time, including pH, dissolved oxygen (DO), turbidity, conductivity, ammonia nitrogen, COD (chemical oxygen demand), chlorophyll a, and heavy metals. The collected data is uploaded in real time to a cloud platform or monitoring center via a wireless network for real-time water quality monitoring.

[0004] Existing smart buoys are typically deployed at single points, and the data they collect can only reflect the water quality of a local area. It is difficult to trace the trajectory of pollution spread in a spatial dimension, making it difficult to trace the source of pollution. On the other hand, existing smart buoy sampling strategies are mostly fixed modes, which cannot be dynamically adjusted according to changes in water quality. In the event of a sudden pollution incident, this fixed sampling mode either misses key pollution data due to the low sampling frequency, or fails to capture the dynamic process of pollution spread due to the fixed sampling range, which seriously affects the efficiency and accuracy of emergency response.

[0005] Therefore, there is a need for an intelligent buoy-based online water quality monitoring system and method based on multimodal data fusion to solve the problems in existing technologies, such as the difficulty in delineating the trajectory of pollution diffusion in a spatial dimension and the inability to dynamically adjust according to changes in water quality. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent buoy-based online water quality monitoring system and method based on multimodal data fusion, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent buoy water quality online monitoring system based on multimodal data fusion, comprising:

[0008] The intelligent buoy array unit consists of multiple intelligent buoys equipped with diverse in-situ sensors, GPS positioning devices, wireless communication equipment, autonomous motion thrusters, and microprocessors. It adopts a distributed layout to cover the monitored water area and is used to collect in-situ water quality data and receive control commands.

[0009] The multi-source data acquisition unit includes a satellite remote sensing data module and a meteorological and hydrological data module. The satellite remote sensing data module is used to acquire remote sensing images and perform data quality assessment and compensation, while the meteorological and hydrological data module is used to collect meteorological and hydrological data.

[0010] The data fusion processing center includes a data storage module, a spatiotemporal alignment unit, and a multimodal fusion unit. The spatiotemporal alignment unit realizes the temporal and spatial alignment of multi-source data, and the multimodal fusion module mines the coupling mechanism between modes and fuses the data to obtain a fusion feature vector that comprehensively reflects the water quality status.

[0011] The multimodal feedback control unit includes a fusion result analysis module, a sampling strategy adjustment decision module, and an instruction sending and execution module. It is used to analyze pollution events based on fused feature vectors and generate sampling strategy adjustment schemes, send sampling strategy adjustment instructions to smart buoys, and receive and execute them.

[0012] It should be noted in the solution that the spatiotemporal alignment unit includes a time alignment module and a spatial alignment module. The time alignment module uses UTC as the standard time base and adds precise timestamps to various types of data. The spatial alignment module is based on the GIS system and converts all data into a unified latitude and longitude coordinate system.

[0013] It is worth noting that the multimodal fusion unit includes a data preprocessing module, a feature extraction and selection module, and a multimodal data fusion module. The data preprocessing module cleans the received data, the feature extraction and selection module extracts features based on the characteristics of different modal data, and the multimodal data fusion module projects single-modal features onto a shared semantic subspace and dynamically allocates weights to generate a fused feature vector.

[0014] Furthermore, it should be noted that the fusion result analysis module uses a support vector machine model to determine the water quality status, and in case of anomalies, it uses a multiple linear regression model to predict pollution trends and constructs a three-level index to assess pollution events.

[0015] In a preferred embodiment, the instruction sending and execution module sends the instructions generated by the sampling strategy adjustment decision module to the corresponding smart buoy via the wireless communication module. After receiving the instructions, the smart buoy dynamically adjusts the sampling strategy and feeds back the adjusted execution status to the data fusion and processing center.

[0016] As a preferred embodiment, the intelligent buoy water quality online monitoring method based on multimodal data fusion is used to implement an intelligent buoy water quality online monitoring system based on multimodal data fusion, comprising:

[0017] Relevant data are collected through intelligent buoy arrays, satellite remote sensing data modules, and meteorological and hydrological data modules and transmitted to the data fusion and processing center.

