Multi-modal water quality monitoring data intelligent processing method and system

Abstract

The invention relates to the field of water quality monitoring analysis, in particular to a multi-modal water quality monitoring data intelligent processing method and system. The method comprises the following steps: obtaining a water quality monitoring parameter stream and a water quality monitoring image, and carrying out adaptive processing to obtain a filtering optimization parameter stream and an optimization monitoring image; performing water quality state evolution perception according to the adaptive filtering optimization parameter flow, and constructing a water quality state evolution graph; and performing biological index response analysis and water quality ecological influence evaluation based on the water quality state trend evolution graph to obtain a water quality evaluation report. And carrying out pollutant spatial distribution change analysis based on the optimized monitoring image, and executing final water quality monitoring evaluation according to a water quality evaluation report. The invention provides efficient and accurate water quality status analysis.

Description

Technical Field

[0001] This invention relates to the field of water quality monitoring and analysis, and in particular to a method and system for intelligent processing of multimodal water quality monitoring data. Background Technology

[0002] With the continuous advancement of ecological environmental protection and sustainable development concepts, water resource security and quality have become a global focus. Water quality monitoring, as a crucial means of ensuring drinking water safety, maintaining the ecological environment, and supporting water resource management, is developing towards higher frequency, wider coverage, and higher precision. Especially against the backdrop of increasingly prominent environmental problems such as industrial emissions, water source pollution, and agricultural non-point source pollution, traditional water quality monitoring methods are no longer sufficient to meet the real-time, comprehensive, and scientific requirements of modern water environment governance.

[0003] Traditional water quality monitoring methods primarily rely on manual sampling and laboratory analysis. While offering high accuracy, these methods suffer from long monitoring cycles, high costs, and limited coverage, making it difficult to reflect the dynamic changes of pollutants in water bodies in a timely manner. Meanwhile, with the development of the Internet of Things (IoT) and sensor technology, online water quality monitoring systems based on sensor networks are gradually being adopted, enabling continuous collection of multiple indicators such as water temperature, pH, dissolved oxygen, conductivity, turbidity, and heavy metal ions. However, faced with massive, heterogeneous, and time-series-dependent multimodal water quality data, relying solely on a single sensor or static analysis method is insufficient to achieve a comprehensive perception and in-depth understanding of complex water quality change trends.

[0004] Furthermore, water quality monitoring data is highly complex in terms of space, time, and attributes. Data characteristics vary significantly across different water bodies, pollution sources, and environmental conditions, leading to problems such as limited analytical dimensions, limited predictive capabilities, and untimely responses in practical applications of traditional data analysis methods. Simultaneously, with the increase in the number of monitoring points and the higher frequency of data collection, the amount of multimodal data generated by water quality monitoring systems is exploding, necessitating more efficient and intelligent data processing and visualization analysis methods. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes a multimodal water quality monitoring data intelligent processing method and system, thereby resolving at least one of the aforementioned technical issues.

[0006] To achieve the above objectives, the present invention provides an intelligent processing method for multimodal water quality monitoring data, comprising the following steps: Step S1: Acquire water quality monitoring parameter stream and water quality monitoring image, and perform adaptive processing to obtain filtered and optimized parameter stream and optimized monitoring image; Step S2: Based on the adaptive filtering optimization parameter flow, sense the evolution of water quality status and construct a water quality status evolution map; Step S3: Based on the water quality status trend evolution map, conduct biological indicator response analysis and water quality ecological impact assessment to obtain a water quality assessment report; Step S4: Analyze the spatial distribution changes of pollutants based on the optimized monitoring images, and perform the final water quality monitoring assessment according to the water quality assessment report.

[0007] This specification provides a multimodal water quality monitoring data intelligent processing system. The system includes four analyzers, four quality control instruments, six electrodes, one general-purpose input / output control board, a UPS, an air conditioner, and an access control system. The six electrodes, the general-purpose input / output control board, the UPS, the air conditioner, and the access control system are directly connected to the central control system via an RS485 bus. The four analyzers and four quality control instruments are connected to the central control system via eight RS232 physical interfaces. The electrodes, the general-purpose control board, the UPS, the air conditioner, and the access control system use RS232 physical interfaces. A camera is connected to a hard disk recorder via a network cable for storing video recordings. The hard disk recorder is connected to the central control system via a network cable. A remote platform can obtain the video stream via remote commands, thereby viewing historical recordings and real-time video remotely. The system is used to execute the multimodal water quality monitoring data intelligent processing method as described in claim 1, including: The multimodal processing module is used to acquire water quality monitoring parameter streams and water quality monitoring images, and perform adaptive processing to obtain filtered and optimized parameter streams and optimized monitoring images. The data processing module is used to acquire water quality monitoring parameter streams and water quality monitoring images, and perform adaptive processing to obtain filtered and optimized parameter streams and optimized monitoring images. The state evolution module is used to perceive the water quality state evolution based on the adaptive filtering optimization parameter flow and construct a water quality state evolution map. The response analysis module is used to perform biological indicator response analysis and water quality ecological impact assessment based on the water quality status trend evolution map to obtain a water quality assessment report. The monitoring and assessment module is used to analyze the spatial distribution changes of pollutants based on optimized monitoring images and to perform the final water quality monitoring and assessment according to the water quality assessment report.

[0008] The beneficial effects of this invention are as follows: Through multimodal data fusion technology, water quality monitoring parameters from different sources (such as pH, dissolved oxygen, turbidity, and temperature) are effectively integrated with high-resolution images, achieving comprehensive perception of water body conditions. The global adaptive filtering and denoising algorithm dynamically adjusts the filtering intensity and range according to the real-time water quality environment, significantly improving the signal-to-noise ratio of monitoring data and enhancing the accuracy and stability of parameter data. Simultaneously, the super-resolution reconstruction technology uses a deep learning model to perform pixel-level enhancement on low-quality images, greatly improving the detail restoration capability of monitoring images and helping to capture subtle water body changes and visualize pollutant particles or microbial structures. Through time series analysis and machine learning algorithms, trend extraction and evolution modeling are performed on the denoised parameter stream. By introducing a multimodal perception mechanism (including chemical, physical, and ecological parameters), the system can identify the nonlinear relationships and coupling effects between various parameters, thereby effectively judging the potential factors of water quality changes. The constructed water quality state trend evolution map not only visualizes the trajectory of water quality changes but also reveals short-term fluctuations and long-term evolution trends, providing important support for sudden pollution monitoring and aquatic ecosystem evolution analysis. By combining water quality parameters with biological response indicators (such as phytoplankton abundance, algal population dynamics, and aquatic organism behavior), a deeper analysis of water health can be conducted from an ecological perspective. The biological indicator data stream extracted based on filtered and optimized parameter streams exhibits higher accuracy and representativeness, effectively identifying anomalous ecosystem responses. Multidimensional response analysis using trend evolution maps reveals the sensitive response mechanisms of biological communities to water quality changes and assesses the specific impact of different pollutants on the ecosystem. Inputting high-precision images into deep visual recognition models (such as YOLO and Mask R-CNN) allows the system to accurately identify and label fluorescent pollutants, microplastic particles, and other visualized pollutants. The high detail rendering capabilities of super-resolution images significantly improve the identification rate and classification accuracy of small pollutants. Utilizing the temporal dimension of image sequences for inter-frame change detection and spatial distribution modeling, a dynamic evolution map of pollutant diffusion over time can be constructed. This map not only shows the spatial migration paths of pollutants but also reveals their diffusion speed and trends, providing a quantitative analytical basis for pollution control. By leveraging the spatiotemporal mapping relationship between multi-dimensional trend maps and spatial pollution distribution maps, spatiotemporal data mining algorithms (such as spatiotemporal clustering, path backtracking analysis, and causal network modeling) are used to achieve intelligent location of pollution sources and inversion of diffusion paths. Reverse inference technology can trace the initial time and location of pollution occurrence, clarifying the behavioral patterns and development trajectories of pollution sources. The pollution diffusion evolution prediction map constructed on this basis has strong foresight and accuracy, capable of predicting future pollutant diffusion trends and potential risk areas. Integrating the prediction map with ecological assessment reports, and using visualization rendering technologies (such as GIS 3D rendering, heatmap overlay, and dynamic pollution cloud map display) provides a dynamic visual demonstration of the pollution diffusion process.The ecological assessment results comprehensively determine the ecological restoration capacity and ecological sensitivity of polluted areas, providing a precise basis for environmental governance measures. The final monitoring and assessment not only covers the diagnosis of the current pollution status but also includes trend warnings, ecological risk assessments, and governance recommendations, providing a systematic solution for achieving sustainable water resource management. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the steps of an intelligent processing method for multimodal water quality monitoring data according to the present invention; Figure 2 A flowchart for a multimodal water quality monitoring data intelligent processing system; Figure 3 This is a structural diagram of a multimodal water quality monitoring data intelligent processing system. Detailed Implementation

[0010] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0011] This application provides a method and system for intelligent processing of multimodal water quality monitoring data. The executing entities of the method and system include, but are not limited to, mechanical equipment, data processing platforms, cloud server nodes, and network upload devices that can be considered general computing nodes in this application. The data processing platform includes, but is not limited to, at least one of an audio-visual management system, an information management system, and a cloud-based data management system.

[0012] Please see Figures 1 to 3 This invention provides an intelligent processing method for multimodal water quality monitoring data, which includes the following steps: Step S1: Acquire water quality monitoring parameter stream and water quality monitoring image, and perform adaptive processing to obtain filtered and optimized parameter stream and optimized monitoring image; Step S2: Based on the adaptive filtering optimization parameter flow, sense the evolution of water quality status and construct a water quality status evolution map; Step S3: Based on the water quality status trend evolution map, conduct biological indicator response analysis and water quality ecological impact assessment to obtain a water quality assessment report; Step S4: Perform depth target visual recognition on the super-resolution optimized monitoring images, and analyze the temporal frame spatial distribution changes to construct a pollutant spatial distribution map; Step S5: Based on the water quality status trend evolution map, perform spatiotemporal correlation mining of pollutant spatial distribution map and spatiotemporal reverse inference of pollution sources to obtain pollution diffusion evolution prediction map. Step S6: Perform dynamic pollution visualization rendering based on the pollution diffusion evolution prediction map, and conduct the final water quality monitoring assessment based on the water quality ecological assessment report.

