Island ecological monitoring and early warning method, system and equipment and storage medium

Abstract

The invention provides an island ecological monitoring and early warning method, system and device and a storage medium, and the method comprises the steps: collecting geological data, water quality data and meteorological data of a target region, and combining the geological data, the water quality data and the meteorological data according to a preset sequence to obtain an observation vector; performing data assimilation on the observation vector and the priori state predicted value through a Kalman filtering algorithm to obtain an optimal posteriori state estimated value at the current moment; wherein the priori state prediction value is obtained by calculation based on a posteriori state estimation value at a previous moment and a pre-constructed state transition matrix; and performing anomaly detection on the target area based on the optimal posterior state estimation value, when a detection result is abnormal, determining a risk level based on a risk assessment model, and generating a corresponding execution strategy according to the risk level, thereby improving the accuracy of island ecological anomaly detection and intelligent early warning.

Description

Technical Field

[0001] This invention relates to the field of marine ecological monitoring technology, and in particular to a method, system, equipment and storage medium for island and reef ecological monitoring and early warning. Background Technology

[0002] In the field of island ecological monitoring, real-time and accurate monitoring of the dynamic changes in multi-dimensional environmental factors such as geology, water quality, and meteorology is a prerequisite for effective ecological assessment and risk early warning. However, existing monitoring systems face three major technological bottlenecks: First, insufficient data reliability, as data from single sensors is easily affected by noise interference and instantaneous errors, leading to distorted state perception; second, the existence of "information silos," where multi-source heterogeneous data from different sources and with varying precision differ in spatiotemporal benchmarks and accuracy, making it difficult to unify, integrate, and analyze, and thus unable to provide a holistic and quantitative description and tracking of the health status of marine ecosystems. These bottlenecks collectively result in the current marine ecological monitoring system's high uncertainty in state estimation and persistently high false alarm and missed alarm rates. Summary of the Invention

[0003] This invention provides a method, system, device, and storage medium for island and reef ecological monitoring and early warning, to solve the problems existing in related technologies. The technical solution is as follows: In a first aspect, embodiments of the present invention provide a method for monitoring and early warning of island and reef ecosystems, including: Collect geological data, water quality data, and meteorological data of the target area, and combine the geological data, water quality data, and meteorological data in a preset order to obtain the observation vector; The observation vector and the prior state prediction are assimilated by the Kalman filter algorithm to obtain the optimal posterior state estimate at the current time. The prior state prediction is calculated based on the posterior state estimate at the previous time and the pre-constructed state transition matrix. Anomaly detection is performed on the target area based on the optimal posterior state estimate. When anomalies are detected, the risk level is determined based on the risk assessment model, and a corresponding execution strategy is generated according to the risk level.

[0004] In one implementation, collecting geological data, water quality data, and meteorological data of the target area includes: The deployment scheme of the scanning equipment is determined based on a pre-built multi-objective optimization model. The scanning equipment is then used to scan the island and reef area to obtain geological data. Based on the spatial constraint model, the deployable area of ​​the target sea area is determined, and water quality data of the target water area is collected by underwater sensors deployed in the deployable area. Meteorological data is obtained through weather stations or drone cameras; the drone camera takes pictures of the target area at a specified angle.

[0005] In one implementation, the method for constructing a multi-objective optimization model includes: Construct a first objective function that aims to minimize the total cost of equipment deployment, and construct a second objective function that aims to maximize the overall terrain coverage of the target area; Define the set of devices to be deployed as decision variables, and impose spacing constraints on the decision variables. The spacing constraints require that the spatial distance between any two different devices in the set of devices is not less than the preset minimum spacing. The first objective function, the second objective function, and the spacing constraint are integrated into a multi-objective optimization model.

[0006] In one implementation, data assimilation of the observation vector and the prior state prediction value using the Kalman filter algorithm includes: The predicted state is calculated based on the prior state prediction value and the pre-constructed observation matrix; Calculate the difference between the observed vector and the predicted state to obtain the observation residual; The Kalman gain is calculated based on the prior error covariance matrix and the predefined observation noise covariance matrix; wherein, the prior error covariance matrix is ​​calculated based on the error covariance matrix of the previous time step through the covariance prediction equation; The prior state prediction is corrected by using Kalman gain and observation residuals to obtain the optimal posterior state estimate at the current time.

[0007] In one implementation, the method for constructing the state transition matrix includes: The system state vector is defined based on the environmental elements of the target area, including geological spatial coordinates, water quality parameters, and meteorological elements. Based on the temporal evolution of the system state vector from the previous moment to the current moment, a block-diagonal state transition matrix is ​​constructed to quantify the dynamic changes of geological data, water quality data, and meteorological data.

[0008] In one implementation, determining the risk level based on a risk assessment model includes: A risk assessment indicator system was constructed based on the analytic hierarchy process, and the indicator weights were calculated. The fuzzy comprehensive evaluation method was used to calculate the membership degree of each indicator to different risk states, and the fuzzy evaluation results were synthesized. A Bayesian network model is used to calculate the posterior probability of the fuzzy evaluation results in order to correct the early warning results and output the final risk level.

[0009] In one implementation, it further includes: Real-time acquisition of the status parameters of the execution device, including at least the device power consumption, device location, and images of the repair area; Based on the state parameters, calculate the key performance indicators for repair execution. The key performance indicators include at least the risk reduction rate and the execution cost. An optimization model is constructed based on the Q-learning algorithm. Key performance indicators are used to determine the reward function of the optimization model. The optimization model continuously optimizes and repairs decisions, and outputs the optimal execution strategy.

