Method and system for collaborative monitoring of aquatic ecological environment

Abstract

The application discloses a water ecological environment cooperative supervision method and system, and belongs to the technical field of ecological environment supervision. The method comprises the following steps: performing global monitoring on a target water area, obtaining a monitoring image sequence of the target water area, calculating a pollution confidence, identifying a pollution area and quantifying an area, generating a diffusion driving signal, and obtaining a category label of a target pollutant; obtaining a motion trajectory of the target pollutant, backstepping a local flow rate and flow direction of a water surface, and identifying an abnormal target in the target water area; combining the pollution area to generate an abnormal pollution correlation table; collecting local data of the water area in real time, generating a water area situation base map, constructing a water quality diffusion model, and generating a pollution diffusion simulation result in multiple time dimensions, performing plan deduction and risk assessment, upgrading supervision decision from passive disposal to active prevention and control, and fully meeting the decision demand of risk prediction and active prevention and control.

Description

TECHNICAL FIELD

[0001] The present application relates to a water ecological environment collaborative supervision method and system, and belongs to the technical field of ecological environment supervision. BACKGROUND

[0002] As an important carrier of the earth's ecological system, water areas bear key functions such as water quality guarantee, biological habitat, and water resource supply, and their ecological environment status is directly related to regional ecological security and people's well-being. With the advancement of industrialization and the intensification of human activities, the pollution risks and ecological imbalance problems faced by water areas are becoming increasingly complex. The traditional supervision mode relying on manual sampling monitoring and post-tracing investigation has shortcomings such as data collection lag, limited coverage, and low tracing efficiency, and is difficult to meet the current needs of fine and real-time supervision of water ecological environment. It is urgent to build an efficient supervision system with the help of intelligent technology.

[0003] The existing Chinese patent application with publication number CN120494735A discloses a water ecological environment supervision system based on artificial intelligence, which includes a pollution monitoring module for obtaining geographical information of the water area, establishing multiple water quality monitoring points, and obtaining pollution data; a backtracking module for obtaining hydrological models, meteorological data, and surrounding enterprise data of the water area to determine suspected pollution sources; a investigation and reporting module for investigating suspected pollution sources by sending underwater robots and drones to generate investigation reports and send them to the supervision department; an ecological collection module for calling underwater robots and drones to conduct regular patrols in the water area, building a biodiversity database, and building a hydrological and water quality database; and an ecological visualization module for building a water twin map and generating a visual ecological environment supervision map.

[0004] Although the existing technology has achieved the effect of improving the efficiency and real-time of water ecological environment supervision, it does not consider dynamic simulation and scenario deduction, resulting in weak decision support capability. In the water supervision scenario, pollution is not static, its diffusion path is dynamically affected by factors such as water flow and weather, and ecological risk also changes over time. For example, after a certain area detects water quality abnormalities, only suspected pollution sources are determined and present situation data is presented, which cannot predict which direction the pollution will spread, how fast it will spread, and cannot plan for potential crises such as pollution spreading to drinking water sources and ecological organisms being threatened. Therefore, the supervision decision can only stay at the passive disposal level after the problem is found, and it is difficult to meet the decision-making needs of risk prediction and active prevention in water ecological environment supervision. SUMMARY

[0005] In view of the deficiencies of the prior art, the purpose of the present application is to provide a water ecological environment collaborative management method and system, which obtains multi-source data through observation equipment and Internet of Things sensors, combines image analysis to locate pollution, back-propagates flow field, constructs situation base map, drives coupling model with base map data, adapts to pollution dynamic changes by optimizing parameters with Kalman filtering, deduces pollution trend and evaluates risk according to the model, matches and simulates the effectiveness of the plan, realizes active prediction and planning response from passive discovery, and solves the technical problem of weak decision support.

[0006] To achieve the above purpose, the present application provides the following technical solutions:

[0007] The water ecological environment collaborative management method comprises:

[0008] The target water area is monitored globally, the monitoring image sequence of the target water area is obtained, the pollution confidence is calculated, the pollution area is identified and quantified, the diffusion driving signal is generated, and the category label of the target pollutant is obtained;

[0009] The motion trajectory of the target pollutant is obtained, the local flow rate and flow direction of the water surface are back-propagated, and the abnormal target in the target water area is identified, combined with the pollution area, to generate an abnormal pollution association table;

[0010] Local water area data is collected in real time, a water area situation base map is generated, a water quality diffusion model is constructed, and a multi-time-dimension pollution diffusion simulation result is generated for plan deduction and risk assessment.

[0011] Specifically, the step of calculating the pollution confidence comprises:

[0012] The monitoring image frame is scanned pixel by pixel, the HSV value of each pixel point is obtained, the real-time similarity of each pixel point is calculated combined with the conventional pixel features ;

[0013] Set a candidate similarity threshold , judge whether the pixel point in the monitoring image frame is a normal water body, once , mark the pixel point as a candidate pixel;

[0014] A candidate pixel label map of the monitoring image frame is constructed, connected region analysis is performed, and a plurality of local candidate regions are generated;

[0015] The total number of pixels of each local candidate region is counted, and the actual area of the local candidate region is calculated combined with the pixel scale;

[0016] If the actual area is greater than the minimum size threshold, it is determined as an effective candidate region, and an effective candidate region information table containing the region ID is constructed; Otherwise, the local candidate region is removed.

[0017] Specifically, the step of calculating the pollution confidence further comprises:

[0018] statistically averaging the HSV values of all pixels in the effective candidate region, calling the normal water body average characteristics of the corresponding scene, and calculating the deviation value of a single characteristic;

[0019] performing scene adaptation adjustment to obtain a water body deviation value after scene adaptation;

[0020] obtaining the HSV feature extreme value of the historical pollution sample, calculating the water body difference range after scene adaptation, and thus obtaining the color difference coefficient;

[0021] calculating the real-time texture variance of the effective candidate region, combining the historical maximum texture variance, and calculating the texture difference coefficient;

[0022] extracting the scene adaptation target band ratio of the pixels in the effective candidate region, calling the corresponding band ratio of the normal water body, and calculating the spectral difference coefficient;

[0023] calculating the pollution confidence of each effective candidate region based on the feature importance weight .

[0024] Specifically, the step of generating the diffusion driving signal comprises:

[0025] setting a secondary confidence threshold 、 , and ;

[0026] if , determining that it is a pollution region, and retaining the initial pixel mask; if , determining that it is a non-pollution region, and deleting the initial pixel mask;

[0027] if , determining that it is a suspected pollution region, triggering secondary verification, if any one of the consecutive three frames is determined to be a non-pollution region, updating it to a non-pollution region, otherwise updating it to a pollution region;

[0028] constructing a pollution region pair, calculating a boundary distance, and marking the pollution region pair with a boundary distance less than an adjacent boundary threshold as adjacent pollution regions;

[0029] calculating the visual feature similarity of adjacent pollution regions, merging the pollution regions once the visual feature similarity is greater than a visual similarity threshold, reassigning the ID, and updating the region boundary coordinates and the actual area.

