Topographic feature recognition method, device and equipment based on remote sensing technology and medium

Abstract

The application relates to the technical field of remote sensing image analysis. A topographic feature recognition method, device, equipment and medium based on remote sensing technology are provided, wherein the method comprises the following steps: performing feature response analysis on a landform structure unit set to obtain semantic initial judgment confidence and a fracture boundary identifier of each landform structure unit; a landform morphology knowledge graph including a micro-landform topology rule and a morphology constraint is constructed based on a preset micro-landform morphology rule; based on the fracture boundary identifier and the landform morphology knowledge graph, topology connection of fracture perception is performed on the landform structure unit with the fracture boundary identifier, and a landform entity object with a complete geometric boundary is generated; based on the semantic initial judgment confidence, morphology-semantic consistency verification is performed on the landform entity object, and a topographic micro-landform recognition result with a semantic label and confidence is obtained, so that the technical effects of improving micro-landform boundary definition, enhancing fracture weak edge reconstruction capability and optimizing semantic-morphology consistency are achieved.

Description

Technical Field

[0001] This invention relates to the field of remote sensing image analysis technology, and in particular to methods, devices, equipment and media for terrain feature recognition based on remote sensing technology. Background Technology

[0002] With the rapid development of remote sensing technology, high-resolution remote sensing imagery plays an increasingly important role in fields such as topographic mapping, geological exploration, and environmental monitoring. Micro-topography identification, as a core component of topographic feature analysis, directly impacts the reliability of critical applications such as geological disaster early warning and mineral resource exploration. Effective micro-topography identification aims to extract geomorphic entities from complex remote sensing data and assign them semantic labels that conform to geomorphological principles, providing structured spatial data support for geographic information systems.

[0003] Traditional techniques employ multispectral feature fusion to segment geomorphic units from remote sensing images. However, this method fails to effectively address the decoupling of spectral and textural features caused by sub-pixel mixing, resulting in blurred geomorphic boundaries. When dealing with fractured and weakly edged geomorphic structures, traditional segmentation techniques generate severely fragmented units and lack fracture boundary reconstruction mechanisms, leading to broken geometric structures of geomorphic entities. Furthermore, traditional methods often neglect consistency verification between morphological features and semantic labels, relying solely on statistical similarity matching, resulting in topological logical contradictions in the micro-geomorphic identification results. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, device, equipment and medium for terrain feature recognition based on remote sensing technology to address the above-mentioned technical problems, so as to improve the clarity of micro-topographic boundaries, enhance the reconstruction capability of weak fracture edges and optimize semantic-morphological consistency.

[0005] In a first aspect, this application provides a terrain feature recognition method based on remote sensing technology, the method comprising:

[0006] The system acquires input multi-source remote sensing images and auxiliary terrain data, performs data preprocessing on the multi-source remote sensing images and auxiliary terrain data, and obtains fused data with radiometric correction and geometric registration.

[0007] Anti-mixing pixel feature extraction is performed on the fused data to generate a multimodal feature map;

[0008] Weak edge response segmentation is performed on the multimodal feature map to generate a fragmented set of geomorphic structural units;

[0009] Feature response analysis was performed on the set of geomorphic structural units to obtain the initial semantic confidence and fault boundary identifiers of each geomorphic structural unit.

[0010] Based on pre-defined micro-topographic morphology rules, a geomorphological knowledge graph including micro-topographic topology rules and morphological constraints is constructed.

[0011] Based on fault boundary markers and geomorphological knowledge graphs, fault-aware topological connections are made to geomorphological structural units with fault boundary markers to generate geomorphological entity objects with complete geometric boundaries.

[0012] Based on the initial semantic confidence level, morphological-semantic consistency verification is performed on the landform entity objects to obtain the terrain micro-landform recognition results with semantic labels and confidence levels.

[0013] In one embodiment, based on fault boundary markers and a geomorphological knowledge graph, fault-aware topological connections are performed on geomorphic structural units with fault boundary markers to generate geomorphic entity objects with complete geometric boundaries, including:

[0014] Based on the direction vector in the fault boundary identifier and the extension rules in the geomorphological knowledge graph, the path derivation for fault perception is performed to obtain a set of candidate connection paths.

[0015] A multi-dimensional compatibility evaluation is performed on the candidate connection path set to generate a topology-optimized path;

[0016] Based on topology optimization path, knowledge-guided geometric reconstruction of the set of geomorphic structural units is performed to obtain geomorphic entity objects with complete geometric boundaries.

[0017] In one embodiment, path derivation for fracture sensing is performed based on the direction vector in the fracture boundary identifier and the extension rules in the geomorphological knowledge graph, resulting in a set of candidate connection paths, including:

[0018] Trajectory prediction based on fracture sensing is performed on the direction vector in the fracture boundary marker to obtain the initial extension trajectory. The expression for the initial extension trajectory is:

[0019] in, Indicates the initial extended trajectory. This represents the direction vector in the fracture boundary identifier. The direction vector represents the Gaussian diffusion kernel. Indicates the parameter of directional uncertainty. Indicates position Direction Edge gradient field, This represents the morphological interruption probability mapping function. Indicates the direction of the search range;

[0020] The initial extended trajectory is corrected for topological compliance based on the extension rules in the geomorphological knowledge graph, and a rule-constrained trajectory is generated.

[0021] Multiple hypothesis path generation is performed on the rule-constrained trajectory to obtain a set of candidate connection paths.

[0022] In one embodiment, a knowledge-guided geometric reconstruction of a set of geomorphic structural units is performed based on a topology optimization path to obtain a geomorphic entity object with complete geometric boundaries, including:

[0023] Based on the topology optimization path and the morphological constraints in the geomorphological knowledge graph, the geometric structure of the geomorphological structural unit set is optimized to obtain the geometrically optimized geomorphological structural unit set.

[0024] Knowledge-guided boundary reconstruction is performed on the geometrically optimized set of geomorphic structural units to generate a complete geometric boundary. The expression for the complete geometric boundary is:

[0025]

[0026] in, Represents the complete geometric boundary. This represents the set of geomorphic structural units after geometric optimization. A knowledge graph representing landform morphology. Represents the energy term of the geometric fit. Represents the energy term of the topological rule. Describe the candidate boundary solution space. This represents the candidate boundary during the optimization process. Represents the geometric constraint weights. Indicates the topology constraint weights;

[0027] Based on the complete geometric boundaries and topological rules in the geomorphological knowledge graph, geomorphological entity objects are constructed to obtain geomorphological entity objects with complete geometric boundaries.

