Method and System for Extracting Seawall Feature Information Based on Multi-Source Remote Sensing Data

Abstract

The present application relates to the technical field of data processing, and discloses a method and system for extracting seawall feature information based on multi-source remote sensing data. The method includes: obtaining standardized data by performing band combination enhancement, geometric correction, and radiometric correction on multi-source remote sensing images; extracting seawall targets using multi-scale segmentation; identifying structural features through profile analysis and spectral matching; calculating vegetation indices and texture features to obtain ecological information; establishing an accuracy evaluation system; and finally constructing a feature information database. The present application improves the accuracy and efficiency of seawall feature information extraction through technical means such as multi-source data fusion processing, multi-scale feature extraction, and ecological feature evaluation, and solves the problems existing in data acquisition, processing, and evaluation of traditional methods.

Description

Technical Field

[0001] This application relates to the field of data processing, and particularly to a method and system for extracting seawall feature information based on multi-source remote sensing data. Background Art

[0002] As an important flood and wave prevention engineering facility, the construction and management of seawalls have received extensive attention. At present, the extraction of seawall feature information mainly relies on manual on-site surveys and single remote sensing data analysis. Among them, manual on-site surveys obtain basic information such as the location and elevation of seawalls through measuring instruments, while remote sensing data analysis mainly uses satellite remote sensing images for visual interpretation and feature extraction of seawalls. These methods have achieved certain results in obtaining seawall information and provided basic data support for seawall management and maintenance. Traditional remote sensing data processing methods mainly include steps such as image preprocessing, feature enhancement, and target extraction, and complete the recognition of seawall features and the extraction of attribute information through a man-machine interaction method.

[0003] However, the traditional methods for extracting seawall feature information have obvious limitations. First, the information extraction method based on a single data source is difficult to comprehensively reflect the structural and ecological features of seawalls. Especially in complex terrains and diverse ecological environments, the expression ability of single-source data is limited. Second, the manual survey method is time-consuming and laborious, and is greatly affected by factors such as weather and terrain, making it difficult to achieve large-scale and high-efficiency information acquisition. Third, the existing remote sensing data processing methods lack standardized and automated processing procedures, with low data processing efficiency and strong subjectivity, making it difficult to ensure the consistency and reliability of extraction results. In addition, there are still technical gaps in the extraction and evaluation of the ecological features of seawalls in the existing methods, which cannot meet the needs of seawall ecological construction and management. Summary of the Invention

[0004] This application provides a method and system for extracting seawall feature information based on multi-source remote sensing data, which improves the accuracy and efficiency of seawall feature information extraction through technical means such as multi-source data fusion processing, multi-scale feature extraction, and ecological feature evaluation, and solves the problems existing in data acquisition, processing, and evaluation of traditional methods.

[0005] In the first aspect, this application provides a method for extracting seawall feature information based on multi-source remote sensing data. The method for extracting seawall feature information based on multi-source remote sensing data includes:

[0006] For the collected multi-temporal satellite remote sensing images, aerial remote sensing images, and unmanned aerial vehicle oblique image data, perform multi-spectral information enhancement through band combination, perform geometric correction and orthorectification based on ground control points, and perform radiometric consistency processing of multi-source images in combination with the histogram matching method to obtain standardized multi-temporal remote sensing data;

[0007] Based on the standardized multi-temporal remote sensing data, the multi-scale segmentation algorithm is used to extract the seawall target area, the three-dimensional elevation information is calculated by combining the oblique photogrammetry data, the boundary range and changes of the seawall are determined through the comparison and analysis of temporal features, and the spatio-temporal feature information of the seawall is obtained;

[0008] According to the spatio-temporal feature information of the seawall, the slope information of the seawall is extracted through cross-section analysis, the slope protection structure is identified based on the material spectral feature library, and the sectional structure features of the seawall are determined by combining the calculation of terrain undulation, and a seawall structure feature dataset is generated;

[0009] For the seawall structure feature dataset, the normalized difference vegetation index is calculated using multi-spectral data, the vegetation distribution density and type are extracted by texture analysis method, and the degree of slope protection ecologicalization is determined by spectral curve matching, and the seawall ecologicalization feature information is generated;

[0010] Based on the seawall ecologicalization feature information, an accuracy evaluation sample set is established according to the field measurement data, the position accuracy and attribute accuracy are calculated through multi-level cross-validation, and dynamic change inspection is carried out in combination with the temporal images, and a feature extraction quality evaluation report is generated;

[0011] The seawall structure feature dataset, the seawall ecologicalization feature information and the feature extraction quality evaluation report are used to construct a seawall feature information database through a hierarchical data organization method.

[0012] In a second aspect, the present application provides a seawall feature information extraction system based on multi-source remote sensing data, and the seawall feature information extraction system based on multi-source remote sensing data includes:

[0013] An enhancement module for enhancing the multi-spectral information of the collected multi-temporal satellite remote sensing images, aerial remote sensing images and unmanned aerial vehicle oblique image data through band combination, performing geometric correction and orthorectification based on ground control points, and performing radiation consistency processing of multi-source images in combination with the histogram matching method to obtain standardized multi-temporal remote sensing data;

[0014] An extraction module for extracting the seawall target area according to the standardized multi-temporal remote sensing data, calculating the three-dimensional elevation information by combining the oblique photogrammetry data, determining the boundary range and changes of the seawall through the comparison and analysis of temporal features, and obtaining the spatio-temporal feature information of the seawall;

[0015] An identification module for extracting the slope information of the seawall through cross-section analysis according to the spatio-temporal feature information of the seawall, identifying the slope protection structure based on the material spectral feature library, and determining the sectional structure features of the seawall by combining the calculation of terrain undulation, and generating a seawall structure feature dataset;

[0016] A generation module, configured to calculate a normalized difference vegetation index using multi-spectral data for the seawall structure feature dataset, extract vegetation distribution density and types using a texture analysis method, determine the degree of ecologicalization of the slope protection through spectral curve matching, and generate seawall ecologicalization feature information;

[0017] An establishment module, configured to establish an accuracy evaluation sample set based on the seawall ecologicalization feature information according to field measurement data, calculate position accuracy and attribute accuracy through multi-level cross-validation, perform dynamic change inspection in combination with time-series images, and generate a feature extraction quality evaluation report;

[0018] A construction module, configured to construct a seawall feature information database by organizing the seawall structure feature dataset, the seawall ecologicalization feature information, and the feature extraction quality evaluation report in a hierarchical data organization manner.

[0019] In the technical solution provided by this application, by performing band combination and multi-spectral information enhancement on multi-temporal satellite remote sensing images, aerial remote sensing images, and UAV oblique image data, the recognition accuracy of seawall features is improved, making the seawall structure and ecological features more clearly distinguishable. Geometric correction and orthorectification based on ground control points, combined with histogram matching method for multi-source image radiation consistency processing, ensure the spatial position accuracy and radiation feature consistency between different data sources, laying a foundation for subsequent feature extraction. At the same time, using a multi-scale segmentation algorithm to extract the seawall target area and calculating three-dimensional elevation information in combination with oblique photogrammetry data not only improves the accuracy of target extraction but also obtains the three-dimensional structure information of the seawall. By comparing and analyzing time-series features to determine the boundary range and changes of the seawall, dynamic monitoring of seawall features is achieved. Based on the spatio-temporal feature information of the seawall, cross-section profile analysis and material spectral feature recognition can accurately extract the slope information and slope protection structure type of the seawall, and further determine the segmented structure features of the seawall in combination with the calculation of terrain undulation, making the extraction of seawall structure features more comprehensive and accurate. In addition, by calculating the normalized difference vegetation index, extracting vegetation distribution density and types using a texture analysis method, and then determining the degree of ecologicalization of the slope protection through spectral curve matching, quantitative evaluation of seawall ecological features is achieved. Finally, based on field measurement data, an accuracy evaluation sample set is established, and through multi-level cross-validation and dynamic change inspection of time-series images, not only the reliability of the feature extraction results is ensured, but also a quality evaluation system is established. The entire technical solution constructs a seawall feature information database through a hierarchical data organization method, realizing the standardized management and efficient utilization of seawall feature information. Brief Description of the Drawings

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0021] Figure 1 FIG. is a schematic diagram of an embodiment of a method for extracting seawall feature information based on multi-source remote sensing data in an embodiment of the present application;

[0022] Figure 2 FIG. is a schematic diagram of an embodiment of a system for extracting seawall feature information based on multi-source remote sensing data in an embodiment of the present application. Detailed implementation manners

