Satellite interferometric measurement and remote sensing image-based reservoir dam identification method, device, equipment, medium and product

Abstract

The application discloses a reservoir dam identification method and device based on satellite interferometric height measurement and remote sensing images, equipment, medium and product, and relates to the field of image recognition. The method comprises the following steps: acquiring water surface elevation data; preprocessing the water surface elevation data to obtain longitude-latitude-height sequence, and determining dam potential point longitude-latitude coordinate information based on a preset elevation mutation threshold; performing vectorization processing on a CSV coordinate file, and performing image interception, buffer zone generation and band selection processing based on satellite remote sensing images to determine an image file; inputting the image file into a dam body identification model, and outputting a reservoir dam prediction point to realize the identification of the reservoir dam; the dam body identification model is determined by using a Bayesian optimization algorithm, iteratively optimizing the hyperparameters of an improved YOLO11 model according to the image file with labeled information. The application aims to efficiently and reliably realize automatic dam identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image recognition, in particular to a reservoir dam identification method and device based on satellite interferometric height measurement and remote sensing images, equipment, medium and product. BACKGROUND

[0002] In the core scenarios of water resources regulation, flood control and disaster reduction, and basin ecological protection, dams as key water conservancy infrastructure, their accurate identification and dynamic monitoring are the core prerequisite for realizing fine management of the basin. The known technology has the following problems:

[0003] There is a significant gap in the dam database, and the identification of new dams and small dams is seriously insufficient:

[0004] The relevant mainstream dam data set generally has structural defects of "emphasizing history and ignoring new additions" and "emphasizing large dams and ignoring small dams". Although some databases cover thousands of large dams, the data sources are highly dependent on the registration information of regional water agencies, and new dams are delayed in construction information registration, and some regional data is not transparent. After supplementing the identification through remote sensing technology, the number of new additions can be increased several times, which shows that the relevant database is insufficient in covering new dams. More importantly, the lack of small-sized dams is more prominent: although some databases include tens of thousands of dams, through statistical models, the actual number of small dams that meet the basic monitoring standards is much higher than the existing records. These data gaps result in the fact that in the management of the basin, the hydrological regulation effect and flood risk of a large number of newly added dams and small dams cannot be included in the evaluation system, greatly increasing the difficulty of decision-making for water resources regulation and flood control and disaster reduction in the basin.

[0005] Traditional identification technology relies on manual and low-precision data, with low efficiency and accuracy:

[0006] The related dam identification technology mainly relies on two paths, both of which have obvious limitations. One is manual traversal of satellite images for interpretation. The coordinates of some dam data sets need to be obtained by professional personnel through frame-by-frame interpretation of 1°x1° satellite image patches, and manual subjective judgment is easily disturbed by ground objects, resulting in missed detection of small dams and hidden dams. The second is to lock the candidate area based on low-precision hydrological data, such as early research through 1km grid precipitation data and rough basin boundaries to lock potential dam areas. The candidate range often covers a large area, resulting in a surge in platform resource consumption to deal with massive irrelevant information during subsequent image processing and data storage.

[0007] Satellite image utilization and model optimization are insufficient, and the adaptability to complex scenarios is poor:

[0008] In the application level of satellite images, the related technologies generally use RGB (red, green, blue) bands to carry out identification, without combining the spectral feature differences of the dam and the surrounding ground objects to optimize the band combination. Related research shows that the RGB band is easy to confuse the dam body with the surrounding mountain shadow and vegetation coverage area in the dense vegetation area. In the model performance level, the problem of weak feature extraction ability of small-sized dam bodies has not been solved, and the identification accuracy of the basic identification model for small dams is generally low, which is difficult to meet the demand of "full-size dam coverage" in actual basin monitoring.

[0009] The large basin verification is missing, and the performance of the model is seriously degraded:

[0010] The related dam identification technology depends on small sample sets and simple scenes for verification, and lacks large-scale testing links across regions and multiple topographies. The verification set data of some data sets covers a small area and is concentrated in simple terrain scenes such as plains and irrigation areas with less ground object interference, without covering complex topographies such as mountain gorges and high-altitude basins. In actual application, the model in the mountainous area has a low reasoning speed due to large elevation difference and complex ground object types, and the false detection rate increases due to terrain obstruction and ground object confusion.

[0011] In summary, the three shortcomings of "database integrity", "identification efficiency" and "complex scene adaptability" lead to the fact that dam identification cannot meet the actual demand of "full-size, full-area and high-precision" in global basin management. Therefore, it is crucial to break through the core bottlenecks of new dam and small dam identification, large candidate area range and low precision in complex scenes, and provide an efficient and reliable technical path for large-scale automatic dam identification. SUMMARY

[0012] The purpose of the present application is to provide a reservoir dam identification method, device, equipment, medium and product based on satellite interferometric height and remote sensing image, which can efficiently and reliably realize automatic dam identification.

[0013] To achieve the above purpose, the present application provides the following solutions:

[0014] In a first aspect, the present application provides a reservoir dam identification method based on satellite interferometric height and remote sensing image, comprising:

[0015] Obtaining water surface elevation data; the water surface elevation data is the water body elevation data of each observation point at the basin scale obtained by SWOT satellite using satellite radar interferometric measurement technology;

[0016] Pretreating the water surface elevation data to obtain latitude-longitude-height sequence, and determining the latitude and longitude coordinate information of the dam potential point based on a preset elevation mutation threshold;

[0017] The CSV coordinate file is subjected to vectorization processing, and image interception, buffer zone generation and band selection processing based on satellite remote sensing images are performed to determine an image file; the image file includes information of dam number, satellite type and band combination; the band combination adopts a near-infrared enhancement combination; the CSV coordinate file is obtained by storing the latitude and longitude coordinate information of the dam potential point in the CSV format in a coordinate data standardization manner;

[0018] The image file is input into a dam body recognition model to output a reservoir dam prediction point to realize recognition of the reservoir dam; the dam body recognition model is determined by iteratively optimizing the hyperparameters of an improved YOLO11 model according to an image file with labeled information by using a Bayesian optimization algorithm; the labeled information includes dam point; the improved YOLO11 model is integrated with an ACmix attention module, a BiFPN feature network and a reparameterization detection head structure; wherein the ACmix attention module is used to enhance the feature extraction capability of the dam linear edge and the gray difference; the BiFPN feature network is used to realize the fusion of multi-scale features and improve the detection accuracy; the reparameterization detection head structure is used to improve the regression accuracy of the dam body and the model convergence stability.