[0013] In the embodiments of the present invention, see Figure 1 The diagram below illustrates the steps of a multimodal water quality monitoring data intelligent processing method according to the present invention. In this example, the steps of the multimodal water quality monitoring data intelligent processing method include: Step S1: Acquire water quality monitoring parameter stream and water quality monitoring image, and perform adaptive processing to obtain filtered and optimized parameter stream and optimized monitoring image; In this embodiment, multiple sensors (such as pH sensors, turbidity sensors, dissolved oxygen sensors, etc.) are used to monitor water quality in real time during the parameter stream acquisition process. Each sensor should be connected to a central processing unit to ensure real-time data transmission. Whenever a sensor collects new data, the system will automatically record relevant information, including timestamp, sensor type, measurement value, etc., and store the data in a database. To ensure data accuracy, sensor calibration is performed regularly, and the sensor's operating status is monitored to ensure its normal operation. Data redundancy measures are taken to prevent data loss and ensure that monitoring data at each time point is completely recorded. The result of this process will generate real-time stream data containing various water quality parameters, providing a basis for subsequent analysis. After acquiring the multimodal water quality monitoring parameter stream, real-time water quality monitoring images are acquired. Real-time images of the water quality pool are captured using a high-resolution camera for subsequent image processing and analysis. The image resolution is set to 1920x1080 pixels to ensure image quality.

[0014] During image acquisition, high-resolution cameras periodically capture images of the water quality pool. The image acquisition frequency is set to once per minute to ensure dynamic changes in water quality are captured. Each image should include a timestamp for subsequent correlation with parameter stream data. Image data will be stored in a central database and backed up regularly to prevent data loss. During image acquisition, the stability and cleanliness of the camera are ensured to avoid image quality issues caused by blurring or contamination. The resulting real-time water quality monitoring images will provide the necessary foundational data for subsequent image processing and analysis. After acquiring the multimodal water quality monitoring parameter stream and real-time water quality monitoring images, global adaptive filtering and noise reduction are performed to improve the signal-to-noise ratio and data quality of the parameter stream. Noise in the monitoring data is removed to ensure the accuracy of subsequent analysis. The filter window size is set to 5x5 to achieve a smooth filtering effect. The implementation of global adaptive filtering includes selecting an appropriate filtering algorithm, such as adaptive mean filtering or Wiener filtering. Each monitoring parameter is preprocessed to remove outliers and noise. The adaptive filter is applied to the monitoring parameter stream to update the optimized parameter data in real time. When applying filters, the filter parameters are dynamically adjusted based on the local characteristics of each data point. When a parameter fluctuates drastically within a short period, the filter will increase its intensity to reduce the impact of noise on the results. The resulting adaptive filter optimization parameter stream will significantly improve data quality and provide reliable data support for subsequent analysis.

[0015] After completing global adaptive filtering, real-time water quality monitoring images undergo super-resolution reconstruction to improve image resolution and clarity. Reconstruction techniques yield higher-quality images for subsequent image analysis and target recognition. A reconstruction factor of 2x is set to achieve higher image clarity. The super-resolution reconstruction process involves selecting a suitable reconstruction algorithm, such as SRCNN (a deep learning technique) or ESPCN (an efficient super-resolution convolutional network). The acquired real-time water quality monitoring images are input into a trained super-resolution model. Through model inference, high-resolution image results are generated. During reconstruction, the algorithm is ensured to effectively capture detailed information in the image, and quality is evaluated after reconstruction. Metrics such as PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity Index) can be used to evaluate the quality of the reconstructed image. The final generated super-resolution optimized monitoring images will provide high-quality visual information for subsequent image processing and analysis. After completing the adaptive filtering parameter optimization and generating the super-resolution optimized monitoring images, results are integrated and data is stored. The processed data is organized and archived for subsequent analysis and use. Data storage formats are set to CSV and JPEG for easy data access and processing. During the integration process, the adaptive filtering optimization parameter stream is correlated with the super-resolution optimized monitoring images to ensure that each data point corresponds to its respective image data. A database table is established to record the parameter stream and corresponding image data for each time point, facilitating subsequent querying and analysis.

[0016] Step S2: Based on the adaptive filtering optimization parameter flow, sense the evolution of water quality status and construct a water quality status evolution map; In this embodiment, water quality fluctuation trend evolution is performed based on the adaptive filtering optimization parameter stream. Filtered water quality data is used to identify and analyze the changing trends of water quality status. The analysis time window is set to the past 72 hours to capture short-term fluctuations in water quality. During the water quality fluctuation trend evolution, various water quality monitoring data (such as pH, turbidity, dissolved oxygen, etc.) are extracted from the adaptive filtering optimization parameter stream. Data analysis tools (such as Python's Pandas and NumPy) are used to process the extracted data, calculating statistical characteristics within each time window, such as mean, standard deviation, and magnitude of change. Time series analysis techniques (such as moving average or exponential smoothing) are used to smooth the water quality data for clearer observation of water quality fluctuation trends. By setting appropriate smoothing parameters (such as a window size of 5 data points), the influence of short-term fluctuations on trend analysis can be effectively removed. During the trend evolution analysis, visualization tools (such as Matplotlib) are used to generate water quality fluctuation trend graphs, showing the changes of different water quality parameters over time. By drawing trend lines, significant water quality fluctuation points are identified, providing visualization support for subsequent analysis. The results of this process will provide the necessary data foundation for multimodal water quality state evolution perception. After completing the evolution of water quality fluctuation trends, multimodal water quality state evolution perception is performed. The relationships between different water quality parameters are comprehensively analyzed to construct a water quality state trend evolution map. The dimensions of the water quality parameters are set to 6 (e.g., pH, dissolved oxygen, turbidity, temperature, ammonia nitrogen, total phosphorus) to comprehensively reflect the water quality state. During multimodal water quality state evolution perception, the water quality parameters optimized by adaptive filtering need to be normalized to ensure that the numerical ranges of different indicators are consistent. Min-Max normalization or Z-score normalization methods can be used to transform the values ​​of various water quality parameters to the range of 0 to 1. Multidimensional data visualization techniques (such as radar charts or 3D scatter plots) are used to display the evolution of water quality state. By displaying different water quality parameters in the same graph, the interrelationships of water quality state and their changes over time can be observed intuitively. Using radar charts to display the changes of various water quality parameters helps to identify the overall trend of water quality changes. During visualization, ensure that changes in each water quality parameter are accurately reflected in the graph, using different colors or markers to distinguish different time periods. Animation effects can be incorporated to showcase the dynamic changes in water quality parameters, making the analysis results more vivid and intuitive. After completing the perception of multimodal water quality state evolution, construct a water quality state trend evolution map. Integrate the previously obtained water quality fluctuation trends and multimodal state evolution information into a comprehensive chart to facilitate more in-depth analysis and decision support. Set the chart's update frequency to once per hour to ensure data timeliness. The process of constructing the water quality state trend evolution map includes integrating the analyzed water quality fluctuation trends with the visualization results of multimodal water quality parameters.Composite charts (such as those with multiple subplots) can be used to display the changing trends of different water quality parameters. Each subplot can display different water quality parameters separately, while the overall trend is shown in the main chart. When constructing charts, readability and aesthetics should be ensured. Interactivity and information content can be improved by adjusting axis labels, adding legends and annotations. Data labels should be used to display specific values ​​at each time point, allowing decision-makers to obtain key information.

[0017] Step S3: Based on the water quality status trend evolution map, conduct biological indicator response analysis and water quality ecological impact assessment to obtain a water quality assessment report; In this embodiment, data analysis tools (such as Python's Pandas library) are used to filter the adaptive filtering optimization parameter stream. Relevant data are extracted based on preset bioindicator types (such as phytoplankton, bacterial count, dissolved oxygen, etc.). Logical conditional filtering is used to ensure that only data points meeting the bioindicator requirements are extracted. Subsequently, the extracted data is cleaned and preprocessed. Duplicate and missing values ​​are removed to ensure data integrity and accuracy. Interpolation methods can be used to handle missing data, ensuring the continuity of bioindicator data. The final generated bioindicator data stream will provide necessary foundational data for subsequent response analysis and ecological assessment. After extracting the bioindicator data stream, bioindicator response analysis is performed. The impact of water quality changes on bioindicators is analyzed to identify the relationship between water quality status and biological ecology. The analysis time window is set to the past 72 hours to capture dynamic changes in bioindicators. During bioindicator response analysis, the bioindicator data stream is combined with a water quality status trend evolution graph. Water quality parameters and bioindicator data are paired using timestamps to form a comprehensive dataset. Regression analysis or correlation analysis methods (such as Pearson correlation coefficient) are used to assess the impact of water quality parameters (such as pH, dissolved oxygen, and ammonia nitrogen) on bioindicators. Visualization tools (such as Seaborn or Matplotlib) are used to plot scatter plots or regression curves to demonstrate the relationship between water quality parameters and bioindicators. The R² value of the regression model is analyzed to assess the model's fit and determine the strength of the impact of water quality changes on bioindicators. The final analytical results will provide necessary data support for subsequent water quality ecological impact assessments.

[0018] After completing the biomarker response analysis, a water quality ecological impact assessment is conducted. A comprehensive analysis of water quality changes and biomarker responses is used to assess the potential impact of water quality on the ecosystem. A one-week baseline timeframe is set for the ecological assessment model to allow for long-term observation. During the water quality ecological impact assessment, the results of the biomarker response analysis are combined with water quality trends to identify significant water quality changes and biomarker anomalies. This can be assessed by setting thresholds (such as national water quality standards) to determine whether water quality exceeds the standards, thereby evaluating its impact on the ecological environment. Quantitative assessments are performed using ecological models (such as ecological risk assessment models or species habitat models). Based on changes in water quality parameters and biomarkers, an ecological impact index is calculated to determine the degree of impact of water quality changes on the ecosystem. Statistical analysis methods (such as analysis of variance) are used to compare the ecological impacts across different time periods to identify potential risk factors.

[0019] Step S4: Analyze the spatial distribution changes of pollutants based on the optimized monitoring images, and perform the final water quality monitoring assessment according to the water quality assessment report.