[0010] Secondly, embodiments of the present invention provide an island and reef ecological monitoring and early warning system, comprising: The data acquisition layer is used to collect geological data, water quality data, and meteorological data of the target area, and to combine the geological data, water quality data, and meteorological data in a preset order to obtain the observation vector. The data processing layer is used to assimilate the observation vector and the prior state prediction value using the Kalman filter algorithm to obtain the optimal posterior state estimate value at the current time. The prior state prediction value is calculated based on the posterior state estimate value at the previous time and the pre-constructed state transition matrix. The early warning decision layer is used to detect anomalies in the target area based on the optimal posterior state estimate. When anomalies are detected, the risk level is determined based on the risk assessment model. The control execution layer is used to generate corresponding execution strategies based on the risk level.

[0011] Thirdly, embodiments of the present invention provide an electronic device comprising a memory and a processor. The memory and the processor communicate with each other via an internal connection path. The memory stores instructions, and the processor executes the instructions stored in the memory. When the processor executes the instructions stored in the memory, it causes the processor to perform the method described in any of the above embodiments.

[0012] Fourthly, embodiments of the present invention provide a computer-readable storage medium that stores a computer program, wherein when the computer program is run on a computer, the methods in any of the embodiments described above are executed.

[0013] The advantages or beneficial effects of the above technical solutions include at least the following: This invention collects geological, water quality, and meteorological data from the target area, constructing observation vectors from heterogeneous data of different sources and with different physical meanings, such as geology, water quality, and meteorology. This achieves deep fusion and unified representation of multi-source heterogeneous data, solving the problem of information silos. This invention uses Kalman filtering to optimally fuse the observation vectors with prior state predictions, effectively overcoming the problems of high noise and poor reliability of single sensor data. This significantly reduces the uncertainty of the generated posterior state estimates, achieving a significant improvement in the accuracy and reliability of state estimation, thereby improving the accuracy of anomaly detection and intelligent early warning.

[0014] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0015] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in the invention and should not be construed as limiting the scope of the invention.

[0016] Figure 1 This is a flowchart illustrating the island and reef ecological monitoring and early warning method of the present invention; Figure 2 This is a schematic diagram of the entire process of the island and reef ecological monitoring and early warning system of the present invention; Figure 3 This is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0017] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0018] With the increasing global demand for marine ecological protection and sustainable development, the importance of environmental monitoring and early warning technologies for island and reef ecosystems, as key nodes in marine ecology, is becoming increasingly prominent. However, traditional island and reef ecological monitoring methods suffer from problems such as insufficient sensor coverage, low data acquisition efficiency, and poor accuracy in multi-source data fusion. This is particularly true in complex terrain and dynamic marine environments, making real-time and comprehensive monitoring of ecological elements difficult. For example, existing systems often use fixed-point sensors or manual inspections, which cannot adapt to changes in island and reef topography and ecology, leading to data gaps or delayed early warnings.

[0019] To address the aforementioned issues, this embodiment provides a method for monitoring and early warning of island and reef ecosystems. This method can effectively solve technical defects such as low sensor deployment coverage and insufficient accuracy of multi-source data fusion, and realize full-process management of island and reef ecological data from "collection-processing-decision-execution".

[0020] Combination Figure 1 As shown, the island and reef ecological monitoring and early warning method in this embodiment specifically includes: Step S1: Collect geological data, water quality data, and meteorological data of the target area, and combine the geological data, water quality data, and meteorological data in a preset order to obtain the observation vector.

[0021] To address the issue of low sensor coverage, this embodiment utilizes multiple types of sensors and intelligent deployment algorithms to cover key indicators of island and reef ecology, including topography, water quality, and meteorology, when deploying sensors in the target area (including the ocean, land, and sky where the islands and reefs are located). This constructs a three-dimensional integrated "air-ground-water" sensing network, enabling full-area, high-precision, and synchronized data collection.

[0022] In terms of terrain mapping, the core components for terrain mapping of the target area consist of airborne LiDAR scanning equipment, a GPS / IMU integrated positioning system, and an unmanned aerial vehicle (UAV) platform. The LiDAR system and GPS / IMU module are integrated and installed inside the UAV cabin, with the laser emitter and scanning mirror at a 45° angle to ensure the laser beam covers a 120° range on both sides of the UAV's flight path. The GPS antenna and IMU sensor error is controlled within ±0.5mm to ensure the spatiotemporal consistency of position and attitude data.

[0023] The drone cruises along a preset route. The laser emitter of the airborne LiDAR emits laser pulses, which are reflected by a scanning mirror and cover the surface of the islands and reefs. The reflected light signal is captured by the receiver, converted into an electrical signal, and the flight time t of the laser is calculated. Combined with the speed of light c, the distance value d = c * (t / 2) is obtained.

[0024] The GPS / IMU integrated positioning system outputs the UAV's position (latitude, longitude, elevation, etc.) and attitude (azimuth, heading, etc.) in real time, and converts the distance values ​​into three-dimensional spatial coordinates in the geodetic coordinate system using coordinate transformation formulas. ; in, Let R be the position of the UAV, and R be the attitude transformation matrix. The elevation angle of the laser beam. It is the azimuth angle.

