Multi-source data fused underwater robot dynamic cooperative control method and system

Abstract

The invention discloses an underwater robot dynamic cooperative control method and system fusing multi-source data, and relates to the field of general control or adjustment systems. The method comprises the following steps: controlling an inspection unmanned aerial vehicle to obtain laser point cloud and multispectral data of a reservoir dam and a water meter, and constructing a centimeter-level global space reference; adopting an anomaly detection algorithm based on mahalanobis distance to identify an abnormal water surface area, and generating a collaborative exploration task package; after the underwater robot receives the task package, centimeter-level accurate positioning is achieved through a permanent magnet mark array and visual positioning, and veneering observation is conducted on the underwater dam face; the water quality sampling unmanned aerial vehicle automatically collects stratified water body samples based on vertical water quality profile data; and collecting multivariate heterogeneous data, constructing a causal network model through principal component analysis, Bayesian reasoning and Granger causal test, and generating a comprehensive diagnosis report. The method greatly improves the positioning precision and diagnosis efficiency, and reduces the operation and maintenance cost.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of general control or regulation systems, in particular to a dynamic cooperative control method and system for underwater robots fusing multi-source data. BACKGROUND

[0002] As a key water conservancy infrastructure, the structural safety of reservoir dams is directly related to the downstream public safety, drinking water safety and sustainable use of water resources. However, in the current operation and maintenance practice of reservoir dams, there are still several outstanding technical bottlenecks: First, the positioning accuracy of underwater inspection is insufficient. Traditionally, underwater structure exploration relies on underwater robots, but due to the fact that global positioning system signals cannot be used underwater, robots can only rely on inertial navigation systems and Doppler log to perform dead reckoning. This method inevitably produces cumulative errors, and practice shows that the error can reach 0.1%-0.5% of the hourly navigation distance. When performing a long-distance inspection task of hundreds of meters in a reservoir dam, such errors cause the structural defects such as cracks and leaks discovered by the exploration to fail to establish an accurate spatial correspondence with the design drawings or BIM model, causing great difficulties in subsequent maintenance positioning and construction.

[0003] Second, it is difficult to trace the source of water ecological abnormalities. Existing water quality monitoring mostly uses a combination of fixed monitoring stations on the shore and periodic manual boat sampling. When a fixed monitoring station detects abnormal parameters such as temperature, chlorophyll, and dissolved oxygen in a certain area of water, it is difficult to quickly and accurately determine whether the anomaly is caused by underwater structural defects. The current technology lacks the ability to accurately and quickly trace the source of underwater problems.

[0004] In addition, the current unmanned system coordination is low. Although unmanned aerial vehicles, underwater robots and other equipment have been widely used, their coordination is mostly limited to simple command transmission and independent data collection, lacking a true intelligent coordination mechanism. The data collected by each platform is inconsistent in format and spatiotemporal reference, requiring a lot of manual intervention for data alignment and comprehensive analysis, which severely restricts the operation and maintenance efficiency and automation level.

[0005] Therefore, there is an urgent need for an innovative technology and system that can achieve high-precision underwater positioning, air-water integrated intelligent coordination, and causal relationship analysis of structural defects and ecological abnormalities. SUMMARY

[0006] The purpose of the present application is to provide a dynamic cooperative control method and system for underwater robots fusing multi-source data, which realizes the correlation analysis and accurate diagnosis of dam structure state and water ecological environment by establishing a unified spatial reference and data fusion mechanism, and overcomes the above-mentioned problems existing in the prior art.

[0007] To achieve the purpose, the present application provides the following technical solutions: A dynamic cooperative control method of an underwater robot fusing multi-source data, comprising the following steps: Step S1: controlling a patrol unmanned aerial vehicle to fly along a preset flight path of a reservoir dam and adjacent water area, acquiring laser point cloud and multispectral data of the reservoir dam and water surface; registering the laser point cloud with a BIM model of the reservoir dam to construct a global spatial reference; Step S2: using an abnormality detection algorithm based on Mahalanobis distance to analyze the multispectral data, identifying an abnormal water surface area; automatically generating a cooperative exploration task package containing the geographic location, range, type and priority of the abnormal water surface area; Step S3: after receiving the cooperative exploration task package, the underwater robot autonomously navigates to the abnormal water surface area specified in the cooperative exploration task package; when the underwater robot arrives at the abnormal water surface area, performing visual feature matching and pose solving on the underwater dam surface image corresponding to the abnormal water surface area captured by the underwater robot and the reference image of the BIM model to achieve centimeter-level accurate positioning; under the guidance of the centimeter-level accurate positioning information, controlling the underwater robot to perform surface observation on the underwater dam surface of the abnormal water surface area, and performing multi-angle shooting and measurement on the discovered defects, recording the accurate position, length and width thereof; Step S4: the water quality sampling unmanned aerial vehicle receives vertical water quality profile data from the underwater robot in real time, automatically identifies a characteristic depth based on a preset triggering rule, and collects a water sample at the characteristic depth; Step S5: collecting the multi-element heterogeneous data of steps S1 to S4, performing data dimension reduction and feature extraction using a principal component analysis method; combining Bayesian inference and Granger causality test methods, constructing a causal network model between structural defect parameters and water body ecological abnormality parameters, quantitatively identifying key associated factors, and automatically generating a comprehensive diagnosis report.

[0008] Further, in the step S1, the laser point cloud is registered with the BIM model of the reservoir dam to construct a global spatial reference, specifically comprising the following sub-steps: step S1.1: performing point cloud solving and denoising on the laser point cloud; Step S1.2: selecting a plurality of common feature points in the laser point cloud and the BIM model; through these common feature points, a coordinate transformation matrix is calculated to preliminarily align the laser point cloud and the BIM model; step S1.3: finding the nearest point pair between the triangular mesh surfaces of the laser point cloud and the BIM model, and iteratively calculating the optimal rigid body transformation through the least squares method to minimize the root mean square error between the two point sets, finally establishing a centimeter-level global spatial reference.