[0020] In this embodiment, a suitable deep learning model (such as YOLOv5 or Faster R-CNN) is selected for training. These models, trained on a large amount of labeled data, can accurately identify and locate fluorescent pollutants in images. To this end, a labeled image dataset containing fluorescent pollutants needs to be prepared, and data augmentation techniques (such as rotation, scaling, and flipping) are used to increase the diversity of training samples to improve the model's generalization ability. After training, super-resolution optimized monitoring images are input into the trained model. The model outputs bounding boxes and class labels for all fluorescent pollutants in the image. Post-processing steps (such as non-maximum suppression) eliminate overlapping detection results, ensuring that each pollutant is labeled only once. The generated results will include the location information and class label of each pollutant, providing necessary data support for subsequent temporal frame spatial distribution analysis. After completing deep target visual recognition, temporal frame spatial distribution change analysis is performed. The distribution changes of fluorescent pollutants in different time frames are analyzed to identify their diffusion trends. The analysis period is set to the past 48 hours to capture the dynamic changes of pollutants. During temporal frame spatial distribution change analysis, the location information of fluorescent pollutants in each frame image is compiled into a dataset. The locations of contaminants in each frame are correlated using timestamps, forming a comprehensive dataset containing contaminant location, category, and time. Data processing tools (such as Pandas) are then used to clean and organize this data to ensure its integrity.

[0021] Spatial analysis techniques (such as spatial clustering algorithms) are employed to analyze the spatial distribution of pollutants. K-means clustering or DBSCAN algorithms can be used to identify pollutant clustering regions and analyze their changes across different time frames. Spatial distribution maps are generated using visualization tools (such as Matplotlib or Plotly) to display the distribution of pollutants at different time points. During the analysis, it is ensured that the spatial distribution of each time frame accurately reflects the dynamic changes of pollutants and identifies significant change points or trends. This process provides necessary data support for the subsequent construction of pollutant spatial distribution maps. After completing the time-series spatial distribution change analysis, a pollutant spatial distribution map is constructed. The analysis results are presented in an intuitive graphical format to facilitate the identification of pollutant distribution and diffusion trends. The map resolution is set to 1920x1080 pixels to ensure graphic clarity. The process of constructing the spatial distribution change map includes integrating the results of the time-series change analysis. By overlaying pollutant distribution information from different time frames onto the same map, the diffusion and change trends of pollutants can be observed intuitively. Interactive maps are generated using map visualization tools (such as Folium or GeoPandas) to display the spatial distribution of pollutants. When constructing the map, it is ensured that the location, concentration, and time information of each pollutant are accurately reflected on the map. Different colors or markers can be used to represent changes in pollutant concentration, and animation effects can be used to demonstrate the dynamic changes of pollutants over time. The final generated spatial distribution map of pollutants will provide important data support for subsequent environmental monitoring and management. After completing the construction of the spatial distribution map of pollutants, the results are integrated and stored. The analysis and visualization results are organized and backed up for subsequent analysis and use. The data storage format is set to CSV (for the main data) and PNG / JPEG (for the map) to facilitate subsequent data access and processing. During the integration process, the results of the time-series frame spatial distribution analysis are correlated with the constructed spatial distribution change map to ensure that the spatial information of each pollutant matches its temporal data. A database table is established to record the pollutant distribution information and its corresponding graphical data at each time point for subsequent querying and analysis.

[0022] In this embodiment, the detailed implementation steps of step S1 include: The water quality monitoring system performs real-time multimodal monitoring of the water quality pool, extracting multimodal water quality monitoring parameters and real-time water quality monitoring images. Global adaptive filtering and noise reduction are performed on multimodal water quality monitoring parameters to obtain the adaptive filtering optimized parameter stream; Differential local brightness enhancement processing is performed on real-time water quality monitoring images to construct locally enhanced and sharpened images; Super-resolution reconstruction is performed on the locally enhanced and sharpened image to construct a super-resolution optimized monitoring image.

[0023] In this embodiment, real-time multimodal monitoring of the water quality monitoring system is performed, extracting multimodal monitoring parameter streams and real-time water quality monitoring images from the water quality pool. Comprehensive water quality data is acquired for subsequent analysis and processing. A sensor sampling frequency of once per second is selected to ensure the real-time nature and accuracy of the water quality data. During multimodal monitoring, different types of sensors (such as chemical sensors, optical sensors, and temperature sensors) are used to monitor water quality parameters in real time. These sensors can measure key indicators in the water, such as pH, dissolved oxygen, turbidity, and temperature, while a high-resolution camera captures real-time images of the water quality pool. The monitoring data and images are transmitted in real time to the central processing unit for data storage and management. Each data point should be timestamped to ensure traceability for subsequent analysis. The monitoring system should have real-time monitoring capabilities to promptly issue alarms and record abnormal events when water quality parameters exceed standards. After acquiring the multimodal water quality monitoring parameters, global adaptive filtering and noise reduction processing is performed. This reduces noise in the monitoring data, improves the signal-to-noise ratio, and thus obtains an adaptively filtered optimized parameter stream. The filter window size is set to 5x5 to ensure the smoothness of the filtering effect. The global adaptive filtering process requires selecting an appropriate filtering algorithm, such as mean filtering or median filtering. For water quality parameter streams, adaptive filters can be used, dynamically adjusting filtering parameters based on local data characteristics. Local variance is used to determine the filter strength, applying stronger filtering in high-noise areas. During filtering, each monitoring parameter is preprocessed to remove outliers caused by sensor malfunctions or environmental changes. The filter is applied to the monitoring parameter stream, with parameters updated and optimized in real time. By comparing data before and after filtering, the filtering effect is evaluated to ensure the effectiveness and reliability of noise reduction. After adaptive filtering, differentiated local brightness enhancement processing is performed on the real-time water quality monitoring image. This improves the visibility of important areas in the image for subsequent analysis and interpretation. A brightness enhancement margin of 30% is set to ensure image clarity without distortion. The local brightness enhancement process requires segmenting the real-time monitoring image to identify key areas (such as the water surface, plankton, etc.). Image segmentation algorithms (such as Canny edge detection or thresholding) can be used to identify these areas. For the identified local areas, contrast enhancement techniques (such as histogram equalization or adaptive histogram equalization) are applied to enhance brightness. By adjusting brightness and contrast parameters, details in these areas are made more apparent. This process can be implemented using the OpenCV library in Python, ensuring the enhancement effect meets practical requirements. After enhancement, image quality is evaluated to ensure the processed image has high visual clarity and readability. User feedback or expert review verifies whether the enhancement effect achieves the expected goals. Super-resolution reconstruction is then performed on the locally enhanced and sharpened image to construct a super-resolution optimized monitoring image. This increases the image resolution, making details in the water quality monitoring image more clearly visible.A reconstruction factor of 2x was set to achieve higher image clarity. The super-resolution reconstruction process requires selecting a suitable reconstruction algorithm, such as deep learning-based super-resolution reconstruction (e.g., SRCNN or ESPCN). These algorithms can synthesize a high-resolution image from multiple low-resolution images, extracting detailed information. During reconstruction, locally enhanced and sharpened images are used as input data and processed by a trained super-resolution model. The reconstruction effect is evaluated by comparing the images before and after reconstruction, ensuring that details are clearly presented. Image quality evaluation metrics (such as PSNR and SSIM) are used to quantitatively assess the reconstruction effect. Based on the evaluation results, the parameters of the super-resolution reconstruction model are further optimized to achieve optimal image clarity. Through these steps, the constructed super-resolution optimized monitoring image will significantly improve the visualization effect of water quality monitoring, providing more accurate information support for subsequent data analysis and decision-making. This process not only improves the efficiency of water quality monitoring but also provides new ideas for the development of environmental monitoring technology.

[0024] In this embodiment, the specific steps for performing global adaptive filtering and noise reduction on multimodal water quality monitoring parameters to obtain the adaptive filtering optimized parameter flow are as follows: Real-time data stream missing detection is performed on multimodal water quality monitoring parameter streams to identify abnormally missing parameters; Extract the location and timestamp of abnormally missing parameters; Based on the location and timestamp, mark the neighboring parameters of surrounding time points; The average value of the nearby parameters at the surrounding time points is calculated to obtain the average value of the missing parameter type; The average value is used to perform average interpolation filling to obtain the interpolation filling optimized monitoring parameter stream; Define a long time period and divide the interpolation filling optimization monitoring parameter stream into multiple time windows to obtain sub-parameter streams for multiple time windows; Perform a window-by-window Fourier transform on the sub-parameter stream to obtain the frequency domain diagram of the parameter stream for each window; High-frequency noise is identified based on the frequency domain diagram of the parameters, and noise points in each window are marked. Calculate the noise frequency and power of each noise point in the window; Global adaptive filtering and noise reduction are performed based on the noise frequency and power to obtain the adaptive filtering optimization parameter stream.

[0025] In this embodiment, real-time data stream missing detection is performed on the multimodal water quality monitoring parameter stream to identify abnormally missing parameters. This timely detection of potential data loss issues in water quality monitoring facilitates subsequent processing. In the specific implementation, a data integrity check algorithm is used, and a reasonable data integrity threshold is set. Assuming a sampling frequency of once per minute, the water quality monitoring parameter stream per minute should include three parameters: pH, dissolved oxygen, and turbidity. If a parameter is not collected within a certain minute, the system will mark it as missing. A data stream processing script (such as using Python's pandas library) is written to traverse the monitoring data and record the location and timestamp of missing parameters. For example, if the dissolved oxygen parameter is missing at 11:05 on May 13, 2023, the system will record it as: Timestamp: 2023-05-13 11:05 Missing parameter: Dissolved oxygen After identifying missing parameters, the location and timestamp of the abnormally missing parameters are extracted. Detailed information about the missing data is collected to facilitate subsequent interpolation and imputation.