[0025] It should be noted that the attitude transformation matrix R, also known as the rotation matrix, can transform a vector in one coordinate system to another, and is commonly represented by the basic rotation matrix. Matrices for rotation about a single coordinate axis (right-handed system) include: Rotation around the X-axis (roll angle) ):

[0026] Rotation around the Y-axis (pitch angle) ):

[0027] Rotation around the Z-axis (yaw angle) ): .

[0028] To ensure full coverage scanning of the entire island and reef, in this embodiment, a deployment scheme for the scanning equipment (i.e., airborne lidar) set is determined based on a pre-built multi-objective optimization model. The scanning equipment is then controlled to perform terrain scanning of the island and reef area according to the deployment scheme, thereby obtaining geological data of the entire island and reef area, including island and reef terrain point cloud data, etc.

[0029] The methods for constructing multi-objective optimization models include: Step S11: Define the decision variable, that is, define the set of scanning devices to be deployed as the decision variable S, where S = {s1, s2, ..., s} n}, s i Let represent the spatial coordinates of the i-th scanning device, and n be the total number of scanning devices.

[0030] Step S12: Construct the first objective function to minimize costs, that is, construct the first objective function f1(S) with the goal of minimizing the total cost of equipment deployment, and its expression is: f1(S) = ; in, The cost of deploying the i-th scanning device; when all scanning devices have the same cost, the first objective function f1(S) degenerates into minimizing the total number of devices n.

[0031] Step S13: Construct the second objective function to maximize the terrain coverage, that is, construct the second objective function f2(S) with the objective of maximizing the overall terrain coverage of the target area, and its expression is: f2(S) = Coverage(S); Where Coverage(S) is a quantitative calculation function for the topographic coverage of the island and reef area by the set of scanning devices S.

[0032] Step S14: Define device spacing constraints. Apply spacing constraints to the decision variables, requiring that for any two different devices s within the scanned device set... i With s j The spatial distance between them is not less than the preset minimum spacing d min Its constraints are expressed as follows: Distance(si, sj) ≥ d min , i ≠ j Step S15: Integrate and solve the multi-objective optimization model, that is, integrate the first objective function, the second objective function, and the spacing constraint into a multi-objective optimization model, the expression of which is as follows: .

[0033] Step S16: Solve the multi-objective optimization model using the multi-objective genetic algorithm (NSGA-II) and finally output the Pareto optimal solution set that achieves the best balance between cost and coverage. The Pareto optimal solution set corresponds to a series of optimal equipment deployment coordinate schemes.

[0034] After the airborne lidar acquires geological data, i.e., the raw terrain point cloud data, each terrain point cloud data point p generated by the airborne lidar is processed. i Calculate the mean elevation μ and standard deviation σ of points within its local neighborhood. Outlier detection using the 3σ criterion: when |p i When μ∣>3σ, the point is marked as an anomaly and removed, effectively eliminating outlier noise points in the terrain point cloud, such as abnormal elevation points caused by birds, vegetation, transient reflections, etc.

[0035] In terms of water quality monitoring, water quality data is detected through underwater water quality sensors (including pH sensors, dissolved oxygen sensors, and turbidity sensors), a waterproof data acquisition terminal, and an anchored buoy platform. The underwater water quality sensor array is encapsulated in a waterproof chamber, which is slightly larger than the sensors. The probes of the pH and dissolved oxygen sensors extend outside the waterproof chamber, while the turbidity sensor is integrated into the surface of the chamber. The waterproof data acquisition terminal is connected to the sensors inside the waterproof chamber via waterproof cables and is fixed below the buoy.

[0036] This dimension requires the deployment of various underwater water quality sensors in the waters surrounding islands and reefs. However, the underwater topography is complex and varied. When encountering areas where installation is not possible, it is necessary to use the spatial constraint + grid method. That is, the deployable areas in the target area are determined by the spatial constraint model, and underwater sensors deployed in the deployable areas collect water quality data of the target waters, including dissolved oxygen (DO), pH value, turbidity, etc.

[0037] In other words, the waters around the islands and reefs are designated as deployable areas. Non-deployable areas Using a 0-1 matrix This indicates the deployable area. The deployment coordinates of the underwater water quality sensor are located within the deployable area, and its expression is: ; A deploy = {(i,j) | M i,j = 1, i∈[1, N x ], j∈[1, N y ]}; Among them, deployable areas N is a set consisting of all grid coordinates (i, j) that satisfy a specific condition. x N y This represents the number of rows and columns after the sea area is rasterized.

[0038] This embodiment also employs a 5-minute moving average filter for water quality data such as pH, that is, averaging five data points based on the current time and the previous four time points to eliminate high-frequency fluctuations caused by natural factors such as tides and waves, while preserving the long-term trend of water quality parameters. The calculation formula is as follows: .

[0039] In terms of meteorology, the meteorological monitoring equipment consists of a high-definition camera, a weather station (capable of measuring wind speed, wind direction, temperature, and humidity), and a solar power system. The high-definition camera is mounted on the drone's gimbal and needs to be at a certain angle to the horizontal plane during shooting, ideally between 30° and 60°. If the angle is less than 30°, the camera lens is too flat, easily resulting in the sky occupying too much of the image, compressing the effective monitoring area of ​​the ground / sea surface (such as cloud formations and near-surface airflow disturbances), thus hindering the capture of crucial information such as coastal meteorological interactions. If the angle is greater than 60°, the excessive downward angle of the lens will reduce the field of view and is easily obstructed by the drone's body, creating shadows. It may also cause blurring and distortion of distant meteorological elements due to the excessive lens tilt.