[0030] Specifically, the step of generating the diffusion driving signal further comprises:

[0031] counting the total number of pixels of each pollution region after optimization, combining the pixel scale, and calculating the actual merged area of the pollution region ;

[0032] The actual merging area of the call pollution area is calculated for three consecutive frames, and the area growth rate is calculated , the scene adaptation area threshold is set synchronously , the diffusion trend threshold ;

[0033] If And , a diffusion driving signal is generated, and the region ID and region geographic coordinates are associated; otherwise, continue to monitor the target water area globally;

[0034] A preliminary classification model is constructed, a pollutant label of the pollution area and corresponding bounding box coordinates are generated, a region ID is associated, and a target pollutant classification table is generated;

[0035] Based on the major labels in the target pollutant classification table, one by one processing is combined with industry specifications and sample statistical characteristics to extract fine-grained index judgment sub-classes, and the target pollutant classification table is updated.

[0036] Specifically, the step of identifying an abnormal target comprises:

[0037] Based on the bounding box coordinates of the target pollutant, the center point image coordinates are calculated, and the gray feature points and texture feature points are obtained;

[0038] For the target pollutant in two consecutive frames, feature points are extracted and a pixel matching window is set, and the pixel displacement vector of each feature point between frames is calculated;

[0039] The inter-frame average displacement vector is calculated, and the frame timestamp is combined to generate the motion trajectory of the target pollutant;

[0040] The pixel displacement size of the target pollutant is calculated, and the pixel scale is combined to obtain the actual displacement of the target pollutant, and the initial flow rate is calculated;

[0041] The flow direction angle of the target pollutant is calculated, and the horizontal turning angle calibration value is combined to obtain the geographic azimuth angle;

[0042] The flow rate confidence of the target pollutant is calculated, and once the flow rate confidence is less than the flow rate confidence threshold, it is labeled as low confidence and the current frame is removed, and a verification report is generated.

[0043] Specifically, the step of identifying an abnormal target further comprises:

[0044] An abnormal target sample library is obtained, an abnormality recognition model is constructed, and an initial abnormal target list is generated;

[0045] The abnormal target state label is determined to generate a global abnormal target list;

[0046] Based on the observation device calibration parameter, the geographical coordinates of the pollution area are obtained, and the direction coefficient is obtained in combination with the geographical azimuth angle;

[0047] The actual straight-line distance between the abnormal target and the pollution area is calculated, the diffusion coefficient is calculated in combination with the effective pollution diffusion distance, the label coefficient is obtained, and the pollution source confidence is calculated;

[0048] A secondary determination threshold is set, the associated label of the abnormal target is divided in combination with the pollution source confidence, and an abnormal pollution association table is constructed.

[0049] Specifically, the step of constructing a water quality diffusion model comprises:

[0050] The dedicated parameters of the target water area are configured, an initial parameter table is generated, and a coupled model is initialized;

[0051] The predicted pollution area boundary is obtained, the visual deviation is calculated, the predicted flow rate and flow direction are obtained synchronously, and the dynamic deviation is calculated;

[0052] Based on the visual deviation and the historical maximum visual deviation, an initial visual weight is set, the flow rate confidence is taken as an initial dynamic weight, and normalization processing is performed to obtain the visual weight and the dynamic weight;

[0053] The fusion deviation is calculated, the parameters in the initial parameter table are adjusted using the fusion deviation as a feedback signal, and a coupled model after parameter assimilation is generated;

[0054] Based on historical pollution events, historical pollution diffusion prediction results are obtained, spatial recurrence accuracy and concentration deviation rate are calculated;

[0055] If the spatial recurrence accuracy is greater than the accuracy threshold and the concentration deviation rate is less than the concentration threshold, it is determined that the coupled model verification is passed, and the water quality diffusion model is output; otherwise, return to the assimilation process of the fusion deviation.

[0056] Specifically, the step of pre-plan deduction comprises:

[0057] A deduction scenario parameter table is constructed, a multi-time dimension deduction is performed in combination with the water quality diffusion model, and a pollution diffusion simulation result is generated;

[0058] For the simulation result, a civil life risk level and an ecological risk level are divided to obtain an impact degree coefficient;

[0059] A time weight coefficient is calculated, the comprehensive risk degree is calculated in combination with the impact degree coefficient, and the risk level is divided through a secondary risk threshold;

[0060] For the simulation result, based on the pollutant sub-class label and the risk level, the corresponding pre-plan is matched respectively, and based on the matching degree, a candidate pre-plan set is generated;

[0061] For each set of candidate plans, calculate key performance indicators, thereby calculate comprehensive scores, and sort, to generate candidate plan performance evaluation table.

[0062] The collaborative monitoring system of the water ecological environment comprises a data acquisition module, an identification module, a prediction module and a deduction module.

[0063] The data acquisition module is used for monitoring a target water area globally and acquiring a monitoring image sequence, and collecting water area data in real time, and combining meteorological data to construct a water area situation base map.

[0064] The identification module is used for screening candidate pixels, performing connected region analysis, calculating pollution confidence, generating diffusion driving signals, backstepping flow velocity and flow direction, identifying abnormal targets and judging states, and generating an abnormal pollution association table.

[0065] The prediction module is used for configuring exclusive parameters to generate an initialized coupling model, calculating fusion deviation, dynamically optimizing model parameters, combining historical pollution event data to verify the model, and generating a water quality diffusion model.

[0066] The deduction module is used for multi-time dimension deduction, combining a sensitive target list to judge pollution influence degree, adopting a risk matrix method to divide risk grades and generate an early warning report, matching plans according to pollution types and risk grades, simulating plan performance, calculating comprehensive scores and recommending optimal plans.

[0067] The beneficial effects of the present application are as follows:

[0068] Through pixel-by-pixel scanning and connected region analysis of the observation equipment, combined with HSV space similarity calculation, local pollution is accurately located, avoiding that global water bodies cover pollution characteristics, and water area situation base maps are constructed by synchronously fusing Internet of Things sensor data and meteorological data, to provide accurate basic data for dynamic simulation, solve the problem of simulation distortion caused by one-sided traditional monitoring data, drive the coupling model by the situation base map data, dynamically optimize the model parameters by the set Kalman filtering algorithm combined with the deviation feedback of the abnormal pollution association table, so that the model can adapt to the dynamic changes of pollution along with water flow and weather in real time, break through the limitation that the static model cannot track pollution diffusion, carry out multi-time dimension deduction based on the optimized model, predict pollution cluster concentration distribution, sensitive target arrival time and peak concentration, combine the risk matrix method to divide grades, realize the change from passive pollution discovery to active risk prediction, match plans intelligently according to pollution types and risk grades, simulate plan performance by the model and recommend optimal schemes, solve the problem that measures cannot be planned in advance, upgrade the supervision decision from passive disposal to active prevention and control, and fully meet the decision-making needs of risk prediction and active prevention and control. BRIEF DESCRIPTION OF DRAWINGS

[0069] Fig. 1A schematic diagram of a water ecological environment collaborative monitoring method is shown in the figure.