[0026] Create a list to record information about missing parameters, including the timestamp and location of each missing parameter. Assume multiple missing parameters are found during monitoring, and record them as follows: Missing item 1: Timestamp: 2023-05-13 11:05, Parameter: Dissolved Oxygen Missing item 2: Timestamp: 2023-05-13 11:15, Parameter: Turbidity Using iteration and conditional judgment, the timestamps and locations of all missing parameters are extracted. Information about missing parameters is filtered out by iterating through all monitoring data. After extracting the missing parameter information, neighboring parameters at surrounding time points are marked based on the location and timestamp. This provides reference data for calculating missing parameters. A time window (e.g., 5 minutes before and after) is set, and surrounding time points are marked based on the timestamp of the missing parameter. For a missing parameter with a timestamp of 2023-05-13 11:05, all parameters between 2023-05-13 10:55 and 2023-05-13 11:15 are marked. All parameters within the marked time window are extracted by iterating through the monitoring data. Assume the parameters recorded within this time period are as follows: 10:55, pH: 7.1, Dissolved Oxygen: 8.0, Turbidity: 9.0 11:00, pH: 7.2, Dissolved Oxygen: 8.5, Turbidity: 10.0 11:10, pH: 7.3, Dissolved Oxygen: 8.3, Turbidity: 9.5 11:15, pH: 7.2, Turbidity: 10.5 The neighboring parameters of the missing parameters at surrounding time points were successfully identified, providing a basis for subsequent averaging calculations. After identifying the neighboring parameters at surrounding time points, these neighboring parameters were averaged to obtain the average value for the missing parameter type. The identified surrounding parameter data were extracted, and the average value for each parameter was calculated. Assuming that the extracted dissolved oxygen parameter within the aforementioned time period was: 8.0 mg / L (10:55) 8.5 mg / L (11:00) 8.3 mg / L (11:10) The average value is calculated as follows: Average dissolved oxygen concentration = (8.0 + 8.5 + 8.3) / 3 = 8.27 mg / L; The same calculation is performed for other parameters (such as turbidity). For example, the average value for the turbidity parameter is: The average turbidity value is calculated as (10.0 + 9.0 + 9.5 + 10.5) / 4 = 9.75 NTU. This yields the average value of the missing parameters, providing a basis for subsequent interpolation. After calculating the average value of the missing parameter types, interpolation is performed based on these average values ​​to obtain an interpolated and optimized monitoring parameter stream. Missing data points are filled in, making the monitoring data more complete. Linear interpolation is used to fill in the missing parameters. Assuming dissolved oxygen is missing at 11:05 on 2023-05-13, interpolation can be performed using the average values ​​of parameters before and after the missing value. Setting the dissolved oxygen at 11:00 as 8.5 mg / L and at 11:10 as 8.3 mg / L, linear interpolation can be calculated: Interpolation result = (8.5 + 8.3) / 2 = 8.4 mg / L. The calculated interpolation result is then filled into the original data stream, replacing the missing values. This forms an interpolated and optimized monitoring parameter stream, ensuring data continuity and completeness. After obtaining the interpolation-filled optimized monitoring parameter stream, a constant time period is defined, and the parameter stream is divided into multiple time windows. The data stream is divided into multiple time windows for frequency domain analysis. The length of the time window is set (e.g., 5 minutes), and the data is divided into multiple windows based on the timestamps of the monitoring data. If the monitoring data starts from 10:55 on 2023-05-13, the time window division is as follows: Window 1: 10:55 - 11:00 Window 2: 11:00 - 11:05 Window 3: 11:05 - 11:10 Window 4: 11:10 - 11:15 Data processing tools (such as pandas) were used to divide the interpolation-filled optimized monitoring parameter stream into multiple sub-parameter streams by window. Each sub-parameter stream will be used for subsequent Fourier transform analysis. Multiple time-window sub-parameter streams were successfully divided, preparing for subsequent frequency domain analysis. After completing the time window division, a window-by-window Fourier transform was performed on each sub-parameter stream to obtain the frequency domain plot of the parameter stream for each window. The frequency domain variation characteristics of water quality parameters were analyzed. The Fast Fourier Transform (FFT) algorithm was used to perform frequency domain transformation on each sub-parameter stream. Assume the sub-parameter stream for window 1 is: pH values: 7.1, 7.2, 7.3, 7.2; The frequency domain graph is obtained through FFT calculation, and the frequency and amplitude information are recorded. Assuming the obtained frequency domain graph shows an amplitude of 0.5 at 10 Hz, it indicates the presence of a significant signal component at that frequency. This process is repeated for each window, generating the frequency domain graph for each sub-parameter stream and recording the relevant frequency and amplitude. Multiple window parameter stream frequency domain graphs are formed, providing data support for subsequent noise identification. After obtaining the parameter stream frequency domain graph for each window, high-frequency noise is identified based on the parameter stream frequency domain graph, and noise points in each window are marked. Noise signals in the frequency domain are identified for subsequent processing. A threshold for noise identification is set; for example, in the frequency domain graph, if the amplitude exceeds a certain value (e.g., 0.4), it is marked as a noise point. High-frequency noise points are identified by traversing the frequency domain graph of each window. In the frequency domain graph of window 1, assuming an amplitude of 0.6 at 15 Hz, it is marked as a noise point. The noise point information for all windows is recorded, including frequency and amplitude. The noise points in each window are successfully marked, providing a basis for subsequent noise frequency and power calculations. After marking the noise points in each window, calculate the noise frequency and power of each noise point. Assess the impact of noise on the monitoring data. Calculate the noise frequency and power. The noise frequency comes directly from the marked points in the frequency domain plot, and the noise power can be calculated by squared amplitude. Assuming that in window 1, the noise frequency is 15 Hz and the amplitude is 0.6, the power is calculated as follows: Noise power = (0.6)² = 0.36; The noise frequency and power of all windows are recorded to form a noise feature dataset. The noise frequency and power of each noise point in each window are successfully calculated, providing data support for subsequent adaptive filtering and noise reduction. Global adaptive filtering and noise reduction are performed based on the noise frequency and power, resulting in an adaptive filtering optimization parameter stream. This reduces noise interference in the monitoring data and improves data quality. A suitable adaptive filtering algorithm is selected, such as Kalman filtering or Wiener filtering. Based on the previously calculated noise feature data, noise model parameters are set to achieve noise suppression. When using Wiener filtering, the ratio of noise power to signal power is used to set the filter weights. The filtering algorithm is applied within each window to remove noise components and retain useful signals. The resulting adaptive filtering optimization parameter stream ensures the accuracy and stability of the monitoring data, providing reliable support for subsequent analysis and decision-making.

[0027] In this embodiment, the specific steps for performing differentiated local brightness enhancement processing on the real-time water quality monitoring image to construct a locally enhanced and sharpened image are as follows: Real-time water quality monitoring images are divided into multiple sub-region images by image region thresholding; Calculate the dark channel value of the sub-region image; The brightness of each region is evaluated based on the dark channel value to obtain the brightness evaluation value of each sub-region. Based on the brightness evaluation value, the region is divided to obtain overly dark and overly bright regions; Perform detailed analysis on multiple sub-region images to identify image details in each sub-region; The detail complexity is quantified based on the image details of each sub-region to obtain the image detail complexity of each sub-region; Differential local brightness enhancement processing is performed on the overly dark and overly bright areas, and dynamic sharpening parameters are adjusted according to the complexity of image details to construct a locally enhanced and sharpened image.

[0028] In this embodiment, image region thresholding is performed on real-time water quality monitoring images to obtain multiple sub-region images. The image is segmented into different regions for subsequent brightness assessment and detail analysis. In the specific implementation, image processing techniques (such as the Otsu thresholding method) are used to segment the water quality monitoring image. Assuming the original image resolution is 1920x1080 pixels, the image is converted to grayscale for thresholding. The threshold is calculated using the Otsu algorithm, which determines the optimal threshold by maximizing the inter-class variance. The calculated optimal threshold is assumed to be 128 (range 0-255). This threshold is used to segment the image into multiple sub-regions, marking the different regions. The image may be divided into water areas, debris areas, and background areas. Successfully obtaining multiple sub-region images provides a basis for subsequent brightness assessment and detail analysis. After completing the image region thresholding, the dark channel values ​​of the sub-region images are calculated. The dark channel characteristics of each sub-region in the image are evaluated for brightness assessment. The core idea of ​​dark channel value calculation is to find the minimum value of each pixel in the image to reflect its dark channel characteristics. In the specific implementation process, assuming that the RGB values ​​of a pixel in a sub-region image are (R, G, B), its dark channel value is calculated as: Dark channel value = min(R, G, B). Pixel-by-pixel calculation is performed for each sub-region image to obtain the dark channel map of each sub-region. Assuming the dark channel value matrix of a certain sub-region is: Sub-region 1 dark channel value matrix = [[30, 40, 20], [25, 50, 35], [10, 20, 15]], the dark channel values ​​of each sub-region are obtained, providing reference data for subsequent brightness evaluation. After calculating the dark channel values ​​of each sub-region, brightness evaluation is performed for each region to obtain the brightness evaluation value of each sub-region. The brightness characteristics of each region are quantified for region division. For the dark channel values ​​of each sub-region, its average brightness evaluation value is calculated. For sub-region 1, the average value of its dark channel values ​​is calculated: Average dark channel value = (30 + 40 + 20 + 25 + 50 + 35 + 10 + 20 + 15) / 9 ≈ 25.56 The same method was used to evaluate the brightness of other sub-regions. Assume the average dark channel value for sub-region 2 is: Sub-region 2 average dark channel value = 60; By averaging the dark channel values ​​of each sub-region, the final brightness evaluation value for all sub-regions is obtained. For example: The estimated brightness value for sub-region 1 is approximately 25.56. Sub-region 2 brightness assessment value = 60 These brightness assessment values ​​will provide a basis for subsequent zoning.

[0029] After obtaining the brightness assessment value for each sub-region, the regions are divided according to these assessment values ​​to identify overly dark and overly bright areas. Special areas requiring processing are then identified. A brightness threshold is set. Assuming a threshold of 30, any region with a brightness assessment value below this value is considered an overly dark region, and any region with a brightness assessment value above the threshold is considered an overly bright region.

[0030] Based on the brightness assessment values, each sub-region is classified as follows: If the brightness assessment value of sub-region 1 is approximately 25.56, it is marked as an overly dark region.

[0031] If the brightness evaluation value of sub-region 2 is 60, it is marked as an overly bright region.