[0023] The embodiments of the present application provide a method and system for extracting seawall feature information based on multi-source remote sensing data. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims, and the above accompanying drawings of the present application are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device comprising a series of steps or units does not necessarily limit to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0024] For ease of understanding, the following describes the specific process of the embodiments of the present application. Please refer to Figure 1 , an embodiment of the method for extracting seawall feature information based on multi-source remote sensing data in an embodiment of the present application includes:

[0025] Step S101: For the collected multi-temporal satellite remote sensing images, aerial remote sensing images, and unmanned aerial vehicle (UAV) oblique image data, perform multi-spectral information enhancement through band combination, perform geometric correction and orthorectification based on ground control points, and perform radiation consistency processing of multi-source images in combination with the histogram matching method to obtain standardized multi-temporal remote sensing data;

[0026] Step S102: According to the standardized multi-temporal remote sensing data, use the multi-scale segmentation algorithm to extract the seawall target area, calculate the three-dimensional elevation information in combination with the oblique photogrammetry data, and determine the boundary range and change situation of the seawall through temporal and spatial feature contrast analysis to obtain the spatio-temporal feature information of the seawall;

[0027] Step S103: Based on the spatio-temporal characteristic information of the seawall, extract the slope information of the seawall through cross-section profile analysis, identify the slope protection structure based on the material spectral feature library, calculate and determine the sectional structure characteristics of the seawall in combination with the terrain undulation degree, and generate a seawall structure feature dataset;

[0028] Step S104: For the seawall structure feature dataset, calculate the normalized difference vegetation index using multi-spectral data, extract the vegetation distribution density and type by using the texture analysis method, determine the degree of slope protection ecologicalization through spectral curve matching, and generate the seawall ecologicalization feature information;

[0029] Step S105: Based on the seawall ecologicalization feature information, establish an accuracy evaluation sample set according to the field measurement data, calculate the position accuracy and attribute accuracy through multi-level cross-validation, and conduct dynamic change inspection in combination with the time-series images to generate a feature extraction quality evaluation report;

[0030] Step S106: Construct a seawall feature information database by organizing the seawall structure feature dataset, the seawall ecologicalization feature information, and the feature extraction quality evaluation report in a hierarchical data organization manner.

[0031] It can be understood that the execution subject of this application can be a seawall feature information extraction system based on multi-source remote sensing data, or a terminal or a server. Specifically, it is not limited here. In this embodiment of the application, the server is taken as the execution subject for illustration.

[0032] Specifically, data preprocessing is carried out. This process starts from multi-temporal satellite remote sensing images, aerial remote sensing images, and UAV oblique image data, and performs multi-spectral information enhancement processing through band combination. Multi-spectral information enhancement refers to the process of combining remote sensing images of different bands to highlight the features of ground objects. Specifically, for the extraction of seawall features, a combination of near-infrared band, red band, and green band is mainly used. In this combination, the near-infrared band is particularly sensitive to vegetation information and can clearly reflect the vegetation coverage on the seawall; the red band can better reflect the characteristics of surface buildings and bare surfaces; the green band has a good effect on the identification of water body boundaries. During the band combination process, by adjusting the band weight coefficients, the seawall structure becomes more clearly distinguishable in the image. After obtaining the image data enhanced by band combination, geometric correction and ortho-rectification need to be carried out based on ground control points. Ground control points are feature points with accurate coordinate information obtained through field measurement. Through these points, the corresponding relationship between the image space and the geographic space is established. Geometric correction uses a polynomial correction model, and the transformation parameters are calculated by the least squares method to eliminate geometric deformations in the image caused by factors such as terrain undulation and lens distortion. Ortho-rectification is the process of converting a central projection image into an ortho-projection image using digital elevation model (DEM) data, which eliminates the position deviation caused by terrain undulation.

[0033] After completing geometric correction and ortho-rectification, radiometric consistency processing of multi-source images is carried out. Radiometric consistency processing uses the histogram matching method, which adjusts the gray-scale distribution of images acquired at different times and by different sensors to make them have similar statistical characteristics. In specific operations, first, a reference image is selected, and then the histograms of other images are adjusted to be similar to the reference image through non-linear transformation, so as to achieve the standardization of multi-source remote sensing data. Next, using the standardized multi-temporal remote sensing data, a multi-scale segmentation algorithm is used to extract the seawall target area. The multi-scale segmentation algorithm is a segmentation method based on graph theory. By calculating the heterogeneity and homogeneity between image objects, the image is segmented into regions with similar characteristics. During the extraction process of the seawall target area, first, appropriate segmentation scale parameters are set, including shape factor, compactness factor, etc., and then through iterative calculation, the image is segmented into objects of appropriate size. After segmentation, the shape features (linear, strip-shaped), spectral features (reflection features of building materials such as concrete and stone), and texture features (regular artificial building features) of the seawall are used for target extraction.

[0034] After obtaining the target area of the seawall, the three-dimensional elevation information is calculated in combination with the oblique photogrammetry data. Oblique photogrammetry obtains multi-view images of the target by taking pictures from multiple angles, and then reconstructs the three-dimensional structure of the target using photogrammetry principles. In actual operation, first use an unmanned aerial vehicle to obtain multi-angle images of the seawall, and then construct a three-dimensional point cloud model of the seawall through steps such as homologous point matching and aerial triangulation. Based on the point cloud model, three-dimensional features such as the elevation information and slope information of the seawall can be extracted. Temporal feature comparison and analysis determine the boundary range and changes of the seawall by comparing the feature changes of the seawall in images of different time phases. The images of different time phases are registered, and then by calculating the normalized difference index (such as NDVI, NDWI, etc.), the change characteristics of the seawall and its surrounding environment are analyzed. Temporal analysis can effectively identify the actual range of the seawall and monitor the changes of the seawall.

[0035] After obtaining the spatio-temporal feature information of the seawall, the slope information of the seawall is extracted through cross-section profile analysis. Cross-section profile analysis extracts the elevation change curve along the cross-section direction of the seawall, and by analyzing the characteristics of the elevation change, structural characteristics such as the slope and steps of the seawall are identified. In actual operation, first set a sampling line along the cross-section direction of the seawall, extract the elevation values on the sampling line, and then obtain the slope information of the seawall by calculating the elevation difference and horizontal distance between adjacent points.

[0036] The material spectral feature library is a dataset containing the spectral reflection characteristics of various slope protection materials (such as concrete, stone, vegetation, etc.). By matching the spectral characteristics in the seawall image with the standard spectra in the feature library, the structural type of the seawall slope protection can be identified. In practical applications, first extract the spectral curve in the seawall image, then calculate the similarity with the spectra of various materials in the feature library, and select the material type with the highest similarity as the recognition result of the slope protection structure. The terrain undulation calculation identifies the segmented structure of the seawall by analyzing the elevation change characteristics of the seawall surface. First, calculate the elevation standard deviation of the local area based on the three-dimensional point cloud data, and then by setting a threshold, identify the areas with significant elevation changes, which usually correspond to the structural segmentation points of the seawall.

[0037] After obtaining the seawall structure feature dataset, the normalized difference vegetation index (NDVI) is calculated using multi-spectral data. NDVI is a vegetation index calculated using the difference in the reflection characteristics of vegetation in the red and near-infrared bands. The larger its value, the higher the vegetation coverage. Through the statistical analysis of the NDVI value, the vegetation distribution density information of the seawall slope protection can be obtained. Texture analysis methods are used to extract the vegetation distribution density and type. Texture analysis includes methods such as gray-level co-occurrence matrix (GLCM) feature calculation and edge density analysis, which can identify the spatial distribution patterns of different types of vegetation. In practical applications, first calculate the texture feature parameters of the image, and then classify and identify the vegetation types based on these parameters.

[0038] Spectral curve matching is used to determine the degree of slope protection ecologicalization. By comparing the similarity between the spectral characteristics of the slope protection area and the standard vegetation spectral characteristics, the ecologicalization level of the slope protection can be evaluated. In specific operations, first, the spectral curve of the slope protection area is extracted, and then the correlation coefficient with the standard vegetation spectrum is calculated. The higher the correlation coefficient, the higher the degree of ecologicalization. After feature extraction, it is necessary to establish an accuracy evaluation sample set for multi-level cross-validation. The accuracy evaluation sample set includes true value data such as on-site measured location data and attribute data. The cross-validation adopts the K-fold cross-validation method, randomly dividing the samples into K parts, taking one part as the validation set each time, and the rest as the training set, and calculating the average accuracy through multiple validations.

