Remote Sensing Monitoring Classification System for Ecological Characteristics of Seawalls and Construction Method of Sample Library

Abstract

The present application relates to the technical field of data processing, and discloses a method for constructing a remote sensing monitoring classification system and an example library for the ecological characteristics of seawalls. The method includes: fusing and processing multi-source remote sensing images to obtain a standardized data set; extracting the characteristic parameters of the seawall cross-section to generate preliminary type distribution data; fusing geographical and ecological information to form characteristic element data; calculating the vegetation coverage and material characteristics to obtain quantitative indicators; establishing a three-level example library structure; updating the examples through similarity matching to construct a classification system. The present application realizes the standardized expression and intelligent classification of the remote sensing monitoring of the ecological characteristics of seawalls.

Description

Technical Field

[0001] This application relates to the field of data processing, and particularly to a remote sensing monitoring classification system for the ecological characteristics of seawalls and a method for constructing a sample library. Background Art

[0002] As an important coastal protection engineering facility, the ecological construction of seawalls is of great significance for maintaining the stability of the coastal ecosystem. At present, the investigation and evaluation of the ecological characteristics of seawalls mainly rely on the combination of on-site manual investigation and remote sensing image interpretation. By collecting data such as the basic information of seawalls, vegetation coverage, slope protection structure characteristics, and surrounding ecological environment, a basic database of seawalls is established. In terms of data processing, technical means such as GIS spatial analysis and remote sensing image processing are used to classify, count, and evaluate and analyze the spatial distribution, structure type, and ecological characteristics of seawalls, providing decision-making support for the ecological construction and management of seawalls.

[0003] However, the existing methods for analyzing the ecological characteristics of seawalls have the following deficiencies: First, the data acquisition method is single, making it difficult to comprehensively reflect the multi-dimensional characteristics of seawalls; second, the data processing process lacks standardized and automated means, with a large amount of manual interpretation work and strong subjectivity; third, the quantitative evaluation of ecological characteristics lacks a systematic index system, making it difficult to conduct objective comparisons; finally, there is a lack of support from a complete sample library, and no effective reference basis can be provided for newly built and renovated seawalls. Summary of the Invention

[0004] This application provides a remote sensing monitoring classification system for the ecological characteristics of seawalls and a method for constructing a sample library, which is used to realize the standardized expression and intelligent classification of the remote sensing monitoring of the ecological characteristics of seawalls.

[0005] In a first aspect, the present application provides a method for constructing a remote sensing monitoring classification system and an example library for the ecological characteristics of seawalls. The method for constructing the remote sensing monitoring classification system and the example library for the ecological characteristics of seawalls includes: performing multi-source data fusion processing on high-resolution satellite remote sensing images, aerial remote sensing images, unmanned aerial vehicle orthophotos, and on-site measurement data, obtaining a digital orthophoto map through aerial triangulation processing, and performing plane position and elevation rectification on the digital orthophoto map according to the control point coordinates to obtain a standardized multi-source data set; extracting the slope value, height value, and width value of the seawall section from the standardized multi-source data set, substituting the obtained characteristic parameter values into a preset expert rule for type recognition, and generating preliminary seawall type distribution data; using the preliminary seawall type distribution data to perform characteristic correlation processing on basic geographic information, crest ecological information, slope protection structure information, and beach ecosystem information to form seawall ecological characteristic element data; calculating the vegetation coverage of the multi-spectral image based on the seawall ecological characteristic element data, determining the crest greening type of the image texture, and performing scoring operations according to the slope protection material and ecological permeability to obtain quantitative characteristic indicators; performing standardized storage coding and multi-dimensional index construction on the example data according to the three-level structure based on the quantitative characteristic indicators to generate a hierarchical seawall example library; performing similarity matching operations on the spatial characteristics, spectral characteristics, and structural characteristics in the hierarchical seawall example library, and performing example update through dynamic feedback processing to construct a classification system for the ecological characteristics of seawalls.

[0006] In the technical solution provided by the present application, through the fusion processing of multi-source remote sensing data and aerial triangulation technology, the accuracy and integrity of the extraction of seawall characteristic information are improved, and the information loss problem caused by a single data source is reduced; through standardized cross-section characteristic extraction and expert rule discrimination, automatic identification of seawall types is realized, and the classification efficiency is improved; the feature correlation processing technology is adopted to perform multi-dimensional fusion of basic geographic information, ecological information, structural information, and environmental information, comprehensively reflecting the ecological characteristics of the seawall; through quantitative vegetation coverage calculation, material characteristic analysis, and ecological scoring, an objective evaluation index system is established, eliminating the uncertainty of subjective judgment; based on the three-level structure example library design, standardized storage and efficient retrieval of seawall characteristic information are realized, providing a reliable reference basis for subsequent seawall construction; through similarity matching and dynamic feedback mechanisms, continuous update and optimization of the example library are ensured, and the adaptability and accuracy of the classification system are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0008] Figure 1 It is a schematic diagram of an embodiment of the remote sensing monitoring classification system and sample library construction method for the ecological characteristics of sea dikes in the embodiments of this application. Specific implementation manners

[0009] The embodiments of this application provide a remote sensing monitoring classification system and a sample library construction method for the ecological characteristics of sea dikes. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned drawings of this application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order other than those illustrated or described here. In addition, the term "including" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0010] For ease of understanding, the following describes the specific process of the embodiments of this application. Please refer to Figure 1 An embodiment of the remote sensing monitoring classification system and sample library construction method for the ecological characteristics of sea dikes in the embodiments of this application includes:

[0011] Step S101: Perform multi-source data fusion processing on high-resolution satellite remote sensing images, aerial remote sensing images, unmanned aerial vehicle orthophoto images, and on-site measurement data. Obtain a digital orthophoto image through aerial triangulation processing, and perform plane position and elevation correction on the digital orthophoto image according to the control point coordinates to obtain a standardized multi-source data set;

[0012] Step S102: Extract the slope value, height value, and width value of the sea dike cross-section according to the standardized multi-source data set, substitute the obtained characteristic parameter values into the preset expert rules for type recognition, and generate preliminary sea dike type distribution data;

[0013] Step S103: Use the preliminary sea dike type distribution data to perform feature association processing on basic geographic information, ecological information on the dike top, slope protection structure information, and beach ecosystem information to form sea dike ecological characteristic element data;

[0014] Step S104: Calculate the vegetation coverage of the multi-spectral image based on the ecological characteristic element data of the seawall, determine the greening type of the dike top according to the image texture, and perform a scoring operation based on the slope protection material and ecological permeability to obtain quantitative characteristic indicators;

[0015] Step S105: Standardize the storage coding and construct multi-dimensional indexes for the sample data according to the three-level structure based on the quantitative characteristic indicators, and generate a hierarchical seawall sample library;

[0016] Step S106: Perform similarity matching operations on the spatial characteristics, spectral characteristics, and structural characteristics in the hierarchical seawall sample library, update the samples through dynamic feedback processing, and construct a classification system for the ecological characteristics of the seawall.