[0070] Fig. 2 A flowchart for generating a diffusion driving signal in the present application is shown in the figure.

[0071] Fig. 3 A flowchart for identifying an abnormal target in the present application is shown in the figure.

[0072] Fig. 4 A flowchart for constructing a water quality diffusion model in the present application is shown in the figure. DETAILED DESCRIPTION

[0073] The technical solutions of the present application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solutions of the present application, and are not limitations of the technical solutions of the present application. In the case of no conflict, the technical features in the embodiments and the embodiments can be combined with each other.

[0074] Embodiment 1

[0075] Reference Figs. 1 to 4 The present embodiment introduces a water ecological environment collaborative monitoring method, which includes the following steps:

[0076] Step S1: The target water area is monitored by the observation equipment deployed at the water monitoring point, and the monitoring image sequence of the target water area is obtained in real time, including continuous frame video and timing snapshot image. The image quality is optimized by image enhancement algorithm to solve the image blur problem in rainy, foggy and night light environment. At the same time, based on the time stamp of image frame, combined with linear difference method, the frame synchronization alignment is carried out on the monitoring image sequence to ensure the uniformity of time interval of continuous frame. Combined with image segmentation technology, the water boundary extraction is carried out on the monitoring image sequence, the pollution confidence is calculated, and the pollution area is identified and quantified to obtain the pixel level mask and coordinate range of the pollution area, so as to generate diffusion driving signal; wherein the water monitoring point is deployed on the high tower, bridge or building on the shore, and the observation equipment includes but is not limited to high-definition camera and pan-tilt camera.

[0077] Step S2: Identify the target pollutant to generate the category label of the target pollutant, analyze the interframe pixel motion vector of the monitoring image sequence, track the motion trajectory of the target pollutant, and inversely deduce the local flow rate and flow direction of the water surface. The target pollutant and the pollution area are combined to identify the abnormal target in the target water area, so as to screen out the abnormal target causing pollution or aggravating pollution. The coordinates of the identified abnormal target are superimposed with the coordinates of the pollution area, and it is judged whether the abnormal target is located upstream or in the pollution area, so as to generate an abnormal pollution association table.

[0078] Step S3: Local monitoring of the target water area is performed by Internet of Things sensors deployed at water collection points, real-time collection of water area data, real-time acquisition of meteorological data through the meteorological department API, combination of the electronic map of the target water area, combination of the abnormal pollution correlation table, and construction of a water area situation base map; wherein the water collection points are deployed downstream of sewage outlets, at the entrances of drinking water sources, and at river sections, and the Internet of Things sensors include but are not limited to water quality buoys and underwater detectors, and the water area data includes but is not limited to water temperature, pH value, dissolved oxygen, chemical oxygen demand, ammonia nitrogen, total phosphorus, and water flow speed, direction, and level;

[0079] Step S4: The water area data and meteorological data in the water area situation base map are used as driving sources to import a water dynamics-water quality coupling model such as EFDC to calculate the water flow field distribution, and a water quality diffusion model is constructed in combination with the water quality data, the deviation is calculated with the abnormal pollution correlation table as the observation value and the output of the water quality diffusion model as the simulation value, the key parameters of the water quality diffusion model are automatically adjusted with the deviation value as the feedback signal by using the ensemble Kalman filter algorithm;

[0080] Step S5: Once the diffusion driving signal is monitored or manually triggered, the water quality diffusion model is used to generate pollution diffusion simulation results in multiple time dimensions to output the concentration distribution, diffusion boundary, estimated time of arrival at each sensitive target, and peak concentration of the pollution plume in different time periods, the plan is deduced, the prediction results in different scenarios are generated, the deduced results are superimposed with the sensitive target attributes to determine whether the pollution plume will affect the sensitive targets and the degree of influence, the risk level is divided by using the risk matrix method according to the toxicity of the pollutant and the deduced concentration, and a risk warning report is generated.

[0081] Specifically, the step of generating the diffusion driving signal includes:

[0082] Directly calculating the features of the global image can easily mask the local pollution features due to the high proportion of normal water body pixels, based on the latest monitoring image frame in the monitoring image sequence, each non-normal water body pixel is accurately located by pixel-by-pixel scanning, the HSV value of each pixel point is obtained, and the real-time similarity of each pixel point is calculated based on the HSV space Euclidean distance formula by calling the conventional pixel features of the corresponding scene in the normal water body visual feature library , avoiding local pollution missed judgment caused by global average effect; wherein the normal water body visual feature library is obtained by feature extraction of normal water bodies labeled by those skilled in the art;

[0083] A candidate similarity threshold is set to determine whether the pixel point in the monitoring image frame is a normal water body; if , it indicates that the pixel point is a non-normal water body pixel, and the pixel point is marked as a candidate pixel; if , indicating that the pixel point is a normal water body pixel at this time;

[0084] Based on the form of a binary image, 1 is a candidate pixel and 0 is a normal water body pixel. A candidate pixel label map of the monitoring image frame is constructed, a connected region analysis is performed, adjacent candidate pixels are clustered into a local candidate region, such as candidate pixels in up, down, left and right and diagonal lines, each region is assigned a unique ID, and the total number of pixels in each local candidate region is counted. Combined with the pixel scale, the actual area of the local candidate region is calculated.

[0085] A minimum size threshold of the pollution region is set, and combined with the actual area of each local candidate region, an effective candidate region is screened out. If the actual area is greater than the minimum size threshold, it is determined to be an effective candidate region, and the initial pixel mask, region boundary image coordinates and actual area of the effective candidate region are obtained to construct an effective candidate region information table containing the region ID to locate all potential local pollution regions in the global image. If the actual area is not greater than the minimum size threshold, it is determined to be image noise and is removed.

[0086] For each effective candidate region, the ID is processed one by one, the average value of all pixels in the effective candidate region is counted, including the average value of hue, saturation and brightness, the average features of the normal water body in the corresponding scene are called, the deviation value of a single feature is calculated, and scene adaptation adjustment is performed. The sum average operation is performed on each feature deviation value to obtain the water body deviation value after scene adaptation to reflect the color difference degree between the effective candidate region and the normal water body. For example, in the river scene, only hue and saturation are counted, and the water body deviation value is calculated by the hue deviation value and the saturation deviation value, because the brightness value of the river is greatly affected by the fluctuation of the silt, and the reference value is low.