[0028] In one embodiment, feature response analysis is performed on the set of geomorphic structural units to obtain the initial semantic confidence level and fault boundary identifiers of each geomorphic structural unit, including:

[0029] Multi-source feature synergistic analysis is performed on the set of geomorphic structural units to obtain the fused response vector of spectrum-texture-topography. The expression of the fused response vector is:

[0030]

[0031] in, Indicates the first The fused response vector of each geomorphic structural unit Represents the spectral eigenvector. Represents texture feature vectors, Represents the terrain feature vector. This represents the spectral variation feature extraction operator. This represents a multi-scale texture feature extraction operator. This represents the terrain-derived feature extraction operator. Represents the cross-modal feature transformation function. Indicates the feature importance weight. This represents the total dimension of the feature vector. Indicates the feature dimension index;

[0032] The similarity of the fused response vector is measured based on a pre-set micro-topographic prototype feature library to generate a semantic initial judgment confidence score.

[0033] A fracture-sensitive edge field analysis is performed on the set of geomorphic structural units to generate fracture boundary markers including direction vectors.

[0034] In one embodiment, weak edge response segmentation is performed on the multimodal feature map to generate a fragmented set of geomorphic structural units, including:

[0035] Multi-scale edge gradient calculation is performed on the multimodal feature map to obtain the multi-scale edge response map;

[0036] Adaptive edge connection with fracture detection is performed on multi-scale edge response maps to generate continuous edge contour maps;

[0037] A watershed transformation with morphological constraints is performed on the multimodal feature map based on the continuous edge contour map to obtain a set of fragmented geomorphic structural units.

[0038] In one embodiment, the expression for the continuous edge contour map is:

[0039]

[0040] in, Represents a continuous edge contour map. Represents a multi-scale edge response map. Represents the coordinates of the edge points. Point Edge direction angle, This represents the kernel function for directional consistency. express arrive The connection path, Indicates the maximum fracture distance threshold. It represents Euclidean distance.

[0041] Secondly, this application also provides a terrain feature recognition device based on remote sensing technology, the device comprising:

[0042] The data preprocessing module is used to acquire input multi-source remote sensing images and auxiliary terrain data, perform data preprocessing on the multi-source remote sensing images and auxiliary terrain data, and obtain fused data with radiometric correction and geometric registration.

[0043] The anti-aliasing feature extraction module is used to extract anti-aliasing pixel features from the fused data and generate a multimodal feature map.

[0044] The weak edge segmentation module is used to perform weak edge response segmentation on the multimodal feature map to generate a fragmented set of geomorphic structure units;

[0045] The feature response analysis module is used to perform feature response analysis on the set of geomorphic structural units to obtain the semantic preliminary confidence and fault boundary identifiers of each geomorphic structural unit.

[0046] The knowledge graph construction module is used to construct a geomorphological knowledge graph that includes micro-topographic rules and morphological constraints based on pre-defined micro-topographic morphology rules.

[0047] The fracture-aware connection module is used to perform fracture-aware topological connections on geomorphic structural units with fracture boundary markers based on fracture boundary identifiers and geomorphological knowledge graphs, generating geomorphic entity objects with complete geometric boundaries.

[0048] The morphological and semantic verification module is used to perform morphological-semantic consistency verification on landform entities based on the initial semantic confidence level, and obtain the terrain micro-landform recognition results with semantic labels and confidence levels.

[0049] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods in the first aspect of this application.

[0050] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods in the first aspect of this application.

[0051] The method, apparatus, equipment, and medium for terrain feature recognition based on remote sensing technology provided in this application include: The method acquires multi-source remote sensing images and auxiliary terrain data, and performs data preprocessing such as radiometric correction and geometric registration to provide fused data for feature extraction; anti-mixing pixel feature extraction is performed on the fused data to generate a multimodal feature map, which can reduce the interference of spectral-texture feature coupling caused by sub-pixel mixing and improve the clarity of micro-landform boundaries; after weak edge response segmentation of the multimodal feature map to obtain a set of fragmented landform structure units, the semantic initial judgment confidence and fracture boundary identification are obtained by combining feature response analysis; at the same time, a landform morphology knowledge graph is constructed based on pre-set micro-landform morphology rules, which can provide standardized rule support for subsequent landform structure processing.

[0052] By using fault boundary markers and geomorphological knowledge graphs to perform fault-aware topological connections on geomorphological structural units with fault boundaries, the identification and geometric reconstruction of weak fault edges can be enhanced, thereby improving the reconstruction capability of weak fault edges. By verifying the morphological-semantic consistency of geomorphic entities based on the initial semantic confidence level, topological logical contradictions in the recognition results can be reduced, and the semantic and morphological matching consistency in the process of topographic micro-geomorphological recognition can be optimized. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 A flowchart of a terrain feature recognition method based on remote sensing technology in one embodiment of the present invention;

[0055] Figure 2 This is a flowchart illustrating how, in one embodiment of the present invention, topological connections are made to geomorphic structural units with fracture boundary markers based on fracture boundary identifiers and geomorphological knowledge graphs to generate geomorphic entity objects with complete geometric boundaries.

[0056] Figure 3 This is a structural diagram of a terrain feature recognition device based on remote sensing technology in one embodiment of the present invention. Detailed Implementation

[0057] To make the above-mentioned objects, features, and advantages of this application more apparent and understandable, the specific embodiments of this application will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0058] First, the application scenarios of the embodiments of this application are described. In the embodiments of this application, a method, apparatus, equipment, and medium for terrain feature recognition based on remote sensing technology are provided, applicable to scenarios such as terrain mapping, geological exploration, and environmental monitoring.

[0059] In illustrative purposes, the terrain feature recognition method, device, equipment and medium based on remote sensing technology provided in the embodiments of this application can also be applied to application scenarios such as geological disaster early warning, mineral resource exploration, and geographic information system data update. This is only an example and does not limit the specific application scenarios.

[0060] like Figure 1 As shown, this application provides a terrain feature recognition method based on remote sensing technology, which includes:

[0061] S101: Acquire the input multi-source remote sensing images and auxiliary topographic data, perform data preprocessing on the multi-source remote sensing images and auxiliary topographic data, and obtain fused data with radiometric correction and geometric registration.

[0062] For example, the recognition and processing terminal receives externally input multi-source remote sensing images and auxiliary terrain data. The multi-source remote sensing images include images acquired by different sensors at different times, and the auxiliary terrain data includes digital elevation models, slope aspect, and other geomorphological data. If the relevant data is already stored in a designated storage device, the recognition and processing terminal retrieves the corresponding multi-source remote sensing images and auxiliary terrain data from that storage device.

[0063] The identification and processing terminal performs radiometric correction on the acquired multi-source remote sensing images. Through operations such as radiometric calibration and atmospheric correction, it eliminates radiometric errors caused by factors such as atmospheric scattering and sensor response differences, thereby ensuring the accuracy of radiometric information in multi-source remote sensing images.

[0064] After radiometric correction is completed, the identification and processing terminal uses the geographic coordinate system of the auxiliary terrain data as a reference to perform geometric registration processing on the radiometrically corrected multi-source remote sensing images and the auxiliary terrain data. Through geometric fine correction, image registration and other operations, geometric distortion in the data is eliminated, so that the spatial position of the multi-source remote sensing images and the auxiliary terrain data is consistent.