[0040] The meteorological station can be fixed at a high point on the island or reef, and the probes of various sensors (such as wind speed, wind direction, temperature and humidity sensors) are higher than the ground. The data is transmitted to nearby computing nodes through waterproof cables.

[0041] When the drone is equipped with a high-definition camera for patrol, the camera captures one frame of image every 2 seconds according to preset parameters, and simultaneously records the GPS position and flight altitude. At the same time, wind speed, wind direction, and temperature and humidity sensors collect data every 10 seconds to correct for atmospheric scattering effects on the optical images.

[0042] This dimension utilizes optical imaging systems in conjunction with weather stations to obtain meteorological data, including wind speed, wind direction, temperature, and humidity. The equipment in this dimension needs to be triggered at the same time as the equipment in the topographic mapping dimension to ensure the time synchronization of multi-source data.

[0043] Subsequently, this embodiment utilizes queue management, multi-source fusion, and noise suppression techniques to transform the raw heterogeneous data (such as geological data, water quality data, and meteorological data) acquired by the data acquisition layer into a standardized, highly reliable ecological monitoring dataset. This provides accurate input for early warning decision-making and solves the problems of "disordered timing and insufficient accuracy" in traditional data processing. Specifically: First, the data is divided into two categories based on its urgency, and different priority coefficients are assigned to them, for example: Emergency data, including critical data such as sudden changes in water quality (pH < 6.5 or pH > 9.0) and extreme weather (wind speed ≥ 10), has the highest priority (P = 1.0) and requires immediate processing.

[0044] Regular data, including non-urgent data such as terrain changes and slow fluctuations in water quality, has a lower priority (P=0.5) and is allowed a certain processing delay.

[0045] The M / M / 1 model from queuing theory is used to perform differentiated scheduling processing for data with different priorities, ensuring system stability and rapid response for high-priority data. The core principle of dynamic optimization scheduling based on the M / M / 1 model lies in virtually allocating independent processing channels to different categories of data, and ensuring that the total system processing capacity is greater than the total data arrival rate (λ). g <μ g This prevents the queue from growing indefinitely. The priority mechanism, on the other hand, allocates a higher processing service rate (μ1) to urgent data to achieve priority scheduling.

[0046] For example, suppose the data arrival rate is... The processing speed is .

[0047] Input parameter: arrival rate of emergency category data Items / second, processing speed is Items / second; arrival rate of regular data Items / second, processing speed is Items per second.

[0048] For mixed data queues, the following conditions must be met: ; Priority scheduling mechanism: Differentiated processing is achieved by adjusting the service rate. The average waiting time for urgent data is: ; The average waiting time for regular data is: ; Calculations show that the average waiting time for emergency data (0.067 seconds) is significantly shorter than that for regular data (0.083 seconds), thus ensuring low-latency processing of critical data.

[0049] Meanwhile, to handle data influx and untimely processing, a FIFO (First-In, First-Out) cache queue is employed. Urgent data has a dedicated cache area with a capacity of 1000 records to prevent high-priority data from being lost due to queue fullness. If urgent data times out in the cache without being processed, the system automatically triggers an alarm to prompt manual intervention. All historical data is archived every 24 hours and transferred to cloud storage for long-term storage; only the most recent 7 days of data are retained locally to free up storage space and ensure system performance.

[0050] In this embodiment, after collecting geological data, water quality data, and meteorological data, the geological data, water quality data, and meteorological data are organized according to the order of state vectors to form the observation vector Z. k : Z k = ; in, , , The three-dimensional coordinates of island and reef terrain measured in real time by lidar; Water quality parameters collected in real time by underwater sensors; Wind speed is collected in real time by weather stations or drones.

[0051] Step S2: Calculate the predicted prior state value at the current time based on the posterior state estimate of the previous time step and the pre-constructed state transition matrix.

[0052] In this embodiment, the object of state prediction is the system state vector X, which integrates multi-dimensional information from geology, water quality, and meteorology. k The system state vector X k It is defined based on the environmental elements of the target area, which include geological spatial coordinates, water quality parameters, and meteorological elements, namely: ; in, The three-dimensional coordinates of the island and reef topography are derived from geological data. These are pH value and dissolved oxygen content, respectively, derived from water quality data; Wind speed, from meteorological data.

[0053] Construct a block diagonal state transition matrix Fk Used to model the system state vector X k The temporal variation patterns of each component are shown below, and their structure is as follows: ; Among them, F k1 It is a 3×3 identity matrix, representing that the spatial coordinates (X, Y, Z) remain unchanged under undisturbed external conditions. F k4 It is a 3×3 diagonal matrix, where the non-zero elements on the diagonal correspond to the variation coefficients of pH, DO, and wind, respectively. For example, 0.99 represents the pH decay coefficient over a short period of time, when the pH value changes slowly; 0.95 represents the daily variation coefficient of dissolved oxygen concentration; and 0.8 represents the short-term fluctuation coefficient of wind speed. k2 and F k3 Set it to zero.

[0054] Based on the posterior state estimate from the previous time step The prior state prediction value at the current moment is calculated using the state transition equation. The state transition equation is as follows: ; In the formula, For the control matrix, it should be set to 0 if there is no external control. To control the quantity.

[0055] Meanwhile, based on the posterior error covariance P of the previous time step k-1 The prior error covariance matrix P at the current time is calculated using the covariance prediction equation. k - The covariance prediction equation is as follows: ; In the formula, Q k The diagonal element of the process noise can be set according to the sensor accuracy, for example, it can be set to... ; It is the state transition matrix F k The transpose of .