[0087] The HSV feature extreme values of the historical pollution samples are obtained, such as the maximum and minimum values of hue, and scene adaptation is also considered. The difference between the extreme values of each feature is calculated, and the sum operation is performed to calculate the scene-adapted water body difference range. Based on the ratio of the water body deviation value to the water body difference range, the color difference coefficient is calculated.

[0088] The texture features of the pixels in the effective candidate region are extracted using local binary pattern, the real-time texture variance of all pixel LBP values in the region is calculated to reflect the unevenness of the texture in the region, the maximum texture variance of the historical pollution samples is called, and based on the ratio of the real-time texture variance to the maximum texture variance, the texture difference coefficient is calculated.

[0089] The scene-adapted target band ratio of the pixels in the effective candidate region is extracted, the corresponding band ratio of the normal water body is called, and the spectral difference coefficient is calculated by absolute difference operation.

[0090] Based on the feature importance weight, combined with color, texture, spectral difference coefficient, the pollution confidence of each effective candidate region is calculated And set the secondary confidence threshold, including the first confidence threshold , the second confidence threshold , To effectively distinguish between real pollution areas and suspected areas;

[0091] If , it is determined as a pollution area, the initial pixel mask is retained, and the initial list of pollution areas is recorded; if , it is determined as a non-pollution area, the initial pixel mask is deleted, and the non-pollution area exclusion list is recorded; if , it is determined as a suspected pollution area, the subsequent secondary verification of 3 frames of images is triggered immediately to avoid pollution omissions caused by global average effect, if any one of the continuous 3 frames is determined as a non-pollution area, it is updated as a non-pollution area to avoid false determination caused by transient interference, otherwise it is updated as a pollution area;

[0092] The initial pixel mask of each pollution area is subjected to morphological closing operation to fill the small holes in the mask whose pixel number is less than the minimum hole pixel threshold, and the edge of the mask is smoothed to avoid area calculation error caused by edge jaggedness;

[0093] In the pollution area identification process, affected by various factors such as image noise, pollution edge blur, algorithm segmentation precision limitation, the same physical pollutant is easily misidentified as multiple independent pollution areas, based on any two pollution areas, a plurality of pollution area pairs are constructed, the boundary distance of the pollution area pairs is counted, the pollution area pairs with boundary distance less than the adjacent boundary threshold are marked as adjacent pollution areas, and the visual feature similarity of the adjacent pollution areas is calculated, once the visual feature similarity is greater than the visual similarity threshold, it is determined as different parts of the same pollutant, and the initial pixel mask is merged into a complete pollution area, the ID is re-allocated, and the area boundary coordinates and actual area are updated;

[0094] The total number of pixels of each complete pollution area after optimization is counted, combined with the pixel scale, the actual merged area of the pollution area is calculated , the actual merged area of the pollution area in the continuous 3 frames is called 、 、 , the difference between 、 is the numerator, and the product of the time interval between and 、 is the denominator, the area growth rate is calculated , the scene adaptive area threshold , the diffusion trend threshold is set simultaneously ;

[0095] If and , a diffusion driving signal is generated, and the region ID is associated with the region geographic coordinates; otherwise, continue to monitor the target water area globally;

[0096] A historical pollution sample data set is obtained, including a pollution label image and a corresponding mask, and a YOLOv8 model is trained by combining transfer learning and scene fine-tuning to generate a preliminary classification model. The pollution categories include oil, biology, chemistry, and solid floating. At this time, the pollution categories are arranged in order of priority for disposal;

[0097] The pixel mask of the pollution area is input into the preliminary classification model to perform coarse classification on the pollutants in the mask, generate a pollution label and corresponding bounding box coordinates for the pollution area, associate the region ID, and generate a target pollution classification table;

[0098] Based on the major labels in the target pollution classification table, each is processed, combined with industry standards and sample statistical characteristics, and fine-grained indicators are extracted to determine the subcategory, and the target pollution classification table is updated. For oil pollutants, the gradient direction histogram is extracted. If the intensity gradient is unidirectionally distributed, it is marked as a light oil film, and if it is multidirectionally distributed, it is marked as a heavy oil film. For biological pollutants, the proportion of green pixels in the pollution area is counted to divide blue-green algae blooms and aquatic plant rot bodies. For chemical pollutants, based on a mapping table containing visual features and water quality indicators, such as dark brown patches corresponding to high COD and gray-white flocculent corresponding to high ammonia nitrogen, combined with sensor data or a pre-trained regression model to associate water quality indicators, for solid floating pollutants, the circularity is calculated based on the pollutant profile to divide it into plastic fragments and plastic bottles and cans.

[0099] Specifically, the step of identifying an abnormal target includes:

[0100] Since single feature point is easy to cause trajectory breakage due to local occlusion, such as shadow covering the center point of the bridge pier, and floating object occluding the edge point, by combining the basic anchor point with the local detail point, the trajectory continuity can be maintained by other two points when a certain feature point is occluded, the tracking stability and the calculation efficiency are balanced, based on the bounding box coordinates of the target pollutant, the image coordinates of the center point are calculated as the basic anchor point of the trajectory, and the reflection area of the oil pollutant and the highlight edge of the solid pollutant are often the maximum gray value points, all the pixels in the bounding box are traversed, the gray value of each pixel is calculated, the pixel point with the maximum gray value is determined as the gray feature point, so as to capture the local motion details, and the flocculent texture of the chemical pollutant and the irregular edge of the biological pollutant often correspond to high texture variance, the pixels in the bounding box are divided into sub-regions according to the preset pixel block, the LBP texture variance of each sub-region is calculated, the center point of the sub-region with the maximum texture variance is determined as the texture feature point, so as to supplement the edge motion deviation, and the feature points of the target pollutant are saved in the target pollutant classification table;

[0101] For the target pollutant in the two continuous frames, the feature point matching between the frames is carried out based on the Lucas-Kanade optical flow method, the feature points are extracted as seed points, a pixel matching window is set, the pixel displacement vectors of each feature point between the frames are calculated, including the geometric center point displacement, the gray feature point displacement and the texture feature point displacement, if a certain feature point is not extracted due to occlusion, the Kalman filter is started, the coordinates of the occluded feature point in the current frame are predicted based on the displacement trend of the feature points in the previous three frames, and the displacement vector is completed;