[0019] In an embodiment, the expression corresponding to the water body elevation data is:

[0020] ;

[0021] Wherein, is the water body elevation data; is the initial elevation data inverted from the radar signal; is the atmospheric delay error correction amount; is the surface roughness error correction amount; is the satellite orbit deviation error correction amount.

[0022] In an embodiment, the water surface elevation data is preprocessed to obtain a latitude-longitude-elevation sequence, and the latitude and longitude coordinate information of the dam potential point is determined based on a preset elevation mutation threshold, specifically including:

[0023] The water surface elevation data is subjected to interference elimination processing based on the vector boundary of the target research area to cut off the non-target water body data outside the basin, and the interference-eliminated water surface elevation data is obtained;

[0024] Based on the interference-eliminated water surface elevation data, the discrete elevation observation points are sorted according to the water flow direction to form a latitude-longitude-elevation sequence;

[0025] For the latitude-longitude-elevation sequence, the latitude and longitude coordinate information of the dam potential point is determined based on a preset elevation mutation threshold.

[0026] In an embodiment, for the latitude-longitude-height sequence, the latitude and longitude coordinate information of the dam potential point is determined based on a preset elevation mutation threshold, specifically including:

[0027] The preset elevation mutation threshold is determined based on the dam water level difference ;

[0028] For the latitude-longitude-height sequence, when the elevation difference between adjacent elevation observation points along the water flow direction is greater than the elevation mutation threshold , the center point of the line connecting the two observation points is taken as a suspected sudden drop point;

[0029] The spatial continuity of the suspected sudden drop point is verified to determine the latitude and longitude coordinate information of the dam potential point; wherein, when the elevation difference between the three observation points before and after the suspected sudden drop point is not greater than the elevation mutation threshold , the single-point mutation caused by natural water drop of the river channel or rock blockage is excluded, and the suspected sudden drop point is determined as a dam potential point.

[0030] In an embodiment, the CSV coordinate file is vectorized and processed, and image interception, buffer zone generation and band selection processing based on satellite remote sensing images are performed to determine the image file, specifically including:

[0031] The CSV coordinate file is vectorized and processed, and a buffer radius is set according to the dam body scale and the resolution of the satellite remote sensing image with each potential dam coordinate point as the center to generate a clipping boundary;

[0032] Based on the Sentinel-2 SR Harmonized multi-spectral dataset, the satellite remote sensing image under the condition of no cloud or little cloud is selected by setting the time range and cloud cover threshold with the dam center coordinate as the spatial constraint, and the near-infrared enhanced image is obtained by using near-infrared enhancement combination for visual enhancement processing;

[0033] The selected near-infrared enhanced image is clipped according to the clipping boundary to remove the redundant area outside the boundary, and the image file is determined.

[0034] In an embodiment, it further includes:

[0035] The dam point recognition model is verified for dam point recognition.

[0036] In a second aspect, the application provides a reservoir dam identification device based on satellite interferometric altimetry and remote sensing images, comprising:

[0037] A data acquisition module for acquiring water surface elevation data; the water surface elevation data is the water body elevation data of each observation point at the basin scale based on the SWOT satellite using satellite radar interferometric measurement technology;

[0038] The processing module is configured to preprocess the water surface elevation data to obtain a latitude-longitude-elevation sequence, and determine latitude and longitude coordinate information of a dam potential point based on a preset elevation mutation threshold.

[0039] The image acquisition module is configured to perform vectorization processing on the CSV coordinate file, and perform image interception, buffer zone generation, and band selection processing based on satellite remote sensing images to determine an image file. The image file includes information of a dam number, a satellite type, and a band combination. The band combination adopts a near-infrared enhancement combination. The CSV coordinate file is obtained by storing the latitude and longitude coordinate information of the dam potential point in a CSV format.

[0040] The recognition module is configured to input the image file into a dam body recognition model to output a reservoir dam prediction point to realize recognition of the reservoir dam. The dam body recognition model is determined by iteratively optimizing hyperparameters of an improved YOLO11 model based on image files with labeled information using a Bayesian optimization algorithm. The labeled information includes dam point information. The improved YOLO11 model integrates an ACmix attention module, a BiFPN feature network, and a reparameterization detection head structure. The ACmix attention module is used to enhance the feature extraction capability of dam linear edges and gray difference. The BiFPN feature network is used to realize multi-scale feature fusion to improve detection accuracy. The reparameterization detection head structure is used to improve the regression accuracy of the dam body and the stability of model convergence.

[0041] In a third aspect, the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the above-mentioned satellite interferometric altimetry and remote sensing image-based reservoir dam recognition method.

[0042] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon. The computer program is executed by a processor to implement the above-mentioned satellite interferometric altimetry and remote sensing image-based reservoir dam recognition method.

[0043] In a fifth aspect, the present application provides a computer program product comprising a computer program. The computer program is executed by a processor to implement the above-mentioned satellite interferometric altimetry and remote sensing image-based reservoir dam recognition method.

[0044] According to the embodiments provided in the present application, the following technical effects are disclosed:

[0045] The application provides a reservoir dam identification method and device based on satellite interferometric height measurement and remote sensing images, a medium and product, pre-processes obtained water surface elevation data to obtain longitude-latitude-height sequence, and determines dam potential point longitude and latitude coordinate information based on a preset elevation mutation threshold. In other words, the water surface elevation data ensures centimeter-level precision, based on the set of high-precision data, by analyzing mutation points in the river height sequence, i.e., the longitude-latitude-height sequence, the water level drop phenomenon is automatically detected to inversely lock the dam position. Then, the CSV coordinate file is vectorized and processed, and image interception, buffer zone generation and band selection processing based on satellite remote sensing images are performed to determine the image file. In combination with the near-infrared enhanced band combination, the spectral difference between the dam body and the vegetation and water body can be strengthened, and the separability and training accuracy can be improved. The image file is input into the dam body identification model, and the reservoir dam prediction point is output to realize the identification of the reservoir dam and improve the identification robustness of small dams and complex backgrounds. Therefore, the application can efficiently and reliably realize automatic dam identification. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0047] Figure 1 The flowchart of the reservoir dam identification method based on satellite interferometric height measurement and remote sensing images;

[0048] Figure 2 The dam identification flowchart;

[0049] Figure 3 The RGB combination and near-infrared enhanced combination satellite image comparison chart; wherein, Figure 3 (a) in the figure is an RGB satellite image schematic diagram of different scenes; Figure 3 (b) in the figure is a near-infrared enhanced combination image schematic diagram of different scenes;

[0050] Figure 4 The YOLO11 model structure schematic diagram;

[0051] Figure 5 The structure diagram of the reservoir dam identification device based on satellite interferometric height measurement and remote sensing images;

[0052] Figure 6 The structure schematic diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0053] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the scope of protection of the present application.