[0032] Successfully segmenting overly dark and overly bright regions provides data support for subsequent detail analysis and local enhancement processing. After region segmentation, detail analysis is performed on multiple sub-region images to identify the image details of each sub-region. Detail information in the image is extracted and quantified for subsequent processing. Edge detection algorithms (such as Canny edge detection) are used to process each sub-region image to extract detail features. Assuming the Canny algorithm is applied to the overly dark region image, the resulting edge image shows obvious detail features. Detail information for each sub-region is recorded, such as the number and intensity of edges. Assuming 100 edges are detected in the overly dark region, with an intensity distribution range of [0, 255]. Image details in each sub-region are identified, providing basic data for subsequent detail complexity quantification. After identifying the image details in each sub-region, these details are quantified to obtain the image detail complexity of each sub-region. The richness of image details is evaluated to guide subsequent enhancement processing. A complexity calculation formula is defined, for example, image complexity can be represented by a combination of edge number and edge intensity: Complexity = Edge Number × Average Intensity; Assuming the number of edges in the darkest region is 100 and the average intensity is 150, the complexity is calculated as follows: Complexity = 100 × 150 = 15000; The same method is used to quantify the complexity of all sub-regions. The image detail complexity of each sub-region is obtained, forming a list of complexity values ​​for subsequent processing. After completing the detail complexity quantification, differential local brightness enhancement is performed on overly dark and overly bright areas. This improves the image quality of these areas to enhance visibility. For overly dark areas, a local brightness enhancement algorithm (such as CLAHE) is used. This algorithm adaptively enhances contrast within local areas, thereby improving detail. For example, applying CLAHE to overly dark areas increases the enhanced brightness range from 25.56 to 40. For overly bright areas, exposure compensation techniques are used to reduce the brightness to an appropriate level. For example, reducing the brightness of overly bright areas from 60 to 45 ensures no loss of detail. This differential local brightness enhancement provides the basis for subsequent dynamic sharpening parameter adjustments. After completing the local brightness enhancement, dynamic sharpening parameter adjustments are performed based on the image detail complexity of each sub-region. The sharpening level is flexibly adjusted according to the detail complexity to achieve the best results. A strategy for adjusting the sharpening parameters is defined. If the detail complexity exceeds a certain threshold (e.g., 10000), the sharpening intensity is increased; if it falls below this threshold, the sharpening intensity is decreased. For example, for excessively dark areas with a detail complexity of 15000, the sharpening intensity is set to strong (e.g., 1.5x sharpening). For excessively bright areas with a detail complexity of 9000, the sharpening intensity is set to medium (e.g., 1.2x sharpening). The image is processed using a sharpening algorithm (e.g., Laplacian sharpening), dynamically adjusted according to the set intensity. This generates a locally enhanced sharpened image, ensuring detail is preserved and visual effects are improved.

[0033] In this embodiment, the detailed implementation steps of step S2 include: The temporal changes of water conductivity, pH, dissolved oxygen and temperature are calculated based on the adaptive filtering optimization parameter flow to obtain the parameter flow difference at different time points; Calculate the rate and magnitude of change of the parameter flow difference; The evolution of water quality fluctuation trends is carried out based on the rate and magnitude of change in order to obtain multimodal water quality fluctuation trend characteristics. Multi-time point trend prediction is performed on the multimodal water quality fluctuation trend characteristics to obtain multi-time point water quality trend prediction values; Multimodal water quality state evolution perception is performed on the predicted water quality trends at multiple time points, and a water quality state trend evolution map is constructed.

[0034] In this embodiment, the temporal changes of water conductivity, pH, dissolved oxygen, and temperature are calculated based on adaptive filtering to optimize the parameter flow, thereby obtaining the parameter flow differences at different time points. Water quality parameter flow data recorded by a real-time monitoring system is utilized. Assume that the recorded water quality parameter flows during a certain monitoring period are as follows: Time point 1: Conductivity 300 µS / cm, pH 7.2, dissolved oxygen 8.5 mg / L, temperature 25°C Time point 2: conductivity 310 µS / cm, pH 7.1, dissolved oxygen 8.3 mg / L, temperature 26°C Time point 3: conductivity 320 µS / cm, pH 7.3, dissolved oxygen 8.7 mg / L, temperature 24°C The parameter flow difference at different time points is calculated as follows: Conductivity difference = 310 - 300 = 10 µS / cm pH difference = 7.1 - 7.2 = -0.1 Dissolved oxygen difference = 8.3 - 8.5 = -0.2 Temperature difference = 26 - 25 = 1°C By performing the above calculations at all time points, a complete dataset of parameter flow difference values ​​is obtained, providing a foundation for subsequent calculations of change rate and magnitude. After calculating the parameter flow difference values ​​at different time points, the change rate and magnitude are calculated. The variation characteristics of water quality parameters are quantified to assess the fluctuation trend of water quality.

[0035] The rate of change of each parameter is calculated based on the difference in flow parameters. The calculation formula is assumed to be: Rate of change = Difference / Time interval; If the time interval is 1 minute, the rate of change is calculated as follows: Rate of change of conductivity = 10 µS / cm / 1 min = 10 µS / cm / min pH change rate = -0.1 / 1 min = -0.1 pH / min Dissolved oxygen change rate = -0.2 / 1 min = -0.2 mg / L / min Rate of temperature change = 1 / 1 min = 1°C / min The magnitude of the change in the parameter flow difference is calculated and defined as the absolute value of the difference: The change in conductivity = |10| µS / cm = 10 µS / cm pH change range = |-0.1| = 0.1 Dissolved oxygen change = |-0.2| = 0.2 mg / L Temperature change range = |1| = 1°C The rate and magnitude of change for each water quality parameter were determined, providing data support for subsequent water quality fluctuation trend evolution. After obtaining the rate and magnitude of change, water quality fluctuation trend evolution analysis was performed based on this data. The trend evolution characteristics of water quality parameters were identified for effective monitoring and early warning. By performing aggregate analysis of parameter flow change rates, the trend of water quality fluctuations was identified. The average and standard deviation of the change rate over each time period can be statistically analyzed to determine the stability of the fluctuations. For example, if the average rate of change of conductivity is 10 µS / cm / min with a standard deviation of 2 over a certain period, it indicates that the conductivity change is relatively stable. The same analysis was performed for pH, dissolved oxygen, and temperature parameters. The evolution trend of the change rate was visualized using graphing tools (such as Matplotlib) to form trend charts, facilitating the observation of the changes in each water quality parameter over time. Multimodal water quality fluctuation trend characteristics were successfully obtained, providing basic data for subsequent trend prediction. After completing the water quality fluctuation trend evolution analysis, multi-time-point trend predictions were performed on the multimodal water quality fluctuation trend characteristics. Historical data was used to predict future water quality parameters. Choose a suitable time series forecasting model, such as the ARIMA model or an LSTM neural network. Build the forecasting model by training it on historical data. Use water quality data from the past 24 hours for training to predict water quality parameters for the next 2 hours. After model building, validate the model using a test set to ensure its prediction accuracy. Assume the model predicts a future conductivity of 330 µS / cm, pH of 7.4, dissolved oxygen of 8.6 mg / L, and temperature of 25.5°C. Record the prediction results to form multi-time-point water quality trend predictions. These multi-time-point water quality trend predictions provide data support for subsequent water quality state evolution sensing. After obtaining the multi-time-point water quality trend predictions, perform multi-modal water quality state evolution sensing on these values. Comprehensively analyze the changes in different water quality parameters to construct a comprehensive water quality state evolution map. Integrate the multi-time-point predictions into a single data framework for convenient comprehensive analysis. A data table is generated as follows: Time point: 2023-05-13 11:00, Conductivity: 330 µS / cm, pH: 7.4, Dissolved oxygen: 8.6 mg / L, Temperature: 25.5°C. A water quality state evolution graph is generated using visualization tools (such as Plotly or Matplotlib) to display the changing trends of different parameters on the time axis. Multiple curves can be selected to be displayed on the same graph, facilitating the observation of the relationships and trends between various parameters. This water quality state trend evolution graph provides intuitive data support for subsequent water quality monitoring and management.

[0036] In this embodiment, the detailed implementation steps of step S3 include: Biological indicator data streams are extracted based on adaptive filtering and optimized parameter streams, and time-series variation fitting is performed to construct time-series variation sequences of biological indicators. Multi-state characteristic evolution analysis was performed on the water quality state trend evolution map to extract the diffusion, migration, accumulation and transformation characteristics of water quality state; Spatiotemporal feature mapping is performed based on the diffusion, migration, accumulation and transformation characteristics of water quality state to construct the spatiotemporal evolution flow field of water quality state; Dynamic calculation of water quality state gradient is performed on the spatiotemporal evolution flow field of water quality state to generate dynamic gradient field of water quality state. Based on the time-series change sequence of biological indicators, the response of biological indicators to the dynamic gradient field of water quality is analyzed to generate the sensitivity law of biological indicator response under changes in water quality. A water quality ecological impact assessment is conducted based on the sensitivity patterns of the aforementioned biological indicators to obtain a water quality ecological assessment report.