[0009] Further, in step S2, the multispectral data is analyzed by using a Mahalanobis distance-based anomaly detection algorithm to identify abnormal water surface areas, specifically including the following sub-steps: step S2.1: performing aerial triangulation and orthophoto mosaic on the multispectral original image obtained in step S1 to generate reflectivity images of each band; step S2.2: for the thermal infrared band, using the radiative transfer equation method combined with real-time meteorological data to obtain the water surface absolute temperature distribution map; using the reflectivity of the red edge and near-infrared bands, the distribution map of chlorophyll-a concentration is generated by using the normalized difference vegetation index algorithm; using the reflectivity ratio of blue light and red light bands, the turbidity distribution map is obtained according to the empirical or semi-empirical model; step S2.3: reading three distribution maps, for each pixel point with the same geographical position, constructing a three-dimensional feature vector with the values of its three bands, that is, obtaining a three-dimensional vector set representing the entire inspection area; randomly extracting a set of pixel points from the three-dimensional vector set as training samples, calculating the mean vector and covariance matrix, and constructing a statistical model for representing the characteristics of normal water area, the mean vector represents the average level of temperature, chlorophyll, and turbidity in the entire area, and the covariance matrix represents the mutual relationship between the three parameters; traversing each pixel point in the image, calculating the Mahalanobis distance of the pixel point to the distribution defined by the statistical model; determining the pixel points with a Mahalanobis distance exceeding a preset value as abnormal pixels; step S2.4: performing DBSCAN spatial clustering on the abnormal pixels to obtain the abnormal water surface area; extracting detailed information of the abnormal water surface area from the data set of steps S2.1 to S2.3, and packaging it into a task package.

[0010] Further, in step S3, a plurality of permanent magnet markers are arranged on the underwater wall surface of the dam in advance by using a topological structure combining equal-interval grids and key point encryption to form a permanent magnet marker array; the step S3 specifically includes the following sub-steps: step S3.1: launching the underwater robot from the mother ship and giving the initial position by the GNSS of the mother ship; after the underwater robot enters the water, it mainly relies on the combined navigation of the inertial navigation system and the Doppler log; step S3.2: when the inertial navigation system indicates that the underwater robot has traveled to a distance of a preset distance from the last calibration point and approaches a pre-arranged magnetic marker, the three-axis magnetometer array starts to monitor the regularity of the spatial magnetic field strength and direction, and obtains the magnetic field vector diagram; the inertial navigation system matches the real-time measured magnetic field vector diagram with the pre-stored magnetic field grid map; by using the least square algorithm, the current underwater robot is uniquely determined to be located near a permanent magnet marker; when the matching is successful, the navigation system takes the accurate coordinates of the permanent magnet marker recorded in the BIM model as a high- confidence observation, and feeds back and corrects the system state of the inertial navigation system, thereby suppressing the cumulative error of the inertial navigation system.

[0011] Further, the step S3 specifically comprises the following sub-steps: step S3.3: when the underwater robot reaches the abnormal water surface area, start the visual auxiliary navigation system, capture the underwater dam surface image corresponding to the abnormal water surface area by the camera, and extract a plurality of key points and their multi-dimensional feature descriptors; match these multi-dimensional feature descriptors with the reference image feature library generated by pre-rendering the BIM model in this area, and solve the coordinates and pose of the current underwater robot by the EPnP algorithm to achieve centimeter-level accurate positioning; step S3.4: the operator controls the underwater robot to observe the underwater dam surface of the abnormal water surface area under the guidance of the centimeter-level accurate positioning information, and multi-angle photographs and measures the discovered defects through the camera and the imaging sonar, records the accurate position, length, width and other information.

[0012] Further, in the step S4, the characteristic depth is automatically identified based on the preset trigger rule, specifically: analyzing the vertical water quality profile data, when detecting that the absolute value of temperature gradient ≥8.0 ℃ / m, or the depth layer of chlorophyll concentration mutation rate exceeds 30%, the depth layer is determined as the characteristic depth needing sampling.

[0013] Further, the step S5 comprises the following sub-steps: step S5.1: collect the laser point cloud, abnormal water surface area information, underwater structure defect parameters and position information, water quality profile data and layered sampling laboratory analysis results from steps S1 to S4, and construct a structure-ecological correlation data table under the unified space-time reference; Step S5.2: After standardizing the structure-ecological correlation data table, principal component analysis is performed to select principal components with eigenvalues greater than 1 and cumulative variance contribution rate greater than or equal to 85%; by analyzing the principal component loading matrix, the original variables with absolute value greater than 0.7 are defined as core representation variables, and the physical meaning label is assigned to the corresponding principal component according to the commonality of the core representation variables; Step S5.3: Based on the principal components obtained in step S5.2, and the structure semantic information and historical maintenance records extracted from the BIM model of the reservoir dam, a Bayesian network structure is initialized, and the observation evidence in the structure-ecological correlation data table is input into the Bayesian network structure, the posterior probability is calculated, the probability of different structure defects leading to specific ecological abnormalities is quantitatively evaluated, and the key associated factors are preliminarily identified; for the preliminarily associated variables, the time series data thereof is extracted, and Granger causality test is performed, when the p value of the test result is less than 0.05, it is determined that there is a statistically significant causal relationship, so as to determine the causal direction; the probabilistic correlation strength of the Bayesian inference and the causal direction of the Granger causality test are fused to construct a quantitative structure defect-water body ecological anomaly causal network model, and the quantitative identification of the key associated factors is completed; Step S5.4: Based on the analysis result of the causal network model, a comprehensive diagnosis report is automatically generated, and the report at least includes the accurate positioning of the defect in the BIM model, the cause mechanism analysis based on the causal relationship, the list of key associated factors identified quantitatively, and the targeted maintenance strategy suggestion.

[0014] A kind of underwater robot dynamic collaborative control system of fusion multi-source data, for implementing any above-mentioned underwater robot dynamic collaborative control method of fusion multi-source data, it includes: inspection unmanned plane, for executing step S1 and step S2;Underwater robot, for executing step S3;Water quality sampling unmanned plane, for executing step S4;Ground control and intelligent analysis center, for collaborative control the inspection unmanned plane, underwater robot and water quality sampling unmanned plane, and execute step S5.

[0015] Further, it further includes permanent magnet mark array for underwater robot correction navigation, the permanent magnet mark array is made of several permanent magnet marks fixed on the underwater wall of dam, and the spatial coordinates of each permanent magnet mark are recorded in the BIM model.

[0016] Compared with the prior art, the present application has the following beneficial effects: Firstly, the application constructs a global spatial reference by high-precision registration of laser point cloud obtained by a patrol unmanned aerial vehicle and a BIM model, and combines a pre-laid permanent magnet marker array for magnetic field matching positioning, and a pose solution of an underwater robot based on visual features and an EPnP algorithm, to form a multi-source fusion positioning correction chain, which reduces the absolute positioning error of the underwater robot from the meter level of traditional inertial navigation to the centimeter level, and provides reliable spatial reference guarantee for accurate maintenance of dam defects.

[0017] Secondly, the application automatically identifies and generates a cooperative exploration task package based on a Mahalanobis distance multi-spectral data anomaly detection algorithm, drives the underwater robot and the water quality sampling unmanned aerial vehicle to perform dynamic cooperative operation, and realizes adaptive layered sampling by using a trigger rule of vertical water quality profile data, so that the traditional anomaly investigation process which depends on manual operation and takes several weeks is shortened to several hours, and the overall efficiency is greatly improved.