[0065] The identification and processing terminal spatially fuses multi-source remote sensing images that have undergone radiometric correction and geometric registration with auxiliary terrain data to obtain fused data with radiometric correction and geometric registration.

[0066] S102: Extract anti-mixing pixel features from the fused data to generate a multimodal feature map.

[0067] For example, the recognition processing terminal performs anti-mixing pixel feature extraction on the fused data of radiometric correction and geometric registration, eliminates feature interference caused by sub-pixel mixing through mixed pixel decomposition, completes the decoupling analysis of spectral features, texture features and terrain features, and performs normalization and multi-dimensional integration of various features to generate multimodal feature maps.

[0068] S103: Perform weak edge response segmentation on the multimodal feature map to generate a fragmented set of geomorphic structural units.

[0069] For example, the recognition processing terminal performs multi-scale edge gradient calculation on the multimodal feature map to extract weak edge response information, forms a continuous edge contour through fracture-aware adaptive edge connection, and performs image segmentation under morphological constraints on the multimodal feature map based on the edge contour to generate a fragmented set of geomorphic structural units.

[0070] S104: Perform feature response analysis on the set of geomorphic structural units to obtain the initial semantic confidence and fault boundary identifiers of each geomorphic structural unit.

[0071] For example, the recognition and processing terminal performs multi-source feature collaborative analysis on the set of geomorphic structural units, extracts comprehensive response information of spectral features, texture features and topographic features, measures the similarity between the comprehensive response information and the preset micro-geomorphic prototype features to determine the degree of semantic matching, and performs fracture-sensitive edge field detection on the set of geomorphic structural units to locate the edge fracture position and determine the corresponding direction information, thereby obtaining the semantic preliminary judgment confidence and fracture boundary identification of each geomorphic structural unit.

[0072] S105: Based on pre-defined micro-topographic morphology rules, construct a geomorphological knowledge graph that includes micro-topographic topology rules and morphological constraints.

[0073] For example, the recognition and processing terminal summarizes and organizes the spatial relationships between micro-landforms according to the preset micro-landform morphology rules to form micro-landform topology rules, standardizes and defines the geometric and extended attributes of micro-landforms to form morphological constraints, and integrates the micro-landform topology rules and morphological constraints in a structured way to construct a landform morphology knowledge graph that includes micro-landform topology rules and morphological constraints.

[0074] S106: Based on fault boundary markers and geomorphological knowledge graphs, perform fault-aware topological connections on geomorphological structural units with fault boundary markers to generate geomorphological entity objects with complete geometric boundaries.

[0075] For example, the identification and processing terminal performs rule matching between the fault boundary markers of each landform structural unit and the landform morphology knowledge graph. Based on the directional information in the fault boundary markers and the extension rules in the landform morphology knowledge graph, it derives candidate connection paths, evaluates the geometric compatibility and topological compliance of the candidate connection paths, and selects the optimal connection path that meets the rule constraints.

[0076] The identification and processing terminal performs topological association of landform structural units with fracture boundary markers based on the optimal connection path, optimizes the spatial positional relationship of landform structural units through geometric structure optimization, and then performs boundary reconstruction to complete the fracture parts, thereby completing the integration of landform structural units and generating landform entity objects with complete geometric boundaries.

[0077] S107: Based on the initial semantic confidence level, perform morphological-semantic consistency verification on the landform entity object to obtain the terrain micro-landform recognition result with semantic label and confidence level.

[0078] For example, the recognition and processing terminal performs consistency verification of the form and semantics of the landform entity object based on the initial semantic confidence level and the relevant rules in the landform knowledge graph. It checks the degree of matching between the geometric morphological features of the landform entity object and the corresponding semantic information, judges the verification results, and optimizes the corresponding semantic labels and confidence levels to obtain the terrain micro-landform recognition results with semantic labels and confidence levels.

[0079] One embodiment of this application provides a terrain feature recognition method based on remote sensing technology. By acquiring multi-source remote sensing images and auxiliary terrain data and performing data preprocessing such as radiometric correction and geometric registration, fused data can be provided for feature extraction. Anti-mixing pixel feature extraction is performed on the fused data to generate a multimodal feature map, which can reduce the interference of spectral-texture feature coupling caused by sub-pixel mixing and improve the clarity of micro-landform boundaries. After weak edge response segmentation of the multimodal feature map to obtain a set of fragmented landform structure units, the semantic preliminary judgment confidence and fracture boundary identification are obtained by combining feature response analysis. At the same time, a landform morphology knowledge graph is constructed based on pre-set micro-landform morphology rules, which can provide standardized rule support for subsequent landform structure processing.

[0080] By using fault boundary markers and geomorphological knowledge graphs to perform fault-aware topological connections on geomorphological structural units with fault boundaries, the identification and geometric reconstruction of weak fault edges can be enhanced, thereby improving the reconstruction capability of weak fault edges. By verifying the morphological-semantic consistency of geomorphic entities based on the initial semantic confidence level, topological logical contradictions in the recognition results can be reduced, and the semantic and morphological matching consistency in the process of topographic micro-geomorphological recognition can be optimized.

[0081] like Figure 2 As shown, based on fault boundary markers and a geomorphological knowledge graph, fault-aware topological connections are performed on geomorphic structural units with fault boundary markers to generate geomorphic entity objects with complete geometric boundaries, including:

[0082] S201: Based on the direction vector in the fault boundary identifier and the extension rules in the geomorphological knowledge graph, the path derivation for fault perception is performed to obtain a set of candidate connection paths.

[0083] For example, the identification and processing terminal performs fault boundary identification parsing on each geomorphic structural unit with fault boundary markings and extracts the direction vector corresponding to each fault boundary.

[0084] Based on the type of geomorphic structural unit, the identification and processing terminal retrieves the corresponding extension rules from the geomorphological knowledge graph, aligns and associates the direction vectors with the extension rules one by one, and clarifies the constraints for path derivation. Based on these constraints, the terminal simulates the boundary extension trajectory using a fracture sensing algorithm, deduces multiple logically possible paths, and summarizes all deduced paths to form a candidate connection path set.

[0085] The directional vector includes vector information related to the extension direction and tilt angle of the fault boundary. The extension rules include the boundary continuity extension specifications formed by micro-topography in natural evolution and spatial distribution.

[0086] S202: Perform multi-dimensional compatibility evaluation on the candidate connection path set to generate topology-optimized paths.

[0087] For example, the identification and processing terminal defines specific indicators for multi-dimensional compatibility evaluation and verifies each candidate connection path in the candidate connection path set according to the indicators. In the geometric morphology adaptation evaluation, the identification and processing terminal compares the degree of fit between the candidate connection path and the contour curve and boundary endpoints of the geomorphic structure unit.