[0054] The present application breaks through the core bottlenecks of difficult identification of new and small dams, large range of candidate area and low precision in complex scenes by integrating high-resolution satellite elevation data (i.e. water surface elevation data), accurate image interception technology and optimized identification model (i.e. dam body identification model), and provides an efficient and reliable technical path for large-scale and large-scale automatic dam identification. The present application accurately obtains the water surface elevation data of the basin by satellite interferometric measurement technology, and realizes the automatic positioning of potential dams by using the water level mutation feature. This method breaks through the limitations of traditional manual image screening or low-precision hydrological data positioning, and realizes the efficiency and automation of dam location identification.

[0055] After positioning, the image interception, buffer zone generation and band selection are automatically executed based on the positioning coordinates using the Google Earth Engine platform, and the near-infrared enhanced combination (B8+B4+B3) is selected to realize the strong expression of dam body features. This band combination significantly improves the discrimination between dam body and surrounding water and vegetation through strong contrast between “bright vegetation” and “dark dam body”, thereby effectively solving the problem of dam body features being covered in complex terrain.

[0056] In the model training link, the present application uses an improved YOLO11 deep learning model, which integrates ACmix attention module, BiFPN feature network and reparameterization detection head structure to solve the problems of small target missed detection and weak feature transmission; further, the Bayesian optimization algorithm is used to intelligently optimize the hyperparameters, thereby improving the model performance stability and generalization ability.

[0057] The present application builds an integrated identification system from water level data acquisition-automatic positioning-image interception-band optimization-model training-wide range verification, which provides a feasible technical solution for basin water resource management, flood control and dispatching and infrastructure monitoring.

[0058] Automatic dam positioning method based on high-precision water elevation data: Interferometric measurement technology of SWOT satellite is used to obtain water elevation data at basin scale. These data are corrected by atmospheric delay, surface roughness and orbit deviation to ensure centimeter-level accuracy. Based on this set of high-precision data, the algorithm automatically detects the "water level drop" phenomenon by analyzing the mutation points in the river elevation sequence, and reversely locates the dam position. At the same time, the algorithm effectively excludes natural waterfalls and riverbed disturbances through threshold judgment and spatial continuity verification, realizes the accurate positioning of potential dams, and provides a high-reliability solution for dam positioning.

[0059] Targeted multi-band image extraction and optimization processing: Google Earth Engine (GEE) platform is used to realize automatic extraction and standardized storage of dam images. Near-infrared enhanced band combination is used to strengthen the spectral difference between dam body and vegetation and water body, improve the separability of the model and the training accuracy.

[0060] Dam body recognition model construction based on improved YOLO11: In the YOLO11 framework, several innovative modules are integrated to improve the recognition robustness of small dams and complex backgrounds.

[0061] Intelligent super parameter optimization and large basin verification mechanism: Bayesian optimization algorithm is used to realize intelligent optimization of model training parameters, avoiding the randomness of manual parameter tuning, and finally forming a complete dam body recognition model.

[0062] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0063] In an exemplary embodiment, as shown in Figure 1 a satellite interferometric measurement and remote sensing image-based reservoir dam identification method is provided, comprising:

[0064] Step 100: Obtain water surface elevation data. The water surface elevation data is the water elevation data at each observation point at basin scale obtained by satellite radar interferometric measurement technology based on SWOT satellite.

[0065] The expression corresponding to the water elevation data is:

[0066] .

[0067] wherein, is the water elevation data; is the radar signal inversion initial elevation data; is the atmospheric delay error correction; is the surface roughness error correction; Satellite orbit deviation error correction quantity.

[0068] Step 200: preprocessing the water surface elevation data to obtain a longitude-latitude-elevation sequence, and determining the longitude and latitude coordinate information of the dam potential point based on a preset elevation mutation threshold.

[0069] In the method, the water surface elevation data is preprocessed to obtain a longitude-latitude-elevation sequence, and the longitude and latitude coordinate information of the dam potential point is determined based on a preset elevation mutation threshold, specifically including:

[0070] The water surface elevation data is subjected to interference elimination processing based on the vector boundary of the target research area to eliminate non-target water body data outside the basin, so as to obtain the water surface elevation data after interference elimination.

[0071] Based on the water surface elevation data after interference elimination, the discrete elevation observation points are sorted according to the water flow direction to form a longitude-latitude-elevation sequence.

[0072] For the longitude-latitude-elevation sequence, the longitude and latitude coordinate information of the dam potential point is determined based on a preset elevation mutation threshold.

[0073] As an optional implementation, for the longitude-latitude-elevation sequence, the longitude and latitude coordinate information of the dam potential point is determined based on a preset elevation mutation threshold, specifically including:

[0074] Determine the preset elevation mutation threshold based on the dam water level difference For the longitude-latitude-elevation sequence, when the elevation difference between adjacent elevation observation points along the water flow direction is greater than the elevation mutation threshold , the center point of the line connecting the two observation points is regarded as a suspected sudden drop point; the suspected sudden drop point is subjected to spatial continuity verification to determine the longitude and latitude coordinate information of the dam potential point; when the three observation points before and after the suspected sudden drop point do not appear the case that the elevation difference is greater than the elevation mutation threshold , the single-point mutation caused by natural water drop of the river channel or rock blockage is excluded, and the dam potential point is determined.