[0018] In the data fusion and processing center, the collected data is aligned in time and space to unify the spatiotemporal scale, and the data after unifying the spatiotemporal scale is preprocessed;

[0019] Feature extraction is performed on the preprocessed data, and the single-modal representation of each modality is projected onto the shared semantic subspace. Weights are dynamically assigned according to the importance of each modality to water quality monitoring, thereby achieving multimodal data fusion and obtaining a fused feature vector.

[0020] The fused feature vectors are analyzed using machine learning classification and regression models to determine the water quality status, predict the trend of water quality parameter changes and the scope and degree of pollution impact, and generate an assessment report.

[0021] Based on the analysis of the evaluation report, and combined with the preset rule base and optimization algorithm, a scheme for adjusting the smart buoy sampling strategy was determined;

[0022] The sampling strategy adjustment command is sent to the smart buoy, and after the smart buoy executes the command, it feeds back the execution status to the data fusion and processing center.

[0023] As a preferred implementation, the fusion result analysis module performs in-depth analysis of the fused feature vectors, uses a support vector machine model to determine whether the current water quality status is normal, and if abnormal, further uses a multiple linear regression model to predict the trend of water quality parameter changes and the possible scope and degree of pollution impact; through comparison with historical data and trend analysis, it assesses the severity and development trend of pollution events; wherein, the steps of constructing a multiple linear regression pollution trend prediction model include:

[0024] The input independent variable (X) includes:

[0025] Time variable: Elapsed time (t) since the anomaly occurred;

[0026] Meteorological variable: wind speed ( ), wind direction angle ( ), precipitation (p);

[0027] Hydrological variables: water flow velocity ( ), water flow direction angle ( ), water level (h);

[0028] Initial pollution variable: COD concentration at the time of the anomaly ( ), NH3-N concentration ( ), initial area of ​​the contaminated area ( );

[0029] The dependent variable (Y) includes:

[0030] COD concentration at time t in the future ( ), NH3-N concentration ( );

[0031] The extent of pollution impact at time t in the future ( );

[0032] Construct an independent multiple linear regression equation for each dependent variable:

[0033]

[0034] in For constant terms, For regression coefficients, It is a random error (following a normal distribution).

[0035] The pollution trend prediction process is as follows:

[0036] Initial data when an input error occurs ( , , ) and meteorological and hydrological forecast data for the next 6 hours ( , p , h);

[0037] The model outputs predictions for the next 6 hours at 1-hour intervals: water quality parameter change trends, pollution impact range, and peak concentrations.

[0038] By comparing the prediction results with historical data, a three-level evaluation index is constructed.

[0039] By comparing the predicted results with similar historical events and combining them with the slope of the trend curve, a comprehensive evaluation report is generated.

[0040] Compared with existing technologies, the intelligent buoy water quality online monitoring system and method based on multimodal data fusion provided by this invention has at least the following beneficial effects:

[0041] (1) By integrating multimodal data, combining in-situ data of smart buoys, satellite remote sensing images and meteorological and hydrological data, a smart buoy sampling strategy adjustment scheme is generated, which can more accurately determine the location of pollution sources, overcome the problem that traditional single-point data of buoys is difficult to trace, and win valuable time for emergency response to sudden pollution events.

[0042] (2) By establishing a unified spatiotemporal alignment framework, multi-source heterogeneous data of different sources, formats and spatiotemporal scales are accurately spatiotemporally aligned. Through advanced multimodal data fusion algorithms, the coupling mechanism between various modal data is deeply explored, which comprehensively improves the accuracy and comprehensiveness of water quality monitoring and provides richer and more accurate data support for comprehensive water environment analysis.