[0102] Different weights are configured for each feature point, the average displacement vector between the frames is calculated in combination with the pixel displacement vector, the average displacement coordinates of each frame are recorded in time sequence in combination with the frame time stamp, the time-stamped motion trajectory of the target pollutant is generated, and the trajectory is subjected to 5-frame sliding average filtering, so as to eliminate the single-frame displacement jump caused by sudden gust and image noise;

[0103] In the motion trajectory, the average displacement vectors of the continuous 10 frames are extracted, the pixel displacement size of the target pollutant is calculated based on the Euclidean distance formula, and the actual displacement of the target pollutant is converted in combination with the pixel scale, the initial flow rate of the target pollutant is calculated with the total time of the continuous 10 frames as the denominator;

[0104] According to the latest average displacement vector in the motion trajectory, the flow direction angle of the target pollutant is calculated by using the arctangent formula, and the geographic azimuth angle is obtained in combination with the horizontal rotation angle calibration value of the observation equipment;

[0105] The number of deviated feature points of each frame that fails to match is obtained, a deviation coefficient is calculated by multiplying a preset matching coefficient and the number of deviated feature points, a flow rate confidence of the target pollutant is calculated by multiplying a preset basic coefficient and the deviation coefficient, once the flow rate confidence is less than a flow rate confidence threshold, the current frame is marked as low confidence and removed, and a verification report is finally generated, including the region ID, the category of the target pollutant, the initial flow rate (including the confidence) and the geographic azimuth angle; wherein the basic coefficient and the matching coefficient are set by a person skilled in the art, in this embodiment, the basic coefficient is 1 and the matching coefficient is 0.2;

[0106] A scene-based abnormal target sample library is obtained, and sub-libraries are divided according to scene labels, such as a river scene sub-library containing illegal industrial sewage outlets and small cargo ships for illegal dumping, and a lake scene sub-library containing dam sewage outlets and sightseeing ships for illegal dumping. Each sub-library sample contains a labeled image, a bounding box, and a functional subtype label. A Faster R-CNN model is combined with a transfer learning and scene fine-tuning training model. For the river scene, the identification weights of industrial sewage outlets and small cargo ships are optimized. For the lake scene, the weights of dam sewage outlets and sightseeing ships are optimized to improve scene adaptability. An abnormality recognition model is finally generated, and a monitoring image sequence is input into the abnormality recognition model to generate an initial abnormal target list containing a target temporary ID, an abnormal target type, and bounding box coordinates.

[0107] An abnormal target state label determination is performed. For sewage outlet type targets, color features of the abnormal target in multiple consecutive images are extracted. If non-normal water bodies are recognized in multiple consecutive images, the state label is marked as continuous discharge. Otherwise, the state label is marked as intermittent discharge. For ship type targets, motion trajectories in multiple frames are tracked. If the displacement distance is less than a preset distance threshold, the state label is marked as stay overtime. Otherwise, the state label is marked as normal navigation. For the identified drainage outlets and ships, if the color features and motion states match the normal water body visual feature library, they are determined to be normal facilities and are removed from the list. A global abnormal target list is finally generated, containing a target temporary ID, a functional subtype, bounding box coordinates, and a state label. The state label includes continuous discharge, intermittent discharge, stay overtime, and normal navigation.

[0108] Based on the observation device calibration parameters, the central point image coordinates are converted into geographic coordinates using the perspective projection formula to ensure uniformity with the pollution area coordinate format. The geographic coordinates of the pollution area are used as the origin, and the inflow direction, i.e., the upstream direction, is determined based on the geographic azimuth angle. If the abnormal target is located on the inflow direction side, it is marked as upstream, otherwise it is marked as downstream. The Haversine formula is used to calculate the actual straight-line distance between the abnormal target and the pollution area to reflect the spatial correlation between the two.

[0109] An effective pollution diffusion distance is set based on the actual straight-line distance, whether the abnormal target is located in the upstream direction of the target pollutant, and the state label, a diffusion coefficient is calculated based on the base coefficient minus the ratio of the actual straight-line distance and the effective pollution diffusion distance, when the actual straight-line distance is less than the effective pollution diffusion distance, whether the abnormal target is located in the upstream direction of the target pollutant is quantified into specific data, the abnormal target is located in the upstream direction as 1, and the downstream direction as 0, to obtain a direction coefficient, and the state label is also quantified into specific data to obtain a label coefficient, and the pollution source confidence is calculated by weighted summation of the diffusion coefficient, the direction coefficient, and the label coefficient;

[0110] A two-level decision threshold is set, including a first decision threshold and a second decision threshold, wherein the first decision threshold is greater than the second decision threshold; if the pollution source confidence is greater than the first decision threshold, the associated label of the abnormal target is marked as a suspected pollution source; if the pollution source confidence is less than the second decision threshold, the abnormal target is marked as an irrelevant abnormality and is removed; if the pollution source confidence is within the second decision threshold and the first decision threshold, the abnormal target is marked as a pollution aggravating factor;

[0111] Finally, an abnormal pollution association table is constructed, including a target temporary ID, a pollution area ID, a functional subtype, geographic coordinates, a pollution source confidence, and an associated label.

[0112] Specifically, the steps of constructing the water quality diffusion model include:

[0113] A known hydrodynamic-water quality coupling model, such as EFDC, is called, based on the collected / calculated multi-dimensional data, the exclusive parameters of the target water pollution scene are configured, the general model is adapted to the specific regulatory requirements, for the hydrodynamic parameter configuration, the initial flow rate and the geographic azimuth angle are used as the initial flow rate field, the model calculation grid is generated based on the topographic data of the target water area, such as water depth and shoreline boundary, the upstream and downstream boundary conditions of the model are set based on the hydrological monitoring data, for the water quality parameter configuration, the pollutant subclass label is matched with the differentiated water quality parameters, such as the diffusion coefficient and the degradation rate, the associated label is used as the initial pollution concentration field, and the parameters are bound through the shared calculation grid, ensuring that the hydrodynamic parameters drive the pollutant diffusion calculation of the water quality module in real time, thereby generating an initial parameter table and initializing the coupling model;

[0114] The latest pollution area boundary geographic coordinates, initial flow rate, and geographic azimuth angle in the target water area are obtained, and based on the initialized coupling model, the predicted pollution area boundary is calculated, the visual deviation is calculated using the Euclidean distance formula, and the predicted flow rate and flow direction are obtained, the flow rate absolute deviation and the flow direction absolute deviation are calculated respectively, and the flow direction absolute deviation is configured with a weight coefficient to calculate the dynamic deviation;

[0115] The initial visual weight is set by the ratio of the visual deviation and the historical maximum visual deviation, and the initial dynamic weight is set by the flow confidence. The initial visual weight and the initial dynamic weight are normalized to obtain the visual weight and the dynamic weight, and the fusion deviation of the pollution area is calculated.