[0075] Step 300: vectorizing the CSV coordinate file, and performing image interception, buffer generation and band selection processing based on satellite remote sensing images to determine an image file. The image file includes information of dam number, satellite type and band combination; the band combination adopts near-infrared enhancement combination; the CSV coordinate file is obtained by storing the longitude and latitude coordinate information of the dam potential point in a coordinate data standardization manner in a comma-separated value (CSV) format.

[0076] Wherein, the CSV coordinate file is vectorized, and image interception, buffer zone generation and band selection processing based on satellite remote sensing image are performed to determine the image file, specifically including:

[0077] The CSV coordinate file is vectorized, and a clipping boundary is generated according to the dam scale and the buffer zone radius of the satellite remote sensing image resolution set with each potential dam coordinate point as the center.

[0078] Based on the Sentinel-2 SR Harmonized multi-spectral dataset, the satellite remote sensing image under the condition of no cloud or little cloud is selected by setting the dam center coordinate as the spatial constraint, setting the time range and cloud amount threshold, and using near-infrared enhancement combination for visual enhancement processing to obtain the selected near-infrared enhancement image.

[0079] The selected near-infrared enhancement image is cropped according to the clipping boundary to remove the redundant area outside the boundary and determine the image file.

[0080] Step 400: input the image file into the dam body recognition model, and output the reservoir dam prediction point to realize the identification of the reservoir dam. The dam body recognition model is determined by using the Bayesian optimization algorithm, iterating the super parameters of the improved YOLO11 model according to the image file with labeled information; the labeled information includes dam point; the improved YOLO11 model integrates ACmix attention module, BiFPN feature network and reparameterization detection head structure; wherein, the ACmix attention module is used to enhance the feature extraction ability of dam linear edge and gray difference; the BiFPN feature network is used to realize the fusion of multi-scale features and improve the detection accuracy; the reparameterization detection head structure is used to improve the regression accuracy of the dam body and the stability of the model convergence.

[0081] As an optional implementation, the reservoir dam identification method based on satellite interferometric measurement and remote sensing image further includes: performing dam point identification verification on the dam body recognition model.

[0082] The present application surrounds the demand for accurate dam identification, and constructs a whole process solution from data acquisition, positioning, modeling to verification, and the specific steps are as shown in Figure 2 Figure 2 The dam detection model in the above means the dam body recognition model.

[0083] Step 1, obtain water surface elevation data and process.

[0084] ​The present application systematically obtains SWOT (Surface Water and Ocean Topography) satellite remote sensing series data through the standardized data acquisition channel of the NASA website. This data system not only covers the core characteristics of the water level height and spatial distribution range of a large range of natural water bodies (such as rivers, lakes, reservoirs, etc.), but also accurately reflects the dynamic change law of the water body with its centimeter-level precision observation data, providing a basis for the subsequent exploration of potential dam locations.

[0085] The SWOT satellite relies on advanced radar interferometry technology to accurately invert the water elevation of each observation point (x, y) at the basin scale. The basic principle is centered on radar altimetry: the satellite-borne dual-frequency radar transmits microwave signals to the surface water body, and the initial elevation is calculated by the round-trip time of the received signal. Subsequently, a series of targeted correction factors are introduced to eliminate various errors in the radar measurement process one by one, including atmospheric delay error (ionospheric and tropospheric interference with microwave signal propagation), surface roughness error (influence of water surface ripples on signal reflection), satellite orbit deviation error (measurement deviation caused by slight deviation of satellite flight trajectory), etc. Finally, high-reliability water elevation data is output, and the calculation formula is as follows:

[0086] .

[0087] The water surface elevation data obtained based on the SWOT satellite at the basin scale needs to be preprocessed to eliminate interference and regularize the data to lay a foundation for subsequent detection: first, the vector boundary of the study area is combined to crop the basin range, and the data of non-target water bodies (such as isolated lakes) outside the basin is removed, and only the elevation observation points of the main channel of the river and the associated water area are retained; then the discrete points are sorted according to the flow direction to form a continuous sequence of latitude-longitude-elevation, and the direction reference for subsequent sudden drop detection is clarified.

[0088] Step 2, obtain the latitude and longitude of the potential dam point through the elevation data.

[0089] Based on the water surface elevation data obtained at the basin scale by the SWOT satellite, the latitude and longitude of the potential dam can be locked by capturing the "water level sudden drop point". The core logic is that the dam, as an artificial water retaining structure, will cause a significant water level difference between its upstream and downstream, which is manifested as a "sudden drop feature" in the elevation data along the flow direction, i.e. there is a mutation in the elevation values of adjacent observation points in the same river axis direction, and the mutation point is most likely to correspond to the dam location.

[0090] For the latitude-longitude-elevation sequence after preprocessing, the latitude and longitude coordinates of the potential dam are locked through multiple steps of detection: first, set the elevation mutation threshold (Reference to the global typical dam water level difference, small dam 1m-5m, medium dam 5m-15m, large dam >15m, combined with the basin size adjustment), when the elevation difference between adjacent observation points along the flow direction is greater than the elevation threshold of the sudden drop, the center point of the line connecting the two observation points is the "suspected sudden drop point"; the next step is to verify the spatial continuity, which requires that the three observation points before and after the suspected sudden drop point do not appear the case of elevation difference greater than . This operation can exclude the single-point mutation caused by natural water drop and rock blockage.

[0091] Step 3, intercept potential point remote sensing satellite image.

[0092] After completing the identification of the latitude and longitude coordinates of potential dams, CSV format is used for coordinate data standardization storage to adapt to the subsequent data import and management requirements. The CSV file needs to set the core field and ensure that it matches the accuracy of the SWOT satellite elevation data to avoid deviation of the subsequent image interception position due to insufficient accuracy; additional dam type size field can be supplemented to provide basis for subsequent buffer setting, and all field names need to be simple and without special symbols to facilitate direct reading and analysis on GEE platform.

[0093] Based on the above CSV coordinate file, the core process of satellite image interception of dams on the GEE platform is divided into four steps:

[0094] First step: coordinate file import and vectorization.

[0095] Upload the generated CSV file in the "Assets" module of the GEE interface, and specify the longitude column and latitude column as geometric fields. GEE will automatically convert table data into a vector layer with geographic coordinates. At the same time, the attribute information in the CSV will be retained as feature attributes, realizing one-to-one correspondence between each dam point and its unique number.

[0096] Second step: buffer and square boundary setting.