[0056] It should be noted that the posterior state estimate and the posterior error covariance of the previous time step both come from the complete filter calculation output of the previous filter cycle in the iterative mechanism of the Kalman filter.

[0057] Step S3: Assimilate the observation vector and the prior state prediction value using the Kalman filter algorithm to obtain the optimal posterior state estimate at the current time.

[0058] To obtain the observation vector Z at the current moment k Compared with the prior state prediction value at the current moment Then, the optimal posterior state estimate for the current time is calculated through the following steps. : Step S31: Calculate the predicted prior state value at the current time. and the pre-constructed observation matrix H k The product of these two terms yields the predicted state H. k .

[0059] It should be noted that the observation matrix H k It is a matrix that maps the state space to the observation space, and it describes the system state vector X. k How do the components affect the observation vector Z? k Its expression is: .

[0060] Step S32: Calculate the observation vector Z k With predicted state H k The difference between them yields the observed residual. This residual reflects the degree of inconsistency between the system's prior predictions and the actual situation.

[0061] Step S33: Calculate the Kalman gain K based on the prior error covariance matrix and the predefined observation noise covariance matrix. k The Kalman gain is dynamically adjusted based on the relative uncertainty of the prediction and observation, and its expression is: ; Among them, P k - R is the prior error covariance matrix at the current time; k The predefined observation noise covariance matrix measures the impact of the observed data Z. k The noise level, R k The diagonal elements are preset according to the sensor's measurement accuracy; H k For the observation matrix, For H k The transpose of .

[0062] Step S34: Utilize Kalman gain K k With observation residuals Predicted value of the prior state at the current moment After making corrections, we obtain the optimal posterior state estimate for the current time step. Its update equation is: .

[0063] It should be noted that the optimal posterior state estimate Content and state vector X k Consistent with the definition, it is an optimal estimate of the true state of the system: ; in, It is a smoother and more reliable estimate of terrain coordinates after filtering and correction (eliminating the influence of observation noise); These are filtered and corrected water quality parameter estimates that are smoother and more reliable. These are filtered and corrected meteorological parameter estimates that are smoother and more reliable.

[0064] At the same time, update the error covariance through the equation Calculate the posterior error covariance P at the current time. k .

[0065] This embodiment outputs the optimal posterior state estimate at the current time. As the final integrated state vector, With P k As input for the next time step of the filtering calculation, steps S2 to S3 are repeated to achieve continuous dynamic state estimation in time.

[0066] Step S4: Perform anomaly detection on the target area based on the optimal posterior state estimate. If anomalies are detected, determine the risk level based on the risk assessment model.

[0067] For the optimal posterior state estimation vector value For each component, an anomaly detection mechanism based on statistical principles is applied: For the geological coordinate components, the historical mean μ of their elevation components is calculated based on the historical best state estimation sequence. 地 and historical standard deviation σ 地 Based on the estimated values ​​of terrain coordinates Determine the current elevation component h i When h i Satisfy |h i μ 地 |>3σ 地 Points marked as abnormal are temporarily removed, triggering a second scan of the area by lidar. If the second scan still reveals abnormalities, the data is sent to the "early warning decision layer" to determine whether a geological lesion or disaster has occurred.

[0068] For each water quality parameter component, a healthy range is established based on its dynamic variation coefficient. When the estimated water quality parameter value continuously deviates from the healthy range, it is marked as an abnormal water quality condition. Specifically, the mean value of estimated water quality parameters such as pH and dissolved oxygen over a certain period (e.g., 5 minutes) is calculated based on the historical best-state estimation sequence. and standard deviation When the estimated water quality parameter y t Satisfy |y t μ 水 |>3σ 水 The time stamp is marked as abnormal, which automatically triggers a second data collection. If the second data collection still shows an abnormality, it is pushed to the "early warning decision layer" to determine whether water pollution has occurred.

[0069] For meteorological element components, extreme event thresholds are set. When the estimated value of a meteorological parameter exceeds the threshold, it is marked as a meteorological anomaly. Specifically, based on... The criteria involve calculating the average value of meteorological parameters such as wind speed and cloud morphology over a certain period of time (e.g., 10 minutes). and standard deviation When meteorological parameter estimates satisfy When the meteorological data is initially marked as abnormal, it triggers automatic calibration of the meteorological sensors. If the abnormality persists after calibration, it is pushed to the "early warning decision layer" to determine whether a meteorological change or disaster has occurred.

[0070] Step S5: Based on the prediction decision layer, the corresponding execution strategy is generated according to the risk level.

[0071] In this embodiment, the predictive decision layer uses the Analytic Hierarchy Process (AHP) + fuzzy comprehensive evaluation to construct a multi-dimensional risk assessment model, accurately determine the risk level and output decision instructions, so that the risk level can be accurately determined.

[0072] Among them, determining the risk level based on the risk assessment model includes: Step S51: Construct a risk assessment indicator system: Based on the analytic hierarchy process, construct a risk assessment indicator system that includes a target layer, a criterion layer, and an indicator layer; among which, the criterion layer includes at least three indicators: water quality (U1: pH, dissolved oxygen, etc.), topographic change (U2: coastline erosion rate, elevation change, etc.) and ecological community dimension (U3: coral coverage, species diversity); the target layer is "island and reef ecological risk U".