[0018] Thirdly, the application constructs a quantitative causal network model between structural defect parameters and water body ecological anomaly parameters by using principal component analysis for data dimension reduction and feature extraction, and combining Bayesian inference and Granger causality test method, which can quantitatively identify key associated factors and analyze the cause mechanism, realizes the leap from phenomenon description to causal diagnosis, and identifies early hidden dangers that are difficult to detect by naked eyes, thereby promoting the transformation of the operation and maintenance mode to early warning maintenance.

[0019] Fourthly, the automatic cooperative operation and intelligent analysis and diagnosis of the above-mentioned air-space-water multi-platform greatly reduce the demand for high-risk manual diving exploration and ship operation, reduce the intensity of manual intervention and safety risk, and practical application shows that the comprehensive operation and maintenance cost can be greatly reduced. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The method flowchart of the application.

[0021] Figure 2 The system structure block diagram of the application. DETAILED DESCRIPTION

[0022] The specific embodiments of the application will be described below with reference to the accompanying drawings. In order to fully understand the application, many details are described below, but the application can be implemented without these details for those skilled in the art.

[0023] As shown in Figure 1 and Figure 2 , a dynamic cooperative control method of an underwater robot fusing multi-source data includes the following steps: Step S1: controlling the inspection unmanned aerial vehicle to fly along a preset flight route of the reservoir dam and the adjacent water area, to obtain laser point cloud and multispectral data of the reservoir dam and the water surface; registering the laser point cloud with a BIM model of the reservoir dam, to construct a global spatial reference with centimeter-level precision.

[0024] In a specific embodiment, in step S1, flying along the preset flight route of the reservoir dam and the adjacent water area specifically refers to extending 200 meters upstream of the reservoir dam based on the dam axis, to plan an arch-shaped flight route.

[0025] In step S1, registering the laser point cloud with the BIM model of the reservoir dam to construct a global spatial reference with centimeter-level precision specifically includes the following sub-steps: Step S1.1: performing point cloud solving and denoising on the laser point cloud.

[0026] Step S1.2: selecting a plurality of common feature points in the laser point cloud and the BIM model; calculating a coordinate transformation matrix through the common feature points, to preliminarily align the laser point cloud and the BIM model.

[0027] Step S1.3: finding the nearest point pairs between the triangular mesh surfaces of the laser point cloud and the BIM model, and iteratively calculating the optimal rigid body transformation through the least squares method, to minimize the root mean square error between the two point sets, to finally establish a centimeter-level global spatial reference.

[0028] Step S2: analyzing the multispectral data by using a Mahalanobis distance-based anomaly detection algorithm, to identify an abnormal water surface area; automatically generating a collaborative investigation task package containing the geographic location, range, type and priority of the abnormal water surface area.

[0029] In a specific embodiment, in step S2, analyzing the multispectral data by using the Mahalanobis distance-based anomaly detection algorithm to identify the abnormal water surface area specifically includes the following sub-steps: Step S2.1: performing aerial triangulation and orthophoto mosaic on the multispectral original images obtained in step S1, to generate reflectivity images of each band, and performing atmospheric correction by using a FLAASH model, to convert apparent reflectivity to ground true reflectivity.

[0030] Step S2.2: for the thermal infrared band, using a radiation transfer equation method combined with real-time weather data, to obtain an absolute temperature distribution map of the water surface; using reflectivity of the red edge and near-infrared bands, to generate a distribution map of chlorophyll a concentration by using a variant algorithm of the normalized difference vegetation index; using reflectivity ratio of the blue light (such as 450 nm) and red light (such as 650 nm) bands, to obtain a turbidity distribution map according to an empirical or semi-empirical model (such as the Ne Chad algorithm).

[0031] Step S2.3: Read three distribution maps, for each pixel point with the same geographical position, construct a three-dimensional feature vector with its three band values, that is, obtain a three-dimensional vector set representing the entire inspection area; randomly extract a pixel point set from the three-dimensional vector set as a training sample, calculate its mean vector and covariance matrix, and construct a statistical model for representing the characteristics of normal water area, the mean vector represents the average level of temperature, chlorophyll, and turbidity in the entire area, and the covariance matrix represents the mutual relationship between the three parameters; traverse each pixel point in the image, calculate its Mahalanobis distance to the distribution defined by the statistical model; determine the pixel points with Mahalanobis distance exceeding the preset value as abnormal pixels.

[0032] Step S2.4: DBSCAN spatial clustering is performed on the abnormal pixels to obtain an abnormal water surface area; detailed information of the abnormal water surface area is extracted from the data set of steps S2.1 to S2.3, and is packaged into a task package.

[0033] Step S3: After receiving the cooperative exploration task package, the underwater robot autonomously navigates to the abnormal water surface area specified in the cooperative exploration task package; when the underwater robot arrives at the abnormal water surface area, the camera of the underwater robot captures the underwater dam surface image corresponding to the abnormal water surface area, and performs visual feature matching and pose solving on the underwater dam surface image and the reference image of the BIM model, to realize centimeter-level accurate positioning; under the guidance of the centimeter-level accurate positioning information, the underwater robot is controlled to perform surface observation on the underwater dam surface of the abnormal water surface area, and to perform multi-angle shooting and measurement on the discovered defects, and to record information such as accurate position, length, width, etc.

[0034] In a specific embodiment, a plurality of permanent magnet markers are pre-arranged on the underwater wall surface of the dam in a topological structure combining equal-interval grid and key point encryption, to form a permanent magnet marker array. In step S3, the following sub-steps are specifically included: Step S3.1: The underwater robot is launched from the mother ship, and the initial position is given by the GNSS of the mother ship; after entering the water, the underwater robot performs combined navigation relying on the inertial navigation system and the Doppler log. However, due to the scale factor error (about 0.2%) of the inertial navigation system and the instability of the gyroscope zero offset of the Doppler log, the position error accumulates at a rate of about 0.3% navigation distance / hour. Therefore, the position of the underwater robot is further magnetically grid-absolute calibrated by means of the permanent magnet marker matrix.

[0035] Step S3.2: When the inertial navigation system indicates that the underwater robot travels to a distance of a preset distance from the last calibration point and approaches a pre-deployed magnetic marker, the three-axis magnetometer array starts to monitor the regularity of the spatial magnetic field strength and direction, and obtains the magnetic field vector diagram; the inertial navigation system matches the real-time measured magnetic field vector diagram with the pre-stored magnetic field grid map; through the least square algorithm, the current underwater robot is uniquely determined to be located near a permanent magnet marker; when the matching is successful, the navigation system takes the accurate coordinates of the permanent magnet marker recorded in the BIM model as a high- confidence observation, and feeds back and corrects the system state of the inertial navigation system, thereby suppressing the cumulative error of the inertial navigation system.