[0097] Set the buffer radius according to the dam size and satellite image resolution, and generate an approximately square cutting boundary with each potential dam coordinate point as the center. This boundary should cover the dam body and key areas upstream and downstream (such as reservoirs and spillways), ensuring complete inclusion of the dam body while avoiding the inclusion of too much irrelevant background.

[0098] Third step: image screening and band processing (determine scheme).

[0099] In the image data screening stage, based on the Sentinel-2 SR Harmonized multispectral dataset, the application sets the dam center coordinates as the spatial constraint, sets the time range and cloud cover threshold, screens high-quality images under the condition of no cloud or little cloud, and automatically removes residual clouds and mists through the cloud mask function to ensure the clarity and consistency of the input images.

[0100] In terms of band selection, combined with the spectral characteristics and identification requirements of the dam scene, the near-infrared enhanced combination (B8+B4+B3) is used as the main band combination. This combination can form a significant spectral contrast between water, dam body and vegetation. The near-infrared band (B8) is strongly reflected by vegetation and weakly reflected by water, which can effectively highlight the feature distribution of "bright vegetation, dark dam body, and darker water", and fundamentally reduce the interference of vegetation shielding on dam body identification. At the same time, the introduction of red band (B4) and green band (B3) further enhances the boundary clarity of dam body and background objects, making the dam body outline more prominent in the image. Compared with other schemes such as true color combination (RGB), the near-infrared enhanced combination has stronger discrimination ability and universality in river valley type or high vegetation coverage areas, can stably highlight the dam structure features in complex topographic environment, and maintain good visual consistency and automatic identification performance under different seasons and lighting conditions. The comparison diagram of near-infrared enhanced combination and RGB combination is shown in Figure 3 . Among them, Figure 3 (a) of FIG. 1 is a schematic diagram of RGB satellite images of different scenes; Figure 3 (b) of FIG. 1 is a schematic diagram of near-infrared enhanced combination images of different scenes.

[0101] Step 4: Image cropping and export.

[0102] The screened near-infrared enhanced images are cropped according to the square boundary to remove the redundant area outside the boundary. When exporting, the dam number, satellite type and band combination information need to be reflected in the file name to ensure data traceability and batch management convenience. The resolution of the exported image is consistent with the original accuracy of the selected satellite, and the projection coordinate system adopts World Geodetic System 1984, WGS84 (EPSG: 4326) to ensure spatial reference consistency. The finally exported image file will be automatically saved to the specified Google Drive folder and can be directly used for subsequent dam sample labeling, feature extraction and identification model training.

[0103] Step 4: Dam dataset establishment and optimization.

[0104] It covers four major links of "image labeling, basic training, model improvement, and hyperparameter optimization", forming a full link from data to optimized model, as follows:

[0105] (1) Dam multi-band satellite image annotation.

[0106] Based on the multi-band satellite images obtained from GEE, open-source annotation tools supporting YOLO format output such as LabelImg and LabelStudio are used. All image sizes are unified to 640x640 pixels to match the default input size of YOLO11, and the training set, validation set, and test set are randomly divided in a ratio of 8:1:1. During annotation, only the "dam body" is targeted, and the "minimum bounding rectangle" principle is used for boxing to ensure that the box completely covers the dam body without excessive redundant background. The annotation file follows the YOLO standard format (one.txt file corresponds to one image, and the content includes "class index, normalized x-coordinate of center, normalized y-coordinate of center, normalized width, and normalized height"). After annotation, "manual sampling inspection + tool verification" is used to ensure quality, and low-quality images with cloud cover exceeding 30% or blurred dam bodies are excluded. Finally, a complete dam annotation dataset is formed.

[0107] (2) Basic training of dam recognition model based on YOLO11.

[0108] YOLO11 is a multiple iteration version of the single-stage real-time target detection algorithm. Its core principle is to convert the target detection task into a single regression problem. Through one-time neural network forward propagation, it directly predicts the bounding box coordinates, class probability, and other information on the input image, without generating candidate regions for classification like two-stage algorithms (such as Faster R-CNN), thus having higher inference efficiency.

[0109] The YOLO11 network structure is composed of Backbone (backbone network), Neck (neck network), and Head (detection head): it uses an improved backbone and neck architecture, enhancing feature extraction capability and improving object detection accuracy and complex task performance. Compared with the YOLOv8 model, it changes the CF2 module to C3K2, adds a C2PSA module after the SPPF module, and introduces the head idea of YOLOv10 into the head of YOLO11, using a deep separable method to improve efficiency. The YOLO11 network structure diagram is shown in Figure 4 .

[0110] During training, a.yaml configuration file is created to specify the dataset path and "dam" single-class information, and "image-annotation file" is stored in the standard directory structure. During the training process, the following core indicators are used to evaluate the model performance:

[0111] Precision is used to measure the proportion of actual positive examples among the samples predicted as positive examples by the model, reflecting the model's ability to avoid misidentification:

[0112] .

[0113] where TP (True Positives) represents correctly identified dams, and FP (False Positives) represents misidentified non-dam targets.

[0114] Recall is used to measure the proportion of actual positive samples successfully identified by the model, reflecting the model's ability to avoid missed identification:

[0115] .

[0116] where FN (False Negatives) represents missed dams.

[0117] mAP50 refers to the average precision (AP) of all classes when the intersection over union (IoU) threshold is 0.5:

[0118] .

[0119] where N is the total number of classes, and IoU is a core indicator for measuring the overlap between the predicted box and the true box in target detection, with a value range of [0, 1]. The closer the value is to 1, the higher the overlap degree and the more accurate the positioning. is the average precision of the i th class when IoU = 0.5, calculated by integrating the precision of different recall rates for this class:

[0120] .

[0121] R is the recall rate, and the integration interval is 0 to 1.

[0122] mAP50-95 is the average of mAP at each threshold from 0.5 to 0.95 (interval 0.05, total 10 thresholds: 0.5, 0.55,..., 0.95):

[0123] .

[0124] where is the mAP value when the IoU threshold is t, which comprehensively reflects the model's performance under different positioning accuracy requirements.

[0125] (3) Targeted improvements for YOLO11 model.