[0017] Specifically, it should be noted that in this application, the implementation process of multi-source data fusion processing for high-resolution satellite remote sensing images, aerial remote sensing images, unmanned aerial vehicle orthophoto images, and on-site measurement data. In the specific implementation process, the multi-source remote sensing monitoring data not only includes high-resolution satellite remote sensing images, aerial remote sensing images, unmanned aerial vehicle orthophoto images, and on-site measurement data, but may also include: panchromatic and multi-spectral fusion images, hyperspectral remote sensing images, lidar point cloud data, synthetic aperture radar images, thermal infrared remote sensing images, ground real-time monitoring station data, and three-dimensional laser scanning data, etc.

[0018] Among them, the high-resolution satellite remote sensing image (spatial resolution better than 2 meters) is subjected to band resampling, and the multi-spectral band data is fused through Gram-schmidt mathematical transformation. The Gram-schmidt transformation orthogonalizes multiple bands according to their correlation and retains the independent information between bands. For example, the near-infrared band and the red band are orthogonally transformed to highlight vegetation information. The fused data is registered with the aerial remote sensing image (spatial resolution better than 1 meter) according to the geographical coordinates. By selecting obvious ground feature points such as road intersections and building corners as homologous points, precise alignment is achieved using the affine transformation and bilinear interpolation methods. The registered image is fused with the unmanned aerial vehicle orthophoto image to enhance the local area detail expression. The unmanned aerial vehicle image first needs to perform aerial triangulation. The connection between the images is established through the matching of connection points, and the exterior orientation elements are solved by bundle adjustment. During this process, ground control points are selected for constraint, and the coordinates of the control points are obtained by RTK measurement. For example, control points are arranged at the inflection points and characteristic points of the seawall, and their three-dimensional coordinates are measured by RTK. The orthophoto image is corrected based on the coordinates of the control points to eliminate the geometric deformation caused by terrain undulation and lens distortion.

[0019] After data fusion and geometric correction, a standardized multi-source dataset is obtained, which contains spatial geometric information and spectral information. Based on this dataset, the cross-section features of the seawall are extracted, edge detection and region segmentation are performed on the seawall image, and the seawall contour is extracted. The seawall contour data is overlaid with the cross-section point cloud data measured on-site, and the three-dimensional coordinates of each point on the cross-section are calculated through spatial interpolation. Profile analysis is carried out in the direction perpendicular to the coastline, and the slope value is obtained by calculating the ratio of the elevation difference between adjacent points to the horizontal distance, the height value is obtained from the elevation difference between the top and bottom of the seawall, and the width value is obtained from the horizontal distance between the points on both sides of the top road.

[0020] The extracted cross-section feature parameters are matched with the expert rules. The expert rules specify the threshold values of the feature parameters for different types of seawalls. For example, the slope of a slope-type seawall is usually between 15° and 45°, and a composite seawall includes a vertical section and a slope section. By comparing the feature parameters with the rule thresholds, the seawalls are classified into two major categories: slope type and composite type. Further combined with the spatial topological relationship, a preliminary distribution layer of the seawalls is generated.

[0021] Based on the preliminary classification results, a multi-dimensional feature correlation analysis of the seawalls is carried out. The seawall trajectory line is overlaid with the geographic base map to extract basic geographic information such as the longitude and latitude coordinates of the start and end points, the projected length, and the trend angle. Spectral analysis is performed on the top area of the seawall to extract ecological element information such as vegetation and flower beds. The surface material of the slope protection is identified through texture analysis, such as structural types like fence panels, tetrapods, and hollow square blocks at the corners. The ecosystem types of the surrounding areas of the seawall are identified, including mangroves, bedrock beaches, sandy beaches, muddy beaches, and salt marshes. Quantitative evaluation is carried out based on the feature correlation analysis. The normalized difference vegetation index (NDVI) is calculated for the multi-spectral image, and the vegetation coverage is quantified by setting classification thresholds. Texture features are extracted from the top area of the seawall, the edge density and direction consistency are calculated, and the greening types such as green belts and flower beds are discriminated. Based on the material characteristics of the slope protection, its ecological permeability score is calculated, and the spatial distribution of the surrounding ecosystem is statistically analyzed to form a quantitative feature index system.

[0022] When establishing a hierarchical sample library, a three-level organizational structure is adopted. The first level is the seawall type (sloping type, composite type), the second level is the specific structural form (such as slope-platform-slope type, straight-up-and-sloping type, etc.), and the third level is the combination of ecological characteristics. A unique identification code is assigned to each sample, and multimedia materials such as multi-angle images and point cloud data are associated with the identification code. A multi-dimensional retrieval mechanism is established to support queries according to conditions such as geographical location, type characteristics, and ecological indicators. Finally, dynamic update of samples is achieved through similarity matching. Morphological analysis is carried out on the spatial characteristics in the sample library to calculate the contour curvature and continuity parameters. Band combination analysis is carried out on the spectral characteristics to calculate the vegetation index and material reflectivity. Texture analysis is carried out on the structural characteristics to evaluate the matching degree of the slope protection type and the embankment top morphology. The similarity of multi-dimensional characteristics is normalized and then comprehensively scored. Based on the scoring results, the samples are clustered and grouped, and the characteristic statistical values of each group are extracted to form a classification system.

[0023] For example: For a certain coastal zone area, high-resolution multi-spectral images of the GF-2 satellite, aerial remote sensing DOM images, and UAV oblique photography data are obtained. The multi-spectral bands are fused through Gram-schmidt transformation, and 20 ground control points are selected for geometric correction. The extraction of the seawall cross-section characteristics shows that the average slope of this section of the seawall is 35°, the height is 4.5 meters, and the top width is 6 meters, which conforms to the characteristics of the sloping seawall. Further analysis reveals that there is a continuous green belt on the embankment top, the slope protection adopts a structure of concrete blocks plus ecological blocks, and mangroves and sandy beaches are distributed around. After quantitative evaluation, the vegetation coverage is 0.6, the ecological score of the slope protection is 0.7, and the proportion of the ecosystem distribution is 0.4. This sample is encoded and stored in the library.