[0036] Step S3.3: When the underwater robot reaches the abnormal water surface area, the magnetic marker positioning of step S3.2 is used to periodically correct the error of the inertial navigation system, and at the same time, the visual auxiliary navigation system is started for accurate positioning. The visual auxiliary navigation system is specifically: capturing the underwater dam surface image corresponding to the abnormal water surface area by the camera, and extracting a plurality of key points and their multi-dimensional feature descriptors; matching these multi-dimensional feature descriptors with the reference image feature library generated by the BIM model in the area in advance, and solving the coordinates and attitude of the current underwater robot through the EPnP algorithm to realize centimeter-level accurate positioning.

[0037] Step S3.4: The operator controls the underwater robot to observe the underwater dam surface of the abnormal water surface area under the guidance of the centimeter-level accurate positioning information, and takes multi-angle photographs and measurements of the discovered defects through the camera and the imaging sonar, and records the accurate position, length and width and other information of the discovered defects.

[0038] Step S4: The water quality sampling unmanned aerial vehicle receives the vertical water quality profile data from the underwater robot in real time through the wireless network, automatically identifies the characteristic depth based on the preset trigger rule, and collects the water sample at the characteristic depth.

[0039] In a specific embodiment, in the step S4, the characteristic depth is automatically identified based on the preset trigger rule, specifically: analyzing the vertical water quality profile data, when detecting that the absolute value of the temperature gradient is ≥8.0 ℃ / m, or the depth layer with a chlorophyll concentration mutation rate exceeding 30% is detected, the depth layer is determined as the characteristic depth needing to be sampled. The chlorophyll concentration mutation rate = |(C2-C1)| / C1x100%, wherein C1 and C2 are the concentration values of adjacent depth layers.

[0040] Step S5: Collecting the multi-element isomerism data of steps S1-S4, using principal component analysis method for data dimension reduction and feature extraction; combining Bayesian inference and Granger causality test method, constructing a causal network model between structure defect parameters and water body ecological anomaly parameters, quantitatively identifying key associated factors, and automatically generating a comprehensive diagnostic report including defect positioning, cause analysis and maintenance suggestions. It includes but is not limited to laser point cloud, spectral image, water quality profile data, structure defect parameters, etc.

[0041] In a specific embodiment, the step S5 includes the following sub-steps: Step S5.1: Collecting the laser point cloud, abnormal water surface area information, underwater structure defect parameters and location information, water quality profile data and layered sampling laboratory analysis results from steps S1-S4, and constructing a structure-ecological correlation data table under a unified space-time reference, wherein the rows represent different observation samples and the columns include structure defect parameters and water body ecological anomaly parameters.

[0042] Step S5.2: After standardizing the structure-ecological correlation data table, performing principal component analysis, selecting principal components with eigenvalues greater than 1 and cumulative variance contribution rate ≥ 85%; by analyzing the principal component loading matrix, defining the original variables with absolute load value greater than 0.7 as core representation variables, and assigning physical meaning labels to the corresponding principal components according to the commonality of the core representation variables.

[0043] Step S5.3: Based on the principal components with physical meaning labels obtained in step S5.2, and the structure semantic information and historical maintenance records extracted from the BIM model of the reservoir dam, initializing a Bayesian network structure, and inputting the observation evidence in the structure-ecological correlation data table into the Bayesian network structure, quantitatively evaluating the probability of different structure defects leading to specific ecological anomalies by calculating the posterior probability, and preliminarily identifying key associated factors; For the preliminarily associated variables, extract their time series data and perform Granger causality test. When the test result p value is less than the threshold value (such as 0.05), it is determined that there is a statistically significant causal relationship, thereby determining the causal direction; Fusing the probabilistic correlation strength of the Bayesian inference and the causal direction of the Granger causality test, constructing a quantitative structure defect-water body ecological anomaly causal network model, and completing the quantitative identification of key associated factors.

[0044] Step S5.4: Based on the analysis results of the causal network model, automatically generating a comprehensive diagnostic report, which at least includes the precise positioning of defects in the BIM model, the cause mechanism analysis based on the causal relationship, the list of key associated factors identified quantitatively, and the targeted maintenance strategy suggestions.

[0045] As Figure 1 and Figure 2 shown, a dynamic cooperative control system of underwater robots fusing multi-source data is used to implement the dynamic cooperative control method of underwater robots fusing multi-source data, and the system comprises: The inspection unmanned aerial vehicle comprises a laser radar, a multispectral camera, a high-precision GNSS / IMU combined navigation system, and a first wireless communication unit, and is used to execute steps S1 and S2.

[0046] The underwater robot is used to execute step S3.

[0047] The water quality sampling unmanned aerial vehicle is used to execute step S4.

[0048] The ground control and intelligent analysis center is used to cooperatively control the inspection unmanned aerial vehicle, the underwater robot, and the water quality sampling unmanned aerial vehicle, and execute step S5.

[0049] In addition, the system further comprises a permanent magnet marker array for underwater robot correction navigation, the permanent magnet marker array is composed of a plurality of permanent magnet markers fixed on the underwater wall surface of the dam, and the spatial coordinates of each permanent magnet marker are recorded in the BIM model. The permanent magnet markers are fixed on the underwater wall surface of the dam in an embedded or anchored manner.

[0050] Embodiment 1 Taking the periodic comprehensive inspection of a certain large concrete gravity dam as an example, the implementation process of the system and method is comprehensively demonstrated.

[0051] (I) System configuration 1. Inspection unmanned aerial vehicle: select DJI Matrice350RTK platform, and mount Zenmuse L1 laser radar and Zenmuse P1 multispectral camera.

[0052] 2. Underwater robot: custom Observation-Class ROV, frame size is 1.5m (length) x 1.0m (width) x 0.8m (height), 6 brushless DC thrusters are used to realize six degrees of freedom maneuvering. It is equipped with: 1) a three-axis magnetometer array composed of 12 sensors; 2) 1080P high-definition low-light camera; 3) SoundMetrics ARISExplorer1800 imaging sonar as underwater acoustic beacon receiver; 4) Kearfott T24 INS / DVL combined navigation system.

[0053] 3. Water quality sampling unmanned aerial vehicle: select DRAGONFLY 2 water sampling unmanned aerial vehicle (model M400-FC100) of Guangzhou Feichuang Intelligent Technology Co., Ltd.

[0054] 4. Ground control and intelligent analysis center: a high-performance mobile workstation running task planning software, data management software, and the core analysis algorithm of the present invention.