[0126] After the model-based training is completed, in view of the characteristics of high resolution of satellite images and much interference of ground objects, the YOLO11 model is improved in multiple modules to improve the dam identification performance: the ACmix module (i.e. ACmix attention module) is introduced on the basis of the original C2f structure of YOLO11, the convolution features and attention features are fused to enhance the feature extraction capability of the dam linear edge and gray difference; the original PANet structure is replaced by BiFPN (bi-directional feature pyramid network), i.e. BiFPN feature network, to realize efficient fusion of multi-scale features and significantly improve the detection accuracy of small size dams; the RepHead structure is used in the detection head part, and the SIoU loss function and dynamic task alignment mechanism (Dynamic Task Alignment) are combined to improve the regression accuracy of rectangular dams and the convergence stability of the model while keeping the model lightweight. The improved combination takes into account the detection accuracy and real-time performance, and can realize stable identification of dams of different scales in complex water environment.

[0127] (4) Bayesian hyperparameter intelligent optimization.

[0128] The Bayesian optimization algorithm is used to iteratively optimize the key hyperparameters, and the specific steps of the algorithm are as follows:

[0129] ① Initialization: In the hyperparameter search space, a small number of initial parameter combinations are randomly selected for experiments to obtain initial "parameter-performance" data pairs.

[0130] ② Constructing a probability model: using a Gaussian process or a tree Parzen estimator model, a probability model is established for the mapping relationship between "hyperparameters and model performance", and the mean and variance of the objective function at any parameter point are output.

[0131] ③ Selecting a collection function: through the collection function (such as expected improvement EI, probability improvement PI, and upper confidence bound UCB), the "exploration" and "exploitation" are balanced to generate the next round of hyperparameters to be tested.

[0132] ④ Iterative update: use the selected hyperparameters for experiments, add new "parameter-performance" data to the data set, and update the probability model again, repeat steps ②-③ until the preset number of iterations or performance convergence is reached.

[0133] The parameters to be optimized include learning rate, batch size, regularization coefficient, and data augmentation strength, and the mAP@0.5:0.95 of the validation set is used as the optimization target, and the optimal hyperparameter combination is finally selected, and the optimized YOLO11 dam identification model is obtained based on the combination.

[0134] Step 5, dam point identification verification in large-scale range.

[0135] (1) Large basin satellite image verification set construction.

[0136] To ensure that the optimized model can work stably in real large basin scenarios, satellite images of typical large basins (such as the middle reaches of the Yangtze River basin and the Amazon River basin) are selected for verification: high-resolution satellite images covering an area of not less than 100,000 km 2 are obtained through Google Earth Engine (consistent with the training set data source), the optimal band combination (B8+B4+B3) is used for processing, the areas already used in the training set are excluded to avoid data duplication, and the dam targets in the images are manually labeled to construct a large dam-labeled basin verification set, covering samples of different scales, different dam types (concrete dams, earth-rock dams, and rockfill dams), and different environments (reservoir areas, river valleys, and plain irrigation areas); the optimized YOLO11 model is deployed to the inference platform to batch detect the verification set images, output the dam location (labeling box), class (“dam”), and confidence (threshold 0.5), and evaluate the performance from three dimensions: overall detection accuracy, dam type adaptability, and environmental robustness; attribution analysis is conducted for the missed and false samples, and finally the optimized YOLO11 model for large basin-scale dam automatic identification is formed, providing technical support for basin water resources management, flood control and disaster reduction, etc.

[0137] (2) Auxiliary verification with public data sets.

[0138] This auxiliary verification selects four high-credibility public data sets covering global and key basins in China, supplementing verification samples of different dam types and scenarios. The specific information is as follows:

[0139] Georeferenced global Dams And Reservoirs (GeoDAR): 24783 geocoded dam points, 21515 reservoir polygons, covering 155 countries, more than 60% of small and medium-sized dams, providing dam height, reservoir capacity and other attributes; GDAT (Global Dam tracker.pdf): 35140 dam records, 31780 geocoded, covering six continents, recording construction year, function, etc., emerging dams in Asia and South America are highlighted; Global Georeferenced Database of Dams (GOODD): 38667 manually digitized dam points, focusing on low-latitude areas, providing upstream catchment area attributes, high coverage of small and medium-sized dams in Southeast Asia and Africa; China's Large Dams, Reservoirs, and Lakes (China-LDRL): 2418 large dams in China (624 off-river dams, 1794 river dams), 2194 reservoirs, and 3051 lakes, high coordinate accuracy, distinguish dam type and reservoir boundary.

[0140] From the above 4 datasets, a total of 4 verification samples were selected: covering large river basins in the satellite image verification set in the previous step. Through the SWOT data output dam latitude and longitude coordinate points, and through GEE to obtain the corresponding position of all coordinate points Sentinel-2 near-infrared enhanced band image, uniform size 640x640 pixels, cloud cover control≤10%, consistent with the training data preprocessing process. After model prediction, the dam point position is obtained. Compare this point with the data set, take the data set dam coordinates as the reference, calculate the Euclidean distance between the model detection box center and the data set coordinates, and the deviation≤50m (adapt to the resolution of Sentinel-2) is judged as the position is effective. This method can verify the model and enhance the credibility of the dam point data.

[0141] The above dam identification method based on SWOT data, GEE image cutting and YOLO11 optimization, compared with the prior art, the core advantages are reflected in the positioning efficiency, training accuracy and application adaptability of three dimensions:

[0142] I. Based on the water level drop point, the dam is quickly positioned, and the resource consumption is greatly reduced.

[0143] The prior art relies on manual traversal of large basin images to find dams, or uses low-precision hydrological data to delimit candidate regions, which has the problems of low positioning efficiency, serious waste of resources, and easy omission of small dams due to subjective judgment; the range of the candidate region delimited by low-precision data is too large, resulting in a large amount of irrelevant data needing to be processed in subsequent image interception and labeling. The application directly captures the water level drop feature of the upstream and downstream of the dam through the high-precision water surface elevation data of the SWOT satellite, which is equivalent to accurately delimiting the region for subsequent work: on the one hand, only the images around the coordinates need to be intercepted, avoiding indiscriminate processing of the entire basin and reducing the resource consumption of GEE image acquisition and storage; on the other hand, the positioning process does not require manual intervention, greatly improving the positioning efficiency in the early stage and saving time and cost for subsequent labeling and training.