[0037] In this embodiment, a bioindicator data stream is extracted based on an adaptive filtering-optimized parameter stream, and time-series variation fitting is performed. The bioindicator data is collected hourly to ensure timeliness and accuracy. During the extraction process, relevant data related to bioindicators (such as phytoplankton, bacterial count, dissolved oxygen, etc.) are selected from the adaptively filtered parameter stream. Data processing tools (such as Python's Pandas library) are used for data cleaning and preparation to ensure data integrity. Time-series variation fitting is performed on the extracted bioindicator data stream. A suitable fitting model (such as linear regression or polynomial fitting) is used, and the fitting parameters are optimized using the least squares method to capture the changing trends of bioindicators. The fitting time range is set to the past month to observe long-term changes in bioindicators. During the fitting process, cross-validation is used to evaluate the accuracy of the model and ensure the reliability of the fitting results. The constructed bioindicator time-series variation sequence will provide basic data for subsequent analysis, helping to identify the changing patterns of bioindicators under different environmental conditions. After constructing the bioindicator time-series variation sequence, multi-dimensional water quality state trend evolution analysis is performed. This study extracts diffusion, migration, accumulation, and transformation characteristics of water quality states to understand their dynamic changes. A 24-hour analysis window is set to comprehensively reflect water quality trends. Historical data on water quality parameters (such as pH, dissolved oxygen, and turbidity) are collected to construct a water quality state characteristic matrix during multi-dimensional water quality state trend evolution analysis. Principal component analysis (PCA) and other dimensionality reduction techniques are employed to extract key water quality features, reducing dimensionality and improving the efficiency of subsequent analyses. Time-series analysis methods (such as Dynamic Time Warping (DTW)) are used to compare water quality states and identify diffusion, migration, accumulation, and transformation characteristics. The rate of change of water quality parameters over different time periods is analyzed to reveal migration trends in water quality states. Visualization tools (such as Matplotlib) are used to generate water quality state trend evolution maps, visually displaying the evolution of different water quality characteristics. These analyses not only identify water quality state trends but also provide a clear direction for subsequent spatiotemporal feature mapping. After multi-dimensional water quality state trend evolution analysis, spatiotemporal feature mapping is performed based on the diffusion, migration, accumulation, and transformation characteristics of water quality states. To better understand water quality dynamics, a spatiotemporal evolution flow field of water quality status is constructed. The temporal granularity of the spatiotemporal mapping is set to once per hour to ensure the granularity of the spatiotemporal data. The process of spatiotemporal feature mapping involves combining water quality status characteristics with temporal and spatial information, and using interpolation algorithms (such as Kriging interpolation) to construct a spatial and temporal distribution model of water quality status. By mapping the spatial coordinates of water quality monitoring points, a spatiotemporal evolution flow field of water quality status is generated. During the construction process, it is ensured that the mapping results can reflect the changing patterns of water quality status, and flow field visualization tools (such as streamline diagrams) are used to display the spatiotemporal evolution characteristics of water quality status.This process provides foundational data for subsequent dynamic calculations, making the spatial distribution and temporal evolution of water quality changes clearer. After constructing the spatiotemporal evolution flow field of water quality states, dynamic calculations of water quality state gradients are performed. The rate and direction of change of water quality states are identified, generating a dynamic gradient field of water quality states. The spatial resolution of the gradient calculation is set to 1 meter to ensure accuracy. The process of dynamic calculation of water quality state gradients includes using numerical differentiation methods to calculate the gradient of the water quality state flow field. The finite difference method is used to calculate the rate of change of water quality parameters in space and time to obtain gradient information of water quality states.

[0038] During the calculation process, it was ensured that the gradient information at each monitoring point accurately reflected water quality changes, and the calculation results were recorded. This process provides necessary data support for subsequent bioindicator response analysis, ensuring the comprehensiveness and accuracy of the analysis. After generating the dynamic gradient field of water quality status, bioindicator response analysis was performed based on the time-series change sequence of bioindicators. The impact of water quality status changes on bioindicators was analyzed, and the sensitivity law of bioindicator response under water quality status changes was generated. The time window for response analysis was set to 10 minutes to capture short-term changes in bioindicators. The process of bioindicator response analysis included combining the time-series change sequence of bioindicators with the dynamic gradient field of water quality status to identify the impact of water quality status changes on bioindicators. Regression analysis could be used to assess the degree of influence of water quality parameter changes on bioindicators and calculate the sensitivity coefficient. During the analysis, it was ensured that the response of each bioindicator was accurately recorded, and visualization tools were used to display the changing trends of bioindicators. This process will provide a basis for subsequent ecological impact assessment, ensuring the scientific nature and accuracy of the assessment. Water quality ecological impact assessment was conducted based on the described bioindicator response sensitivity law. A water quality ecological assessment report is generated, providing a comprehensive analysis of the impact of water quality status on the ecological environment. The report is updated weekly to ensure the timeliness of the results. The process of conducting the water quality ecological impact assessment includes a comprehensive analysis of water quality status, dynamic gradient fields, and bioindicator responses to identify the potential impacts of water quality changes on the ecosystem. Quantitative assessments are conducted using ecological models (such as ecological risk assessment models) to calculate the degree of impact of water quality changes on the ecological environment. During the assessment process, the results are presented visually, summarizing the analysis through charts and reports to form a complete water quality ecological assessment report. This process not only provides a scientific basis for water quality management and governance but also provides data support for the formulation of ecological protection measures, promoting sustainable environmental development.

[0039] In this embodiment, step S4 includes the following steps: Perform deep target visual recognition on super-resolution optimized monitoring images to mark all fluorescent pollutants; The number of fluorescent pollutants is calculated frame by frame to obtain the number of pollutants in each frame; The time-series quantity change is calculated based on the quantity of pollutants to obtain the pollutant quantity change curve; All fluorescent pollutants were spatially located to obtain the spatial coordinates of each fluorescent pollutant. Based on the pollutant quantity change curve, a temporal frame spatial distribution change analysis is performed on the spatial positioning coordinates to construct a pollutant spatial distribution map. The final water quality monitoring and assessment will be conducted based on the spatial distribution map of pollutants and the water quality assessment report.

[0040] In this embodiment, deep target visual recognition is performed on the super-resolution optimized monitoring images to label all fluorescent pollutants. All fluorescent pollutants are automatically identified and labeled from the images, providing basic data for subsequent analysis. In the specific implementation, a suitable deep learning model is selected for target detection, such as YOLO (You Only Look Once) or Faster R-CNN, which perform well in object detection. Assuming the YOLOv5 model is used, it needs to be trained using a labeled fluorescent pollutant image dataset for training and validation. After training, target detection is performed using the optimized monitoring images. By inputting the images into the trained model, the model outputs the bounding box and its confidence score for each fluorescent pollutant. Assuming that five fluorescent pollutants are detected in one frame, they are labeled as follows: Pollutant 1: Bounding box coordinates (x1, y1, x2, y2), confidence level 0.95 Pollutant 2: Bounding box coordinates (x3, y3, x4, y4), confidence level 0.92 ... All fluorescent contaminants were successfully labeled, providing data support for subsequent quantity calculations. After labeling, a frame-by-frame quantity calculation was performed to obtain the number of contaminants in each frame. The number of fluorescent contaminants present in each frame was quantified for time-series variation analysis. The labeled fluorescent contaminants were counted. A programming language (such as Python) was used to iterate through the detection results of each frame and count the number of labeled contaminants in each frame. Assuming that the number of fluorescent contaminants detected in a certain frame is: Frame 1: 5 Frame 2: 7 Frame 3: 4 The pollutant quantity for each frame is recorded in a list or data frame, forming a time-series dataset of pollutant quantities. This dataset will be used for subsequent quantity change calculations. Obtaining the pollutant quantities for all frames provides the basic data for time-series quantity change calculations. After obtaining the pollutant quantities for each frame, time-series quantity change calculations are performed to obtain the pollutant quantity change curve. The trend of fluorescent pollutant quantity changes over time is analyzed. A time series is defined. Assuming the time interval between each frame is 1 second, the frame sequence and the corresponding time series can be established as follows: Time: t1, t2, t3, ... Pollutant quantities: 5, 7, 4, ... Use data analysis tools (such as NumPy or Pandas) to calculate the rate of change in pollutant quantities. Calculate the difference in quantities between each frame and the previous frame: Δ quantity = Current frame quantity - Previous frame quantity; By calculating the rate of change, a change curve can be constructed. Assume the results are as follows: The rate of change from frame 1 to frame 2: 7 - 5 = 2 The rate of change from frame 2 to frame 3: 4 - 7 = -3 Generate pollutant quantity change curves and plot them using visualization tools (such as Matplotlib) to show the trend of quantity changes. After completing the quantity change calculation, spatially locate all fluorescent pollutants to obtain the spatial coordinates of each pollutant. Record the specific location of each pollutant in the image for spatial distribution analysis. Calculate the center coordinates of each pollutant using the previously marked bounding box information. Assuming the bounding box coordinates are (x1, y1, x2, y2), the formula for calculating the center coordinates is: The center coordinates (cx, cy) = ((x1 + x2) / 2, (y1 + y2) / 2); Iterate through the detection results of each frame and record the center coordinates of all contaminants. Assume the center coordinates of the contaminants detected in frame 1 are: Pollutant 1: (cx1, cy1) Pollutant 2: (cx2, cy2) The spatial coordinates of each fluorescent pollutant were successfully acquired, forming a coordinate dataset that provides a foundation for subsequent distribution change analysis. Based on the pollutant quantity change curve, a time-series frame spatial distribution change analysis was performed on the spatial coordinates to construct a pollutant spatial distribution map. This visually displays the spatial changes of pollutants. The pollutant center coordinates of each frame were combined with the corresponding quantity change data to form a comprehensive dataset. Using visualization tools (such as Plotly or Matplotlib), the pollutant coordinates of each frame were plotted as a scatter plot, and different colors or sizes were used to indicate the quantity changes. If the quantity of pollutants in a frame is large, the corresponding coordinate points can be represented by larger dots. By plotting the spatial distribution of all frames, a dynamic spatial distribution change map is finally formed, showing the distribution trend of pollutants over time. The constructed pollutant spatial distribution map provides a scientific basis for subsequent environmental monitoring and remediation.

[0041] In this embodiment, the specific steps for performing the final water quality monitoring and assessment based on the spatial distribution map of pollutants and the water quality assessment report are as follows: Detect abrupt changes in water quality by analyzing the trend evolution map of water quality status and mark the characteristic points of abrupt changes in water quality. The degree of anomalous change is quantitatively assessed for the characteristic points of abnormal water quality changes in order to obtain the current degree of anomalous change. Spatiotemporal correlation mining is performed on the spatial distribution map of pollutants based on the current abnormal mutation degree value to obtain the correlation between pollutants and abnormal mutations; Based on the correlation between pollutants and abnormal mutations, the spatiotemporal inverse extrapolation of pollution sources is performed to generate the location and emission characteristics of potential pollution sources; Based on the location and emission characteristics of potential pollution sources, the duration of pollution sources and diffusion paths are predicted to obtain a pollution diffusion evolution prediction map; The final water quality monitoring and assessment will be conducted based on the pollution diffusion evolution prediction map and water quality assessment report.