[0024] In the embodiments of the present application, through the fusion processing of multi-source remote sensing data and aerial triangulation technology, the accuracy and integrity of the extraction of seawall characteristic information are improved, and the information missing problem caused by a single data source is reduced; through standardized cross-section characteristic extraction and expert rule discrimination, automatic identification of seawall types is realized, and the classification efficiency is improved; the feature correlation processing technology is adopted to conduct multi-dimensional fusion of basic geographic information, ecological information, structural information, and environmental information, comprehensively reflecting the ecological characteristics of the seawall; through quantitative calculation of vegetation coverage, material characteristic analysis, and ecological scoring, an objective evaluation index system is established, eliminating the uncertainty of subjective judgment; based on the design of the sample library with a three-level structure, standardized storage and efficient retrieval of seawall characteristic information are realized, providing a reliable reference basis for subsequent seawall construction; through similarity matching and dynamic feedback mechanisms, continuous update and optimization of the sample library are ensured, and the adaptability and accuracy of the classification system are improved.

[0025] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0026] (1) Resample the bands of the high - resolution satellite remote sensing image, and input the resampled band data into the Gram - schmidt mathematical transformation to obtain spectral fusion data;

[0027] (2) Register the spectral fusion data and the aerial remote sensing image according to the geographical coordinate correspondence to obtain the first registered image;

[0028] (3) Perform affine transformation on the first registered image and the UAV orthophoto image according to the homologous ground feature points, and obtain the second registered image through bilinear interpolation;

[0029] (4) Perform image stretching, equalization and filtering enhancement processing on the second registered image through on - site measurement data to generate enhanced image data;

[0030] (5) Input the enhanced image data into the aerial triangulation program, and obtain the digital orthophoto map through connection point matching and bundle adjustment operations;

[0031] (6) Substitute the control point coordinates into the rectification equation, perform plane position correction and elevation correction calculations on the digital orthophoto map, and output a standardized multi - source data set.

[0032] Specifically, perform band resampling processing on the high - resolution satellite remote sensing image. The high - resolution remote sensing image contains multiple spectral bands, such as the blue - light band (0.45 - 0.52μm), the green - light band (0.52 - 0.60μm), the red - light band (0.63 - 0.69μm) and the near - infrared band (0.76 - 0.90μm). The resampling process unifies the data of each band to the same spatial resolution, and uses the Gram - schmidt orthogonal transformation for spectral fusion. The key of the Gram - schmidt transformation is to gradually orthogonalize multiple bands according to their correlation and retain the independent information between the bands. For example, perform an orthogonal transformation on the near - infrared band and the red - light band, which particularly highlights the vegetation information and helps to identify the greening situation on the seawall in the subsequent process.

[0033] Subsequently, the spectral fusion data is spatially registered with the aerial remote sensing image. The registration process is based on the geographical coordinate correspondence and selects ground control points as references. Ground control points are usually selected at obvious feature locations such as the corners of seawalls and the boundaries of slope protection structures, and their precise three-dimensional coordinates are obtained by RTK measurement. During registration, an affine transformation model is used to transform the spectral fusion image into the same geographical coordinate system as the aerial remote sensing image. The registration accuracy requirement is controlled within one pixel to ensure that the two types of image data can be precisely superimposed. After obtaining the first registered image, it is necessary to further register it with the UAV orthophoto image. The UAV image has a higher spatial resolution but a relatively smaller coverage area and is mainly used to obtain local detailed features of the seawall. During registration, first, homologous ground points are selected in the two types of images, such as the feature points of the slope protection structure and the intersection points of the top-of-dike roads. Based on these homologous points, an affine transformation relationship is established, and the bilinear interpolation method is used to resample the image to obtain the second registered image with precise spatial correspondence.

[0034] The second registered image also needs to be subjected to image enhancement processing. The enhancement processing includes three key steps: image stretching, histogram equalization, and filtering enhancement. Image stretching adjusts the gray value range of the image to improve image contrast; histogram equalization improves the overall brightness distribution of the image; filtering enhancement highlights edge details through high-pass filtering, which is beneficial for subsequent extraction of seawall structure features. The parameter settings of the enhancement processing need to be verified and optimized in combination with on-site measurement data. The enhanced image data is input into the aerial triangulation program for processing. The core of the aerial triangulation processing is to establish the connection between photos through the matching of connection points, and then the exterior orientation elements are solved by bundle adjustment. The connection points are usually automatically extracted in the overlapping area of the images, and it is required that the point positions are evenly distributed and have obvious features. The bundle adjustment comprehensively considers the constraints of ground control points and the geometric relationship between photos, and obtains the optimal orientation parameters through iterative calculation. The digital orthophoto map output by this process eliminates the geometric deformation caused by terrain undulation and camera tilt.

[0035] The digital orthophoto map is precisely rectified using the control point coordinates, that is, the rectification equation is as follows:

[0036] ;

[0037] where: is the rectification displacement; , , are the rectification coefficients in each direction; , , are the coordinates of the measurement points; , , are the coordinates of the reference points; is the weight factor. The coordinates of each pixel point are corrected according to the actual geographical location to eliminate systematic errors and distortions, and a standardized multi-source dataset is output.

[0038] For example: First, obtain high-resolution satellite multi-spectral images and panchromatic images, perform band resampling to unify the spatial resolution to 2 meters, and then fuse the multi-spectral bands through Gram-schmidt transformation. Select 20 control points such as road intersections and building corners on the fused image and register them with the aerial remote sensing DOM image. The registered image is then registered with the orthophoto image of the UAV oblique photography, and 30 corresponding points such as the edge of the dike top road and the boundary of the slope protection structure are accurately corresponded. When performing image enhancement, the gray value range is extended to 0-255 through linear stretching, the histogram distribution is made more uniform through equalization processing, and the high-pass filter highlights the contour edges. In the aerial triangulation process, 200 connection points are automatically extracted, and bundle adjustment is performed in combination with 25 ground control points. Substitute into the rectification equation to achieve centimeter-level plane position and elevation accuracy.

[0039] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0040] (1) Perform edge detection on the seawall image in the standardized multi-source dataset, segment the seawall area through region growing, and obtain the seawall contour data;

[0041] (2) Superimpose the seawall contour data on the on-site measurement data, calculate the three-dimensional coordinates of each cross-section point through spatial interpolation, and obtain the seawall cross-section point cloud;

[0042] (3) Intercept the seawall cross-section point cloud in the direction perpendicular to the coastline, divide the height difference between adjacent cross-section points by the horizontal distance, and obtain the slope value;

[0043] (4) Extract the elevation difference between the top and bottom points of the seawall from the seawall cross-section point cloud as the height value, and extract the horizontal distance between the points on both sides of the top road of the seawall as the width value;

[0044] (5) Substitute the slope value, height value, and width value into the hierarchical threshold discriminant formula, classify them into two types of seawalls: slope type and composite type, and obtain the seawall type discrimination result;

[0045] (6) Associate the seawall type discrimination result with the geographical coordinates through spatial topological relations, and output the preliminary seawall type distribution data.