[0055] 5. Permanent magnet marker array: permanent magnet markers are N52 grade neodymium iron boron permanent magnets, with dimensions of 50mm x 50mm x 25mm and a surface magnetic field strength of 0.8±0.1T. Each permanent magnet marker is encapsulated in a laser-welded 316 stainless steel waterproof shell. Multiple permanent magnet markers are arranged on key structural parts of the dam underwater wall, gate slot, etc. according to the topology, forming a permanent magnet marker array. Among them, the topology arrangement: combination of equidistant grid (basic layer) and non-uniform topology based on key points of structure (application layer). Among them, equidistant grid (basic layer): on the main underwater wall of the dam, arranged according to the 80m x 80m grid node. Key point encryption (application layer): on all dam segment transverse joints, 5m range upstream and downstream of the flood discharge gate slot, around the drainage gallery outlet, the arrangement density is encrypted to 20m x 20m. After the completion of the arrangement of each permanent magnet marker, the three-dimensional coordinates (precision: ±2mm) of its center point are recorded into the BIM model.

[0056] (II) Method implementation Step S1: Global spatial reference construction and preliminary screening Task planning and flight: The operator at the ground control and intelligent analysis center opens the DJI Pilot 2 software and imports the boundary KML file of the reservoir dam. A bow-shaped flight route is planned 200 meters upstream of the dam axis. The specific parameter settings are: flight height relative to dam crest 80 meters, flight speed 5 meters / second, heading overlap rate set to 85%, and lateral overlap rate set to 25%. This setting is designed to ensure sufficient density of laser point cloud and sufficient stereo pairs of multispectral images for modeling. Subsequently, the unmanned aerial vehicle automatically flies along the preset route and synchronously collects laser point cloud and multispectral raw images through the laser radar and multispectral camera. The exposure points of the laser radar and multispectral camera are triggered by the high-precision GNSS / IMU system of the unmanned aerial vehicle, ensuring that each data point has accurate longitude, latitude, and height (WGS84 UTM coordinate system) and attitude information (pitch, roll, yaw).

[0057] Point cloud data processing and registration: After the flight is completed, the ground control and intelligent analysis center receives the laser point cloud and imports it into the ContextCapture software. First, the laser point cloud is processed for point cloud solving and denoising, and obvious aircraft noise and abnormal drift points are removed. Then, in the laser point cloud and BIM model, several common feature points are selected. In this inspection, 4 corners of the dam crest, 2 lamp post bases, and the center of the flood discharge gate hoist machine seat are selected.

[0058] Through these common feature points, a preliminary coordinate transformation matrix is calculated to preliminarily align the laser point cloud and the BIM model; then, the nearest point pairs are found between the triangular mesh surface of the laser point cloud and the BIM model, and the optimal rigid body transformation (rotation matrix R and translation vector T) is iteratively calculated by the least square method to minimize the root mean square error between the two point sets, and finally a centimeter-level global spatial reference is established. The iteration termination condition is set as: the root mean square error of two consecutive iterations changes less than 1 cm, or the maximum number of iterations reaches 100 times. The final registration report of this inspection shows that the planar mean error is 2.5 cm, and the height mean error is 4.1 cm, and a centimeter-level global spatial reference is successfully established.

[0059] Step S2: Intelligent identification of water surface anomalies and dynamic generation of tasks Multispectral data preprocessing and distribution map generation: First, the multispectral original image obtained by the multispectral camera is processed in Pix4Dmatic to perform aerial triangulation and orthophoto mosaic to generate reflectivity images of each band. Subsequently, for the thermal infrared band, atmospheric correction and specific radiation correction are performed in ENVI software using the radiation transfer equation method combined with real-time meteorological data to obtain the water surface absolute temperature distribution map; the reflectivity of the red edge (730 nm) and near-infrared (840 nm) bands is used to generate a chlorophyll a concentration distribution map by a variant algorithm of the normalized difference vegetation index; the reflectivity ratio of blue light (450 nm) and red light (650 nm) bands is used to obtain a turbidity distribution map according to an empirical or semi-empirical model (such as the Nechad algorithm); Multi-parameter joint anomaly detection based on Mahalanobis distance: Read the three distribution maps, and for each pixel point with the same geographical location, form a three-dimensional feature vector x i =[temp i ,chl i ,turb i ] of the values of its three bands; where temp i , chl i and turb i represent the temperature, chlorophyll a concentration and turbidity of point i respectively), to obtain a three-dimensional feature vector set representing the entire inspection area; Randomly select a large number of 10,000 pixel points in the entire area as training samples, which represent the normal water area state of the reservoir. Calculate the mean vector μ = [μ temp , μ chl , μ turb ] and the covariance matrix Σ of the three-dimensional feature vector set of these training samples. Where μ temp , μ chl , μ turbThe average level of the whole region temperature, chlorophyll, and turbidity.

[0060] In this inspection, the mean vector μ = [20.2, 13.5, 12.3] of the training sample's three-dimensional feature vector set, and the covariance matrix Σ:

[0061] Wherein, Σ[1,2]=Σ[2,1]=2.16, which is the covariance of temperature and chlorophyll a concentration, and the correlation coefficient ρ=2.16 / √(1.21×9.61)≈0.63, which is positively correlated, indicating that temperature rise usually accompanied by chlorophyll a concentration increase. Σ[1,3]=Σ[3,1]=-0.85, which is the covariance of temperature and turbidity, and the correlation coefficient ρ=-0.85 / √(1.21×2.25)≈-0.52, which is negatively correlated, indicating that the water body with higher temperature usually has lower turbidity. Σ[2,3]=Σ[3,2]=1.23, which is the covariance of chlorophyll a concentration and turbidity, and the correlation coefficient ρ=1.23 / √(9.61×2.25)≈0.27, which is positively correlated, indicating that algae reproduction may slightly increase water turbidity.

[0062] Each pixel point x in the reflectivity image is traversed pixel by pixel i , and its Mahalanobis distance to the normal water area is calculated independently. The calculation formula of Mahalanobis distance D M (x i ) is: D M (x i )=√[(x i μ) T Σ -1 (x i μ)]. Under the assumption that the three-dimensional feature vector obeys multivariate normal distribution, the square of Mahalanobis distance (D M (xi)) 2 obeys chi-square distribution with degree of freedom 3: χ 2 3. Setting 95% confidence interval, χ 2 3distribution table gives P(χ 2 3≤ 7.815)= 0.95. Therefore, the pixel points with Mahalanobis distance D M (x i )>√7.815≈2.796 are determined as abnormal pixels. Among them, the set 95% confidence interval can be dynamically adjusted according to historical data, or a self-adaptive threshold mechanism can be introduced.

[0063] Clustering and task package generation: DBSCAN clustering algorithm (parameters: eps=5 meters, min_samples=10) is used to spatially cluster the abnormal pixels to form an abnormal water surface area, and a collaborative exploration task package corresponding to the abnormal water surface area is generated. The collaborative exploration task package is pushed to the underwater robot control unit on the mother ship in real time.