[0039] Finally, all the extracted feature information and evaluation results are organized in a hierarchical manner to construct a seawall feature information database. The database adopts an object-oriented organization method to associate and store the spatial information, attribute information, and temporal information of the seawall. In specific applications, for example, during the feature extraction process of a certain section of the seawall, the spatial range of the seawall is identified through multi-temporal remote sensing images, and the calculated NDVI value is distributed between 0.2 and 0.6, indicating a medium degree of vegetation coverage. Through cross-section analysis, the average slope is obtained as 25 degrees, and the slope protection structure is a concrete structure. These information are organized and stored in the database for subsequent query and analysis.

[0040] In the embodiments of the present application, by performing band combination and multi-spectral information enhancement on multi-temporal satellite remote sensing images, aerial remote sensing images, and UAV oblique image data, the recognition accuracy of seawall features is improved, making the seawall structure and ecological features more clearly distinguishable. Geometric correction and ortho-rectification are performed based on ground control points, and the radiometric consistency processing of multi-source images is carried out in combination with the histogram matching method, which ensures the spatial position accuracy and radiometric feature consistency between different data sources, laying a foundation for subsequent feature extraction. At the same time, the multi-scale segmentation algorithm is used to extract the seawall target area, and the three-dimensional elevation information is calculated in combination with the oblique photogrammetry data, which not only improves the accuracy of target extraction but also can obtain the three-dimensional structure information of the seawall. The boundary range and changes of the seawall are determined through the contrast analysis of temporal features, realizing the dynamic monitoring of seawall features. Based on the spatio-temporal feature information of the seawall, cross-section profile analysis and material spectral feature recognition are carried out, which can accurately extract the slope information and the type of slope protection structure of the seawall, and further determine the sectional structure features of the seawall in combination with the calculation of terrain undulation, making the extraction of seawall structure features more comprehensive and accurate. In addition, by calculating the normalized difference vegetation index, using the texture analysis method to extract the vegetation distribution density and type, and then determining the degree of slope protection ecologicalization through spectral curve matching, the quantitative evaluation of seawall ecological features is realized. Finally, an accuracy evaluation sample set is established based on field measurement data, and through multi-level cross-validation and dynamic change inspection of temporal images, not only the reliability of the feature extraction results is ensured, but also a quality evaluation system is established. The entire technical solution constructs a seawall feature information database through a hierarchical data organization method, realizing the standardized management and efficient utilization of seawall feature information.

[0041] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0042] (1) Based on multi-temporal satellite remote sensing images, aerial remote sensing images, and UAV oblique image data, perform band separation processing on different data sources, extract near-infrared band, red band, and green band information from them, and perform multi-spectral information enhancement through band combination to obtain multi-spectral enhanced data;

[0043] (2) For the multi-spectral enhanced data, construct a geometric transformation model through the coordinate information of ground control points, perform correction and resampling calculation on the image spatial position to obtain geometric correction data;

[0044] (3) Input the geometric correction data into the ortho-rectification processing algorithm, establish a terrain correction model to eliminate projection deformation, and obtain ortho-rectified data;

[0045] (4) For the ortho-rectified data, extract the gray histogram features of each band image, and perform radiometric consistency processing of multi-source images through the histogram matching method to obtain radiometric consistency data;

[0046] (5) Organize the radiation consistency data in a time series, establish a time-series correlation relationship, and obtain multi-time-series correlation data;

[0047] (6) Standardize the multi-time-series correlation data to obtain standardized multi-time-series remote sensing data.

[0048] Specifically, perform band separation processing on the acquired multi-temporal satellite remote sensing images, aerial remote sensing images, and UAV oblique image data. Band separation processing refers to separating different bands in a multi-spectral remote sensing image to obtain image data of a single band. In remote sensing data, the near-infrared band is particularly sensitive to vegetation information, the red band is sensitive to surface buildings and bare surface features, and the green band is sensitive to water body features. By extracting and combining these three bands, the characteristics of the seawall and its surrounding environment can be better highlighted. Multi-spectrum information enhancement through band combination is a process of assigning different weights to different bands and combining them into a new image. In seawall feature extraction, a combination of the near-infrared band, red band, and green band is used, and by adjusting the weight coefficients of each band, the display effect of the seawall structure is enhanced. The selection of weight coefficients needs to consider factors such as the material characteristics of the seawall, vegetation coverage, and water environment. For example, for a concrete-structured seawall, the weight of the red band will be appropriately increased to highlight the building characteristics; for an ecological seawall, the weight of the near-infrared band will be increased to highlight the vegetation information.

[0049] After obtaining the multi-spectrum enhanced data, geometric correction needs to be performed through ground control points. Ground control points are feature points with accurate coordinate information obtained through on-site measurement, and these points can be accurately identified both on the image and at the actual ground position. The geometric transformation model is a mathematical model that establishes the correspondence between the image space coordinates and the actual ground coordinates, usually using a polynomial model or a rational function model. In practical applications, first select obvious ground feature points on the image as control points, then use the measured coordinate data on-site to calculate the transformation parameters, and finally perform resampling calculations on the entire image to obtain the geometrically corrected image data. For the orthorectification processing of geometric correction data, a terrain correction model needs to be established to eliminate the projection deformation caused by terrain undulation. The mathematical model of orthorectification can be expressed as:

[0050] ;

[0051] Where: represents the coordinate value after orthorectification; is the terrain correction coefficient; is the weight coefficient of the i-th control point; is the coordinate function of the i-th control point; is the projection deformation correction coefficient; is the weight coefficient of the j-th topographic feature point; is the elevation, slope, and azimuth angle function of the j-th topographic feature point; n is the number of control points; m is the number of topographic feature points.

[0052] After obtaining the orthorectified data, radiometric consistency processing is performed. Radiometric consistency processing is to eliminate the radiometric differences between images acquired at different times and by different sensors. Extract the gray histogram features of each band image, and then process the image through histogram matching. Histogram matching is the process of adjusting the gray distribution of the image to be processed to be similar to that of the reference image, so that images at different times have similar statistical features, facilitating subsequent temporal analysis.

[0053] Organizing the radiometric consistency data in a time series is to establish the temporal variation relationship of the seawall features. Establishing the temporal correlation relationship needs to consider factors such as the time interval between image acquisitions and seasonal changes, and associate the image data at different times by establishing a time series index. Finally, perform standardization processing on the multi-temporal correlation data, including unifying data formats, coordinate systems, resolutions, etc., to obtain standardized multi-temporal remote sensing data.

[0054] For example: Separate the bands of the acquired satellite images, and extract data in the near-infrared band (760 - 900 nm), red band (630 - 690 nm), and green band (520 - 600 nm). By setting the combination of the near-infrared band weight to 0.4, the red band weight to 0.35, and the green band weight to 0.25, the structural features of the seawall and the vegetation distribution are highlighted. Then select 5 obvious inflection points within the seawall area as ground control points, obtain the precise coordinates of these points using RTK measurement, and establish a geometric transformation model for correction. During orthorectification, use 1-meter resolution DEM data to establish a terrain correction model to eliminate the position deviation caused by terrain undulation. In radiometric consistency processing, select the summer image with the best lighting conditions as the reference, and perform histogram matching on the images of other seasons to make the gray distributions of images at different times tend to be consistent. Organize the image data of the four seasons in chronological order to establish a temporal analysis data set, and through standardization processing, ensure that the spatial resolution of all data is 2 meters and the projection coordinate system is unified as the UTM projection.

[0055] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0056] (1) According to the standardized multi-temporal remote sensing data, perform gray normalization transformation on the image data, construct a four-level pyramid image according to the scale ratio of 1:2:4:8, calculate the homogeneity index for each level of image through the multi-scale segmentation algorithm, determine the optimal segmentation threshold, extract the initial seawall target area, and generate seawall segmentation data;

[0057] (2) Perform spatial registration on the seawall segmentation data and the oblique photogrammetry data, establish the mapping relationship between geographical coordinates and pixel coordinates, use the bilinear interpolation method to perform dense matching on the corner points, calculate the three-dimensional coordinate values of the feature points, and generate three-dimensional point cloud data;

[0058] (3) Organize the three-dimensional point cloud data using the K-D tree spatial index structure, calculate the elevation value within each grid through least squares adjustment, calculate the local dip value by combining the coordinate differences of adjacent points, and extract the slope value according to the eight-neighborhood relationship to obtain three-dimensional elevation information;

[0059] (4) For the three-dimensional elevation information, establish a cross-section curve based on the elevation mutation points, extract the slope change trend along the cross-section direction, and determine the inner and outer contour lines of the seawall boundary range through the region growing method to obtain the seawall boundary data;

[0060] (5) Project the seawall boundary data onto the standardized multi-temporal remote sensing data of different time phases, calculate the gray difference value, texture change amount, and regional area change rate of adjacent time-phase images, and obtain the time-series change characteristic values through weighted fusion to generate the change situation data;

[0061] (6) Conduct spatio-temporal clustering analysis on the change situation data, extract the dynamic evolution characteristics of the seawall boundary, and combine with the spatial distribution characteristics of the seawall boundary data to obtain the spatio-temporal characteristic information of the seawall.