[0073] Step S52: Determine Indicator Weights: Construct the judgment matrices for the criterion layer and the indicator layer, calculate the weight vector of each indicator relative to the upper-level target, and perform a consistency check. When the consistency ratio CR < 0.1, the weight vector is confirmed to be valid. Specifically, construct the judgment matrix: ; Where n represents the number of indicators to be compared, a ij This indicates the degree of importance of indicator i relative to indicator j.

[0074] This embodiment determines the weight of each indicator by calculating the eigenvector of the judgment matrix. The formula for calculating the eigenvector is as follows: ; Where A is the judgment matrix; w is the eigenvector, also known as the weight vector; The largest eigenvalue of the matrix is ​​denoted as . In this embodiment, the geometric mean of each row of the matrix is ​​calculated, and the geometric mean is normalized to obtain the eigenvector w. Then, Aw is used to... The relationship was verified to confirm the correctness of the calculation, and finally w = [w1, w2, w3] was obtained, which represent the weights of water quality, topographic change and ecological community respectively, and w1+w2+w3=1.

[0075] The consistency check described above verifies the logical consistency of the judgment matrix, ensuring the rationality of the weight allocation. Its formula is as follows: ; ; Where n represents the order of the judgment matrix, that is, the number of indicators that are being compared pairwise in the current level; The average random consistency index (the baseline value obtained by looking up Table 1).

[0076] Table 1. Correspondence between n and RI

[0077] when When CR ≥ 0.1, it means the consistency of the judgment matrix is ​​acceptable and the weight allocation is reasonable; when CR ≥ 0.1, it means the consistency of the judgment matrix is ​​unacceptable and the comparison values ​​in the judgment matrix need to be readjusted.

[0078] Step S53: Perform fuzzy comprehensive evaluation: Construct a membership function for each index. The membership degrees of observed values ​​to different risk levels are calculated to form a membership degree matrix. The membership degree matrix and the corresponding weight vectors are then fuzzily synthesized to obtain the fuzzy evaluation results. The water quality indicators are represented by a trapezoidal membership degree: ; Among them, a, b, c, and d are threshold values ​​determined based on specific indicators for island and reef ecological monitoring, serving to delineate different "risk state intervals." Taking the "water quality" indicator as an example: ① Ideal range (safe state): If the health requirements of the island reef ecosystem for pH are [7.5, 8.5], we can set a=7.5 and b=8.5. correspond This indicates that the pH is in an ideal state of "complete health and no risk".

[0079] ② Transitional interval (early warning state): When the pH is slightly below the healthy value, for example, deviating from 7.0 (let's call it c), and then entering the ideal range at 7.5 (i.e. a), then... At that time, membership degree This reflects the gradual transition from risk to safety; If the upper limit of the warning range for pH above the healthy value is set at 9.0 (i.e., d), when ,use This reflects the transition from safety to risk.

[0080] ③ Abnormal interval (dangerous state): When x < c or x > d, This indicates that the pH level has deviated significantly from the healthy range.

[0081] The trapezoidal membership degree used for water quality indicators can also be defined similarly for topographic indicators, through fuzzy transformation B=w R outputs the risk level, where w is the weight vector and R is the membership matrix. It is a fuzzy operator.

[0082] Step S54: Establish a Bayesian network model: Train a Bayesian network using historical data to learn the conditional probability relationship between various indicator evidence and the actual risk state. .

[0083] Step S55: Output the final risk level: Input the fuzzy evaluation result as evidence into the Bayesian network model, calculate the posterior probability, correct the warning result, and output the execution decision instruction corresponding to the risk level.

[0084] This embodiment pre-establishes a mapping rule between risk levels and execution decision commands. For example, in the event of extreme risk, it triggers a full-area water purification device, and in the event of moderate risk, it initiates a local coral restoration task. See Table 2 for details. Table 2 Correspondence between Risk Level and Execution Action

[0085] In this embodiment, the execution devices for executing decision commands include water purification devices, emergency lighting systems, etc. Each execution device is connected to an industrial controller and is driven by an industrial serial communication protocol such as Modbus. The mathematical expression is as follows: If the risk level is Then execute the set of actions. ,satisfy: (f is a predefined mapping function).

[0086] Step S6: Collect feedback data from the execution device and optimize the execution strategy. The specific method is as follows: Step S61: Collect the status parameters of the execution device in real time. The status parameters include at least the device power consumption, device location, and images of the repair area.

[0087] In this embodiment, the energy consumption of the device is collected by a current sensor, the position of the robot is recorded by a GPS module, and images of the repair area are captured by a high-definition camera. The data is then converted from A / D and uploaded to the industrial controller via Modbus.

[0088] Step S62: Based on the state parameters, calculate the key performance indicators (KPIs) for the repair execution. The KPIs include at least the risk reduction rate and the execution cost.

[0089] The expression for the risk reduction rate is: ; in, These are the comprehensive risk assessment values ​​before and after the repair execution.

[0090] The expression for execution cost is: ; Among them, the energy consumption cost is: .

[0091] Step S63: Construct an optimization model based on the Q-learning algorithm, determine the reward function of the optimization model by using key performance indicators, and continuously optimize the repair decision through the optimization model to output the optimal execution strategy.

[0092] Construct an optimization model based on the Q-learning algorithm, where: The system state s is defined as a multi-dimensional vector containing the current risk level, equipment energy consumption, water temperature, and flow rate, i.e., s = [L k [, P, T, V], where L k The risk level is represented by P, power is represented by T, water temperature is represented by V, and flow rate is represented by V.

[0093] The action 'a' is defined as an instruction to adjust the operating parameters of the equipment, such as the dosage of the purification equipment or the speed of the robot.