[0046] Specifically, edge detection processing is performed on the seawall image. The multi-scale gradient operator is used for edge detection to extract the boundary information between the seawall and the background at different resolutions. The gradient operator calculates the gray-scale change rate of the image in the horizontal and vertical directions, and pixels with gradient values greater than the set threshold are marked as edge points. After obtaining the preliminary edge point set, the seawall area is segmented through the region growing algorithm. The region growing starts from the preselected seed points and gradually merges adjacent similar pixels into the current region. The similarity judgment is based on multiple attributes such as pixel gray-scale values and texture features. This process is iterated until no new pixels that meet the conditions can be found, and the seawall contour data is obtained.

[0047] The seawall contour data is spatially superimposed with the cross-section point data obtained from on-site RTK measurements. Each RTK measurement point contains accurate three-dimensional coordinate information. Through the Kriging spatial interpolation method, the discrete measurement points are extended into a continuous cross-section point cloud. Kriging interpolation takes into account the spatial autocorrelation and estimates the three-dimensional coordinates of unknown points based on the position relationships and attribute values of known points. In this way, dense and uniformly distributed point cloud data can be obtained on each cross-section. After obtaining the seawall cross-section point cloud, the cross-section extraction direction is determined. By analyzing the coastline trend, its normal direction is determined as the cross-section extraction benchmark. In the direction perpendicular to the coastline, cross-sections are extracted at fixed intervals. For the adjacent point cloud data on each cross-section, the ratio of the elevation difference to the horizontal distance is calculated to obtain the slope value of this section. Since the seawall cross-section often shows a stepped feature, it is necessary to calculate the slope in segments and record the start and end positions of each segment.

[0048] When extracting feature parameters from the seawall cross-section point cloud, first, the top and bottom feature points are identified. The top points usually correspond to the edge of the dike top road and have obvious elevation mutation features; the bottom points are located at the junction of the seawall and the beach surface and also show a drastic elevation change. The elevation difference between these two feature points is the seawall height value. For the extraction of the dike top width, the horizontal distance is calculated by identifying the edge points on both sides of the road.

[0049] For seawall type discrimination, the following hierarchical threshold discriminant formula is used in the discrimination process:

[0050] ;

[0051] Where: is the discrimination index; is the slope angle of the i-th segment; is the seawall height value; is the dike top width value; is the structural feature index of the i-th segment; is the weight coefficient of each feature; is the segment correction coefficient; n is the number of cross-section segments.

[0052] When the discrimination index J is greater than the set threshold, it is determined as a composite seawall; otherwise, it is determined as a slope seawall. The discrimination result needs to be associated with the geographic coordinate information to establish a spatial topological relationship. Through the spatial indexing technology, the type attribute of each section of the seawall is bound to its geographical location to form the seawall type distribution data.

[0053] For example: When processing a certain coastal section, first extract the seawall image from the standardized multi-source dataset. Identify the approximate outline of the seawall through edge detection, and then perform region growing based on the pre-selected seed points. The seed points are selected at positions with obvious seawall features, such as the center line of the dike top road or the boundary of the slope protection structure. During the region growing process, pixels with a gray value difference within 10 gray levels and a texture feature similarity exceeding 0.8 are gradually included in the target area to obtain the seawall outline. Combining with the on-site RTK measurement data, set a cross-section measurement point every 50 meters, and arrange 5 - 7 feature points from the bottom to the top of each cross-section. Generate continuous point clouds between cross-sections through Kriging interpolation, and set the cross-section spacing to 5 meters. Conduct profile analysis in the direction perpendicular to the coastline to calculate the slope values of each section. For example, a certain cross-section is divided into three sections from the bottom to the top: the bottom section has a slope of 35 degrees, the middle section is a horizontal platform, and the top section has a slope of 40 degrees. At the same time, the total height of the seawall is measured as 4.5 meters, and the width of the dike top road is 6 meters. Substitute these parameters into the hierarchical threshold discriminant formula. Considering that this cross-section contains an obvious horizontal platform section and has a relatively steep slope change, the calculated result of the discrimination index is greater than the threshold, so it is determined as a composite seawall. Associate this discrimination result with the start and end coordinates of the cross-section through the spatial topological relationship, and perform attribute interpolation between the front and rear cross-sections to obtain the type distribution of this section of the seawall. Such a processing process makes full use of the advantages of multi-source data, ensuring both the accuracy of geometric feature extraction and the automatic discrimination of seawall types.

[0054] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0055] (1) Overlay the seawall trajectory line in the preliminary seawall type distribution data with the geographic base map, and extract the longitude and latitude coordinates, projected length, and trend angle of the start and end points of the seawall to obtain the basic geographic information;

[0056] (2) Conduct band combination analysis on the seawall top area in the standardized multi-source dataset to separate the spectral features of green plants, flower beds, and hardened ground to obtain the ecological information of the dike top;

[0057] (3) Distinguish the surface materials of the slope protection in the preliminary seawall type distribution data through texture analysis, and classify them according to fence panels, acropode blocks, four-corner hollow blocks, concrete smooth surface structures, riprap slope protection, and grouted stone masonry to obtain the slope protection structure information;

[0058] Extract the area around the seawall from the standardized multi-source dataset, and identify the distribution ranges of mangroves, bedrock beaches, sandy beaches, muddy beaches, and salt marshes through spectral separation to obtain the information of the beach ecosystem;

[0059] Combine and correlate the basic geographic information, the ecological information of the seawall top, the slope protection structure information, and the beach ecosystem information according to the spatial position correspondence relationship;

[0060] Structurally organize and uniformly encode and label the associated multi-dimensional feature information, and output the data of the ecological characteristic elements of the seawall.

[0061] Specifically, perform geographic information processing on the preliminary seawall type distribution data, and overlay the seawall track line with a high-precision geographic base map. The geographic base map contains standardized coordinate grids and terrain information, and ensures the precise correspondence between the track line and the base map through spatial registration. The extraction process uses vector data processing methods to obtain the longitude and latitude coordinates of the starting and ending points from the endpoints of the track line, calculate the length of the line segment on the projection plane, and calculate the heading angle through the endpoint coordinates. The basic geographic information obtained in this way describes the spatial distribution characteristics of the seawall. When performing band combination analysis on the seawall top area in the standardized multi-source dataset, a multi-spectral remote sensing image processing method is used. Combine the near-infrared band with the visible light band, and classify using the spectral reflection characteristics of different ground objects. Green plants have the characteristics of high reflection in the near-infrared band and low reflection in the red light band. The flower bed area shows regular green plant arrangements and flower spectral characteristics, while the hardened ground shows a higher reflectivity and a flat spectral curve. Through spectral feature comparison and spatial distribution feature analysis, the seawall top area is classified into types such as green belts, flower beds, scattered green plants, and hardened ground to form the ecological information of the seawall top.