[0064] In this inspection, based on the above pixel-by-pixel anomaly detection results, after using the DBSCAN clustering algorithm to spatially cluster all the pixels marked as abnormal, these abnormal pixels in the water surface area above the fine cracks collectively form an abnormal water surface area. This is because the cracks cause the reservoir water to seep into the dam body, forming a local seepage field, changing the water thermal conduction path, and thus forming a temperature anomaly area on the corresponding water surface. The mean vector of the abnormal water surface area is [18.5, 15.2, 12.5]. Among them, the mean absolute temperature of the water surface is 18.5 ℃, which is about 1.7 ℃ lower than the surrounding normal water area of 20.2 ℃, the mean chlorophyll a concentration is 15.2 μg / L, which is slightly higher than the normal value of 13.5 μg / L, and the mean turbidity is 12.5 NTU, with no obvious abnormalities. The average Mahalanobis distance is calculated to be 3.45, which is much higher than the threshold value of 2.796.

[0065] Step S3: centimeter-level accurate positioning and close-in exploration underwater Initial navigation and INS / DVL calculation: the underwater robot is launched from the mother ship's launch point (coordinates known). The initial position and attitude are provided by the mother ship's GNSS (Global Navigation Satellite System) and attitude instrument. After entering the water, rely on the inertial navigation system and Doppler velocity log (INS / DVL combined navigation system) for combined navigation. The inertial navigation system provides the ground speed, which is combined with the angular rate and acceleration of the Doppler velocity log in the extended Kalman filter to output the optimal navigation solution.

[0066] Magnetic field grid absolute calibration: when the inertial navigation system indicates that the underwater robot has traveled a distance of a preset distance (such as 60 m) from the last calibration point and approaches a permanent magnet marker, the three-axis magnetometer array begins to monitor the regularity of the spatial magnetic field strength and direction, and obtains the magnetic field vector diagram. The inertial navigation system performs least squares matching between the real-time measured magnetic field vector diagram and the pre-stored magnetic field map, and calculates that the current underwater robot is only(Δx=0.15 m, Δy=0.08 m, Δz=0.05 m) different from the theoretical position of the permanent magnet marker.

[0067] The absolute position with uncertainty (X3+Δx, Y3+Δy, Z3+Δz) is input into the extended Kalman filter as an observation update. The filter compares the predicted position of the inertial navigation system with the observation position, and calculates the estimates of the current position error, velocity error, attitude error and sensor bias error of the inertial navigation system. The filter then uses these error estimates to correct the entire system state vector of the inertial navigation system, thereby correcting the navigation path to the correct trajectory.

[0068] Vision-aided centimeter-level repositioning: After the underwater robot arrives at the abnormal water surface area, the high-definition camera is turned on to capture the underwater dam surface image corresponding to the abnormal water surface area. The contrast of the underwater dam surface image is enhanced through the CLAHE algorithm; about 520 scale and rotation invariant key points and their 128-dimensional descriptors are extracted from the underwater dam surface image.

[0069] These 128-dimensional feature descriptors are matched with the reference image feature library pre-rendered by the BIM model in this area, and 162 pairs of points are obtained through initial matching. 1000 RANSAC mis-matching elimination iterations are performed, and 4 pairs of points are randomly selected each time to calculate the homography matrix. Finally, an optimal set containing 135 pairs of inliers is found, and 27 pairs of outliers are eliminated. Using the 135 pairs of inliers, the transformation matrix from the camera coordinate system to the world coordinate system is solved through the EPnP algorithm, thereby realizing the centimeter-level accurate positioning of the underwater robot. The calculation result shows that the position accuracy of the underwater robot relative to the target dam surface is ±2.1 centimeters, and the attitude accuracy is ±0.3 degrees.

[0070] Under the guidance of the centimeter-level accurate positioning information, the operator controls the underwater robot to observe the underwater dam surface of the abnormal water surface area, and through the camera and imaging sonar, the discovered defects are photographed and measured from multiple angles, and the accurate position, length, width and other information are recorded. In this inspection, a fine crack with a length of 1.2 meters and a maximum width of 2.5 millimeters is found.

[0071] Step S4: adaptive hierarchical water quality sampling and verification Data reception and feature depth identification: The water quality sampling drone receives the vertical water quality profile data forwarded by the underwater robot through underwater acoustic communication, analyzes the vertical water quality profile data, and identifies the depth layer as the feature depth when detecting that the absolute value of the temperature gradient is ≥8.0 ℃ / m or the mutation rate of chlorophyll concentration exceeds 30%. During the analysis and calculation of the temperature gradient, it was found that between the depths of 15.5 meters and 16.0 meters, the temperature dropped sharply from 15.2 ℃ to 11.2 ℃, with a gradient of 4.0 ℃ / 0.5m=8.0 ℃ / m, i.e. 0.8 ℃ / 10cm, reaching the threshold value of 8.0 ℃ / m. Accordingly, the system determines 15.75 meters (the center of the thermocline) as the feature depth.

[0072] Precise stratified sampling: The water quality sampling drone flies over the target water area and begins to lower the sampler.

[0073] When the sampler reaches 15.70 meters, the water quality sampling drone begins to collect water samples. After completing the sampling, the water quality sampling drone recovers the sampler.

[0074] Step S5: Multi-source data causal correlation analysis and diagnostic report generation Multivariate heterogeneous data collection and structuring: The ground control and intelligent analysis center collects raw and processed data from the following sources: From step S1: registered BIM model, laser point cloud.

[0075] From step S2: boundary, type (temperature / chlorophyll), confidence of abnormal water surface area.

[0076] From step S3: three-dimensional coordinates of defects, defect type (such as cracks), defect size (length, width, depth), visual image, sonar image.

[0077] All data is integrated into a structure-ecological correlation data table, and the behavior observation samples are listed as variables. The variables include crack length, crack width, crack depth, distance from specific components in the BIM model, water surface absolute temperature, chlorophyll a concentration, turbidity, temperature at feature depth, Cl - concentration at feature depth, and SO4 2- concentration at feature depth, etc.

[0078] In order to perform principal component analysis and causal inspection, the ground control and intelligent analysis center collects this inspection data and historical inspection data, a total of ≥100 observation samples, constructs a structure-ecological correlation data table, and performs modeling. The following table is used as a demonstration, only showing part of the observation samples, of which observation sample 1 is the data of this inspection:

[0079] Principal component analysis for data dimensionality reduction and feature extraction: Z-score standardization is applied to the data.

[0080] Principal component analysis is performed on the 10 samples, and the eigenvalues are obtained: [5.8, 3.2, 1.5, 0.9,...]. The first three principal components with eigenvalues greater than 1 are selected, and the cumulative variance contribution rate is 88%.