[0088] The identification and processing terminal verifies whether candidate connection paths conform to the topological rules in the geomorphological knowledge graph during the topological relationship compliance assessment, and judges the path fit by referring to the natural evolution law of micro-geomorphology during the geomorphological evolution rationality assessment. The identification and processing terminal scores the candidate connection paths based on the comprehensive evaluation results, eliminates low-scoring paths, and selects the highest-scoring path as the topology optimization path.

[0089] Among them, the geometric morphology adaptability assessment is a verification operation on the degree of fit between the candidate connection path and the contour of the geomorphic structure unit. The multi-dimensional compatibility assessment includes a comprehensive judgment on the compliance of the path topology and the rationality of the geomorphic evolution.

[0090] S203: Based on topology optimization path, perform knowledge-guided geometric reconstruction of the set of geomorphic structural units to obtain geomorphic entity objects with complete geometric boundaries.

[0091] For example, the identification and processing terminal uses the topology optimization path as a basis to delineate the spatial connection range of each unit in the set of geomorphic structural units, and clarifies the connection nodes and connection methods between units. The identification and processing terminal retrieves the topological rules and morphological constraints in the geomorphological knowledge graph to determine the specific standards for geometric reconstruction.

[0092] The recognition and processing terminal calibrates the position of the geomorphic structural units according to the reconstruction standards, and adjusts the spatial coordinates of the units to match the connection nodes. The recognition and processing terminal completes the boundaries of the fractured parts, generates missing boundary line segments to fill the fracture gaps, smooths and regularizes the overall outline, and integrates the reconstructed geomorphic structural units to form a complete geometric shape, resulting in a geomorphic entity object with complete geometric boundaries.

[0093] Spatial connectivity refers to the spatial area defined by which geomorphic structural units connect with each other. Knowledge-guided geometric reconstruction includes a series of reconstruction operations on geomorphic structural units, such as position calibration, boundary completion, and morphological regularization, under the rule constraints of the geomorphological knowledge graph.

[0094] In one embodiment, path derivation for fracture sensing is performed based on the direction vector in the fracture boundary identifier and the extension rules in the geomorphological knowledge graph, resulting in a set of candidate connection paths, including:

[0095] (1) The direction vector in the fracture boundary marker is used for fracture sensing trajectory prediction to obtain the initial extension trajectory. The expression of the initial extension trajectory is:

[0096] in, Indicates the initial extended trajectory. This represents the direction vector in the fracture boundary identifier. The direction vector represents the Gaussian diffusion kernel. Indicates the parameter of directional uncertainty. Indicates position Direction Edge gradient field, This represents the morphological interruption probability mapping function. Indicates the search range in terms of direction.

[0097] For example, the identification and processing terminal extracts the direction vectors of each geomorphic structural unit with fracture boundary markings, clarifying the direction vector corresponding to each fracture boundary. The identification and processing terminal determines the specific form of the Gaussian diffusion kernel of the direction vector and configures the basis for the value of the direction uncertainty parameter.

[0098] The identification and processing terminal obtains the edge gradient field at the location, calls the morphological interruption probability mapping function, and combines the direction search range to predict the trajectory of fracture perception, thus obtaining the initial extended trajectory.

[0099] The direction vector includes vector information related to the extension direction and tilt angle of the fracture boundary. The morphological interruption probability mapping function is used to characterize the probability of morphological interruption during trajectory extension, and the direction search range is used to limit the directional interval of trajectory extension.

[0100] (2) Based on the extension rules in the geomorphological knowledge graph, the initial extension trajectory is corrected for topological compliance, and a rule-constrained trajectory is generated.

[0101] For example, the recognition and processing terminal retrieves extension rules matching the current geomorphological structure unit type from the geomorphological knowledge graph, and verifies each segment of the initial extension trajectory against the extension rules point by point. The recognition and processing terminal marks the trajectory segments in the initial extension trajectory that do not conform to the extension rules.

[0102] The recognition and processing terminal corrects and adjusts the marked trajectory segments that do not conform to the rules, so that the trajectory segments conform to the constraints of the extension rules, and generates a rule-constrained trajectory.

[0103] The extension rules in the geomorphological knowledge graph include topological constraints and morphological specifications for the continuous extension of micro-geomorphic boundaries. Topological compliance correction includes adjustments to the direction, length, and curvature of trajectory segments.

[0104] (3) Generate multiple hypothetical paths for the rule-constrained trajectory to obtain a set of candidate connection paths.

[0105] For example, the recognition and processing terminal uses the rule-constrained trajectory as a benchmark and, under the constraint of the direction uncertainty parameter, performs perturbation adjustments on the extension direction of the trajectory to generate multiple different path variants. The recognition and processing terminal performs compliance verification on each path variant to determine whether the path variant conforms to the extension rules in the geomorphological knowledge graph.

[0106] The identification and processing terminal retains path variants that meet the compliance verification standards, removes path variants that do not meet the standards, and summarizes all retained path variants to obtain a set of candidate connection paths.

[0107] The orientation uncertainty parameter characterizes the degree of dispersion of the orientation vector, and path variants include paths with different extension details generated based on the rule-constrained trajectory. Multi-hypothesis path generation is used to cover different connection scenarios and enhance the richness of path selection.

[0108] In one embodiment, a knowledge-guided geometric reconstruction of a set of geomorphic structural units is performed based on a topology optimization path to obtain a geomorphic entity object with complete geometric boundaries, including:

[0109] (1) Based on the topology optimization path and the morphological constraints in the geomorphological knowledge graph, the set of geomorphological structural units is geometrically optimized to obtain the set of geomorphological structural units after geometric optimization.

[0110] For example, the identification and processing terminal analyzes the node distribution and line segment orientation of the topology optimization path to clarify the spatial position requirements of the connecting nodes and connecting line segments of each unit in the geomorphological structure unit set. The identification and processing terminal retrieves morphological constraints that match the type of the geomorphological structure unit set from the geomorphological knowledge graph to clarify the constraints such as the geometric size range, contour curvature limit, and spatial spacing specification of the geomorphological structure unit.

[0111] Based on the connection requirements of the topology optimization path, the recognition and processing terminal performs spatial coordinate fine-tuning on each unit in the set of geomorphic structure units to match the boundary endpoints of the units with the connection nodes. According to the contour curvature limit of the morphological constraint, the recognition and processing terminal smooths the edge curves of the units to eliminate jagged protrusions and depressions in the contour.

[0112] The recognition and processing terminal performs morphological compliance verification on the geomorphic structural units that have undergone spatial coordinate fine-tuning and edge smoothing, confirming that each unit conforms to the morphological constraints in the geomorphological knowledge graph; the recognition and processing terminal summarizes all geomorphic structural units that have passed the compliance verification to obtain a geometrically optimized set of geomorphic structural units.