[0062] The extraction of slope protection structure information mainly relies on texture analysis. Different slope protection materials have characteristic texture patterns: fence panels present regular grid textures, twisted blocks present continuous corrugations, four-corner hollow squares form a regular grid arrangement, concrete smooth structure texture is relatively uniform, riprap slope protection presents irregular blocks, and mortar masonry presents a regular stone splicing texture. The texture feature parameters of each material are extracted through the grayscale co-occurrence matrix, including statistics such as energy, contrast, and entropy, and a material classification feature library is established. Then, the texture features of the area to be analyzed are matched with the feature library to obtain the material type of each section of the slope protection. The identification of the shore ecosystem around the seawall is assisted by spectral separation technology. Mangroves have obvious high reflectivity characteristics in the near-infrared band, and their spatial distribution is irregular and patchy; bedrock beaches have high reflectivity in the visible light band and rough surface texture; the spectral curve of sandy beaches is relatively flat in the visible light band, and the nearshore shows characteristic wave reflection; muddy beaches show low spectral reflectivity and fine surface texture; salt marshes have unique spectral characteristics in the near-infrared and short-wave infrared bands. Through spectral feature analysis and spatial morphological feature identification, the distribution range of various ecosystems can be divided.

[0063] When integrating the above-mentioned information in space, a unified spatial reference system is first established. Basic geographic information is used as a framework to provide a spatial positioning framework for the seawall. The ecological information of the levee crest, the slope protection structure information, and the shore ecosystem information are superimposed on the framework according to their spatial position correspondence. Spatial indexing technology is used to associate information at different levels to ensure that the characteristic information at each spatial position corresponds one to one.

[0064] The multi-dimensional feature information after association is structured and organized. A unified coding rule is designed, including position coding, type coding and feature coding. Position coding reflects the spatial location of the seawall section, type coding indicates the seawall structure type, and feature coding describes the ecological and structural characteristics. Standardized storage and rapid retrieval of information are achieved through coding annotation.

[0065] For example, analyze a sea dike section about 2 kilometers long. Through the extraction of the trajectory line, the starting longitude is 120.5 degrees and the latitude is 30.2 degrees, and the ending longitude is 120.6 degrees and the latitude is 30.3 degrees. The calculated projected length is 2150 meters, and the overall trend is 45 degrees northeast. The multi-spectral image analysis shows that there is a continuous green belt on the dike top, and its reflectance in the near-infrared band is significantly higher than that of the surrounding hardened road surfaces, with three flower bed areas dotted in the middle. The analysis of the slope protection structure finds that this section of the sea dike adopts a mixed structure: the upper section is a fence panel slope protection, showing regular grid textures; the middle section is a Dolos block, presenting characteristic wavy patterns; the lower section is a riprap slope protection, with irregular block textures. The surrounding beach ecosystem includes a salt marsh belt on the inner side, showing typical tidal channel textures and the spectral characteristics of high-salt vegetation, and a sandy beach on the outer side, presenting a uniform banded distribution. After integrating this information according to the spatial position, a unified code is assigned and a characteristic element data set is formed to describe the ecological characteristics of this section of the sea dike.

[0066] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0067] (1) Perform a normalized difference vegetation index operation on the multi-spectral images in the ecological characteristic element data of the sea dike, and quantify and classify the vegetation coverage density by setting grading thresholds to obtain the percentage of vegetation coverage;

[0068] (2) Perform regional segmentation on the ecological information of the dike top, and extract the spatial distribution characteristics of the green belt, flower beds, and sporadic vegetation on the dike top through the calculation of edge density and direction consistency to obtain the dike top greening distribution form data;

[0069] (3) Convert the surface characteristics of different materials in the slope protection structure information into permeability coefficients, and quantitatively score the ecology of the slope protection surface structure to obtain an ecological permeability score value;

[0070] (4) Conduct area statistics and spatial distribution calculations on various ecosystems in the beach ecosystem information to obtain the distribution ratio of the ecosystems around the sea dike;

[0071] (5) Perform data standardization processing on the percentage of vegetation coverage, the dike top greening distribution form data, the ecological permeability score value, and the ecosystem distribution ratio;

[0072] (6) Perform a weighted summation operation on the standardized indicators according to the weights, and output the quantitative characteristic indicators.

[0073] Specifically, for the multi-spectral images in the ecological characteristic element data of the seawall, the vegetation index is calculated. The normalized difference vegetation index (NDVI) is an important parameter characterizing the vegetation growth status. The multi-spectral images contain the near-infrared band (0.76 - 0.90 μm) and the visible red light band (0.63 - 0.69 μm). Vegetation has a high reflectance in the near-infrared band and a low reflectance in the red light band. By calculating the NDVI value and setting the classification threshold, the vegetation coverage density is divided into four levels: no vegetation coverage (NDVI < 0.1), low coverage (0.1 ≤ NDVI < 0.3), medium coverage (0.3 ≤ NDVI < 0.6), and high coverage (NDVI ≥ 0.6), so as to obtain the area proportion of each level.

[0074] When conducting in-depth analysis of the ecological information on the dike top, regional segmentation is carried out. The segmentation process is based on the image edge and region growing algorithms, and the edge density parameter is calculated to identify the boundaries of different types of greening areas. Among them, the green belt shows a continuous linear distribution feature, with uniform edge density and strong directionality; the flower bed area presents a regular geometric shape, with a relatively high edge density and clear boundaries; the sporadic vegetation shows a discrete distribution, with uneven edge density. By calculating the spatial distribution density and the consistency of the main direction of the edge points, the distribution characteristics of various greening forms are extracted to form the dike top greening distribution form data. The ecological evaluation of the slope protection structure mainly considers its permeability. The permeability of different materials varies significantly: the fence panels and four-corner hollow blocks have relatively high permeability, which is beneficial for the habitat of small organisms; the tetrapods and rubble slope protection are next, which can provide space for plant growth; the concrete smooth surface structure has the worst permeability and is not conducive to ecological development. The permeability coefficient is assigned to each type of material through professional judgment criteria, and combined with the surface structure characteristics, the comprehensive ecological permeability score is calculated.

[0075] The spatial distribution analysis of the beach ecosystem needs to consider two aspects: area and spatial configuration, and the following calculation formula is adopted:

[0076] ;

[0077] Where: is the comprehensive ecosystem index; is the area of the i-th type of ecosystem; is the area weight coefficient; is the ecological importance coefficient; is the boundary length between the i-th type and the j-th type of ecosystem; is the ecological value coefficient of the junction zone; is the spatial connection strength; is the overall coordination coefficient; n is the number of ecosystem types.

[0078] Quantify the above quantitative indicators and standardize them to eliminate the differences in dimensions and numerical ranges. The maximum-minimum method is used for standardization to map each indicator value to the interval [0, 1]. Finally, weighted summation is performed on the standardized indicators, and the weight values are determined according to the influence degree of each indicator on the ecological function of the seawall, obtaining the quantitative characteristic indicators.