[0094] Construct a reward function r = α·ΔR - β·C, where α and β are the weighting coefficients for risk reduction and cost control, respectively. For example, α = 0.8 (risk reduction weight) and β = 0.2 (cost weight).

[0095] Based on the optimization model, the strategy is continuously optimized through the following steps: Step S631: In the current state s, select and execute action a according to the Q-value table; Step S632: Observe the new state s' of the system after execution and calculate the immediate reward r; Step S633: Update the formula based on the Q value Update the Q-value table, where γ=0.9 is the discount factor. The new state after the action is performed; Step S631: Update the system state to s' and repeat the above steps to achieve adaptive optimization of the repair strategy. Once the Q-value table converges, output the optimal action strategy π(s) = argmax based on state s. a Q(s,a) is used to guide the adaptive control of the restoration equipment. This method drives the reinforcement learning model through real-time feedback data, and achieves dynamic optimization of the dual objectives of risk control and execution cost in the ecological restoration process, which significantly improves the intelligence level and comprehensive benefits of restoration operations.

[0096] Based on the technical means adopted in this embodiment, the following beneficial effects are achieved: 1. By constructing a state transition matrix to model the system dynamics, and using Kalman gain as the weight to optimally fuse the observation vector with the prior state prediction, the problem of high noise and poor reliability of single sensor data is effectively overcome. The uncertainty of the generated posterior state estimate is significantly reduced, providing the system with more accurate and reliable state perception data.

[0097] 2. By employing a technical approach that constructs a unified system state vector from heterogeneous data such as geology, water quality, and meteorology, and then coordinating the processing of this vector in state prediction and observation updates, we have achieved deep fusion and unified representation of multi-source heterogeneous data, generating a comprehensive and consistent panoramic view of the system state, and providing a unified and authoritative data foundation for decision-making.

[0098] 3. By using the state transition matrix to quantify the evolution law of different types of parameters and realizing continuous estimation in time series through a recursive filtering framework, the system can generate smooth and continuous state evolution trajectories, effectively identify real trend changes, and significantly improve the system's ability to track the dynamics of the ecological environment.

[0099] 4. By using a spatial constraint model to divide the deployable area and combining it with the NSGA-II multi-objective optimization algorithm for site planning, this technique significantly reduces the number of devices and lowers deployment costs while ensuring monitoring coverage. It also avoids installation difficulties in complex terrain and achieves the optimal cost-effectiveness of the monitoring network.

[0100] 5. By combining front-end 3σ statistical filtering with back-end intelligent analysis based on optimal state estimation, and adopting a decision model that integrates analytic hierarchy process (AHP), fuzzy comprehensive evaluation, and Bayesian network, this multi-level technical approach effectively distinguishes between noise interference and real risks. Through quantitative management of uncertainty, it significantly improves the accuracy of early warning and greatly reduces false alarms and missed alarms in the system.

[0101] 6. By adopting the Q-learning algorithm, a reward function is constructed with risk reduction rate and execution cost as key performance indicators. The control strategy is dynamically optimized through online learning. This technology realizes a complete closed loop from state perception to decision execution, enabling the system to have self-learning and self-optimization capabilities, and significantly improving the intelligence level and comprehensive benefits of ecological restoration operations.

[0102] In another embodiment, an island and reef ecological monitoring and early warning system is also provided, which performs the island and reef ecological monitoring and early warning method as described above. Combined with... Figure 2 As shown, this system includes: The data acquisition layer is used to collect geological data, water quality data, and meteorological data of the target area, and to combine the geological data, water quality data, and meteorological data in a preset order to obtain the observation vector. The data processing layer is used to assimilate the observation vector and the prior state prediction value using the Kalman filter algorithm to obtain the optimal posterior state estimate value at the current time. The prior state prediction value is calculated based on the posterior state estimate value at the previous time and the pre-constructed state transition matrix. The early warning decision layer is used to detect anomalies in the target area based on the optimal posterior state estimate. When anomalies are detected, the risk level is determined based on the risk assessment model. The control execution layer is used to generate corresponding execution strategies based on the risk level.

[0103] It should be noted that the functions of each module in the system of this embodiment can be found in the corresponding descriptions in the above methods, and will not be repeated here.

[0104] In another embodiment, an electronic device is also provided. Figure 3 A structural block diagram of an electronic device according to an embodiment of the present invention is shown. Figure 3As shown, the electronic device includes a memory 100 and a processor 200. The memory 100 stores a computer program that can run on the processor 200. When the processor 200 executes the computer program, it implements the island and reef ecological monitoring and early warning method described in the above embodiments. The number of memories 100 and processors 200 can be one or more.

[0105] The electronic device also includes: The communication interface 300 is used to communicate with external devices and perform data exchange and transmission.

[0106] If the memory 100, processor 200, and communication interface 300 are implemented independently, they can be interconnected via a bus to communicate with each other. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc.

[0107] Optionally, in a specific implementation, if the memory 100, processor 200, and communication interface 300 are integrated on a single chip, then the memory 100, processor 200, and communication interface 300 can communicate with each other through an internal interface.

[0108] This invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method provided in this invention.

[0109] This invention also provides a chip, which includes a processor for calling and executing instructions stored in a memory, causing a communication device on which the chip is installed to perform the method provided in this invention.

[0110] This invention also provides a chip, including: an input interface, an output interface, a processor, and a memory. The input interface, output interface, processor, and memory are connected through an internal connection path. The processor is used to execute code in the memory. When the code is executed, the processor is used to execute the method provided in this invention.