[0043] (3) Through the designed multimodal feedback control mechanism, the smart buoy can dynamically adjust the sampling strategy in real time according to the multimodal data fusion results, increase the sampling density in areas with abnormal water quality, and optimize the sampling resources in areas with stable water quality, thereby improving the monitoring efficiency and targeting, making more reasonable use of monitoring resources, and reducing monitoring costs. Attached Figure Description

[0044] Figure 1 This is a block diagram of the intelligent buoy water quality online monitoring system based on multimodal data fusion of the present invention;

[0045] Figure 2 This is a flowchart of the intelligent buoy water quality online monitoring method based on multimodal data fusion according to the present invention. Detailed Implementation

[0046] The present invention will be further described below with reference to embodiments.

[0047] Reference Figure 1 As shown, this invention provides an intelligent buoy-based online water quality monitoring system based on multimodal data fusion, comprising:

[0048] The intelligent buoy array unit consists of multiple intelligent buoys equipped with diverse in-situ sensors, GPS positioning devices, wireless communication equipment, autonomous motion thrusters, and microprocessors. It adopts a distributed layout to cover the monitored water area and is used to collect in-situ water quality data and receive control commands.

[0049] The multi-source data acquisition unit includes a satellite remote sensing data module and a meteorological and hydrological data module. The satellite remote sensing data module is used to acquire remote sensing images and perform data quality assessment and compensation, while the meteorological and hydrological data module is used to collect meteorological and hydrological data.

[0050] The data fusion processing center includes a data storage module, a spatiotemporal alignment unit, and a multimodal fusion unit. The spatiotemporal alignment unit realizes the temporal and spatial alignment of multi-source data, and the multimodal fusion module explores the coupling mechanism between modes and fuses the data.

[0051] The multimodal feedback control unit consists of a fusion result analysis module, a sampling strategy adjustment decision module, and an instruction sending and execution module. It is used to analyze pollution events based on fused feature vectors and generate sampling strategy adjustment schemes, send sampling strategy adjustment instructions to smart buoys, and receive and execute them.

[0052] Understandably, each smart buoy is equipped with multiple in-situ sensors, including but not limited to water quality parameter sensors (measuring dissolved oxygen, pH, chemical oxygen demand, ammonia nitrogen, etc.), flow velocity sensors, and temperature sensors, to collect real-time physical and chemical parameters of the water at its location. The buoy is equipped with an edge computing unit to perform preliminary processing and feature extraction on the collected raw data, reducing data transmission volume. At the same time, it is equipped with a wireless communication module, which can transmit the processed data to the data fusion and processing center, and can also receive control commands from the center. In addition, it is also equipped with a GPS positioning device, an autonomous motion thruster, and a power supply.

[0053] Understandably, the satellite remote sensing data module acquires remote sensing images of the monitoring area through the satellite data receiving terminal, including multispectral and hyperspectral images, and can retrieve parameters such as chlorophyll concentration and suspended solids content in the water body; at the same time, a data quality assessment and compensation module is set up to interpolate and compensate when the remote sensing data is affected by cloud interference or revisit period using historical data from the same period and data from similar water bodies in the surrounding area.

[0054] The meteorological and hydrological data module connects meteorological and hydrological stations distributed within and around the monitoring area to acquire meteorological and hydrological data in real time. It transmits satellite remote sensing data and meteorological and hydrological data to the data fusion and processing center through a dedicated communication network, providing environmental background information for comprehensive analysis of water quality changes.

[0055] Understandably, the data storage module uses a distributed database to classify and store in-situ data collected by the intelligent buoy array unit, remote sensing data acquired by the satellite remote sensing data module, and meteorological and hydrological data, ensuring data integrity and traceability.

[0056] The spatiotemporal alignment unit includes a time alignment module and a spatial alignment module. The time alignment module uses Coordinated Universal Time (UTC) as the standard time reference and adds precise timestamps to various types of data. For in-situ data from smart buoys, it records the precise UTC timestamp for each sampling; for satellite remote sensing image data, it obtains the corresponding UTC time based on the satellite's transit time; for meteorological and hydrological data, it ensures that the acquisition equipment records accurate UTC time; for data with different time resolutions, a resampling method is used to unify the time scale. For example, high-time-resolution smart buoy data is downsampled according to the satellite remote sensing revisit cycle, or low-time-resolution meteorological and hydrological data is upsampled by linear interpolation according to the smart buoy sampling time interval, so that the data of each modality are strictly aligned in time.