[0079] For example: For the ecological evaluation of a certain section of seawall, first extract the data of the near-infrared and red light bands from the multi-spectral image to calculate the NDVI. By statistically analyzing the NDVI value distribution, it is found that the NDVI values in the crest area are concentrated between 0.4 and 0.7, indicating good vegetation growth. Regional segmentation analysis shows that there is a continuous green belt with a width of 3 meters on the crest of this section, with uniform edge density, the directionality is consistent with the seawall trend, and flower beds are set every 200 meters, showing a regular rectangular distribution. The slope protection adopts a composite structure, with ecological blocks (permeability coefficient 0.8) in the upper part and riprap structure (permeability coefficient 0.6) in the lower part. There are three ecosystems of mangroves, sandy beaches, and muddy beaches distributed around, with areas of 10 hectares, 15 hectares, and 8 hectares respectively, and a good ecological transition zone is formed between the mangroves and the sandy beaches. Through standardization processing and weighted calculation, the comprehensive ecological characteristic index of this section of seawall is obtained, reflecting its ecological function status.

[0080] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0081] (1) Perform hierarchical grouping processing on the quantitative characteristic indicators, taking the seawall type as the first level, the seawall structure as the second level, and the ecological characteristics as the third level, to obtain a three-level hierarchical data structure;

[0082] (2) Uniquely encode each seawall sample in the three-level hierarchical data structure, and obtain the sample identification code by splicing the type code, structure code, and ecological characteristic code;

[0083] (3) Associate and bind the sample identification code with the multi-angle images, oblique photography point cloud data, and on-site photos in the seawall ecological characteristic element data to obtain the sample multimedia data set;

[0084] (4) Extract multi-dimensional feature values from the sample multimedia data set according to geographical location, type characteristics, and ecological indicators to obtain the sample retrieval feature vector;

[0085] (5) Convert the sample retrieval feature vector into an inverted index structure, and establish a key-value pair mapping relationship according to the feature values to obtain the sample multi-dimensional retrieval table;

[0086] (6) Set the update period identifier and quality evaluation score for the sample multi-dimensional retrieval table, and output the hierarchical seawall sample library.

[0087] Specifically, the quantitative characteristic indicators are hierarchically grouped to establish a three-level hierarchical data structure. The first level is the seawall type, including two major categories: the slope type and the composite type; the second level is the seawall structure. For the slope-type seawall, it is subdivided into the continuous slope type and the slope-platform-slope segmented type, and for the composite-type seawall, it is subdivided into the upper straight and lower slope type, the straight-flat-slope type, the slope-straight-slope type, the slope-straight-flat type, the upper slope and lower straight type, and the slope-straight-flat type; the third level is the ecological characteristics, including four aspects of characteristics: vegetation coverage, greening form of the dike top, ecological nature of the slope protection structure, and the surrounding ecosystem. For each seawall sample in the three-level hierarchical data structure, a unique coding rule is designed. The code consists of three parts: the type code (2 digits, e.g., 01 represents the slope type, 02 represents the composite type), the structure code (2 digits, e.g., in the slope-type seawall, 01 represents the continuous slope type, 02 represents the segmented type), and the ecological characteristic code (4 digits, respectively representing the vegetation coverage level, the greening type of the dike top, the slope protection ecological level, and the surrounding ecosystem type). By splicing these three parts of the code, an 8-digit sample identification code is formed.

[0088] Associate the generated sample identification codes with the multimedia data. The multimedia data includes multi-angle images (high-resolution images taken from different perspectives), oblique photography point cloud data (point clouds recording the three-dimensional structure information of the seawall), and on-site photos (field photos reflecting the actual condition of the seawall). Establish a data association table to establish a mapping relationship between each sample identification code and the corresponding multimedia file path. Metadata information such as the acquisition time and spatial resolution of the multimedia data is also recorded in the association table. Extract the feature values from the sample multimedia dataset to establish a multi-dimensional retrieval vector. The geographical location features include longitude and latitude coordinates, coastline orientation angles, etc.; the type features include cross-section shape parameters, slope protection structure parameters, etc.; the ecological indicators include vegetation indices, ecological permeability scores, etc. Each feature value is standardized to ensure a unified numerical range. In this way, each sample is transformed into a multi-dimensional feature vector, which is convenient for subsequent retrieval and matching.

[0089] Convert the extracted feature vectors into an inverted index structure. The inverted index is an efficient retrieval method that establishes a key-value pair mapping for each feature value. The key is the value range of the feature value, and the value is the list of sample identification codes with this feature. For example, for the vegetation coverage feature, multiple value intervals are established as keys, and the value corresponding to each interval is all the sample numbers falling within that interval. In this way, a multi-dimensional retrieval table covering all feature dimensions is constructed. Finally, to ensure the timeliness and quality of the sample library, a dynamic maintenance mechanism is introduced for the retrieval table. Set an update cycle identifier to record the storage time and expiration date of each sample. Calculate the quality assessment score according to the quality, accuracy, and representativeness of the sample data. Regularly clean or update the low-score samples to ensure that the sample library always maintains a high-quality reference role.

[0090] For example: A newly built sea dike section is used as a sample for database entry. This sea dike belongs to a composite sea dike (type code 02) and adopts a slant-straight-slant structure (structure code 03). After ecological feature evaluation, the vegetation coverage is high coverage (1), the top of the dike is a continuous green belt (1), the ecological nature of the slope protection is medium (2), and there is a mangrove ecosystem nearby (1). Therefore, the ecological feature code is 1121. After splicing, the sample identification code 02031121 is obtained. Then, the multimedia data of this sea dike is collected, including UAV images taken from three angles of 0°, 45°, and 90°, point cloud data generated by oblique photogrammetry (point density reaching 50 points per square meter), and 20 field photos reflecting details such as the greening of the dike top and the slope protection structure. When extracting feature vectors, the characteristic values in multiple dimensions such as geographical location (120° east longitude, 30° north latitude, and a trend of 45°), type characteristics (a slope of 35°, a height of 4.5 meters, and a width of 6 meters), and ecological indicators (an NDVI value of 0.75 and a permeability score of 0.6) are calculated. Inverted indexes are respectively established for these characteristic values. For example, the key-value pairs in the interval of NDVI value 0.7 - 0.8 contain this sample number. Finally, a 6-month update cycle is set, and a quality score of 85 points is obtained based on data quality and accuracy assessment. Through such a processing flow, the information of this sea dike section is stored in the sample database in a standardized form.