[0113] In this context, node distribution refers to the arrangement of feature points along the topology optimization path used to locate the connection points of geomorphic structural units, while line segment orientation refers to the direction of the lines connecting the feature points in the topology optimization path. Morphological constraints are the standardized requirements for geometric attributes and spatial forms set for different types of micro-landforms in the geomorphological knowledge graph. Spatial coordinate fine-tuning is a precise adjustment operation performed on the planar position and elevation information of geomorphic structural units. Morphological compliance verification is a check operation to verify whether the geometric form of the geomorphic structural units meets the morphological constraint requirements.

[0114] (2) Knowledge-guided boundary reconstruction is performed on the geometrically optimized set of geomorphic structural units to generate a complete geometric boundary. The expression for the complete geometric boundary is:

[0115]

[0116] in, Represents the complete geometric boundary. This represents the set of geomorphic structural units after geometric optimization. A knowledge graph representing landform morphology. Represents the energy term of the geometric fit. Represents the energy term of the topological rule. Describe the candidate boundary solution space. This represents the candidate boundary during the optimization process. Represents the geometric constraint weights. This represents the topology constraint weight.

[0117] For example, the identification and processing terminal retrieves the geometrically optimized set of geomorphic structural units and geomorphic morphology knowledge graph, while determining the range of the candidate boundary solution space, clarifying the calculation basis of the geometric fitting energy term and the topological rule energy term, and setting the assignment rules for the geometric constraint weights and topological constraint weights.

[0118] The recognition and processing terminal generates multiple sets of candidate boundaries with different shapes in the candidate boundary solution space. For each set of candidate boundaries, it extracts the edge features of the geometrically optimized set of geomorphic structural units, calculates the geometric fit between the candidate boundary and the set of units based on the edge features, and converts the fit into the value of the geometric fitting energy term.

[0119] For each group of candidate boundaries, the recognition and processing terminal checks whether the connection method and closure shape of the candidate boundaries meet the rule requirements by referring to the topological rules in the geomorphological knowledge graph. The verification results are converted into the numerical value of the topological rule energy term. According to the assignment rules, the recognition and processing terminal multiplies the geometric fitting energy term with the geometric constraint weight, multiplies the topological rule energy term with the topological constraint weight, and then adds the two product results to obtain the total energy value of each group of candidate boundaries.

[0120] The recognition and processing terminal compares the total energy values ​​of all candidate boundaries and selects the candidate boundary with the smallest total energy value. The recognition and processing terminal performs an integrity check on the candidate boundary and confirms that the boundary is unbroken and unoverlapping, and then determines it as a complete geometric boundary.

[0121] Here, the candidate boundary solution space is the spatial range containing all possible boundary forms that cover the set of geometrically optimized geomorphic structural units. The geometric fit energy term is an energy index that quantifies the degree of fit between the candidate boundary and the edge features of the set of geometrically optimized geomorphic structural units. The topological rule energy term is an energy index that quantifies the degree of fit between the candidate boundary and the topological rules in the geomorphological knowledge graph.

[0122] (3) Based on the complete geometric boundary and topological rules in the geomorphological knowledge graph, geomorphological entity objects are constructed to obtain geomorphological entity objects with complete geometric boundaries.

[0123] For example, the recognition and processing terminal spatially associates the complete geometric boundary with the geometrically optimized set of geomorphic structural units. Using the complete geometric boundary as a contour frame, the set of geometrically optimized geomorphic structural units is embedded within this contour frame to obtain a preliminary geomorphic entity structure. The recognition and processing terminal retrieves topological rules for the current geomorphic type from the geomorphic morphology knowledge graph, clarifying the rule requirements such as the boundary closure of the geomorphic entity, the topological connection method between units, and the spatial hierarchy relationship.

[0124] According to the requirements of topology rules, the identification and processing terminal performs a region-by-region topology verification on the preliminary geomorphic entity structure, marking structural areas with substandard boundary closure, non-compliant inter-unit connection methods, and chaotic spatial hierarchy. For the marked non-compliant areas, the identification and processing terminal adjusts the spatial arrangement of geomorphic structural units, fills in the broken parts of the boundaries, and corrects the connection methods between units to make the structural areas conform to the requirements of topology rules.

[0125] The recognition and processing terminal integrates all structural regions after topological adjustment, confirming that the geometric boundaries of the overall structure are complete and the topological relationships are compliant; the recognition and processing terminal then encapsulates the integrated structure into a physical object, resulting in a terrain entity object with complete geometric boundaries.

[0126] The initial geomorphic entity structure is a structure that has not undergone topological verification, formed by spatial matching of complete geometric boundaries and a set of geometrically optimized geomorphic structural units. Topological rules are a set of rules in the geomorphological knowledge graph that define the spatial relationships and overall structural specifications between different geomorphic units. Entity encapsulation is the operation of transforming a compliant structure into a geomorphic entity with clearly defined geometric boundaries and topological relationships.

[0127] In one embodiment, feature response analysis is performed on the set of geomorphic structural units to obtain the initial semantic confidence level and fault boundary identifiers of each geomorphic structural unit, including:

[0128] (1) Multi-source feature synergistic analysis is performed on the set of geomorphic structural units to obtain the fused response vector of spectrum-texture-topography. The expression of the fused response vector is:

[0129]

[0130] in, Indicates the first The fused response vector of each geomorphic structural unit Represents the spectral eigenvector. Represents texture feature vectors, Represents the terrain feature vector. This represents the spectral variation feature extraction operator. This represents a multi-scale texture feature extraction operator. This represents the terrain-derived feature extraction operator. Represents the cross-modal feature transformation function. Indicates the feature importance weight. This represents the total dimension of the feature vector. Indicates the feature dimension index.

[0131] For example, the recognition and processing terminal extracts spectral feature vectors, texture feature vectors, and terrain feature vectors from each unit of the set of landform structure units. It performs in-depth analysis of the three types of features through spectral variation feature extraction operators, multi-scale texture feature extraction operators, and terrain-derived feature extraction operators, respectively. The spatial alignment and unified expression of multimodal features are achieved through cross-modal feature transformation functions. The unified features are weighted and aggregated by combining feature importance weights to obtain the fused response vector of spectrum-texture-terrain.

[0132] Among them, the spectral feature vector includes the radiation response information of the landform structure unit under different spectral bands, the texture feature vector includes the spatial gray-scale variation information of the surface of the landform structure unit, the terrain feature vector includes the elevation and derived terrain attribute information of the landform structure unit, the cross-modal feature transformation function is used to transform the features of different modes to the same feature space, and the feature importance weight is used to characterize the degree of contribution of different feature dimensions to subsequent analysis.

[0133] (2) Based on the pre-set micro-topographic prototype feature library, the similarity of the fused response vector is measured to generate the initial semantic confidence.