[0144] II. Multi-band combination comparison experiment to ensure optimal accuracy of model training.

[0145] Most existing dam identification models directly use default RGB band images for training without considering the feature differences of multi-band satellite images, which limits the recognition accuracy of the model in complex scenes (such as densely vegetated areas and reservoir areas). For example, the RGB band is easily confused with the surrounding mountains in the densely vegetated mountainous area. The application uses a "near-infrared enhanced combination (B8+B4+B3)". This band combination can fundamentally solve the problem of vegetation obstruction through strong contrast between "bright vegetation and dark dam body"; in addition, this combination can distinguish between vegetation, water body, and dam body through the dual role of "near-infrared distinguishing vegetation and red / green distinguishing water body and dam body", which can stably identify the dam body, avoiding the color confusion problem of visible light combination and the embankment misjudgment problem of short-wave infrared combination, and avoiding the interference sensitivity problem of multispectral fusion; at the same time, the 3-band combination achieves the best balance between "effective feature coverage" and "data efficiency".

[0146] This targeted band selection approach avoids the limitations of existing technologies that use a one-size-fits-all approach to bands, allowing the model to fully utilize the multispectral information of satellite images and improve training accuracy from the source, laying a foundation for high robustness in subsequent large basin verification.

[0147] III. Model improvement combined with basin verification to strengthen practical application value.

[0148] The prior art usually stops at model training of a small range of data sets, does not optimize the model for the special needs of dam identification (such as small size dam missed detection, complex topographic interference), and lacks actual verification of large basin scale, resulting in a significant performance degradation of the model when landing. The present application integrates ACmix attention module, BiFPN feature network and reparameterization detection head structure in YOLO11, specifically solves the problems of small target missed detection and weak feature transmission, and realizes intelligent optimization of hyperparameters through Bayesian optimization to avoid the blindness of manual parameter tuning. More importantly, through large basin verification, it ensures that the model can work stably in different scenes such as mountains, plains and reservoir areas, rather than only performing excellently in small sample sets.

[0149] This "optimization-verification-iteration" closed loop not only has theoretical precision, but also can adapt to the complex environment of real basins, providing a practical technical solution for basin water resource management, flood control and disaster reduction, and making up for the short board of the prior art "high laboratory precision, poor practical application".

[0150] In an exemplary embodiment, as shown in Figure 5 a satellite interferometric height measurement and remote sensing image-based reservoir dam identification device is provided, comprising:

[0151] A data acquisition module is configured to acquire water surface elevation data. The water surface elevation data is water body elevation data of each observation point at a basin scale, which is obtained by a satellite radar interferometric measurement technology based on a SWOT satellite.

[0152] A processing module is configured to preprocess the water surface elevation data to obtain a latitude-longitude-elevation sequence, and determine latitude and longitude coordinate information of a dam potential point based on a preset elevation mutation threshold.

[0153] An image acquisition module is configured to perform vectorization processing on a CSV coordinate file, and perform image interception, buffer generation and band selection processing based on satellite remote sensing images to determine an image file. The image file includes information of dam number, satellite type and band combination. The band combination adopts a near-infrared enhancement combination. The CSV coordinate file is obtained by storing the latitude and longitude coordinate information of the dam potential point in a CSV format.

[0154] The identification module is configured to input the image file into a dam body identification model, and output a reservoir dam prediction point to realize identification of the reservoir dam. The dam body identification model is determined by using a Bayesian optimization algorithm, according to an image file with labeled information, and performing hyperparameter iterative optimization on an improved YOLO11 model. The labeled information includes a dam point. The improved YOLO11 model is integrated with an ACmix attention module, a BiFPN feature network and a reparameterization detection head structure. The ACmix attention module is configured to enhance the feature extraction capability of the dam linear edge and the gray difference. The BiFPN feature network is configured to realize fusion of multi-scale features and improve detection accuracy. The reparameterization detection head structure is configured to improve the regression accuracy of the dam body and the model convergence stability.

[0155] In an exemplary embodiment, a computer device, which can be a server or a terminal, is provided, and an internal structure diagram of the computer device can be as shown in Figure 6 The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store reservoir dam identification data based on satellite interferometric height and remote sensing image. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through network connection. The computer program is executed by the processor to implement the reservoir dam identification method based on satellite interferometric height and remote sensing image.

[0156] Those skilled in the art can understand that Figure 6 the structure shown in the above

[0157] In an exemplary embodiment, a computer device is also provided, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in each of the above method embodiments.

[0158] In an exemplary embodiment, a computer readable storage medium storing a computer program is provided, the computer program, when executed by a processor, implements the steps of any of the above method embodiments.

[0159] In an exemplary embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps of any of the above method embodiments.

[0160] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0161] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (Resistive Random Access Memory, ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc.

[0162] The database involved in each of the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, and the like, without being limited thereto. The processor involved in each of the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device, and the like, without being limited thereto.

[0163] The technical features of the above embodiments can be combined in any manner. To make the description concise, all possible combinations of the technical features in the above embodiments are not described, but as long as there is no contradiction, any combination of the technical features should be considered within the scope of the present disclosure.

[0164] The principles and implementation manners of the present application are described by using specific examples herein, and the above embodiments are only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, the specific implementation manners and application ranges can be changed according to the idea of the present application. In conclusion, the content of the present description should not be understood as a limitation of the present application.