[0091] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0092] (1) Perform morphological processing on the spatial features in the hierarchical sea dike sample database, calculate the curvature value and continuity parameter of the sea dike contour line, and obtain a spatial feature similarity matrix;

[0093] (2) Perform band combination analysis on the spectral features in the hierarchical sea dike sample database, calculate the difference value of vegetation coverage and material reflectivity, and obtain a spectral feature similarity matrix;

[0094] (3) Perform texture analysis on the structural features in the hierarchical sea dike sample database, calculate the matching degree of the slope protection type and the dike top morphology, and obtain a structural feature similarity matrix;

[0095] (4) Normalize the spatial feature similarity matrix, the spectral feature similarity matrix, and the structural feature similarity matrix to obtain a comprehensive similarity score;

[0096] (5) Cluster and group the sample data according to the comprehensive similarity score, and perform feature statistics on the samples within each group to obtain a type feature statistical table;

[0097] (6) Structurally reorganize the feature values in the type feature statistical table to generate a classification system for the ecological features of the sea dike.

[0098] Specifically, in the final stage of constructing the classification system for the ecological characteristics of seawalls, it is necessary to conduct multi-dimensional similarity analysis and classification integration on the examples in the hierarchical seawall example library. Morphological processing is performed on the spatial characteristics, and the core of morphological processing is to analyze the geometric characteristics of the seawall contour line. The curvature value reflects the degree of curvature of the seawall trajectory line. By calculating the curvature of each point on the trajectory line, a curvature sequence is obtained; the continuity parameter reflects the smoothness of the trajectory line and is calculated by analyzing the direction change between adjacent points. For the spatial characteristics of any two examples, calculate the Euclidean distance of their curvature sequences and continuity parameters to construct a spatial characteristics similarity matrix. The analysis of spectral characteristics performs band combination on multi-spectral images. Typical band combinations include: the combination of near-infrared, red, and green bands to highlight vegetation information; the combination of short-wave infrared, near-infrared, and red bands to highlight moisture and material information. By calculating the reflectance characteristics of each band combination, vegetation coverage indicators and material reflectance characteristics are obtained. Calculate the difference values of these characteristics for each pair of examples to form a spectral characteristics similarity matrix. The calculation of difference values needs to consider the weights of each band, as different bands contribute differently to the identification of vegetation and materials.

[0099] The analysis of structural characteristics focuses on the slope protection type and the crest morphology. The slope protection types include various forms such as cribbing blocks, tetrapods, and hollow square blocks at the corners, and each type has its characteristic texture pattern. Texture feature parameters, including statistics such as energy, contrast, and entropy, are extracted through the gray-level co-occurrence matrix. The crest morphology focuses on the spatial distribution patterns of structures such as green belts and flower beds. Convert these characteristics into feature vectors and calculate the cosine similarity between the vectors to obtain a structural characteristics similarity matrix.

[0100] Normalize the three similarity matrices to eliminate the influence of the dimension of different feature indicators. The normalization uses the maximum-minimum method to map all similarity values to the interval [0,1]. Then, perform weighted fusion on the three normalized matrices, and the weight values are determined according to the importance of various features to the ecological function of the seawall to obtain a comprehensive similarity score. Based on the comprehensive similarity score, use the hierarchical clustering algorithm to group the examples. First, regard each example as an independent category, and then gradually merge the two categories with the highest similarity until the preset number of categories is reached. Within each clustering group, statistics such as the mean and standard deviation of each feature are calculated to form a type feature statistical table. The statistical table reflects the typical features and variation ranges of each category.

[0101] Finally, structurally reorganize the type feature statistical table to construct a hierarchical classification system. The first layer is the basic types of seawalls (sloping type, composite type), the second layer is the specific structural forms, and the third layer is the ecological feature combinations. Each category is accompanied by detailed feature descriptions and typical examples.

[0102] For example: Analyze the sample library of a certain coastal zone. First, extract the characteristics of a section of composite seawall. The spatial feature analysis shows that the trajectory line of this section of seawall presents a weak curve shape. The mean value of the calculated curvature sequence is 0.002, and the continuity parameter is 0.95, indicating that the seawall alignment is relatively smooth. In the spectral analysis, the reflectance in the near-infrared band is 0.45, and in the red light band is 0.12. The calculated vegetation index is 0.58, indicating good greening conditions; the reflectance distribution in the short-wave infrared band shows that the slope protection material is mainly concrete. The structural analysis finds that the slope protection adopts a combination form of upper fence panels and lower riprap. The texture feature parameters show that the energy value in the fence panel area is 0.85, the contrast is 0.45, and the entropy is 0.65. These values highly match the characteristics of the same type of slope protection in the sample library. The crest shape shows a continuous green belt structure with a uniform width, and the flower beds are regularly spaced. Compare these characteristic values with other seawall samples in the sample library and calculate the similarity matrix. After cluster analysis, this section of seawall is classified into the category of "composite - ecological slope protection - continuous greening", and it shows high similarity with other samples in this category in terms of spatial form, vegetation coverage, and structural characteristics.

[0103] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for constructing a remote sensing monitoring classification system for ecological characteristics of seawalls, characterized in that: include: Perform multi-source data fusion processing on high-resolution satellite remote sensing images, aerial remote sensing images, drone orthophotos and field measurement data, obtain digital orthophotos through aerial triangulation, perform plane position and elevation correction on digital orthophotos according to control point coordinates, and obtain standardized multi-source data sets; The slope, height and width of the seawall section are extracted based on the standardized multi-source data set, and the obtained characteristic parameter values ​​are substituted into the preset expert rules for type identification to generate preliminary seawall type distribution data; Using the preliminary seawall type distribution data, the basic geographic information, the ecological information of the levee crest, the slope protection structure information and the shore ecosystem information are processed with characteristic association to form the ecological characteristic element data of the seawall; According to the ecological characteristic element data of the seawall, the vegetation coverage of the multispectral image is calculated, the image texture is used to determine the type of greening on the embankment top, and the scoring operation is performed according to the slope protection material and ecological permeability to obtain quantitative characteristic indicators; including normalizing the vegetation index of the multispectral image in the ecological characteristic element data of the seawall, quantitatively grading the vegetation coverage density by setting the grading threshold, and obtaining the vegetation coverage percentage; regional segmentation of the ecological information on the embankment top, and extracting the spatial distribution characteristics of the green belt, flower bed, and scattered vegetation on the embankment top through edge density and direction consistency calculation to obtain the distribution morphological data of the greening on the embankment top; The surface characteristics of different materials in the slope protection structure information are converted into permeability coefficients, and the ecological properties of the slope protection surface structure are quantitatively scored to obtain the ecological permeability score value; the area statistics and spatial distribution calculations of various ecosystems in the shore ecosystem information are performed to obtain the distribution ratio of ecosystems around the seawall; the vegetation coverage percentage, the distribution morphology data of the greening on the top of the embankment, the ecological permeability score value and the ecosystem distribution ratio are standardized. The standardization adopts the maximum and minimum method to map each indicator value to the [0,1] interval, and the weighted sum operation is performed on each standardized indicator according to the weight to output the quantitative characteristic indicator; Based on the quantitative characteristic indicators, the sample data is standardized and stored in a three-level structure, and a multi-dimensional index is constructed to generate a hierarchical seawall sample library; Similarity matching operations are performed on the spatial features, spectral features and structural features in the hierarchical seawall sample library, and samples are updated through dynamic feedback processing to build a seawall ecological feature classification system.