[0111] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that the processor can be a processor supporting the Advanced Reduced Instruction Set Computing (RISC) machine (ARM) architecture.

[0112] Further, optionally, the aforementioned memory may include read-only memory and random access memory, and may also include non-volatile random access memory. The memory may be volatile or non-volatile, or may include both. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache. Many forms of RAM are available by way of example, but not limitation. Examples include static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0113] In the above embodiments, implementation can be achieved, in whole or in part, by software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0114] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0115] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0116] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in the present invention, and these should all be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for monitoring and early warning of island and reef ecology, characterized in that, include: Collect geological data, water quality data, and meteorological data of the target area, and combine the geological data, water quality data, and meteorological data in a preset order to obtain an observation vector; The observation vector and the prior state prediction value are assimilated by the Kalman filter algorithm to obtain the optimal posterior state estimate at the current time. The prior state prediction value is calculated based on the posterior state estimate at the previous time and the pre-constructed state transition matrix. Anomaly detection is performed on the target area based on the optimal posterior state estimate. If anomalies are detected, the risk level is determined based on the risk assessment model, and a corresponding execution strategy is generated according to the risk level.

2. The island and reef ecological monitoring and early warning method according to claim 1, characterized in that, The geological data, water quality data, and meteorological data of the target area to be collected include: The deployment scheme of the scanning equipment is determined based on a pre-built multi-objective optimization model, and the geological data is obtained by scanning the island and reef area using the scanning equipment. Based on the spatial constraint model, the deployable area of ​​the target sea area is determined, and water quality data of the target water area is collected by underwater sensors deployed in the deployable area. Meteorological data is obtained by collecting data from weather stations or drone cameras; wherein the drone camera captures images of the target area at a specified angle.

3. The island and reef ecological monitoring and early warning method according to claim 2, characterized in that, The method for constructing the multi-objective optimization model includes: Construct a first objective function that aims to minimize the total cost of equipment deployment, and construct a second objective function that aims to maximize the overall terrain coverage of the target area; Define the set of devices to be deployed as decision variables, and impose spacing constraints on the decision variables, wherein the spacing constraints require that the spatial distance between any two different devices in the set of devices is not less than a preset minimum spacing; The first objective function, the second objective function, and the spacing constraint are integrated into the multi-objective optimization model.

4. The island and reef ecological monitoring and early warning method according to claim 1, characterized in that, The process of assimilating the observation vector and the prior state prediction value using the Kalman filter algorithm includes: The predicted state is calculated based on the prior state prediction value and the pre-constructed observation matrix; Calculate the difference between the observed vector and the predicted state to obtain the observation residual; The Kalman gain is calculated based on the prior error covariance matrix and the predefined observation noise covariance matrix; wherein the prior error covariance matrix is ​​calculated based on the error covariance matrix of the previous time step through the covariance prediction equation; The prior state prediction is corrected using the Kalman gain and the observation residual to obtain the optimal posterior state estimate at the current time.

5. The island and reef ecological monitoring and early warning method according to claim 1, characterized in that, The method for constructing the state transition matrix includes: The system state vector is defined based on the environmental elements of the target area, where the environmental elements include geological spatial coordinates, water quality parameters, and meteorological elements. Based on the temporal evolution relationship of the system state vector from the previous moment to the current moment, a block-diagonal state transition matrix is ​​constructed to quantify the dynamic change patterns of geological data, water quality data, and meteorological data.

6. The island and reef ecological monitoring and early warning method according to claim 1, characterized in that, The risk level determination based on the risk assessment model includes: A risk assessment indicator system was constructed based on the analytic hierarchy process, and the indicator weights were calculated. The fuzzy comprehensive evaluation method was used to calculate the membership degree of each indicator to different risk states, and the fuzzy evaluation results were synthesized. A Bayesian network model is used to calculate the posterior probability of the fuzzy evaluation results in order to correct the warning results and output the final risk level.

7. The island and reef ecological monitoring and early warning method according to claim 1, characterized in that, Also includes: Real-time acquisition of status parameters of the execution device, including at least device energy consumption, device location, and images of the repair area; Based on the state parameters, calculate the key performance indicators for repair execution, which include at least the risk reduction rate and execution cost. An optimization model is constructed based on the Q-learning algorithm. The key performance indicators are used to determine the reward function of the optimization model. The optimization model is then used to continuously optimize and repair decisions, and output the optimal execution strategy.

8. An island and reef ecological monitoring and early warning system, characterized in that, include: The data acquisition layer is used to collect geological data, water quality data, and meteorological data of the target area, and to combine the geological data, water quality data, and meteorological data in a preset order to obtain an observation vector; The data processing layer is used to assimilate the observation vector and the prior state prediction value using the Kalman filter algorithm to obtain the optimal posterior state estimate value at the current time; wherein, the prior state prediction value is calculated based on the posterior state estimate value at the previous time and the pre-constructed state transition matrix. The early warning decision layer is used to perform anomaly detection on the target area based on the optimal posterior state estimate. When the detection result shows anomalies, the risk level is determined based on the risk assessment model. The control execution layer is used to generate corresponding execution strategies based on the risk level.

9. An electronic device, characterized in that, include: A processor and a memory, wherein the memory stores instructions that are loaded and executed by the processor to implement the island and reef ecological monitoring and early warning method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the island and reef ecological monitoring and early warning method as described in any one of claims 1 to 7.

Images