[0062] Specifically, perform gray normalization transformation on the standardized multi-temporal remote sensing data. Gray normalization transformation refers to the process of mapping the gray values of the image data into the interval [0, 1], aiming to eliminate the gray value differences under different sensors and different imaging conditions. After the transformation, construct a four-level pyramid image according to the ratio of 1:2:4:8. The pyramid image is a multi-resolution image representation method, and the resolution of each level of image is 1 / 2 of the previous level, obtained through downsampling. This structure is conducive to improving the efficiency of image processing. The multi-scale segmentation algorithm is an image segmentation method based on region growing, which determines the optimal segmentation threshold by calculating the homogeneity index. The homogeneity index includes parameters such as spectral feature similarity and shape feature similarity. For each pixel point, calculate the feature difference between it and the adjacent region. When the difference is less than the threshold, merge the pixel point into the corresponding region. In this way, the image is segmented into objects with similar features to form the initial seawall target area.

[0063] The seawall segmentation data needs to be spatially registered with the oblique photogrammetry data, and this process requires establishing a mapping relationship between geographic coordinates and pixel coordinates. The oblique photogrammetry data is image data obtained by taking pictures from multiple angles and contains the three-dimensional information of the target. Spatial registration uses bilinear interpolation to densely match the corner points. Bilinear interpolation is a method of interpolation in a two-dimensional plane, and the interpolation result is obtained by calculating the weighted average of the four nearest points around the point to be calculated. For each feature point, calculate its corresponding position in the images from different perspectives, and then use photogrammetry principles to calculate its three-dimensional coordinate values to generate three-dimensional point cloud data. For the generated three-dimensional point cloud data, a K-D tree spatial index structure is used for organization. The K-D tree is a data structure for indexing multi-dimensional spatial data. By continuously dividing the space to organize data points, it can improve the efficiency of spatial queries and nearest neighbor searches. After constructing the spatial index, calculate the elevation value within each grid through least squares adjustment, and then calculate the local dip value based on the coordinate differences of adjacent points.

[0064] The calculation formulas for the local dip value and the slope value are as follows:

[0065] ;

[0066] Where: is the slope value at point (i, j); is the weight coefficient of the p-th neighborhood point; is the three-dimensional coordinate of the p-th neighborhood point; is the distance between adjacent points; is the terrain undulation correction coefficient; p is the neighborhood point serial number, ranging from 1 to 8.

[0067] Based on the obtained three-dimensional elevation information, establish a cross-section curve with the elevation mutation point as the reference. The elevation mutation point refers to the position where the elevation value changes significantly, usually corresponding to the edge of the seawall or the structural change. Extract the slope change trend along the cross-section direction, and determine the inner and outer contour lines of the seawall boundary range through the region growing method. The region growing method is a method that starts from the seed point and gradually merges adjacent pixels with similar features into the growing region. For the extraction of the seawall boundary, select the elevation mutation point as the seed point and perform region growing according to the similarity of elevation values and slope values. Project the extracted seawall boundary data into the standardized multi-temporal remote sensing data of different time phases, and calculate the gray difference value, texture change amount, and regional area change rate of adjacent time-phase images. The gray difference value reflects the change of image brightness, the texture change amount represents the change degree of image texture features, and the regional area change rate reflects the dynamic change of the seawall range. By weighted fusion of these characteristic parameters, obtain the time-series change characteristic value to characterize the change of the seawall.

[0068] Performing spatio-temporal clustering analysis on the data of change situations is to identify the laws of the change of the seawall boundary over time. Spatio-temporal clustering analysis comprehensively considers the spatial location and time attributes of the data, clusters the regions with similar change characteristics together, so as to extract the dynamic evolution characteristics of the seawall boundary. Combining with the spatial distribution characteristics of the seawall boundary data, the spatio-temporal characteristic information of the seawall is obtained.

[0069] For example: Perform gray normalization on the acquired multi-temporal remote sensing images, convert the DN values of the original images to the interval [0, 1], and then construct a four-level pyramid image. Segment the top-level image (the lowest resolution), set the spectral heterogeneity threshold to 0.3 and the shape heterogeneity threshold to 0.5, and obtain the initial segmentation result through iterative calculation. Then use the multi-angle images obtained by oblique photography, select feature points for matching, and calculate the three-dimensional coordinates of the feature points through aerial triangulation. Organize these three-dimensional point cloud data into a K-D tree structure, perform least squares adjustment calculation on each 5-meter × 5-meter grid, and obtain the grid elevation value. Based on the elevation data, use the eight-neighborhood method to calculate the slope value, and identify the areas with a slope greater than 15 degrees as the candidate areas of the seawall. Extract the seawall boundary through the region growing method, and select the seed points at the positions of elevation mutations (the positions with an elevation difference greater than 2 meters). Finally, project the boundary onto three images of different time phases, calculate the average gray value, texture entropy value and area change rate of the boundary region, and obtain the time series change characteristic value through weighted average. Through spatio-temporal clustering analysis, determine the spatial range and change characteristics of the seawall.

[0070] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0071] (1) According to the spatio-temporal characteristic information of the seawall, set cross-section sampling points at 10-meter intervals in the main area of the seawall, construct a cross-section elevation curve through a triangular network, extract the slope change points of each cross-section from it, and generate slope sampling data;

[0072] (2) Conduct statistical analysis on the slope sampling data, calculate the ratio of the elevation difference and the horizontal distance between adjacent sampling points, screen the slope mutation points with 0.5 degrees as the threshold, connect all the mutation points to generate a slope gradient line, and obtain slope information;

[0073] (3) Compare the slope information with the characteristic curves in the material spectral feature library, calculate the spectral reflectance of each band, extract the material feature parameters, and establish a slope protection structure feature set;

[0074] (4) For the slope protection structure feature set, extract the three-dimensional coordinate values of each discrete point, calculate the elevation difference coefficient between adjacent points, establish a terrain change curve, and obtain terrain undulation data;

[0075] (5) Based on the terrain undulation data, statistically analyze the undulation change amplitude of each slope section, and set the points with a slope change exceeding 30% as segmentation nodes to obtain the segmented structural characteristics;

[0076] (6) Conduct a spatial correlation analysis on the slope information, the slope protection structure feature set, and the segmented structural characteristics, statistically analyze the correlation coefficients between each feature, and generate a seawall structure feature data set.

[0077] Specifically, based on the obtained spatio-temporal characteristics information of the seawall, cross-section sampling is carried out at a fixed interval of 10 meters in the main body area of the seawall. Cross-section sampling refers to setting sampling points on the vertical cross-section of the main body of the seawall, and the positions of these sampling points are determined by the spatial contour of the seawall. Constructing a cross-section elevation curve through a triangulation network means connecting the sampling points using an irregular triangulation network (TIN) to form a continuous surface model. The triangulation network is composed of a series of non-overlapping triangles, and the vertices of each triangle correspond to a sampling point with three-dimensional coordinates. Based on the constructed cross-section elevation curve, extract the slope change points of each cross-section. The slope change points refer to the positions on the elevation curve where the slope changes significantly, and these points usually correspond to the important feature positions of the seawall structure. When statistically analyzing the obtained slope sampling data, it is necessary to calculate the ratio of the elevation difference and the horizontal distance between adjacent sampling points. This ratio is actually the tangent value of the slope, and through this calculation, the slope change situation of each part of the seawall can be obtained. Using 0.5 degrees as the threshold for screening slope mutation points means that when the slope difference between two adjacent calculation points exceeds 0.5 degrees, this point is marked as a slope mutation point. By connecting all the mutation points, a slope gradient line is generated, and this gradient line reflects the change trend of the seawall slope.

[0078] Comparing the obtained slope information with the characteristic curves in the material spectral feature library is an important step. The material spectral feature library is a data set containing the spectral reflection characteristics of various common seawall slope protection materials (such as concrete, stone, vegetation, etc.). When calculating the spectral reflectance of each band, it is necessary to consider the intensity ratio of the incident light and the reflected light. Through this calculation and comparison, material characteristic parameters can be extracted, including the reflectance characteristics, absorption characteristics, etc. of the material, and these parameters together constitute the slope protection structure feature set.

[0079] For the data in the slope protection structure feature set, extract the three-dimensional coordinate values of each discrete point, and calculate the elevation difference coefficient between adjacent points. The calculation formula for the elevation difference coefficient is:

[0080] ;

[0081] Where: represents the elevation difference coefficient at point (r, q); is the weight coefficient of the mth sampling point; is the elevation value of adjacent sampling points; is the terrain undulation correction factor; is the distance attenuation coefficient; is the terrain gradient factor; N is the total number of sampling points.