[0081] Observation of the loading matrix: By analyzing the principal component loading matrix, the original variables with absolute load greater than 0.7 are defined as core representation variables, and the common characteristics of this set of core representation variables are used to assign physical meaning labels to the corresponding principal components; in this inspection, the principal component 1 has high load (>0.85) on crack width, Cl - concentration, SO4 2- concentration, which collectively reflects the characteristics of chemical substance migration caused by leakage, so the physical meaning label of this principal component is defined as leakage-chemical corrosion factor. Principal component 2 has high load (>0.82) on water table absolute temperature and 15.7m depth temperature, which collectively represent the thermodynamic anomalies of the water body, so its label is defined as thermal dynamic factor. Principal component 3 has high load (>0.88) on chlorophyll a, which directly indicates the level of biological activity in the water body, so its label is defined as biological activity factor.

[0082] Construction of causal network model and quantitative identification of key associated factors: a. Bayesian inference: In this embodiment, the Bayesian network uses a "three-node, three-state" naive Bayesian network, and the topology and variable definition are shown in the following table:

[0083] The edge structure is fixed as C→I and C→T, and reverse or circular dependence is prohibited, which meets the physical hypothesis of "structure defect driving ecological anomaly".

[0084] The prior probability P(C) directly references the dam inspection database in the past 5 years: P(C=no)=0.62, P(C=mild)=0.27, P(C=significant)=0.11.

[0085] The conditional probability table (CPT) is obtained by combining expert knowledge (from 3 senior engineers (water conservancy / structure / water environment) two-round Delphi review) and statistical frequency, as shown in the following table: The initial frequency is obtained from 10 samples, and 3 structure / water environment experts are invited to modify the frequency according to the Delphi method for 2 rounds, and the final CPT is shown in the following table, which is stored in the ground control and intelligent analysis center for automatic loading before each inspection by the system.

[0086]

[0087] Posterior probability calculation: observation evidence E = {I = yes, T = yes}, calculate for each state θ ∈ {none, slight, significant}:

[0088] After substituting the above table and prior and normalizing the results are as follows:

[0089] Conclusion: under the condition of observing chemical and temperature double anomalies, the posterior probability of "significant structural defects" is 65%, much higher than the rest of the states, meeting the key correlation factor judgment criteria. Therefore, it can be determined that significant structural defects are the most likely key correlation factor causing the ecological anomaly.

[0090] b. Granger causality test: Extract the underwater robot's crack opening micro-strain time series and synchronous water temperature time series at 0.5 meters above the crack for 1 hour of monitoring (both with 1-minute intervals, a total of 361 data points), and the first 50 data points are shown in the table below.

[0091]

[0092] Take water temperature as the dependent variable and crack opening as the independent variable to test whether crack opening is the Granger cause of water temperature change. The lag order is set to 2.

[0093] Result: the original hypothesis that crack opening is not the cause of water temperature change has a p-value of 0.0028.

[0094] Conclusion: at a confidence level of 99%, the original hypothesis is rejected, confirming that changes in crack opening are the Granger cause of changes in water temperature.

[0095] Causal network model: integrate both to form the final conclusion: structural defects (key correlation factor) with a high probability (65%) and high statistical confidence (p-value < 0.01) caused the chemical and temperature anomalies in the water body.

[0096] Automatically generate a comprehensive diagnosis report: The core content of the comprehensive diagnosis of this inspection includes: Title: Inspection and Diagnosis Report of a Large Reservoir Dam on X Month X Day in 2025.

[0097] Abstract: A 1.2m long crack was found in the spillway section (BIM coordinates: X=505105.123, Y=343205.456, Z=-15.75), quantitative analysis confirmed that it was the root cause of the abnormal temperature and ion concentration of the water above.

[0098] Defect details: attached crack image, size, accurate coordinates.

[0099] Causal analysis: show PCA result chart, Bayesian network chart, Granger test result table.

[0100] Repair suggestion: it is suggested to carry out epoxy resin pressure grouting treatment on the crack during dry season, and to carry out detailed exploration on the surrounding area.

[0101] The above is only a specific embodiment of the present application, but the design concept of the present application is not limited to this, and any non-essential modification of the present application using this concept shall be regarded as an infringement of the protection scope of the present application.

Claims

1. A dynamic cooperative control method for underwater robots that integrates multi-source data, characterized in that, Includes the following steps: Step S1: Control the inspection drone to fly along the preset route of the reservoir dam and adjacent waters to acquire laser point cloud and multispectral data of the reservoir dam and water surface; register the laser point cloud with the BIM model of the reservoir dam to construct a global spatial reference. Step S2: Analyze the multispectral data using an anomaly detection algorithm based on Mahalanobis distance to identify abnormal water surface areas; automatically generate a collaborative exploration task package containing the geographical location, range, type, and priority of the abnormal water surface areas; Step S3: After receiving the collaborative exploration task package, the underwater robot autonomously navigates to the abnormal water surface area specified in the collaborative exploration task package; when the underwater robot arrives at the abnormal water surface area, the underwater dam surface image corresponding to the abnormal water surface area captured by the underwater robot is visually matched and pose calculated with the reference image of the BIM model to achieve centimeter-level precise positioning; under the guidance of the centimeter-level precise positioning information, the underwater robot is controlled to conduct close-up observation of the underwater dam surface in the abnormal water surface area, and the defects found are photographed and measured from multiple angles, recording their precise position, length and width; Step S4: The water quality sampling drone receives vertical water quality profile data from the underwater robot in real time, automatically identifies the characteristic depth based on preset trigger rules, and collects water samples at that characteristic depth. Step S5: Collect the multivariate heterogeneous data from Steps S1 to S4, and use principal component analysis to reduce data dimensionality and extract features; combine Bayesian inference and Granger causality test methods to construct a causal network model between structural defect parameters and aquatic ecological anomaly parameters, quantitatively identify key correlation factors, and automatically generate a comprehensive diagnostic report.

2. The underwater robot dynamic cooperative control method according to claim 1, characterized in that, In step S1, the laser point cloud is registered with the BIM model of the reservoir dam to construct a global spatial reference, which specifically includes the following sub-steps: Step S1.1: Perform point cloud calculation and denoising on the laser point cloud; Step S1.2: Select several common feature points in the laser point cloud and BIM model; calculate the coordinate transformation matrix using these common feature points, and perform preliminary alignment of the laser point cloud and BIM model; Step S1.3: Find the nearest point pair between the laser point cloud and the triangular mesh surface of the BIM model, and calculate the optimal rigid body transformation iteratively using the least squares method to minimize the root mean square error between the two point sets, and finally establish a centimeter-level global spatial benchmark.