[0134] For example, the recognition and processing terminal retrieves a preset micro-landform prototype feature library, extracts each modal feature component from the fused response vector, compares it with the standard feature components of various micro-landforms in the library dimension by dimension and quantifies the overall matching degree, maps the overall matching degree to a preset confidence interval to complete the normalization calibration, and generates the semantic initial judgment confidence degree.

[0135] Among them, the micro-landform prototype feature library includes a standard feature set of different types of micro-landforms. The overall matching degree is a comprehensive index that represents the degree of fit between the fused response vector and the micro-landform prototype features. The semantic initial judgment confidence degree is used to represent the credibility of the landform structure unit belonging to a certain type of micro-landform.

[0136] (3) Perform fracture-sensitive edge field analysis on the set of geomorphic structural units to generate fracture boundary markers including direction vectors.

[0137] For example, the identification and processing terminal constructs an edge gradient field for a set of landform structure units, identifies gradient anomaly regions through fracture sensitivity detection and removes noise interference to lock the real fracture location, extracts the direction vector of the fracture boundary by combining the direction information of the edge gradient field, and integrates the real fracture location with the direction vector to generate a fracture boundary identifier including the direction vector.

[0138] Among them, the edge gradient field is used to characterize the characteristic intensity change information at the boundary of the geomorphic structure unit, the true fault location is the actual discontinuous location at the boundary of the geomorphic structure unit, the direction vector includes the extension direction and tilt angle information of the fault boundary, and the fault boundary marker is used to mark the fault location and extension direction of the boundary of the geomorphic structure unit.

[0139] In one embodiment, weak edge response segmentation is performed on the multimodal feature map to generate a fragmented set of geomorphic structural units, including:

[0140] (1) Perform multi-scale edge gradient calculation on the multi-modal feature map to obtain the multi-scale edge response map.

[0141] For example, the recognition and processing terminal sets multiple different observation scales for the multimodal feature map, traverses all pixels of the multimodal feature map at each observation scale, calculates the horizontal gradient value and vertical gradient value of the corresponding pixel, synthesizes the edge intensity distribution information at a single scale, and superimposes and integrates the edge intensity distribution information corresponding to all observation scales to obtain a multi-scale edge response map.

[0142] The observation scale refers to the different spatial analysis granularities used to extract edge features. Edge intensity distribution information is a set of numerical values ​​characterizing the edge prominence of pixels in the multimodal feature map.

[0143] (2) Perform adaptive edge connection with fracture sensing on the multi-scale edge response map to generate a continuous edge contour map.

[0144] For example, the identification and processing terminal performs edge breakpoint screening on the multi-scale edge response map, locates the breakpoint regions where the edge intensity is discontinuous in the multi-scale edge response map, extracts the edge extension trend information of the breakpoint regions, performs adaptive connection fitting on the breakpoint regions based on the edge extension trend information, eliminates the edge break gaps, integrates all connected edges, and generates a continuous edge contour map.

[0145] Edge breakpoint screening is a detection operation that identifies edge breakpoints in a multi-scale edge response map. Edge extension trend information is feature information that characterizes the edge orientation of the breakpoint region.

[0146] (3) Based on the continuous edge contour map, perform watershed transformation with morphological constraints on the multimodal feature map to obtain a set of fragmented geomorphic structural units.

[0147] For example, the recognition and processing terminal uses the continuous edge contour map as the basis for segmentation, introduces morphological constraints related to landform, performs watershed transformation on the multimodal feature map, completes the independent division of landform regions with the continuous edge contour as the segmentation boundary, and summarizes all the independent landform regions after division to obtain a fragmented set of landform structure units.

[0148] The segmentation constraints are based on the edge contour information used to limit the segmentation range of the watershed transformation. The morphological constraints are segmentation restriction rules that conform to the natural landform.

[0149] In one embodiment, the expression for the continuous edge contour map is:

[0150]

[0151] in, Represents a continuous edge contour map. Represents a multi-scale edge response map. Represents the coordinates of the edge points. Point Edge direction angle, This represents the kernel function for directional consistency. express arrive The connection path, Indicates the maximum fracture distance threshold. It represents Euclidean distance.

[0152] For example, the identification and processing terminal extracts the coordinate information and corresponding orientation angle information of all edge points from the multi-scale edge response map, traverses all edge point pairs and calculates the Euclidean distance between edge point pairs, compares the Euclidean distance with the maximum fracture distance threshold, and filters out edge point pairs whose Euclidean distance is less than the maximum fracture distance threshold.

[0153] The identification and processing terminal extracts the orientation angle information of the filtered edge point pairs, inputs the orientation angle information into the orientation consistency kernel function, calculates the orientation consistency score of the edge point pairs, and applies the connection path function to generate the connection path between the edge point pairs for edge point pairs whose orientation consistency scores meet the preset matching criteria.

[0154] The recognition and processing terminal summarizes all generated connection paths, performs spatial stitching and contour regularization on the scattered connection paths, eliminates gaps at the path joints, and presents the integrated continuous edge structure in an image to obtain a continuous edge contour map.

[0155] Among them, the multi-scale edge response map is an image formed by superimposing edge intensity distribution information at different observation scales, the edge point coordinates are the spatial location parameters of the edge feature points in the multi-scale edge response map, the direction angle is the angular information representing the edge extension direction at the edge point, and the maximum breakage distance threshold is the distance constraint standard for defining whether the edge point pair has the connection condition.

[0156] The orientation consistency kernel function quantifies the degree of angular alignment between edge point pairs. The orientation consistency score is a numerical value representing the orientation matching degree of edge point pairs. The connection path function constructs continuous edge segments between edge point pairs. A connection path is a continuous edge segment formed between edge point pairs, and a continuous edge contour map is an image of unbroken edge contours formed by integrating all compliant connection paths.

[0157] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0158] In one embodiment, such as Figure 3 As shown, this application also provides a terrain feature recognition device 300 based on remote sensing technology, the device 300 comprising:

[0159] The data preprocessing module 301 is used to acquire input multi-source remote sensing images and auxiliary terrain data, perform data preprocessing on the multi-source remote sensing images and auxiliary terrain data, and obtain fused data with radiometric correction and geometric registration.

[0160] The anti-aliasing feature extraction module 302 is used to extract anti-aliasing pixel features from the fused data and generate a multimodal feature map;

[0161] The weak edge segmentation module 303 is used to perform weak edge response segmentation on the multimodal feature map to generate a fragmented set of geomorphic structure units;

[0162] The feature response analysis module 304 is used to perform feature response analysis on the set of geomorphic structural units to obtain the semantic preliminary judgment confidence and fracture boundary identifier of each geomorphic structural unit.

[0163] The knowledge graph construction module 305 is used to construct a geomorphological knowledge graph that includes micro-topographic rules and morphological constraints based on pre-set micro-topographic morphology rules.

[0164] The fracture sensing connection module 306 is used to perform fracture sensing topological connection on geomorphic structural units with fracture boundary markers based on fracture boundary markers and geomorphic morphology knowledge graphs, and generate geomorphic entity objects with complete geometric boundaries.