[0116] The set Kalman filtering algorithm is adopted, the fusion deviation is taken as the feedback signal, the parameters in the initial parameter table are adjusted according to the current real-time pollution scene, the iteration is performed until the deviation meets the threshold judgment rule, the optimized parameters are output, the real-time parameter calibration is performed, the optimized parameters are substituted into the initialized coupling model, the model prediction is more suitable for the actual situation, and thus the coupling model after parameter assimilation is generated.

[0117] The archive historical pollution event data with known diffusion results are obtained, input into the coupling model, and the historical pollution diffusion prediction results are obtained to verify and optimize the historical scene. The predicted pollution area and the actual pollution area in the historical event are extracted, the intersection area of the predicted pollution area and the actual pollution area is taken as the numerator, and the union area of the predicted pollution area and the actual pollution area is taken as the denominator to calculate the spatial reproduction accuracy. The predicted pollutant concentration and the archived measured concentration are extracted synchronously, the absolute difference between the predicted pollutant concentration and the archived measured concentration is taken as the numerator, and the archived measured concentration is taken as the denominator to calculate the concentration deviation rate.

[0118] If the spatial reproduction accuracy is greater than the accuracy threshold and the concentration deviation rate is less than the concentration threshold, it is determined that the coupling model verification is passed, and the water quality diffusion model is output. Otherwise, the assimilation process of the fusion deviation is traced back, the observation weight is adjusted, such as increasing the weight of high-confidence visual data in the historical event, and the assimilation process is re-run until the verification is passed.

[0119] Specifically, the triggering of the pre-plan deduction includes automatic triggering and manual triggering. The automatic triggering is detecting the diffusion driving signal, and the manual triggering is manually setting the deduction scene based on the pollutant sub-class label, such as leaking benzene wastewater from an upstream chemical plant. The historical optimal parameters corresponding to the benzene wastewater, such as the diffusion coefficient and the degradation rate, are automatically matched, and the parameters do not need to be manually input, thereby improving the efficiency.

[0120] The steps of the pre-plan deduction include:

[0121] The scenario parameter table is constructed to input the water quality diffusion model for multi-time dimension deduction. For each time node, the concentration distribution of the pollution group, the diffusion boundary, and the sensitive target arrival time and peak concentration are output. The pollution group diffusion boundary is extracted by the contour method, that is, the contour line formed by selecting the exceeding concentration threshold of the pollutant as the boundary is converted into geographic coordinates. The sensitive target arrival time is calculated by calculating the straight line distance between the pollution group boundary and the sensitive target geographic coordinates at different time nodes. When the distance is zero, the arrival time and the peak concentration at that time are recorded, so as to generate the pollution diffusion simulation result. Based on the electronic map of the target water area, a sensitive target list is constructed. The sensitive targets include the names of drinking water intake, ecological protection area red line, aquaculture area, and swimming area, and also include geographic coordinates, water quality threshold / standard;

[0122] The simulation results are sorted in layers. According to whether the pollution group will affect the sensitive target, the simulation results are divided into two groups. The prediction results with sensitive target arrival time less than the sensitive time threshold and peak concentration greater than the sensitive concentration threshold are selected and classified as the sensitive target group in the corresponding relationship of the target name, arrival time, and peak concentration. Other prediction results are directly classified as the non-sensitive target group.

[0123] By comparing the peak concentration of the pollutant at the sensitive target with the corresponding water quality threshold in the sensitive target list, the livelihood risk level is judged. If the peak concentration does not exceed the threshold, there is no livelihood risk. If the peak concentration exceeds the threshold but does not reach twice the threshold, it is a mild livelihood risk. If the peak concentration exceeds twice the threshold, it is a severe livelihood risk. The biodiversity database is called to compare the peak concentration with the pollutant tolerance threshold of aquatic organisms to judge the ecological risk level. If the peak concentration does not exceed half of the tolerance threshold, there is no ecological risk. If the peak concentration exceeds half of the tolerance threshold but does not reach the tolerance threshold, it is a mild ecological risk. If the peak concentration exceeds the tolerance threshold, it is a severe ecological risk.

[0124] Different numerical values are assigned to different livelihood risks and ecological risks. For each prediction result, the maximum value of the livelihood risk and the ecological risk is selected as the impact degree coefficient. Based on the ratio of the sensitive target arrival time to the preset sensitive time threshold, the time weight coefficient is calculated. Based on the weighted superposition, the comprehensive risk degree is calculated. Through the preset secondary risk threshold, the risk level of the prediction result is divided, so as to generate the risk warning report, including the impact target name, type, sensitive target arrival time, peak concentration, comprehensive risk degree, and risk level.

[0125] Intelligent pre-plan matching is performed, for each prediction result, based on the pollutant sub-class label, a corresponding pre-plan is matched, such as matching the oil containment boom and oil absorption felt related pre-plan for oil pollutant, matching the chemical wastewater related pre-plan for reagent addition and activated carbon adsorption, synchronously matching the corresponding pre-plan based on the risk level, matching the emergency disposal pre-plan for high risk to ensure the arrangement of pollution prevention facilities in a short time, matching the conventional disposal pre-plan for medium risk to arrange the disposal work according to the conventional process, after the matching is completed, a plurality of candidate pre-plans are generated, and the candidate pre-plans are sorted according to the matching degree to generate a candidate pre-plan set; wherein the matching degree is calculated by superimposing the pollutant type adaptation score and the risk level adaptation score according to a certain weight, wherein the risk level adaptation score is determined by the risk level, and different risk levels correspond to different adaptation scores, and the pollutant type adaptation score is called from a preset adaptation degree library, and the adaptation degree library saves the corresponding relationship between the pollutant type pre-plan and the adaptation score;

[0126] For each set of candidate pre-plans, the implementation parameters are input into the water quality diffusion model, and the corresponding parameters of the model are adjusted, such as adding the oil containment boom boundary in the model and setting the constraint condition that the pollutant cannot cross when the oil containment boom pre-plan is implemented, adjusting the degradation rate of the pollutant in the model according to the historical verification data when the reagent addition pre-plan is implemented, calculating the key performance indicators after the implementation of the pre-plan, including the pollution control rate, the ecological recovery period and the disposal cost; wherein the pollution control rate is the change ratio of the peak concentration at the sensitive target before and after the implementation of the pre-plan, the ecological recovery period is the time required for the pollutant concentration to decrease to a certain proportion of the water organism tolerance threshold, and the disposal cost is the total cost calculated according to the pre-plan material list and the market unit price;

[0127] The pollution control rate, the ecological recovery period and the disposal cost are comprehensively considered to calculate the comprehensive score of each candidate pre-plan, the candidate pre-plans are sorted according to the comprehensive score, the pre-plan with the highest score is recommended, and a candidate pre-plan performance evaluation table is generated, which includes the candidate pre-plan name, the implementation parameters, the performance indicators and the comprehensive score.