[0057] The spatial alignment module, based on a Geographic Information System (GIS), converts all data into a unified latitude and longitude coordinate system. For smart buoys, it obtains precise geographic location coordinates through their onboard Global Positioning System (GPS). For satellite remote sensing imagery, it uses the geographic location information and map projection parameters in the imagery to convert image coordinates into latitude and longitude coordinates. For meteorological and hydrological station data, it combines the station's geographic location information to convert it into latitude and longitude coordinates. In terms of spatial scale unification, the satellite remote sensing imagery is cropped according to the distribution of smart buoys and the monitoring area, and a spatial interpolation algorithm is used to interpolate the meteorological and hydrological station data to the same spatial resolution as the smart buoys, so that the data of each modality are accurately aligned in space, which facilitates subsequent coupled analysis.

[0058] The multimodal fusion unit includes a data preprocessing module, a feature extraction and selection module, and a multimodal data fusion module. The data preprocessing module cleans the received data, removes outliers from the in-situ data of the smart buoy, and uses filtering algorithms to remove noise interference. For remote sensing image data, it further checks and corrects the images to ensure that there are no residual stripes or bad pixels. For meteorological and hydrological data, it fills in missing values ​​and corrects erroneous data.

[0059] The feature extraction and selection module extracts features based on the characteristics of different modal data. For in-situ data from smart buoys, in addition to simple statistical features, features reflecting the changing trends of water quality parameters are also extracted. For remote sensing image data, deep learning models are used to extract water color features, texture features, and spectral features related to pollutants. For meteorological and hydrological data, features closely related to water quality changes are extracted, such as precipitation intensity and frequency features, wind speed and direction vector features, and flow rate change features. A feature selection algorithm is used to remove redundant features, reduce data dimensionality, and improve the efficiency of subsequent fusion and analysis.

[0060] The multimodal data fusion module adopts a joint fusion approach. First, each modal data is processed by a specific encoder or feature extractor to obtain a single modal representation. These single modal representations are then projected onto a shared semantic subspace. In this subspace, an attention-based fusion algorithm is used to dynamically allocate weights according to the importance of each modal data to the water quality monitoring task, thereby achieving multimodal data fusion and obtaining a fusion feature vector that comprehensively reflects the water quality status.

[0061] Understandably, the fusion result analysis module performs in-depth analysis on the feature vectors obtained after multimodal data fusion, uses a support vector machine model to determine whether the current water quality status is normal, and if abnormal, further uses a multiple linear regression model to predict the changing trends of water quality parameters and the possible scope and extent of pollution impact; through comparison with historical data and trend analysis, it assesses the severity and development trend of pollution events; the steps for constructing a multiple linear regression pollution trend prediction model include:

[0062] The input independent variable (X) includes:

[0063] Time variable: Elapsed time (t) since the anomaly occurred;

[0064] Meteorological variable: wind speed ( ), wind direction angle ( ), precipitation (p);

[0065] Hydrological variables: water flow velocity ( ), water flow direction angle ( ), water level (h);

[0066] Initial pollution variable: COD concentration at the time of the anomaly ( ), NH3-N concentration ( ), initial area of ​​the contaminated area ( );

[0067] The dependent variable (Y) includes:

[0068] COD concentration at time t in the future ( ), NH3-N concentration ( );

[0069] The extent of pollution impact at time t in the future ( );

[0070] Construct an independent multiple linear regression equation for each dependent variable:

[0071]

[0072] in For constant terms, For regression coefficients, It is a random error (following a normal distribution).

[0073] The pollution trend prediction process is as follows:

[0074] Initial data when an input error occurs ( , , ) and meteorological and hydrological forecast data for the next 6 hours ( , p , h);

[0075] The model outputs predictions for the next 6 hours at 1-hour intervals: water quality parameter change trends, pollution impact range, and peak concentrations.