2. The method for constructing a remote sensing monitoring classification system for ecological characteristics of seawalls according to claim 1 is characterized in that: Data fusion processing is performed on high-resolution satellite remote sensing images, aerial remote sensing images, drone orthophotos and remote sensing images in field measurement data. Digital orthophotos are obtained through aerial triangulation. The plane position and elevation deviation of the digital orthophotos are corrected according to the coordinates of the control points to obtain a standardized multi-source data set, including: band resampling of high-resolution satellite remote sensing images, inputting the resampled band data into Gram-schmidt mathematical transformation to obtain spectral fusion data; The spectral fusion data and the aerial remote sensing image are registered according to the corresponding relationship of geographic coordinates to obtain a first registered image; The first registered image and the drone orthophoto are subjected to affine transformation according to the same-name ground object points, and the second registered image is obtained by bilinear interpolation; Performing image stretching, equalization and filtering enhancement processing on the second registered image through on-site measurement data to generate enhanced image data; The enhanced image data is input into the aerial triangulation program, and the digital orthophoto map is obtained through tie point matching and bundle adjustment operations; Substitute the coordinates of the control points into the correction equation, perform plane position correction and elevation correction calculations on the digital orthophoto, and output a standardized multi-source data set.

3. The method for constructing a remote sensing monitoring classification system for ecological characteristics of seawalls according to claim 1, characterized in that: The slope, height and width of the seawall section are extracted based on the standardized multi-source data set, and the obtained characteristic parameter values ​​are substituted into the preset expert rules for type identification to generate preliminary seawall type distribution data, including: Perform edge detection on the seawall images in the standardized multi-source dataset, segment the seawall area through region growing, and obtain the seawall contour data; The seawall contour data is superimposed on the field measurement data, and the three-dimensional coordinates of each section point are calculated by spatial interpolation to obtain the seawall section point cloud; The seawall cross-section point cloud is cut into sections in a direction perpendicular to the coastline, and the height difference between adjacent cross-section points is divided by the horizontal distance to obtain the slope value; Extract the elevation difference between the top and bottom points of the seawall from the seawall cross-section point cloud as the height value, and extract the horizontal distance between the points on both sides of the road at the top of the seawall as the width value; Substituting the slope value, height value and width value into the layered threshold discriminant formula, the seawall is divided into two types: slope type and composite type, and the seawall type discrimination result is obtained; The seawall type identification results are associated with geographic coordinates through spatial topological relationships, and preliminary seawall type distribution data are output.

4. The method for constructing a remote sensing monitoring classification system for ecological characteristics of seawalls according to claim 1, characterized in that: Using the preliminary seawall type distribution data, basic geographic information, levee top ecological information, slope protection structure information and shore ecosystem information are processed with characteristic association to form seawall ecological characteristic element data, including: superimposing the seawall trajectory line in the preliminary seawall type distribution data with the geographic base map, extracting the latitude and longitude coordinates of the starting and ending points of the seawall, the projection length and the strike angle, and obtaining basic geographic information; Band combination analysis was performed on the top area of ​​the seawall in the standardized multi-source dataset to separate the spectral characteristics of green plants, flower beds, and hardened ground, and obtain the ecological information of the top of the seawall. The surface materials of the slope protection in the preliminary seawall type distribution data were distinguished by texture analysis, and classified into fence panels, twisted blocks, four-corner hollow blocks, concrete smooth structures, riprap slope protection, and mortar masonry to obtain the slope protection structure information; The area around the seawall was extracted from the standardized multi-source data set, and the distribution range of mangroves, bedrock beaches, sandy beaches, muddy beaches and salt marshes was identified through spectral separation to obtain the beach ecosystem information; Combine and associate basic geographic information, embankment ecological information, slope protection structure information and shore ecosystem information according to the spatial location correspondence; The associated multi-dimensional feature information is structured and uniformly coded and labeled to output the ecological feature element data of the seawall.

5. The method for constructing a remote sensing monitoring classification system for ecological characteristics of seawalls according to claim 1, characterized in that: Based on the quantitative characteristic indicators, the sample data is standardized and stored in a three-level structure, and a multidimensional index is constructed to generate a hierarchical seawall sample library, including: hierarchical grouping of the quantitative characteristic indicators, taking the seawall type as the first level, the seawall structure as the second level, and the ecological characteristics as the third level, to obtain a three-level hierarchical data structure; Uniquely encode each seawall sample in the three-level hierarchical data structure, and obtain the sample identification code by splicing the type code, structure code, and ecological characteristic code; The sample identification code is associated and bound with the multi-angle images, oblique photography point cloud data, and field photos in the ecological feature element data of the seawall to obtain a sample multimedia data set; Extract multi-dimensional feature values ​​of sample multimedia datasets according to geographical location, type characteristics, and ecological indicators to obtain sample retrieval feature vectors; Convert the sample retrieval feature vector into an inverted index structure, establish a key-value pair mapping relationship according to the feature value, and obtain a sample multidimensional retrieval table; Set update cycle identification and quality assessment scores for the sample multidimensional retrieval table, and output a hierarchical seawall sample library.

6. The method for constructing a remote sensing monitoring classification system for ecological characteristics of seawalls according to claim 1, characterized in that: The spatial features, spectral features and structural features in the hierarchical seawall sample library are matched similarly, and the samples are updated through dynamic feedback processing to build a seawall ecological feature classification system, including: morphological processing of the spatial features in the hierarchical seawall sample library, calculating the curvature value and continuity parameter of the seawall contour line, and obtaining the spatial feature similarity matrix; The spectral features in the hierarchical seawall sample library were analyzed by band combination, the difference values ​​of vegetation coverage and material reflectance were calculated, and the spectral feature similarity matrix was obtained; The texture analysis of the structural features in the hierarchical seawall sample library is performed to calculate the matching degree between the slope protection type and the embankment top morphology, and the structural feature similarity matrix is ​​obtained. The spatial feature similarity matrix, the spectral feature similarity matrix and the structural feature similarity matrix are normalized to obtain a comprehensive similarity score; Cluster and group the sample data according to the comprehensive similarity score, perform feature statistics on the samples in each group, and obtain a type feature statistics table; The characteristic values ​​in the type characteristic statistical table are restructured to generate a classification system of seawall ecological characteristics.

Images