[0042] In this embodiment, water quality anomaly detection is performed on the water quality status trend evolution chart to mark characteristic points of water quality anomalies. Abrupt changes in water quality parameters are identified to promptly detect potential water quality problems. A detection threshold of three times the standard deviation is set to identify significant anomalies. During anomaly detection, historical data of water quality parameters are collected, and their mean and standard deviation are calculated. Data analysis tools (such as Python's NumPy and Pandas) are used to iterate through the water quality parameter time series, checking whether each data point exceeds the set threshold. This can be achieved using a defined Z-score method; if the Z-score of a data point is greater than 3, it is marked as an anomaly. During the marking process, it is ensured that the timestamp and corresponding water quality parameter value of each anomaly point are accurately recorded. Visualization tools (such as Matplotlib) are used to mark the anomaly characteristic points on the trend chart, making the anomalies immediately apparent. The results of this process provide the necessary data foundation for subsequent quantitative assessment of the anomaly severity. After marking the characteristic points of water quality anomalies, a quantitative assessment of the anomaly severity is performed to obtain the current anomaly severity value. Quantitatively assess the severity of sudden water quality changes to enable appropriate countermeasures. A 10-minute time window is set for anomaly severity calculation to capture short-term variations. The process of quantitatively assessing anomaly severity includes calculating the deviation of each marked anomalous feature point from its normal value. Both absolute and relative deviation methods can be used for assessment, calculating the difference between the anomaly point and the mean, and its proportion of the mean, respectively. In practice, a Python script can be written to extract the water quality values ​​of anomaly points and compare them with historical mean values. The final anomaly severity value will be presented numerically, with higher values ​​indicating more severe changes. This process will provide a quantitative basis for subsequent correlation analysis.

[0043] After obtaining the current anomalous mutation level, spatiotemporal correlation mining is performed on the pollutant spatial distribution map based on this value to obtain the pollutant-anomalous mutation correlation. The potential link between water quality anomalies and pollutant distribution is explored. A 30-minute time window is set for the spatiotemporal analysis to ensure the relevance of the analysis. During spatiotemporal correlation mining, the timestamps of anomalous feature points are matched with the spatial distribution data of fluorescent pollutants. The anomalous data and pollutant distribution data are aligned using the time window to identify the spatial location and concentration of pollutants when the water quality anomaly occurs. Correlation analysis methods (such as Pearson correlation coefficient) can be used to assess the relationship between water quality anomalies and pollutant concentrations. By writing analysis scripts, correlation coefficients are calculated for each time period to identify significantly correlated pollutant-anomalous mutation pairs. A pollutant-anomalous mutation correlation map is generated to visually demonstrate the connection between the two. After identifying the pollutant-anomalous mutation correlation, spatiotemporal inverse extrapolation of pollution sources is performed to generate potential pollution source locations and emission characteristics. The source of pollutants is traced for effective treatment. The base time period for the extrapolation model is set to 1 hour for dynamic analysis. The reverse engineering process involves using fluid dynamics or diffusion models, combined with spatial distribution data of pollutants, to infer their possible sources. By setting known water flow direction and velocity, reverse diffusion algorithms (such as the Lagrangian method) are used to predict the initial location of the pollutants. During the simulation, the diffusion path of the pollutants is calculated using water flow model parameters (such as flow velocity and direction). The final generated potential pollution source locations and emission characteristics will be presented in the form of coordinates and emission intensities, providing a basis for subsequent remediation measures.

[0044] Based on the location and emission characteristics of potential pollution sources, the duration and diffusion paths of pollution sources are predicted to obtain a pollution diffusion evolution prediction map. This provides a scientific basis for pollution prevention and control. The diffusion prediction time span is set at 24 hours to facilitate long-term monitoring. When predicting diffusion paths, historical data and current monitoring results are combined, and numerical simulation methods (such as the finite element method or finite difference method) are used to simulate the diffusion process of pollutants in water bodies. By setting initial conditions (such as pollution source intensity and environmental factors), the concentration changes and diffusion range of pollutants in the next few hours are calculated. During the simulation process, the influence of various parameters (such as water flow velocity, temperature, wind speed, etc.) on the diffusion model is fully considered. The final pollution diffusion evolution prediction map will show the distribution of pollutants over a future period, providing decision support for managers and helping to formulate corresponding environmental protection measures. Through this series of analyses and predictions, the scientific nature and effectiveness of water quality monitoring and control can be effectively improved.

[0045] In this embodiment, the specific steps for performing the final water quality monitoring and assessment based on the pollution diffusion evolution prediction map and water quality assessment report are as follows: The intensity, rate of change, and spatial diffusion radius of pollution events are extracted based on the pollution diffusion evolution prediction map. The color, deformation, and animation frequency parameters are set according to the intensity, rate of change, and spatial diffusion radius. The semantic rendering kernel function is defined based on color, deformation, and animation frequency parameters to obtain a multi-dimensional semantic rendering kernel function; Dynamic pollution visualization rendering is performed on the pollution diffusion evolution prediction map based on a multi-dimensional semantic rendering kernel function, and a dynamic pollution visualization model is constructed. The final water quality monitoring and assessment will be carried out based on the dynamic pollution visualization model and the water quality ecological assessment report.

[0046] In this embodiment, the intensity, rate of change, and spatial diffusion radius of a pollution event are extracted based on a pollution diffusion evolution prediction map. The characteristics of the pollution event are quantified to provide foundational data for subsequent visualization and assessment. The time period is set to the past 24 hours to comprehensively analyze the dynamic changes of the pollution event. When extracting pollution event features, pollutant concentration data needs to be obtained from the prediction map. This can be achieved by using image processing techniques (such as thresholding or region growing algorithms) to identify the distribution area of ​​pollutants. Using the OpenCV library in Python, the distribution of pollution concentration is analyzed, and the highest concentration point of pollutants is identified as the intensity of the pollution event. The rate of change is calculated. By comparing the pollutant concentration at different time points, a simple difference formula (e.g., rate of change = (current concentration - previous concentration) / time interval) is used to evaluate the rate of change of pollutant concentration. The spatial diffusion radius is determined by calculating the distance at which the pollutant concentration decreases to a certain threshold (e.g., 50%). Throughout this process, it is ensured that the calculation results for each feature are accurately recorded to provide necessary parameter support for subsequent steps. After extracting the pollution event features, color, deformation, and animation frequency parameters are set according to the intensity, rate of change, and spatial diffusion radius. This provides an intuitive representation for rendering the pollution visualization model. The color range is set from green (low concentration) to red (high concentration) to represent different levels of pollution. When setting parameters, the intensity of the pollution event is mapped to the color range. Intensity values ​​can be converted to RGB color values ​​using linear interpolation to reflect the severity of pollution in the visualization. Similarly, the rate of change affects the deformation parameter; the faster the rate of change, the greater the shape distortion. The animation frequency parameter can be set according to the rate of change of the pollution event, typically updated 5 to 10 times per second to ensure smooth visualization. These parameters provide the foundational data for defining the semantic rendering kernel function. A multi-dimensional semantic rendering kernel function is defined based on the color, deformation, and animation frequency parameters. A mathematical model is established to map pollution features to the visualization effect. The kernel function is set to have three dimensions (color, deformation, and animation frequency). When defining the kernel function, a suitable rendering model needs to be selected, such as a Gaussian kernel or radial basis functions. Using these functions, the features of the pollution event can be mapped to the visualization space. In practical implementation, a Python function can be written, taking pollution characteristic parameters (such as intensity, rate of change, and diffusion radius) as input and outputting the corresponding rendering results. During the definition process, it is ensured that the impact of each parameter is fully considered to achieve more accurate and aesthetically pleasing visualizations. The generated multi-dimensional semantic rendering kernel function will provide the mathematical foundation for subsequent dynamic pollution visualization rendering.

[0047] Based on a multi-dimensional semantic rendering kernel function, dynamic pollution visualization rendering is performed on the pollution diffusion evolution prediction map to construct a dynamic pollution visualization model. This generates a visualization model that can reflect pollution changes in real time. The rendering update frequency is set to 10 frames per second to ensure smooth dynamic effects. During dynamic rendering, the pollution diffusion prediction map is loaded, and the previously defined semantic rendering kernel function is applied. By iteratively processing each frame of data and combining color, deformation, and animation frequency parameters, the pollution visualization effect is dynamically generated. Visualization tools (such as Matplotlib or Plotly) can be used to display the dynamic effects. During rendering, it is ensured that changes in each parameter are reflected in the visualization graphics in real time to observe the dynamic evolution of pollution. The final generated dynamic pollution visualization model will be able to intuitively display changes in pollution events, providing decision support for managers. A final water quality monitoring assessment is performed based on the dynamic pollution visualization model and the water quality ecological assessment report. A complete water quality assessment report is generated by comprehensively analyzing the water quality monitoring data and visualization results. The assessment report is updated weekly to reflect the latest water quality status. During the final assessment, the pollution information displayed by the dynamic pollution visualization model is integrated with the data in the water quality ecological assessment report. By comparing visualized results with historical monitoring data, the impact of pollution on water quality is analyzed. Statistical analysis methods (such as regression analysis or analysis of variance) are used to assess the impact of pollution on water quality indicators, while visualization effects are combined to generate intuitive report charts. The final water quality monitoring and assessment report will include a detailed analysis of the pollution event, impact assessment, and follow-up recommendations, providing decision-makers with comprehensive reference information.

[0048] In this embodiment, a multimodal water quality monitoring data intelligent processing system is provided. The system includes four analyzers, four quality control instruments, six electrodes, one general-purpose input / output control board, a UPS, an air conditioner, and an access control system. The six electrodes, the general-purpose input / output control board, the UPS, the air conditioner, and the access control system are directly connected to the central control system via an RS485 bus. The four analyzers and four quality control instruments are connected to the central control system via eight RS232 physical interfaces. The electrodes, the general-purpose control board, the UPS, the air conditioner, and the access control system use RS232 physical interfaces. A camera is connected to a hard disk recorder via a network cable for storing video recordings. The hard disk recorder is connected to the central control system via a network cable. A remote platform can obtain the video stream via remote commands, thereby viewing historical recordings and real-time video on a remote end. The system is used to execute the multimodal water quality monitoring data intelligent processing method as described in claim 1, including: The multimodal processing module is used to acquire water quality monitoring parameter streams and water quality monitoring images, and perform adaptive processing to obtain filtered and optimized parameter streams and optimized monitoring images. The data processing module is used to acquire water quality monitoring parameter streams and water quality monitoring images, and perform adaptive processing to obtain filtered and optimized parameter streams and optimized monitoring images. The state evolution module is used to perceive the water quality state evolution based on the adaptive filtering optimization parameter flow and construct a water quality state evolution map. The response analysis module is used to perform biological indicator response analysis and water quality ecological impact assessment based on the water quality status trend evolution map to obtain a water quality assessment report. The monitoring and assessment module is used to analyze the spatial distribution changes of pollutants based on optimized monitoring images and to perform the final water quality monitoring and assessment according to the water quality assessment report.