[0082] By calculating the elevation difference coefficient, a terrain change curve can be established to reflect the undulation characteristics of the seawall surface. The terrain undulation data is an important index to characterize the severity of terrain change, which directly reflects the structural characteristics of the seawall slope protection. Based on the terrain undulation data, the undulation change amplitude of each slope section is statistically analyzed. When the slope change exceeds 30% of the point position, these positions are set as segmentation nodes. The determination of segmentation nodes is of great significance for understanding the structural distribution of the seawall because these nodes often correspond to the conversion positions of the seawall structure types or the positions of important functional areas.

[0083] Finally, a spatial correlation analysis is performed on the slope information, the slope protection structure feature set, and the segmented structure features. The spatial correlation analysis reveals the degree of association between different features by calculating the correlation coefficients between various features. The calculation of the correlation coefficients needs to consider the spatial distribution law and the numerical change trend of the features to generate a seawall structure feature dataset.

[0084] For example: First, within the main body area of a 500-meter-long seawall, 50 cross-section sampling points are set at 10-meter intervals. The three-dimensional coordinates of these points are obtained through RTK measurement, and a continuous cross-section elevation curve is generated using triangular network interpolation. When analyzing the slope change of each cross-section, it is found that there are obvious slope changes at 20 meters and 40 meters from the top of the embankment. The ratios of the elevation difference to the horizontal distance at these positions are 0.6 and 0.8 respectively, exceeding the threshold of 0.5 degrees, so they are marked as slope mutation points. By analyzing the distribution law of these mutation points, the main slope change characteristics of this section of the seawall are determined. Then, using hyperspectral remote sensing images, the spectral reflection characteristics of the slope protection of this section of the seawall are extracted, and by comparing with the material feature library, the structural characteristics of the upper part being concrete slope protection and the lower part being riprap slope protection are identified. When calculating the terrain undulation degree, it is found that the elevation difference coefficient at the position of the slope protection structure conversion reaches 1.5, which is much higher than the average value of 0.8 at other positions, and this position is subsequently determined as the segmentation node. In this way, the structural feature information of this section of the seawall is obtained, including slope distribution, material change, and segmentation features, etc.

[0085] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0086] (1) Extract the near-infrared band and the red band of the multispectral data from the seawall structure feature dataset, calculate the normalized difference vegetation index for each pixel point, and obtain the vegetation index distribution data;

[0087] (2) Conduct sliding window statistics on the vegetation index distribution data, calculate the mean and standard deviation of the vegetation index within each grid, establish a density classification standard, and obtain the vegetation distribution density data;

[0088] (3) Resample the vegetation distribution density data according to a 20×20-meter grid, extract the gray-level co-occurrence matrix features of each grid, and statistically analyze the angular second moment and entropy value parameters to obtain the texture feature data;

[0089] (4) Identify the vegetation types based on the texture feature data, statistically analyze the spatial distribution patterns of various vegetation types, and obtain the vegetation type data;

[0090] (5) Use the vegetation type data to extract the typical spectral curves of each type of vegetation, calculate the similarity with the spectral curves of the slope protection area, and obtain the slope protection ecologicalization degree data;

[0091] (6) Perform feature fusion on the vegetation distribution density data, vegetation type data, and slope protection ecologicalization degree data to generate the seawall ecologicalization feature information.

[0092] Specifically, extracting multi-spectral data from the seawall structure feature dataset is the primary step. The multi-spectral data contains spectral information in different bands. Among them, the near-infrared band and the red band are particularly important for vegetation identification because vegetation has significant differences in reflection characteristics in these two bands. The near-infrared band mainly reflects the biomass information of vegetation, while the red band mainly reflects the absorption characteristics of chlorophyll. The calculation formula for the normalized difference vegetation index (NDVI) of each pixel point is:

[0093] ;

[0094] Where: represents the normalized difference vegetation index value at coordinates (x, y); is the weight coefficient of the k-th spectral band; is the reflection value of the near-infrared band; is the reflection value of the red band; is the vegetation reflection characteristic correction coefficient; is the atmospheric influence correction coefficient; is the terrain influence correction coefficient; M is the number of bands participating in the calculation.

[0095] After obtaining the vegetation index distribution data, it is necessary to perform sliding window statistical analysis on it. The sliding window is a local feature extraction method that calculates the statistical features of the data within a window of a fixed size by moving the window over the data. Calculate the mean and standard deviation of the vegetation index within each grid, and these statistical values reflect the density and uniformity of vegetation cover in the local area. Based on these statistical values, establish a density classification standard, divide the vegetation cover degree into different levels, and form vegetation distribution density data. Resampling the vegetation distribution density data is to standardize the spatial resolution. Resampling according to a grid size of 20×20 meters can make the data have a unified spatial scale. On the resampled data, extract the gray-level co-occurrence matrix features of each grid. The gray-level co-occurrence matrix is an important tool for describing the texture features of an image, and it statistically analyzes the spatial correlation of the gray values in the image. The angular second moment reflects the smoothness of the image, while the entropy value parameter reflects the complexity of the image. These parameters together constitute the texture feature data of the vegetation distribution.

[0096] When identifying vegetation types based on the texture feature data, it is necessary to combine the typical texture features of different vegetation types. Different types of vegetation, such as herbaceous plants, shrubs, trees, etc., show different characteristics in texture features. By statistically analyzing the spatial distribution laws of various types of vegetation, establish the spatial distribution pattern of vegetation types and obtain detailed vegetation type data. For each identified vegetation type, it is necessary to extract its typical spectral curve. The typical spectral curve refers to the standard curve that can represent the spectral characteristics of this type of vegetation, and it contains the reflection characteristics of the vegetation in different bands. Calculate the similarity between these typical spectral curves and the actual spectral curves in the slope protection area, and evaluate the ecologicalization degree of the slope protection by comparing the shape features of the spectral curves and the reflection values of the key bands.

[0097] Integrate the vegetation distribution density data, vegetation type data, and slope protection ecologicalization degree data through feature fusion. The feature fusion process needs to consider the correlation and weight distribution between different data to ensure that the fused result can comprehensively reflect the ecologicalization characteristics of the seawall.

[0098] For example: First, extract data of the near-infrared band (wavelength range: 760 - 900 nm) and the red band (wavelength range: 630 - 690 nm) from the multispectral image. By calculating the normalized difference vegetation index, a vegetation index distribution map is obtained, and it is found that the distribution of vegetation index values in the seawall slope protection area shows obvious stratification characteristics. A 5×5 pixel sliding window is used to statistically analyze the vegetation index, and the mean and standard deviation within each window are calculated. Based on these statistical values, the vegetation density is divided into three levels: high, medium, and low. After resampling the data to a 20×20 m grid, the characteristic parameters of the gray-level co-occurrence matrix are calculated, and it is found that in areas with good vegetation coverage, the angular second moment value is lower and the entropy value is higher, indicating that the vegetation distribution is relatively uniform and has a higher complexity. According to these texture characteristics, two main vegetation types, herbaceous plants and shrubs, are identified on this section of the seawall. Among them, herbaceous plants are mainly distributed in the lower slope protection, while shrubs are mainly distributed in the upper slope protection. The typical spectral curves of these two types of vegetation are extracted and compared with the actual spectral curves in the slope protection area, and the calculated similarity values reflect that the ecologicalization degree of this section of the seawall is relatively good. By integrating all feature information, a seawall ecologicalization feature dataset is generated.

[0099] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0100] (1) According to the seawall ecologicalization feature information, group the field measurement data according to the spatial position and attribute type, construct a feature type comparison table, and establish an accuracy evaluation sample set;

[0101] (2) Divide the accuracy evaluation sample set into a training set and a validation set according to a 7:3 ratio, and select inspection samples through the stratified sampling method to obtain multi-level validation data;

[0102] (3) Use the multi-level validation data to calculate the coordinate deviation and attribute consistency of each inspection point, and statistically analyze the position accuracy index to obtain position accuracy data;

[0103] (4) Establish a confusion matrix based on the position accuracy data, calculate the recognition accuracy of each type of feature, and obtain attribute accuracy data;

[0104] (5) Perform spatio-temporal matching on the attribute accuracy data and the time-series images, calculate the change amount between each time phase, and obtain dynamic change inspection data;

[0105] (6) Conduct a comprehensive evaluation on the position accuracy data, attribute accuracy data, and dynamic change inspection data to generate a feature extraction quality evaluation report.