Claims

1. A method for identifying reservoir dams based on satellite interferometric altimetry and remote sensing imagery, characterized in that, include: Obtain water surface elevation data; The water surface elevation data is based on the water surface elevation data of each observation point at the watershed scale obtained by the SWOT satellite using satellite radar interferometry technology. The water surface elevation data is preprocessed to obtain a latitude-longitude sequence, and the latitude-longitude coordinates of potential dam points are determined based on a preset elevation change threshold. The CSV coordinate file is vectorized, and image cropping, buffer generation, and band filtering are performed based on satellite remote sensing imagery to determine the image file. The image file includes information such as the dam number, satellite type, and band combination. The band combination adopts near-infrared enhancement combination. The CSV coordinate file is obtained by standardizing and storing the latitude and longitude coordinates of the potential dam points in CSV format. The image files are input into the dam identification model, which outputs predicted dam locations to achieve dam identification. The dam identification model is determined by using a Bayesian optimization algorithm and performing hyperparameter iteration optimization on an improved YOLO11 model based on image files with labeled information. The labeled information includes dam locations. The improved YOLO11 model integrates an ACmix attention module, a BiFPN feature network, and a reparameterized detection head structure. The ACmix attention module enhances the feature extraction capability of linear edges and grayscale differences of the dam body. The BiFPN feature network is used to achieve multi-scale feature fusion and improve detection accuracy. The reparameterized detection head structure is used to improve the regression accuracy and model convergence stability of the dam body. The CSV coordinate file is vectorized, and image cropping, buffer generation, and band filtering are performed based on satellite remote sensing imagery to determine the image file. Specifically, this includes: The CSV coordinate file is vectorized, and a clipping boundary is generated by setting the buffer radius based on the dam size and the resolution of the satellite remote sensing image, with each potential dam coordinate point as the center. Based on the Sentinel-2 SR Harmonized multispectral dataset, with the dam center coordinates as spatial constraints, a time range and cloud cover threshold were set to filter satellite remote sensing images under cloudless or low-cloud conditions, and near-infrared enhancement was performed for visual enhancement to obtain the filtered near-infrared enhanced images. The near-infrared enhanced images after screening are cropped according to the cropping boundaries, and redundant areas outside the boundaries are removed to determine the image file.

2. The method for identifying reservoir dams based on satellite interferometric altimetry and remote sensing imagery according to claim 1, characterized in that, The expression corresponding to the water surface elevation data is: ； in, This refers to water surface elevation data; Initial elevation data for radar signal inversion; This is the atmospheric delay error correction amount; This is the correction amount for surface roughness error; This is the correction amount for satellite orbit deviation error.

3. The method for identifying reservoir dams based on satellite interferometric altimetry and remote sensing imagery according to claim 1, characterized in that, The water surface elevation data is preprocessed to obtain a latitude-longitude sequence, and the latitude-longitude coordinates of potential dam points are determined based on a preset elevation change threshold. Specifically, this includes: The watershed range is clipped based on the vector boundary of the target study area to perform interference removal processing on the water surface elevation data, and non-target water body data outside the watershed are removed to obtain the water surface elevation data after interference removal. Based on the water surface elevation data after interference elimination, the discrete elevation observation points are sorted according to the direction of water flow to form a latitude-longitude sequence; For latitude-longitude sequence, the latitude and longitude coordinates of potential dam points are determined based on a preset elevation change threshold.

4. The method for identifying reservoir dams based on satellite interferometric altimetry and remote sensing imagery according to claim 3, characterized in that, For latitude-longitude-height sequences, the latitude and longitude coordinates of potential dam points are determined based on a preset elevation change threshold, specifically including: Determine the preset elevation change threshold based on the dam water level difference. ; For latitude-longitude-height series, when the elevation difference between adjacent elevation observation points along the direction of water flow is greater than the elevation change threshold... At that time, the midpoint of the line connecting the two observation points is taken as the suspected point of sudden drop; Spatial continuity verification is performed on the suspected sudden drop points to determine the latitude and longitude coordinates of potential dam points; wherein, if the elevation difference between the three observation points before and after the suspected sudden drop point does not exceed the elevation change threshold, then the dam is considered to have a potential dam point. In such cases, single-point abrupt changes caused by natural waterfalls or reef obstruction in the river channel are excluded, and these are identified as potential dam points.

5. The method for identifying reservoir dams based on satellite interferometric altimetry and remote sensing imagery according to claim 1, characterized in that, Also includes: The dam identification model was verified by identifying dam points.

6. A reservoir dam identification device based on satellite interferometric altimetry and remote sensing imagery, characterized in that, include: The data acquisition module is used to acquire water surface elevation data; The water surface elevation data is based on the water surface elevation data of each observation point at the watershed scale obtained by the SWOT satellite using satellite radar interferometry technology. The processing module is used to preprocess the water surface elevation data to obtain a latitude-longitude sequence, and determine the latitude-longitude coordinate information of potential dam points based on a preset elevation change threshold. The image acquisition module is used to vectorize the CSV coordinate file and perform image cropping, buffer generation, and band filtering based on satellite remote sensing imagery to determine the image file. The image file includes information such as the dam number, satellite type, and band combination. The band combination adopts near-infrared enhancement combination. The CSV coordinate file is obtained by standardizing and storing the latitude and longitude coordinates of the potential dam points in CSV format. The recognition module is used to input the image file into the dam recognition model and output the predicted location of the reservoir dam to achieve the recognition of the reservoir dam. The dam recognition model is determined by using a Bayesian optimization algorithm and performing hyperparameter iteration optimization on an improved YOLO11 model based on the image file with labeled information. The labeled information includes the dam location. The improved YOLO11 model integrates an ACmix attention module, a BiFPN feature network, and a reparameterized detection head structure. Among them, the ACmix attention module is used to enhance the feature extraction capability of linear edges and gray-level differences of the dam body; the BiFPN feature network is used to achieve multi-scale feature fusion and improve detection accuracy; the reparameterized detection head structure is used to improve the regression accuracy of the dam body and the model convergence stability. The CSV coordinate file is vectorized, and image cropping, buffer generation, and band filtering are performed based on satellite remote sensing imagery to determine the image file. Specifically, this includes: The CSV coordinate file is vectorized, and a clipping boundary is generated by setting the buffer radius based on the dam size and the resolution of the satellite remote sensing image, with each potential dam coordinate point as the center. Based on the Sentinel-2 SR Harmonized multispectral dataset, with the dam center coordinates as spatial constraints, a time range and cloud cover threshold were set to filter satellite remote sensing images under cloudless or low-cloud conditions, and near-infrared enhancement was performed for visual enhancement to obtain the filtered near-infrared enhanced images. The near-infrared enhanced images after screening are cropped according to the cropping boundaries, and redundant areas outside the boundaries are removed to determine the image file.

7. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the reservoir dam identification method based on satellite interferometric altimetry and remote sensing imagery as described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the reservoir dam identification method based on satellite interferometric altimetry and remote sensing imagery as described in any one of claims 1-5.

9. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the reservoir dam identification method based on satellite interferometric altimetry and remote sensing imagery as described in any one of claims 1-5.

Images