[0076] By comparing the prediction results with historical data, a three-level evaluation index is constructed.

[0077] By comparing the predicted results with similar historical events and combining them with the slope of the trend curve, a comprehensive evaluation report is generated.

[0078] The sampling strategy adjustment decision module generates a smart buoy sampling strategy adjustment plan based on the comprehensive evaluation report output by the fusion result analysis unit, combined with a preset rule base and optimization algorithm. If water quality anomalies are detected and there is a possibility of pollution source spread, the optimized particle swarm algorithm is used to determine the area that needs to be sampled more intensively, and the smart buoy is moved to that area (if the buoy has autonomous movement capability, it is achieved by controlling the thruster; if it is a fixed buoy, a nearby movable buoy is arranged to go there), and the sampling frequency is increased. If the water quality is relatively stable, the sampling frequency is appropriately reduced or the sampling parameters are adjusted according to the monitoring objectives and resource optimization principles to save energy and storage space and improve monitoring efficiency.

[0079] The instruction sending and execution module sends the instructions generated by the sampling strategy adjustment decision module to the corresponding smart buoy via the wireless communication module. After receiving the instructions, the smart buoy controls the sensor working mode, sampling frequency and its own movement operation to realize dynamic adjustment of the sampling strategy, and feeds back the adjusted execution status to the data fusion and processing center for subsequent evaluation and further optimization.

[0080] For further details, please refer to Figure 2 As shown, the same inventive concept as the intelligent buoy water quality online monitoring system based on multimodal data fusion is proposed in this solution. This solution proposes an intelligent buoy water quality online monitoring method based on multimodal data fusion, including:

[0081] Relevant data are collected through intelligent buoy arrays, satellite remote sensing data modules, and meteorological and hydrological data modules and transmitted to the data fusion and processing center.

[0082] In the data fusion and processing center, the collected data is aligned in time and space to unify the spatiotemporal scale, and the data after unifying the spatiotemporal scale is preprocessed;

[0083] Feature extraction is performed on the preprocessed data to obtain modal data. The single-modal representation of each modal data is projected onto a shared semantic subspace. Weights are dynamically assigned according to the importance of each modal data to water quality monitoring to achieve multimodal data fusion and obtain a fused feature vector.

[0084] The fused feature vectors are analyzed using machine learning classification and regression models to determine the water quality status, predict the trend of water quality parameter changes and the scope and degree of pollution impact, and generate an assessment report.

[0085] Based on the analysis of the evaluation report, and combined with the preset rule base and optimization algorithm, a scheme for adjusting the smart buoy sampling strategy was determined;

[0086] The sampling strategy adjustment command is sent to the smart buoy, and after the smart buoy executes the command, it feeds back the execution status to the data fusion and processing center.

[0087] In summary, the advantages of this invention are as follows:

[0088] By fusing multimodal data and combining in-situ data from smart buoys, satellite remote sensing images, and meteorological and hydrological data, a smart buoy sampling strategy adjustment scheme is generated. This can more accurately determine the location of pollution sources, overcome the problem that traditional single-point data from buoys is difficult to trace, and win valuable time for emergency response to sudden pollution incidents.

[0089] By establishing a unified spatiotemporal alignment framework, multi-source heterogeneous data from different sources, formats, and spatiotemporal scales are accurately spatiotemporally aligned. Furthermore, through advanced multimodal data fusion algorithms, the coupling mechanisms between various modal data are deeply explored, comprehensively improving the accuracy and comprehensiveness of water quality monitoring and providing richer and more accurate data support for integrated water environment analysis.

[0090] Through the designed multimodal feedback control mechanism, the smart buoy can dynamically adjust the sampling strategy in real time based on the multimodal data fusion results, increase sampling density in areas with abnormal water quality, and optimize sampling resources in areas with stable water quality. This improves monitoring efficiency and targeting, makes more rational use of monitoring resources, and reduces monitoring costs.