[0106] Specifically, the accuracy assessment work is carried out based on the ecological characteristic information of the seawall to reasonably organize and classify the field measurement data. The field measurement data is high-precision geographic information data obtained at the seawall site through measurement equipment such as RTK and total station, including spatial position coordinates and attribute characteristic information. Grouping these data according to spatial position means classifying and sorting the data according to different areas of the seawall (such as the top of the dyke, the slope of the dyke, the toe of the dyke, etc.); while grouping according to attribute types is to classify according to the characteristic attributes of the measurement points (such as elevation, slope, material, etc.). A characteristic type comparison table is constructed through this grouping method. The comparison table contains information such as the spatial coordinates, type attributes, and measurement accuracy of each measurement point, thus establishing a benchmark data set for accuracy assessment. The method of separating the training set and the validation set is adopted for the processing of the accuracy assessment sample set. Dividing the training set and the validation set in a ratio of 7:3 is to ensure that there are enough samples for model training, and at the same time leave a certain proportion of independent samples for verification. Stratified sampling is a sampling method that takes into account the data distribution characteristics, which ensures that samples of different types and different regions can be reasonably selected. In specific operations, the samples are first stratified according to different characteristic types, and then samples are randomly selected in each layer according to a certain proportion, so as to obtain representative multi-level verification data.

[0107] When using multi-level verification data for accuracy assessment, the coordinate deviation of each inspection point needs to be calculated first. The coordinate deviation refers to the difference between the coordinates of the extracted feature points and the field measurement coordinates, including the plane position deviation and the elevation deviation. The attribute consistency refers to the degree of conformity between the extracted feature types and the field survey results. By statistically analyzing these deviation values and consistency indicators, a data set reflecting the position accuracy is obtained.

[0108] A confusion matrix is established based on the position accuracy data, and the recognition accuracy rate of each type of feature is calculated. The confusion matrix is an important tool for evaluating the accuracy of classification results, and its calculation formula is:

[0109] ;

[0110] where: is the correct rate of the i-th type of feature being recognized as the j-th type; is the e-th weight factor; is the value of the confusion matrix element; is the feature weight coefficient; is the normalization coefficient; is the accuracy correction coefficient; K is the total number of feature categories; N is the total number of samples.

[0111] Performing spatio-temporal matching of attribute accuracy data with time-series images is to verify the temporal consistency of feature extraction results. Time-series images refer to remote sensing image data acquired at different time points. By calculating the feature change amounts between different time phases, the temporal stability and reliability of feature extraction results can be evaluated. When performing spatio-temporal matching, the impacts of factors such as image acquisition time, spatial resolution, and imaging conditions need to be considered. Comprehensive evaluation is carried out on position accuracy data, attribute accuracy data, and dynamic change inspection data. Comprehensive evaluation requires considering multiple aspects of indicators, including spatial position accuracy, attribute recognition accuracy, temporal consistency, etc. By comprehensively analyzing these indicators, a feature extraction quality assessment report is generated.

[0112] For example, during the feature extraction quality assessment process of a sea dike, first, data of 100 field measurement points were collected, and these measurement points covered various feature areas of the sea dike. According to spatial positions, these points were divided into four groups: the top of the dike area, the uphill area, the middle slope area, and the toe of the dike area. At the same time, according to attribute types, they were divided into three categories: elevation points, structural points, and vegetation points. According to the ratio of 7:3, 70 points were selected as training samples, and the remaining 30 points were used as verification samples. This ratio was maintained in each grouping to ensure the representativeness of the samples. The three-dimensional coordinates of these points were obtained through RTK measurement, and the measured values were compared with the feature extraction results to calculate the planar position deviation and elevation deviation. According to the calculation results, a confusion matrix was established to analyze the recognition situation of each type of feature. At the same time, remote sensing images of four periods of this section of the sea dike were collected. By comparing and analyzing the changes of feature points in different periods, the temporal consistency of feature extraction was verified. Finally, by integrating all evaluation indicators, a quality assessment report was formed, which detailed the accuracy situation of feature extraction and existing problems.

[0113] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0114] (1) Establish an index for the sea dike structure feature dataset according to spatial position and attribute type, construct a structure feature data table, and obtain structure feature storage data;

[0115] (2) Classify and code the ecological feature information of the sea dike, establish an ecological feature hierarchical relationship, and obtain ecological feature storage data;

[0116] (3) Quantitatively analyze the accuracy indicators in the feature extraction quality assessment report, establish a quality assessment index system, and obtain quality assessment storage data;

[0117] (4) Based on the structure feature storage data, establish a spatial reference framework, construct the core architecture of a geographic database, and obtain the basic database structure;

[0118] (5) Import the ecological feature storage data and quality assessment storage data into the basic database structure, establish data association relationships, and obtain an associated database structure;

[0119] (6) Optimize the hierarchical organization of the associated database structure to construct a seawall feature information database.

[0120] Specifically, establish an index for the seawall structure feature dataset. Index establishment refers to creating a directory structure for rapid data retrieval. Among them, the spatial location index is a spatial retrieval structure established based on the geographic coordinate system, including information such as the starting coordinates, ending coordinates, and spatial scope of the seawall, while the attribute type index is an attribute retrieval structure established for the structural features of the seawall (such as the embankment type, slope protection type, slope, etc.). The structural feature data table needs to include basic information such as field definitions, data types, and field lengths. At the same time, constraint conditions such as primary keys and foreign keys need to be set to ensure data integrity and consistency. The classification and coding of seawall ecological feature information is a process of systematically organizing data. The classification and coding adopt a hierarchical structure, organizing ecological features at different levels. For example, the first level is the vegetation type (herbs, shrubs, trees, etc.), the second level is the vegetation distribution density (sparse, medium, dense, etc.), and the third level is the degree of ecologicalization (low, medium, high, etc.). During the coding process, a unique coding rule needs to be set for each level to ensure the uniqueness and scalability of the coding. The establishment of the hierarchical relationship of ecological features needs to consider the subordinate relationship and association relationship between features, and express these relationships by establishing a hierarchical tree structure.

[0121] The quantitative analysis of the accuracy index in the feature extraction quality assessment report is a process of converting qualitative assessment results into measurable numerical indicators. The quality assessment index system includes position accuracy indicators (such as planar accuracy, elevation accuracy, etc.), attribute accuracy indicators (such as classification accuracy rate, integrity, etc.), and time consistency indicators (such as change detection accuracy rate, etc.). Each indicator needs to clearly define the calculation method, evaluation criteria, and weight coefficient to form an evaluation system. When establishing a spatial reference framework based on the structural feature storage data, it is necessary to first determine a unified coordinate system and projection method. The spatial reference framework is the basis of a geographic database, which defines the spatial positioning rules and measurement standards of data. The construction of the core architecture of a geographic database needs to consider multiple aspects such as data organization methods, storage structures, and index mechanisms. The core architecture includes basic functional modules such as a spatial data engine, attribute data management, and spatial indexing.

[0122] The process of importing ecological feature storage data and quality assessment storage data into the basic database structure requires establishing the association relationships between the data. The data association relationships are mainly achieved through primary key-foreign key constraints. For example, the seawall section number is used as the primary key to associate the structural features, ecological features, and quality assessment data. During the data import process, operations such as data format conversion, coordinate conversion, and attribute mapping are carried out to ensure the consistency and integrity of the data. The hierarchical organization optimization of the associated database structure is the last important step. The purpose of the hierarchical organization optimization is to improve data access efficiency and management convenience. The optimization process includes table structure optimization (such as field design, index optimization, etc.), relationship optimization (such as reducing redundancy, improving normalization, etc.), and storage optimization (such as partitioning strategy, caching mechanism, etc.).

[0123] Take a specific example to illustrate the entire database construction process: For a 2-kilometer-long seawall, establish an R-tree index for its structural feature data according to the spatial location for quickly retrieving seawall information in a specific area; at the same time, establish a B-tree index according to the attribute type (such as the slope protection type) for facilitating querying data according to specific attributes. The structural feature data table contains fields such as "seawall section number" (primary key), "starting point coordinates", "ending point coordinates", "slope protection type", "slope", etc.

[0124] In the ecological feature classification coding, a four-digit coding scheme is adopted. The first digit represents the vegetation type (1 - herbaceous, 2 - shrub, 3 - tree), the second digit represents the distribution density (1 - sparse, 2 - medium, 3 - dense), and the third and fourth digits represent the specific subclass features. For example, the code "2301" represents "densely distributed mangroves". These codes are associated with the structural feature table through foreign keys.