[0128] Embodiment 2

[0129] Another embodiment provided by the application: a water ecological environment collaborative management system, comprising: a data acquisition module, an identification module, a prediction module and a deduction module;

[0130] The data acquisition module is used for monitoring and acquiring a monitoring image sequence of a target water area through observation equipment deployed at a water monitoring point, optimizing image quality through image enhancement and frame synchronization processing, and unifying frame intervals, and simultaneously monitoring the target water area through Internet of Things sensors deployed at a water collection point, collecting water area data in real time, acquiring meteorological data through an API of a meteorological department, and constructing a water situation base map in combination with a target water area electronic map and an abnormal pollution association table;

[0131] The recognition module is used for pixel-by-pixel scanning of the monitoring image sequence, real-time similarity calculation in combination with the Euclidean distance of the HSV space, screening of candidate pixels, construction of a candidate pixel label map of the monitoring image frame, recognition of a pollution area based on connected region analysis and pollution confidence determination, generation of a diffusion driving signal, avoidance of global water body covering local pollution features, training of a YOLOv8 model for rough classification of pollutants in the pollution area and combination with fine-grained indexes for sub-classification, generation of a target pollutant classification table, tracking of a pollutant motion trajectory by combination of multiple feature points to back-propagate flow velocity and flow direction, identification of an abnormal target by a Faster R-CNN model and determination of a state, generation of an abnormal pollution association table by spatial superposition analysis, solution of the problems of easy missed judgment and broken trajectory in traditional recognition, and accurate positioning of a pollution source and a pollutant type;

[0132] The prediction module is used for calling a coupled model, configuring exclusive parameters to generate an initialized coupled model, taking the abnormal pollution association table as an observation value, calculating a fusion bias with a simulation value of the initialized coupled model, dynamically optimizing model parameters by a set Kalman filtering algorithm, verifying the model in combination with historical pollution event data, and outputting a water quality diffusion model, breaking through the limitation of a traditional static model that cannot adapt to dynamic changes of pollution with water flow and weather, and ensuring accurate pollution diffusion prediction;

[0133] The deduction module is used for driving or manually triggering based on the diffusion driving signal, constructing a deduction scenario parameter table and carrying out multi-time dimension deduction, outputting a pollution plume concentration distribution, a diffusion boundary, and arrival time and peak concentration of a sensitive target, judging pollution influence degree in combination with a sensitive target list, dividing risk levels by a risk matrix method to generate an early warning report, matching a plan according to a pollutant type and a risk level, simulating plan effectiveness and calculating a comprehensive score to recommend an optimal plan, realizing whole-process support from pollution discovery to risk prediction and plan programming, solving the problem of passive disposal in traditional supervision, and meeting the needs of risk prediction and active prevention and control of water ecological environment supervision.

[0134] Working principle and effect:

[0135] The target water area monitoring image sequence is obtained by an observation device deployed on the shore, and after image enhancement and frame synchronization processing, non-normal water body pixels are located by pixel-by-pixel scanning in combination with HSV space similarity calculation, effective pollution areas are screened by connected region analysis, target pollutant categories are identified, flow velocity and flow direction are back-propagated by tracking a motion trajectory, and an association table is generated by identifying abnormal targets in combination with pollution areas;

[0136] Synchronously relying on Internet of Things sensor to collect water area data, accessing meteorological data, integrating electronic map to build water area situation base map, solves the problem of traditional monitoring of easy to miss local pollution, data of partiality, provides comprehensive data support for subsequent analysis, and drives coupled model with situation base map data, calculates deviation of abnormal pollution correlation table as observation value and model simulation value, dynamically optimizes model parameters through ensemble Kalman filter, so that the model can adapt to the dynamic change of pollution under the influence of water flow and weather, and break through the limitation of traditional static model simulation distortion;

[0137] When the pollution signal is monitored or the deduction is manually triggered, multi-time dimension pollution diffusion results are generated, the influence degree is judged combined with the sensitive target attribute, the risk level is divided, the plan is matched according to the pollution type and the risk level, the optimal scheme is recommended by simulating the plan efficiency, and finally the whole process support from pollution discovery to risk prediction and plan planning is realized, so that the supervision decision is upgraded from passive disposal to active prevention and control, and the demand of water ecological environment supervision for risk prediction and active response is met.

[0138] The above is only the preferred embodiment of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments. Any technical solution falling within the scope of the present application should be considered as falling within the protection scope of the present application. It should be noted that, for ordinary skilled persons in the art, some improvements and refinements without departing from the principles of the present application should also be considered as falling within the protection scope of the present application.

Claims

1. A method for synergistic monitoring of the ecological environment of water areas, characterized in that, The method comprises the following steps: Conducting global monitoring on a target water area, obtaining a monitoring image sequence of the target water area, calculating a pollution confidence, identifying a pollution area and quantifying an area, generating a diffusion driving signal, and obtaining a category label of a target pollutant; Obtaining a motion trajectory of the target pollutant, backstepping a local flow rate and flow direction on the water surface, and identifying an abnormal target in the target water area, combining the pollution area, and generating an abnormal pollution correlation table; Real-time acquisition of local data of the water area, generation of a water area situation base map, construction of a water quality diffusion model, and generation of a multi-time-dimension pollution diffusion simulation result, plan deduction and risk assessment; The step of calculating the pollution confidence comprises the following steps: The monitoring image frame is scanned pixel by pixel, the HSV value of each pixel point is obtained, and the real-time similarity of each pixel point is calculated in combination with a conventional pixel feature ; Setting a candidate similarity threshold , judging whether the pixel point in the monitoring image frame is a normal water body, and once , marking the pixel point as a candidate pixel; Constructing a candidate pixel label map of the monitoring image frame, conducting connected region analysis, and generating a plurality of local candidate regions; Counting the total number of pixels in each local candidate region, combining a pixel scale, and calculating the actual area of the local candidate region; If the actual area is greater than a minimum size threshold, it is determined to be an effective candidate region, and an effective candidate region information table containing a region ID is constructed; otherwise, the local candidate region is removed; Counting the average HSV value of all pixels in the effective candidate region, calling the average features of the normal water body in the corresponding scene, and calculating the deviation value of a single feature; Scene adaptation adjustment is performed to obtain the water body deviation value after scene adaptation; Obtaining the HSV feature extreme value of the historical pollution sample, calculating the water body difference range after scene adaptation, and thus obtaining the color difference coefficient; Calculating the real-time texture variance of the effective candidate region, combining the historical maximum texture variance, and calculating the texture difference coefficient; Extracting the scene-adapted target band ratio of the pixels in the effective candidate region, calling the corresponding band ratio of the normal water body, and calculating the spectral difference coefficient; Based on the feature importance weight, calculate the pollution confidence of each valid candidate region .