[0091] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. An intelligent buoy-based online water quality monitoring system based on multimodal data fusion, characterized in that, include: The intelligent buoy array unit consists of multiple intelligent buoys equipped with diverse in-situ sensors, GPS positioning devices, wireless communication equipment, autonomous motion thrusters, and microprocessors. It adopts a distributed layout to cover the monitored water area and is used to collect in-situ water quality data and receive control commands. The multi-source data acquisition unit includes a satellite remote sensing data module and a meteorological and hydrological data module. The satellite remote sensing data module is used to acquire remote sensing images and perform data quality assessment and compensation, while the meteorological and hydrological data module is used to collect meteorological and hydrological data. The data fusion processing center includes a data storage module, a spatiotemporal alignment unit, and a multimodal fusion unit. The spatiotemporal alignment unit realizes the temporal and spatial alignment of multi-source data, and the multimodal fusion module mines the coupling mechanism between modes and fuses the data to obtain a fusion feature vector that comprehensively reflects the water quality status. The multimodal feedback control unit includes a fusion result analysis module, a sampling strategy adjustment decision module, and an instruction sending and execution module. It is used to analyze pollution events based on fused feature vectors and generate sampling strategy adjustment schemes, send sampling strategy adjustment instructions to smart buoys, and receive and execute them.

2. The intelligent buoy water quality online monitoring system based on multimodal data fusion according to claim 1, characterized in that: The spatiotemporal alignment unit includes a time alignment module and a spatial alignment module. The time alignment module uses UTC as the standard time base and adds precise timestamps to various types of data. The spatial alignment module is based on a geographic GIS system and converts all data into a unified latitude and longitude coordinate system.

3. The intelligent buoy water quality online monitoring system based on multimodal data fusion according to claim 2, characterized in that: The multimodal fusion unit includes a data preprocessing module, a feature extraction and selection module, and a multimodal data fusion module. The data preprocessing module cleans the received data. The feature extraction and selection module extracts features based on the characteristics of different modal data. The multimodal data fusion module projects single-modal features onto a shared semantic subspace and dynamically assigns weights to generate a fused feature vector.

4. The intelligent buoy water quality online monitoring system based on multimodal data fusion according to claim 3, characterized in that: The fusion result analysis module uses a support vector machine model to determine the water quality status, and in case of anomalies, it uses a multiple linear regression model to predict the pollution trend and constructs a three-level index to assess pollution events.

5. The intelligent buoy water quality online monitoring system based on multimodal data fusion according to claim 3, characterized in that: The instruction sending and execution module sends the instructions generated by the sampling strategy adjustment decision module to the corresponding smart buoy through the wireless communication module. After receiving the instructions, the smart buoy dynamically adjusts the sampling strategy and feeds back the adjusted execution status to the data fusion and processing center.

6. A method for online water quality monitoring of intelligent buoys based on multimodal data fusion is used to implement the online water quality monitoring system for intelligent buoys based on multimodal data fusion as described in any one of claims 1-5, characterized in that, include: Relevant data are collected through intelligent buoy arrays, satellite remote sensing data modules, and meteorological and hydrological data modules and transmitted to the data fusion and processing center. In the data fusion and processing center, the collected data is aligned in time and space to unify the spatiotemporal scale, and the data after unifying the spatiotemporal scale is preprocessed; Feature extraction is performed on the preprocessed data, and the single-modal representation of each modality is projected onto the shared semantic subspace. Weights are dynamically assigned according to the importance of each modality to water quality monitoring, thereby achieving multimodal data fusion and obtaining a fused feature vector. The fused feature vectors are analyzed using machine learning classification and regression models to determine the water quality status, predict the trend of water quality parameter changes and the scope and degree of pollution impact, and generate an assessment report. Based on the analysis of the evaluation report, and combined with the preset rule base and optimization algorithm, a scheme for adjusting the smart buoy sampling strategy was determined; The sampling strategy adjustment command is sent to the smart buoy, and after the smart buoy executes the command, it feeds back the execution status to the data fusion and processing center.

Images