[0125] In the quality assessment index system, the position accuracy is represented by the root mean square error, the attribute accuracy is represented by the discrimination accuracy rate of the confusion matrix, and the time consistency is represented by the correct rate of change detection. These indicators are associated with other data through the seawall section number. When establishing the spatial reference framework, the 2000 National Geodetic Coordinate System and the Gauss-Krüger projection are adopted, and a two-layer storage structure including spatial data and attribute data is established. During the data import process, data conversion and loading are carried out through the ETL tool to ensure the unity of the coordinate system and the correct mapping of attribute fields. Finally, by establishing spatial indexes and attribute indexes, the query performance is optimized, and the seawall feature information database is constructed.

[0126] The above describes the method for extracting seawall feature information based on multi-source remote sensing data in the embodiments of the present application. Next, the system for extracting seawall feature information based on multi-source remote sensing data in the embodiments of the present application will be described. Please refer to Figure 2 One embodiment of the system for extracting seawall feature information based on multi-source remote sensing data in the embodiments of the present application includes:

[0127] Enhancement module 201 is used to enhance multi-spectral information of the collected multi-temporal satellite remote sensing images, aerial remote sensing images and UAV oblique images through band combination, perform geometric correction and ortho-rectification based on ground control points, and perform radiometric consistency processing of multi-source images by combining the histogram matching method to obtain standardized multi-temporal remote sensing data;

[0128] Extraction module 202 is used to extract the seawall target area according to the standardized multi-temporal remote sensing data, calculate the three-dimensional elevation information by combining the oblique photogrammetry data, determine the boundary range and change situation of the seawall through temporal feature comparison and analysis, and obtain the spatio-temporal feature information of the seawall;

[0129] Recognition module 203 is used to extract the slope information of the seawall through cross-section analysis based on the spatio-temporal feature information of the seawall, identify the slope protection structure based on the material spectral feature library, determine the sectional structure characteristics of the seawall by combining the calculation of terrain undulation, and generate a seawall structure feature data set;

[0130] Generation module 204 is used to calculate the normalized difference vegetation index for the seawall structure feature data set by using multi-spectral data, extract the vegetation distribution density and type by using the texture analysis method, determine the degree of slope protection ecologicalization through spectral curve matching, and generate the seawall ecologicalization feature information;

[0131] Establishment module 205 is used to establish an accuracy evaluation sample set based on the seawall ecologicalization feature information according to the field measurement data, calculate the position accuracy and attribute accuracy through multi-level cross-validation, and perform dynamic change inspection by combining the temporal images to generate a feature extraction quality evaluation report;

[0132] Construction module 206 is used to construct a seawall feature information database by organizing the seawall structure feature data set, the seawall ecologicalization feature information and the feature extraction quality evaluation report in a hierarchical data organization manner.

[0133] Through the collaborative cooperation of the above-mentioned various components, by performing band combination and multi-spectral information enhancement on multi-temporal satellite remote sensing images, aerial remote sensing images, and UAV oblique image data, the recognition accuracy of seawall features is improved, making the seawall structure and ecological features more clearly distinguishable. Geometric correction and ortho-rectification based on ground control points, combined with histogram matching method for radiometric consistency processing of multi-source images, ensure the spatial position accuracy and radiometric feature consistency between different data sources, laying a foundation for subsequent feature extraction. At the same time, using multi-scale segmentation algorithm to extract the seawall target area and calculating three-dimensional elevation information in combination with oblique photogrammetry data not only improves the accuracy of target extraction but also obtains the three-dimensional structure information of the seawall. By comparing and analyzing the temporal features, the boundary range and changes of the seawall are determined, realizing the dynamic monitoring of seawall features. Based on the spatio-temporal feature information of the seawall, cross-section profile analysis and material spectral feature recognition can accurately extract the slope information and revetment structure type of the seawall, and further determine the sectional structure features of the seawall in combination with the calculation of terrain undulation degree, making the extraction of seawall structure features more comprehensive and accurate. In addition, by calculating the normalized difference vegetation index, using texture analysis method to extract the vegetation distribution density and type, and then determining the degree of revetment ecologicalization through spectral curve matching, the quantitative evaluation of seawall ecological features is realized. Finally, based on the field measurement data, an accuracy evaluation sample set is established. Through multi-level cross-validation and dynamic change inspection of temporal images, not only the reliability of the feature extraction results is ensured, but also a quality evaluation system is established. The entire technical solution constructs a seawall feature information database through a hierarchical data organization method, realizing the standardized management and efficient utilization of seawall feature information.

[0134] 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 replacement 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 extracting seawall feature information based on multi-source remote sensing data, characterized in that The method for extracting sea dike feature information based on multi-source remote sensing data includes: For the collected multi-temporal satellite remote sensing images, multi-source aerial remote sensing images and UAV oblique image data, perform multi-spectral information enhancement through band combination, perform geometric correction and ortho-rectification based on ground control points, and perform radiometric consistency processing of multi-source images by combining the histogram matching method to obtain standardized multi-temporal multi-source remote sensing data, including: separating the bands of different data sources according to the multi-temporal satellite remote sensing images, multi-source aerial remote sensing images and UAV oblique image data, extracting near-infrared band, red band and green band information therefrom, and performing multi-spectral information enhancement through band combination to obtain multi-spectral enhanced data; for the multi-spectral enhanced data, construct a geometric transformation model through the coordinate information of ground control points, perform correction and resampling calculation on the image spatial position to obtain geometric correction data; input the geometric correction data into the ortho-rectification processing algorithm, establish a terrain correction model to eliminate projection deformation, and obtain ortho-rectification data; for the ortho-rectification data, extract the gray histogram features of each band image, and perform radiometric consistency processing of multi-source images by the histogram matching method to obtain radiometric consistency data; organize the radiometric consistency data in time series, establish a time series correlation relationship to obtain multi-temporal correlation data; perform standardization processing on the multi-temporal correlation data to obtain standardized multi-temporal multi-source remote sensing data; According to the standardized multi-temporal multi-source remote sensing data, use the multi-scale segmentation algorithm to extract the sea dike target area, calculate the three-dimensional elevation information of the sea dike in combination with the UAV oblique photogrammetry data, determine the boundary range and change situation of the sea dike through time series feature comparison and analysis, and obtain the spatio-temporal feature information of the sea dike; Based on the spatio-temporal feature information of the sea dike, extract the slope information of the sea dike through cross-section profile analysis, identify the slope protection structure based on the material spectral feature library, and determine the segmented structure feature of the sea dike in combination with the calculation of terrain undulation to generate a sea dike structure feature data set; For the sea dike structure feature data set, calculate the normalized difference vegetation index using multi-spectral data, extract the vegetation distribution density and type by using the texture analysis method, and determine the degree of slope protection ecologicalization through spectral curve matching to generate sea dike ecologicalization feature information; Based on the sea dike ecologicalization feature information, establish an accuracy evaluation sample set according to the field measurement data, calculate the position accuracy and attribute accuracy through multi-level cross-validation, and perform dynamic change inspection in combination with time series images to generate a feature extraction quality evaluation report; Construct a sea dike feature information database by organizing the sea dike structure feature data set, the sea dike ecologicalization feature information and the feature extraction quality evaluation report in a hierarchical data organization manner.

2. The method for extracting seawall feature information based on multi-source remote sensing data according to claim 1, wherein The step of, according to the standardized multi-temporal multi-source remote sensing data, using the multi-scale segmentation algorithm to extract the sea dike target area, calculating the three-dimensional elevation information of the sea dike in combination with the UAV oblique photogrammetry data, determining the boundary range and change situation of the sea dike through time series feature comparison and analysis, and obtaining the spatio-temporal feature information of the sea dike includes: Based on the standardized multi-temporal and multi-source remote sensing data, perform gray normalization transformation on the image data, construct a four-level pyramid image according to the scale ratio of 1:2:4:8, calculate the homogeneity index for each level of the image through the multi-scale segmentation algorithm, determine the optimal segmentation threshold, extract the initial seawall target area, and generate seawall segmentation data; Perform spatial registration on the seawall segmentation data and the drone oblique photogrammetry data, establish the mapping relationship between geographic coordinates and pixel coordinates, use the bilinear interpolation method to perform dense matching on the corner points, calculate the three-dimensional coordinate values of the feature points, and generate three-dimensional point cloud data; Organize the three-dimensional point cloud data using the K-D tree spatial index structure, calculate the elevation value within each grid through least squares adjustment, calculate the local inclination value by combining the coordinate differences of adjacent points, and extract the slope value according to the eight-neighborhood relationship to obtain the three-dimensional elevation information of the seawall; For the three-dimensional elevation information of the seawall, establish a cross-section curve based on the elevation mutation points, extract the slope change trend along the cross-section direction, and determine the inner and outer contour lines of the seawall boundary range through the region growing method to obtain the seawall boundary data; Project the seawall boundary data onto the standardized multi-temporal and multi-source remote sensing data of different time phases, calculate the gray difference, texture change amount, and regional area change rate of adjacent time-phase images, and obtain the time-series change characteristic value through weighted fusion to generate the change situation data; Perform spatio-temporal clustering analysis on the change situation data, extract the dynamic evolution characteristics of the seawall boundary, and combine the spatial distribution characteristics of the seawall boundary data to obtain the spatio-temporal characteristic information of the seawall.