2. The method for collaborative monitoring of aquatic ecological environment according to claim 1, characterized in that, The step of generating the diffusion driving signal comprises the following steps: Setting a secondary confidence threshold , , and ; If , determine as a contaminated region, retain the initial pixel mask; if , determine as a non-contaminated region, delete the initial pixel mask; If , determine as suspected pollution area, trigger secondary verification, if any one frame in the continuous 3 frames is determined as non-pollution area, update as non-pollution area, otherwise update as pollution area; Constructing a pollution region pair, calculating a boundary distance, and marking the pollution region pair with a boundary distance less than an adjacent boundary threshold as an adjacent pollution region; Calculating the visual feature similarity of adjacent pollution regions, and once the visual feature similarity is greater than a visual similarity threshold, merging the pollution regions, reassigning the ID, and updating the region boundary coordinates and the actual area.

3. The method for collaborative monitoring of aquatic ecological environment according to claim 2, characterized in that, The step of generating the diffusion driving signal further comprises the following steps: After statistics optimization, the total number of pixels of each pollution area is calculated, and the actual combined area of the pollution area is calculated in combination with the pixel scale ; The actual merging area of the call pollution region is calculated for three frames in succession, and the area growth rate is calculated , the scene adaptation area threshold is set synchronously , the diffusion trend threshold ; If and , generate a diffusion driving signal, associate the region ID with the region geographic coordinates; otherwise, continue to monitor the target water area globally; Constructing a preliminary classification model, generating a pollutant label of the pollution region and corresponding boundary box coordinates, associating a region ID, and generating a target pollutant classification table; Processing one by one based on the major labels in the target pollutant classification table, combining industry specifications and sample statistical features, extracting fine-grained index judgment sub-classes, and updating the target pollutant classification table.

4. The method for collaborative monitoring of aquatic ecological environment according to claim 3, characterized in that, The step of identifying an abnormal target comprises the following steps: Based on the boundary box coordinates of the target pollutant, calculating the center point image coordinates, and obtaining the gray feature points and texture feature points; For the target pollutant in two consecutive frames, extracting feature points and setting a pixel matching window, and calculating the pixel displacement vector of each feature point between frames; Calculating the average displacement vector between frames, combining the frame time stamp, and generating the motion trajectory of the target pollutant; Calculating the pixel displacement size of the target pollutant, combining the pixel scale, obtaining the actual displacement of the target pollutant, and calculating the initial flow rate; Calculate the flow direction angle of the target pollutant, combine the horizontal turning angle calibration value, and obtain the geographic azimuth angle; Calculate the flow velocity confidence of the target pollutant, and once the flow velocity confidence is less than the flow velocity confidence threshold, mark it as low confidence and remove the current frame, and generate a verification report.

5. The method for collaborative monitoring of aquatic ecological environment according to claim 4, characterized in that, The step of identifying abnormal targets further includes: Obtain an abnormal target sample library, build an abnormal identification model, and generate an initial abnormal target list; Determine the state label of the abnormal target to generate a global abnormal target list; Based on the observation device calibration parameters, obtain the geographic coordinates of the pollution area, and combine the geographic azimuth angle to obtain the direction coefficient; Calculate the actual straight-line distance between the abnormal target and the pollution area, combine the effective pollution diffusion distance, calculate the diffusion coefficient, obtain the label coefficient, and calculate the pollution source confidence; Set a secondary judgment threshold, combine the pollution source confidence, divide the correlation label of the abnormal target, and build an abnormal pollution correlation table.

6. The method for collaborative monitoring of aquatic ecological environment according to claim 5, characterized in that, The step of building a water quality diffusion model includes: Configure the exclusive parameters of the target water area, generate an initial parameter table, and initialize the coupled model; Obtain the predicted pollution area boundary, calculate the visual deviation, and simultaneously obtain the predicted flow velocity and flow direction to calculate the dynamic deviation; Based on the visual deviation and the historical maximum visual deviation, set the initial visual weight, take the flow velocity confidence as the initial dynamic weight, and perform normalization to obtain the visual weight and the dynamic weight; Calculate the fusion deviation, take the fusion deviation as the feedback signal, adjust the parameters in the initial parameter table, and generate the parameter assimilated coupled model; Based on historical pollution events, obtain historical pollution diffusion prediction results, calculate the spatial reproduction accuracy and concentration deviation rate; If the spatial reproduction accuracy is greater than the accuracy threshold and the concentration deviation rate is less than the concentration threshold, it is determined that the coupled model verification is passed, and the water quality diffusion model is output; otherwise, return to the assimilation process of the fusion deviation.

7. The method for synergic monitoring of aquatic ecological environment according to claim 6, characterized in that, The steps of scenario deduction include: Build a deduction scenario parameter table, combine the water quality diffusion model, and perform multi-time dimension deduction to generate pollution diffusion simulation results; For the simulation results, divide the civil life risk level and the ecological risk level to obtain the impact degree coefficient; Calculate the time weight coefficient, combine the impact degree coefficient, calculate the comprehensive risk degree, and divide the risk level through the secondary risk threshold; For the simulation results, match the corresponding preplan based on the pollutant sub-class label and the risk level, and generate a candidate preplan set based on the matching degree; For each candidate preplan, calculate the key performance indicators to calculate the comprehensive score, sort, and generate a candidate preplan performance evaluation table.

8. A system for the coordinated monitoring of the ecological environment of a water area for implementing the method for the coordinated monitoring of the ecological environment of a water area according to any one of claims 1 to 7, characterized in that, It includes: Data acquisition module, identification module, prediction module and deduction module; The data acquisition module is used for monitoring the target water area and obtaining monitoring image sequences, and real-time acquisition of water area data, combined with meteorological data to build a water area situation base map; The identification module is used for screening candidate pixels, performing connected region analysis, calculating pollution confidence, generating diffusion driving signals, backstepping flow velocity and flow direction, identifying abnormal targets and determining states, and generating an abnormal pollution correlation table; The prediction module is configured to generate an initialized coupling model by using the exclusive parameters, calculate a fusion deviation, dynamically optimize model parameters, verify the model by using historical pollution event data, and generate a water quality diffusion model. The deduction module is configured to perform multi-time dimension deduction, judge a pollution influence degree by using a sensitive target list, divide risk levels by using a risk matrix method to generate an early warning report, match a plan according to a pollution type and a risk level, simulate a plan efficiency, calculate a comprehensive score, and recommend an optimal plan.

Images