[0165] The morphological semantic verification module 307 is used to perform morphological-semantic consistency verification on landform entity objects based on the initial semantic confidence level, and obtain the terrain micro-landform recognition results with semantic labels and confidence levels.

[0166] Specifically, the recognition and processing terminal includes a data preprocessing module 301, an anti-aliasing feature extraction module 302, a weak edge segmentation module 303, a feature response analysis module 304, a knowledge graph construction module 305, a fracture perception connection module 306, and a morphological semantic verification module 307.

[0167] The data preprocessing module acquires the input multi-source remote sensing images and auxiliary terrain data, performs radiometric correction on the multi-source remote sensing images to eliminate atmospheric scattering and sensor noise interference, performs geometric registration on the multi-source remote sensing images and auxiliary terrain data to unify spatial coordinates and projection methods, spatially fuses the multi-source remote sensing images and auxiliary terrain data that have completed radiometric correction and geometric registration, and outputs the fused data of radiometric correction and geometric registration.

[0168] Multi-source remote sensing imagery includes remote sensing image data from different sensors and at different times. Auxiliary topographic data includes digital elevation models and slope / aspect data. Radiometric correction is used to eliminate interference from atmospheric scattering and sensor noise. Geometric registration is used to unify the spatial coordinates and projection system of the data.

[0169] The anti-aliasing feature extraction module receives fused data from radiometric correction and geometric registration, performs hybrid pixel decomposition on the fused data to separate the spectral responses of different land cover types, extracts the feature information of the decomposed spectral responses to generate spectral feature components, performs multi-scale texture analysis on the fused data to generate texture feature components, extracts terrain-derived features from the fused data to generate terrain feature components, and performs cross-modal fusion of spectral feature components, texture feature components, and terrain feature components to output a multimodal feature map.

[0170] Hybrid pixel decomposition is used to separate the response information of multiple land features within a single pixel. Multimodal feature maps are image data containing multi-source features including spectral, texture, and topographic features.

[0171] The weak edge segmentation module performs multi-scale edge gradient calculation on the multimodal feature map to generate a multi-scale edge response map, performs fracture-aware adaptive edge connection on the multi-scale edge response map to generate a continuous edge contour map, and performs morphological constraint watershed transformation on the multimodal feature map with the continuous edge contour map as a constraint to output a fragmented set of geomorphic structural units.

[0172] Multi-scale edge gradient calculation is used to extract edge intensity information from images at different scales. A continuous edge contour map is an edge contour image without any breaks or defects. A fragmented set of geomorphic structural units is a set of independent geomorphic regions after region segmentation.

[0173] The feature response analysis module performs multi-source feature collaborative analysis on the set of geomorphic structural units to generate a fused response vector of spectrum-texture-topography. Based on a pre-set micro-geomorphic prototype feature library, it performs similarity measurement on the fused response vector to generate a semantic preliminary judgment confidence score. It performs fracture-sensitive edge field analysis on the set of geomorphic structural units and outputs fracture boundary markers containing direction vectors.

[0174] The fused response vector is a feature vector obtained by weighted aggregation of multi-source features. The semantic initial judgment confidence score characterizes the credibility of a geomorphic structural unit's attribution to its corresponding micro-geomorphic type. The fault boundary marker is structured information indicating the location and extension direction of faults within geomorphic structural units.

[0175] The knowledge graph construction module summarizes the spatial relationships between micro-landforms to form micro-landform topological rules, standardizes the geometric and extended attributes of micro-landforms to form morphological constraints, and integrates the micro-landform topological rules and morphological constraints in a structured way to construct and output a landform morphology knowledge graph containing micro-landform topological rules and morphological constraints.

[0176] Among them, micro-topographical rules are the morphological norms formed by the natural evolution of micro-topography. Micro-topographical rules are the constraints on the spatial relationships of micro-topography. Morphological constraints are the normative requirements for the geometric and extended attributes of micro-topography.

[0177] The fracture-aware connection module derives a set of candidate connection paths based on the direction vector in the fracture boundary identifier and the extension rules in the geomorphological knowledge graph. It then performs a multi-dimensional compatibility evaluation on the candidate connection path set to generate a topology-optimized path. Based on the topology-optimized path, it performs knowledge-guided geometric reconstruction on the geomorphological structural unit set and outputs a geomorphological entity object with complete geometric boundaries.

[0178] The candidate connection path set consists of multiple alternative connection paths that conform to the extension rules. The topology optimization path is the optimal connection path after multi-dimensional evaluation and screening. Knowledge-guided geometric reconstruction is a geomorphic unit integration operation under the constraints of a geomorphological knowledge graph.

[0179] The morphological and semantic verification module performs morphological-semantic consistency verification on landform entities based on the initial semantic confidence score and the rules of the landform morphology knowledge graph. It checks the degree of matching between the geometric morphological features and semantic information of the landform entities, optimizes the semantic labels and confidence scores corresponding to the verification results, and outputs the terrain micro-landform recognition results with semantic labels and confidence scores.

[0180] Among them, the morphological-semantic consistency check is used to verify the degree of matching between the geometric shape and semantic information of the landform entity object.

[0181] The fracture sensing connection module 306 is also used for:

[0182] Based on the direction vector in the fault boundary identifier and the extension rules in the geomorphological knowledge graph, the path derivation for fault perception is performed to obtain a set of candidate connection paths.

[0183] A multi-dimensional compatibility evaluation is performed on the candidate connection path set to generate a topology-optimized path;

[0184] Based on topology optimization path, knowledge-guided geometric reconstruction of the set of geomorphic structural units is performed to obtain geomorphic entity objects with complete geometric boundaries.

[0185] The fracture sensing connection module 306 is also used for:

[0186] Trajectory prediction based on fracture sensing is performed on the direction vector in the fracture boundary marker to obtain the initial extension trajectory. The expression for the initial extension trajectory is:

[0187] in, Indicates the initial extended trajectory. This represents the direction vector in the fracture boundary identifier. The direction vector represents the Gaussian diffusion kernel. Indicates the parameter of directional uncertainty. Indicates position Direction Edge gradient field, This represents the morphological interruption probability mapping function. Indicates the direction of the search range;

[0188] The initial extended trajectory is corrected for topological compliance based on the extension rules in the geomorphological knowledge graph, and a rule-constrained trajectory is generated.

[0189] Multiple hypothesis path generation is performed on the rule-constrained trajectory to obtain a set of candidate connection paths.

[0190] The fracture sensing connection module 306 is also used for:

[0191] Based on the topology optimization path and the morphological constraints in the geomorphological knowledge graph, the geometric structure of the geomorphological structural unit set is optimized to obtain the geometrically optimized geomorphological structural unit set.