[0049] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0050] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those 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 invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for intelligent processing of multimodal water quality monitoring data, characterized in that, Includes the following steps: Step S1: Acquire water quality monitoring parameter stream and water quality monitoring image, and perform adaptive processing to obtain filtered and optimized parameter stream and optimized monitoring image; Step S2: Based on the adaptive filtering optimization parameter flow, sense the evolution of water quality status and construct a water quality status evolution map; Step S3: Based on the water quality status trend evolution map, conduct biological indicator response analysis and water quality ecological impact assessment to obtain a water quality assessment report; Step S4: Analyze the spatial distribution changes of pollutants based on the optimized monitoring images, and perform the final water quality monitoring assessment according to the water quality assessment report.

2. The intelligent processing method for multimodal water quality monitoring data according to claim 1, characterized in that, The specific steps of step S1 are as follows: The water quality monitoring system performs real-time multimodal monitoring of the water quality pool, extracting multimodal water quality monitoring parameters and real-time water quality monitoring images; Global adaptive filtering and noise reduction are performed on multimodal water quality monitoring parameters to obtain the adaptive filtering optimized parameter stream; Differential local brightness enhancement processing is performed on real-time water quality monitoring images to construct locally enhanced and sharpened images; Super-resolution reconstruction is performed on the locally enhanced and sharpened image to construct a super-resolution optimized monitoring image.

3. The intelligent processing method for multimodal water quality monitoring data according to claim 2, characterized in that, The specific steps for performing global adaptive filtering and noise reduction on multimodal water quality monitoring parameters to obtain the adaptive filtering optimized parameter flow are as follows: Real-time data stream missing detection is performed on multimodal water quality monitoring parameter streams to identify abnormally missing parameters; Extract the location and timestamp of abnormally missing parameters; Based on the location and timestamp, mark the neighboring parameters of surrounding time points; The average value of the nearby parameters at the surrounding time points is calculated to obtain the average value of the missing parameter type; The average value is used to perform average interpolation filling to obtain the interpolation filling optimized monitoring parameter stream; Define a long time period and divide the interpolation filling optimization monitoring parameter stream into multiple time windows to obtain sub-parameter streams for multiple time windows; Perform a window-by-window Fourier transform on the sub-parameter stream to obtain the frequency domain diagram of the parameter stream for each window; High-frequency noise is identified based on the frequency domain diagram of the parameters, and noise points in each window are marked. Calculate the noise frequency and power of each noise point in the window; Global adaptive filtering and noise reduction are performed based on the noise frequency and power to obtain the adaptive filtering optimization parameter stream.

4. The intelligent processing method for multimodal water quality monitoring data according to claim 2, characterized in that, The specific steps for performing differential local brightness enhancement processing on real-time water quality monitoring images to construct locally enhanced and sharpened images are as follows: Real-time water quality monitoring images are divided into multiple sub-region images by image region thresholding; Calculate the dark channel value of the sub-region image; The brightness of each region is evaluated based on the dark channel value to obtain the brightness evaluation value of each sub-region. Based on the brightness evaluation value, the region is divided to obtain overly dark and overly bright regions; Perform detailed analysis on multiple sub-region images to identify image details in each sub-region; The detail complexity is quantified based on the image details of each sub-region to obtain the image detail complexity of each sub-region; Differential local brightness enhancement processing is performed on the overly dark and overly bright areas, and dynamic sharpening parameters are adjusted according to the complexity of image details to construct a locally enhanced and sharpened image.

5. The intelligent processing method for multimodal water quality monitoring data according to claim 1, characterized in that, The specific steps of step S2 are as follows: The temporal changes of water conductivity, pH, dissolved oxygen and temperature are calculated based on the adaptive filtering optimization parameter flow to obtain the parameter flow difference at different time points; Calculate the rate and magnitude of change of the parameter flow difference; The evolution of water quality fluctuation trends is carried out based on the rate and magnitude of change in order to obtain multimodal water quality fluctuation trend characteristics. Multi-time point trend prediction is performed on the multimodal water quality fluctuation trend characteristics to obtain multi-time point water quality trend prediction values; Multimodal water quality state evolution perception is performed on the predicted water quality trends at multiple time points, and a water quality state trend evolution map is constructed.

6. The intelligent processing method for multimodal water quality monitoring data according to claim 1, characterized in that, Step S3 is as follows: Biological indicator data streams are extracted based on adaptive filtering and optimized parameter streams, and time-series variation fitting is performed to construct time-series variation sequences of biological indicators. Multi-state characteristic evolution analysis was performed on the water quality state trend evolution map to extract the diffusion, migration, accumulation and transformation characteristics of water quality state; Spatiotemporal feature mapping is performed based on the diffusion, migration, accumulation and transformation characteristics of water quality state to construct the spatiotemporal evolution flow field of water quality state; Dynamic calculation of water quality state gradient is performed on the spatiotemporal evolution flow field of water quality state to generate dynamic gradient field of water quality state. Based on the time-series change sequence of biological indicators, the response of biological indicators to the dynamic gradient field of water quality is analyzed to generate the sensitivity law of biological indicator response under changes in water quality. A water quality ecological impact assessment is conducted based on the sensitivity patterns of the aforementioned biological indicators to obtain a water quality ecological assessment report.

7. The intelligent processing method for multimodal water quality monitoring data according to claim 1, characterized in that, The specific steps of step S4 are as follows: Perform deep target visual recognition on super-resolution optimized monitoring images to mark all fluorescent pollutants; The number of fluorescent pollutants is calculated frame by frame to obtain the number of pollutants in each frame; The time-series quantity change is calculated based on the quantity of pollutants to obtain the pollutant quantity change curve; All fluorescent pollutants were spatially located to obtain the spatial coordinates of each fluorescent pollutant. Based on the pollutant quantity change curve, a temporal frame spatial distribution change analysis is performed on the spatial positioning coordinates to construct a pollutant spatial distribution map. The final water quality monitoring and assessment will be conducted based on the spatial distribution map of pollutants and the water quality assessment report.

8. The intelligent processing method for multimodal water quality monitoring data according to claim 1, characterized in that, The specific steps for performing the final water quality monitoring and assessment based on the spatial distribution map of pollutants and the water quality assessment report are as follows: Detect abrupt changes in water quality by analyzing the trend evolution map of water quality status and mark the characteristic points of abrupt changes in water quality. The degree of anomalous change is quantitatively assessed for the characteristic points of abnormal water quality changes in order to obtain the current degree of anomalous change. Spatiotemporal correlation mining is performed on the spatial distribution map of pollutants based on the current abnormal mutation degree value to obtain the correlation between pollutants and abnormal mutations; Based on the correlation between pollutants and abnormal mutations, the spatiotemporal inverse extrapolation of pollution sources is performed to generate the location and emission characteristics of potential pollution sources; Based on the location and emission characteristics of potential pollution sources, the duration of pollution sources and diffusion paths are predicted to obtain a pollution diffusion evolution prediction map; The final water quality monitoring and assessment will be conducted based on the pollution diffusion evolution prediction map and water quality assessment report.

9. The intelligent processing method for multimodal water quality monitoring data according to claim 1, characterized in that, The specific steps for performing the final water quality monitoring and assessment based on the pollution diffusion evolution prediction map and water quality assessment report are as follows: The intensity, rate of change, and spatial diffusion radius of pollution events are extracted based on the pollution diffusion evolution prediction map. The color, deformation, and animation frequency parameters are set according to the intensity, rate of change, and spatial diffusion radius. The semantic rendering kernel function is defined based on color, deformation, and animation frequency parameters to obtain a multi-dimensional semantic rendering kernel function; Dynamic pollution visualization rendering is performed on the pollution diffusion evolution prediction map based on a multi-dimensional semantic rendering kernel function, and a dynamic pollution visualization model is constructed. The final water quality monitoring and assessment will be carried out based on the dynamic pollution visualization model and the water quality ecological assessment report.

10. A multimodal water quality monitoring data intelligent processing system, characterized in that, The multimodal water quality monitoring data intelligent processing system includes 4 analyzers, 4 quality control instruments, 6 electrodes, 1 general-purpose input / output control board, UPS, air conditioning, and access control. The 6 electrodes, 1 general-purpose input / output control board, UPS, air conditioning, and access control are directly connected to the central control system via an RS485 bus. The 4 analyzers and 4 quality control instruments are connected to the central control system via 8 RS232 physical interfaces. The electrodes, general-purpose control board, UPS, air conditioning, and access control use RS232 physical interfaces. A camera is connected to a hard disk recorder via a network cable for storing recordings. The hard disk recorder is connected to the central control system via a network cable. A remote platform can obtain the video stream via remote commands, allowing remote viewing of historical recordings and real-time video. The system is used to execute the multimodal water quality monitoring data intelligent processing method as described in claim 1, including: The data processing module is used to acquire water quality monitoring parameter streams and water quality monitoring images, and perform adaptive processing to obtain filtered and optimized parameter streams and optimized monitoring images. The state evolution module is used to perceive the water quality state evolution based on the adaptive filtering optimization parameter flow and construct a water quality state evolution map. The response analysis module is used to perform biological indicator response analysis and water quality ecological impact assessment based on the water quality status trend evolution map to obtain a water quality assessment report. The monitoring and assessment module is used to analyze the spatial distribution changes of pollutants based on optimized monitoring images and to perform the final water quality monitoring and assessment according to the water quality assessment report.

Images