3. The method for extracting seawall feature information based on multi-source remote sensing data according to claim 1, wherein Based on the spatio-temporal characteristic information of the seawall, extract the slope information of the seawall through cross-section profile analysis, identify the slope protection structure based on the material spectral feature library, and combine the calculation of the terrain undulation degree to determine the segmented structure characteristics of the seawall, generating a seawall structure feature data set, including: Based on the spatio-temporal characteristic information of the seawall, set cross-section sampling points at 10-meter intervals in the main area of the seawall, construct a cross-section elevation curve through a triangular network, extract the slope change points of each cross-section from it, and generate slope sampling data; Perform statistical analysis on the slope sampling data, calculate the ratio of the elevation difference and horizontal distance between adjacent sampling points, screen the slope mutation points with a threshold of 0.5 degrees, and connect all the mutation points to generate a slope gradient line to obtain the slope information; Compare the slope information with the characteristic curves in the material spectral feature library, calculate the spectral reflectance of each band, extract the material characteristic parameters, and establish a slope protection structure feature set; For the slope protection structure feature set, extract the three-dimensional coordinate values of each discrete point, calculate the elevation difference coefficient between adjacent points, and establish a terrain change curve to obtain the terrain undulation degree data; Based on the terrain undulation degree data, statistically analyze the undulation change amplitude of each slope section, and set the points with a slope change exceeding 30% as segmented nodes to obtain the segmented structure characteristics; Perform spatial correlation analysis on the slope information, slope protection structure feature set, and segmented structure characteristics, statistically analyze the correlation coefficients between each feature, and generate a seawall structure feature data set.

4. The method for extracting seawall feature information based on multi-source remote sensing data according to claim 1, wherein For the dataset of the seawall structure characteristics, calculate the normalized difference vegetation index (NDVI) using multispectral data, extract the vegetation distribution density and type by texture analysis methods, determine the degree of ecologicalization of the slope protection through spectral curve matching, and generate the ecologicalization characteristic information of the seawall, including: Extract the near-infrared band and red band of the multispectral data from the seawall structure characteristic dataset, calculate the normalized difference vegetation index for each pixel point, and obtain the vegetation index distribution data; Perform sliding window statistics on the vegetation index distribution data, calculate the mean and standard deviation of the vegetation index within each grid, establish a density classification standard, and obtain the vegetation distribution density data; Resample the vegetation distribution density data according to a 10×10 m grid, extract the gray-level co-occurrence matrix features of each grid, and statistically calculate the angular second moment and entropy value parameters to obtain the texture feature data; Based on the texture feature data, identify the vegetation types, and statistically analyze the spatial distribution law of each type of vegetation to obtain the vegetation type data; Use the vegetation type data to extract the typical spectral curve of each type of vegetation, calculate the similarity with the spectral curve of the slope protection area, and obtain the degree of ecologicalization data of the slope protection; Fuse the vegetation distribution density data, vegetation type data, and degree of ecologicalization data of the slope protection to generate the ecologicalization characteristic information of the seawall.

5. The method for extracting seawall feature information based on multi-source remote sensing data according to claim 1, wherein Based on the ecologicalization characteristic information of the seawall, establish an accuracy evaluation sample set according to the field measurement data, calculate the position accuracy and attribute accuracy through multi-level cross-validation, and perform dynamic change inspection in combination with time-series images, and generate a feature extraction quality evaluation report, including: According to the ecologicalization characteristic information of the seawall, group the field measurement data according to the spatial position and attribute type, construct a feature type comparison table, and establish an accuracy evaluation sample set; Divide the accuracy evaluation sample set into a training set and a validation set according to a 7:3 ratio, select inspection samples through stratified sampling, and obtain multi-level verification data; Use the multi-level verification data to calculate the coordinate deviation and attribute consistency of each inspection point, and statistically calculate the position accuracy index to obtain the position accuracy data; Establish a confusion matrix based on the position accuracy data, calculate the recognition accuracy of each type of feature, and obtain the attribute accuracy data; Perform spatio-temporal matching on the attribute accuracy data and the time-series images, calculate the change amount between each time phase, and obtain the dynamic change inspection data; Comprehensively evaluate the position accuracy data, attribute accuracy data, and dynamic change inspection data, and generate a feature extraction quality evaluation report.

6. The method for extracting seawall feature information based on multi-source remote sensing data according to claim 1, wherein Construct a seawall feature information database by organizing the seawall structure characteristic dataset, seawall ecologicalization characteristic information, and feature extraction quality evaluation report in a hierarchical data organization manner, including: Establish an index for the seawall structure characteristic dataset according to the spatial position and attribute type, construct a structure characteristic data table, and obtain the structure characteristic storage data; Classify and code the ecologicalization characteristic information of the seawall, establish an ecological characteristic hierarchical relationship, and obtain the ecological characteristic storage data; Quantitatively analyze the accuracy indicators in the feature extraction quality evaluation report, establish a quality evaluation index system, and obtain the quality evaluation storage data; Based on storing data with the said structural features, a spatial reference framework is established, the core architecture of the geographic database is constructed, and the basic database structure is obtained; The ecological feature storage data and the quality assessment storage data are imported into the basic database structure, and a data association relationship is established to obtain an associated database structure; The hierarchical organization of the associated database structure is optimized to construct a seawall feature information database.

7. A seawall feature information extraction system based on multi-source remote sensing data, which is used to implement the seawall feature information extraction method based on multi-source remote sensing data as described in any one of claims 1-6, characterized in that, The seawall feature information extraction system based on multi-source remote sensing data includes: An enhancement module, which is used to perform multi-spectral information enhancement on the collected multi-temporal satellite remote sensing images, multi-source aerial remote sensing images and unmanned aerial vehicle oblique image data through band combination, perform geometric correction and orthorectification based on ground control points, and perform multi-source image radiation consistency processing in combination with the histogram matching method to obtain standardized multi-temporal multi-source remote sensing data, including: separating the bands of different data sources based on the multi-temporal satellite remote sensing images, multi-source aerial remote sensing images and unmanned aerial vehicle oblique image data, extracting near-infrared band, red band and green band information from them, and performing multi-spectral information enhancement through band combination to obtain multi-spectral enhanced data; for the multi-spectral enhanced data, constructing a geometric transformation model through the coordinate information of the ground control points, performing correction and resampling calculation on the image spatial position to obtain geometric correction data; inputting the geometric correction data into the orthorectification processing algorithm, establishing a terrain correction model to eliminate projection deformation to obtain orthorectification data; for the orthorectification data, extracting the gray histogram features of each band image, performing multi-source image radiation consistency processing through the histogram matching method to obtain radiation consistency data; organizing the radiation consistency data in time series, establishing a time series association relationship to obtain multi-temporal associated data; performing standardization processing on the multi-temporal associated data to obtain standardized multi-temporal multi-source remote sensing data; An extraction module, which is used to extract the seawall target area using the multi-scale segmentation algorithm according to the standardized multi-temporal multi-source remote sensing data, calculate the three-dimensional elevation information of the seawall in combination with the unmanned aerial vehicle oblique photogrammetry data, and determine the boundary range and change situation of the seawall through time series feature contrast analysis to obtain the spatio-temporal feature information of the seawall; An identification module, which is used to extract the slope information of the seawall through cross-section analysis based on the spatio-temporal feature information of the seawall, identify the slope protection structure based on the material spectral feature library, and determine the segmented structural features of the seawall in combination with the calculation of the terrain undulation degree to generate a seawall structure feature data set; A generation module, which is used to calculate the normalized difference vegetation index for the seawall structure feature data set using multi-spectral data, extract the vegetation distribution density and type using the texture analysis method, and determine the degree of slope protection ecologicalization through spectral curve matching to generate seawall ecologicalization feature information; A establishment module, which is used to establish an accuracy assessment sample set based on the seawall ecologicalization feature information according to the field measurement data, calculate the position accuracy and attribute accuracy through multi-level cross-validation, and perform dynamic change inspection in combination with the time series images to generate a feature extraction quality assessment report; A building module for constructing a seawall feature information database by hierarchical data organization of a seawall structure feature dataset, seawall ecological feature information, and a feature extraction quality assessment report.

Images