3. The underwater robot dynamic cooperative control method according to claim 1, characterized in that, In step S2, an anomaly detection algorithm based on Mahalanobis distance is used to analyze the multispectral data and identify abnormal water surface areas. This specifically includes the following sub-steps: Step S2.1: Perform aerial triangulation and orthophoto stitching on the multispectral raw image obtained in step S1 to generate reflectance images for each band; Step S2.2: For the thermal infrared band, the absolute temperature distribution map of the water surface is obtained by using the radiative transfer equation method and combining real-time meteorological data; the chlorophyll a concentration distribution map is generated by using the reflectance of the red and near-infrared bands through a variant algorithm of the normalized differential vegetation index; the turbidity distribution map is obtained by using the reflectance ratio of the blue and red bands according to empirical or semi-empirical models. Step S2.3: Read the three distribution maps. For each pixel with the same geographical location, construct a three-dimensional feature vector from the values ​​of its three bands, thus obtaining a three-dimensional vector set representing the entire inspection area. Randomly select a set of pixels from the three-dimensional vector set as training samples, calculate its mean vector and covariance matrix, and construct a statistical model to characterize the features of normal water bodies. The mean vector represents the average level of temperature, chlorophyll, and turbidity in the entire area, and the covariance matrix represents the relationship between these three parameters. Traverse each pixel in the image and calculate its Mahalanobis distance to the distribution defined by the statistical model. Pixels whose Mahalanobis distance exceeds the preset value are judged as abnormal pixels. Step S2.4: Perform DBSCAN spatial clustering on the abnormal pixels to obtain the abnormal water surface region; Extract detailed information about the abnormal water surface area from the datasets from steps S2.1 to S2.3 and package it into a task package.

4. The underwater robot dynamic cooperative control method according to claim 1, characterized in that, In step S3, multiple permanent magnet markers are pre-deployed on the underwater wall of the dam using a topology structure that combines equally spaced grids with key point encryption, forming a permanent magnet marker array; step S3 specifically includes the following sub-steps: Step S3.1: Deploy the underwater robot from the mother ship, and the mother ship's GNSS will provide the initial position; after the underwater robot enters the water, it mainly relies on the inertial navigation system and the Doppler log for combined navigation; Step S3.2: When the inertial navigation system instructs the underwater robot to navigate to a preset distance from the previous calibration point and approach a pre-placed magnetic marker, the three-axis magnetometer array begins to monitor the regular changes in the strength and direction of the spatial magnetic field and acquire a magnetic field vector map; the inertial navigation system matches the real-time measured magnetic field vector map with a pre-stored magnetic field grid map; through the least squares algorithm, it uniquely determines that the underwater robot is currently located near a permanent magnet marker; When a match is successful, the navigation system uses the precise coordinates of the permanent magnet marker recorded in the BIM model as a high-confidence observation to provide feedback correction to the system state of the inertial navigation system, thereby suppressing the cumulative error of the inertial navigation system.

5. The underwater robot dynamic cooperative control method according to claim 4, characterized in that, Step S3 specifically includes the following sub-steps: Step S3.3: When the underwater robot arrives at the abnormal water surface area, the vision-assisted navigation system is activated. The camera captures the underwater dam surface image corresponding to the abnormal water surface area and extracts multiple key points and their multi-dimensional feature descriptors. These multi-dimensional feature descriptors are matched with the reference image feature library pre-rendered in the area by the BIM model, and the coordinates and attitude of the current underwater robot are calculated by the EPnP algorithm to achieve centimeter-level precise positioning. Step S3.4: Guided by centimeter-level precise positioning information, the operator controls the underwater robot to conduct close-up observation of the underwater dam surface in the abnormal water surface area, and uses cameras and imaging sonar to take pictures and measure the defects found from multiple angles, recording their precise location, length, width and other information.

6. The underwater robot dynamic cooperative control method according to claim 1, characterized in that, In step S4, the feature depth is automatically identified based on the preset triggering rules. Specifically, the vertical water quality profile data is analyzed, and when a depth layer with an absolute temperature gradient value ≥ 8.0 ℃ / m or a chlorophyll concentration mutation rate exceeding 30% is detected, the depth layer is determined as the feature depth that needs to be sampled.

7. The underwater robot dynamic cooperative control method according to claim 1, characterized in that, Step S5 includes the following sub-steps: Step S5.1: Collect the laser point cloud, abnormal water surface area information, underwater structural defect parameters and location information, water quality profile data and stratified sampling laboratory analysis results from steps S1 to S4, and construct a structure-ecology correlation data table under a unified spatiotemporal reference. Step S5.2: After standardizing the structure-ecology association data table, perform principal component analysis and select principal components with eigenvalues ​​greater than 1 and cumulative variance contribution rate ≥ 85%; by analyzing the principal component loading matrix, define the original variables with absolute loading values ​​greater than 0.7 as core characterization variables, and assign physical meaning labels to the corresponding principal components based on the commonalities of this group of core characterization variables. Step S5.3: Based on the principal components obtained in Step S5.2, and the structural semantic information and historical maintenance records extracted from the BIM model of the reservoir dam, a Bayesian network structure is initialized. Observational evidence from the structure-ecology association data table is input into this Bayesian network structure. By calculating the posterior probability, the probability of different structural defects leading to specific ecological anomalies is quantitatively assessed, and key association factors are initially identified. For the variables with preliminary associations, their time series data are extracted, and Granger causality tests are performed. When the p-value of the test result is less than 0.05, a statistically significant causal relationship is determined, thereby determining the causal direction. By integrating the probabilistic association strength of the Bayesian inference and the causal direction of the Granger causality test, a quantitative structural defect-aquatic ecological anomaly causal network model is constructed to complete the quantitative identification of key association factors. Step S5.4: Based on the analysis results of the causal network model, automatically generate a comprehensive diagnostic report. The report includes at least the precise location of the defect in the BIM model, the causal mechanism analysis based on causal relationships, a list of quantitatively identified key related factors, and targeted maintenance strategy recommendations.

8. A dynamic cooperative control system for underwater robots that integrates multi-source data, used to implement the dynamic cooperative control method for underwater robots that integrates multi-source data as described in any one of claims 1-7, characterized in that, include: An inspection drone is used to perform steps S1 and S2. An underwater robot is used to perform step S3; A water quality sampling drone is used to perform step S4; The ground control and intelligent analysis center is used to coordinate the control of the inspection drone, underwater robot and water quality sampling drone, and to execute step S5.

9. The system according to claim 8, characterized in that, It also includes a permanent magnet marker array for underwater robot navigation correction, the permanent magnet marker array consisting of several permanent magnet markers fixed to the underwater wall of the dam, and the spatial coordinates of each permanent magnet marker are recorded in the BIM model.

Images