[0192] Knowledge-guided boundary reconstruction is performed on the geometrically optimized set of geomorphic structural units to generate a complete geometric boundary. The expression for the complete geometric boundary is:

[0193]

[0194] in, Represents the complete geometric boundary. This represents the set of geomorphic structural units after geometric optimization. A knowledge graph representing landform morphology. Represents the energy term of the geometric fit. Represents the energy term of the topological rule. Describe the candidate boundary solution space. This represents the candidate boundary during the optimization process. Represents the geometric constraint weights. Indicates the topology constraint weights;

[0195] Based on the complete geometric boundaries and topological rules in the geomorphological knowledge graph, geomorphological entity objects are constructed to obtain geomorphological entity objects with complete geometric boundaries.

[0196] The characteristic response analysis module 304 is also used for:

[0197] Multi-source feature synergistic analysis is performed on the set of geomorphic structural units to obtain the fused response vector of spectrum-texture-topography. The expression of the fused response vector is:

[0198]

[0199] in, Indicates the first The fused response vector of each geomorphic structural unit Represents the spectral eigenvector. Represents texture feature vectors, Represents the terrain feature vector. This represents the spectral variation feature extraction operator. This represents a multi-scale texture feature extraction operator. This represents the terrain-derived feature extraction operator. Represents the cross-modal feature transformation function. Indicates the feature importance weight. This represents the total dimension of the feature vector. Indicates the feature dimension index;

[0200] The similarity of the fused response vector is measured based on a pre-set micro-topographic prototype feature library to generate a semantic initial judgment confidence score.

[0201] A fracture-sensitive edge field analysis is performed on the set of geomorphic structural units to generate fracture boundary markers including direction vectors.

[0202] The weak edge segmentation module 303 is also used for:

[0203] Multi-scale edge gradient calculation is performed on the multimodal feature map to obtain the multi-scale edge response map;

[0204] Adaptive edge connection with fracture detection is performed on multi-scale edge response maps to generate continuous edge contour maps;

[0205] A watershed transformation with morphological constraints is performed on the multimodal feature map based on the continuous edge contour map to obtain a set of fragmented geomorphic structural units.

[0206] The weak edge segmentation module 303 also includes an expression for the continuous edge contour map:

[0207]

[0208] in, Represents a continuous edge contour map. Represents a multi-scale edge response map. Represents the coordinates of the edge points. Point Edge direction angle, This represents the kernel function for directional consistency. express arrive The connection path, Indicates the maximum fracture distance threshold. It represents Euclidean distance.

[0209] In one embodiment, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0210] In one embodiment, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method embodiments.

[0211] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0212] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for identifying a topographical feature based on remote sensing technology, characterized by, The method includes: The input multi-source remote sensing images and auxiliary terrain data are acquired, and the multi-source remote sensing images and auxiliary terrain data are preprocessed to obtain fused data with radiometric correction and geometric registration. Anti-mixing pixel feature extraction is performed on the fused data to generate a multimodal feature map; Weak edge response segmentation is performed on the multimodal feature map to generate a fragmented set of geomorphic structure units; Feature response analysis is performed on the set of geomorphic structural units to obtain the initial semantic confidence and fracture boundary identifiers of each geomorphic structural unit; Based on pre-defined micro-topographic morphology rules, a geomorphological knowledge graph including micro-topographic topology rules and morphological constraints is constructed. Based on the fracture boundary markers and the geomorphological knowledge graph, fracture-aware topological connections are made to geomorphological structural units with fracture boundary markers to generate geomorphological entity objects with complete geometric boundaries. Based on the initial semantic confidence level, the morphological-semantic consistency of the landform entity object is verified to obtain the terrain micro-landform recognition result with semantic labels and confidence levels.

2. The method of claim 1, wherein, The step of performing fracture-aware topological connections on geomorphic structural units with fracture boundary markers based on the fracture boundary identifiers and the geomorphological knowledge graph to generate geomorphic entity objects with complete geometric boundaries includes: Based on the direction vector in the fracture boundary identifier and the extension rules in the geomorphological knowledge graph, the path derivation for fracture perception is performed to obtain a set of candidate connection paths. A multi-dimensional compatibility evaluation is performed on the candidate connection path set to generate a topology-optimized path; Based on the topology optimization path, the set of geomorphic structural units is subjected to knowledge-guided geometric reconstruction to obtain the geomorphic entity object with complete geometric boundaries.

3. The method according to claim 2, wherein The path derivation for fracture perception based on the direction vector in the fracture boundary identifier and the extension rules in the geomorphological knowledge graph yields a set of candidate connection paths, including: The trajectory prediction of the direction vector in the fracture boundary mark is fracture-aware, and an initial extension trajectory is obtained, and the expression of the initial extension trajectory is: wherein, denotes an initial extension trajectory, denotes a direction vector in a break boundary identification, denotes a direction vector Gaussian spread kernel, denotes a direction uncertainty parameter, denotes a position edge gradient field at a direction at a position, denotes a morphology break probability mapping function, denotes a direction search range; Based on the extension rules in the geomorphological knowledge graph, the initial extension trajectory is corrected for topological compliance to generate a rule-constrained trajectory. The set of candidate connection paths is obtained by generating multiple hypothetical paths from the rule-constrained trajectory.

4. The method according to claim 2, wherein, The knowledge-guided geometric reconstruction of the set of geomorphic structural units based on the topology optimization path, resulting in the geomorphic entity object with complete geometric boundaries, includes: Based on the topology optimization path and the morphological constraints in the geomorphological knowledge graph, the set of geomorphological structural units is geometrically optimized to obtain a geometrically optimized set of geomorphological structural units. Knowledge-guided boundary reconstruction is performed on the geometrically optimized set of geomorphic structural units to generate a complete geometric boundary; Based on the complete geometric boundary and the topological rules in the landform knowledge graph, landform entity objects are constructed to obtain the landform entity objects with complete geometric boundaries.

5. The terrain feature recognition method based on remote sensing technology according to claim 1, characterized in that, The feature response analysis of the set of geomorphic structural units to obtain the initial semantic confidence level and fault boundary identifier of each geomorphic structural unit includes: Multi-source feature collaborative analysis is performed on the set of geomorphic structural units to obtain a fused response vector of spectrum-texture-topography. The expression of the fused response vector is as follows: in, Indicates the first The fused response vector of each geomorphic structural unit Represents the spectral eigenvector. Represents texture feature vectors, Represents the terrain feature vector. This represents the spectral variation feature extraction operator. This represents a multi-scale texture feature extraction operator. This represents the terrain-derived feature extraction operator. Represents the cross-modal feature transformation function. Indicates the feature importance weight. This represents the total dimension of the feature vector. Indicates the feature dimension index; The similarity of the fused response vector is measured based on a pre-set micro-topographic prototype feature library to generate the initial semantic confidence score. A fracture-sensitive edge field analysis is performed on the set of geomorphic structural units to generate fracture boundary markers including direction vectors.

6. The terrain feature recognition method based on remote sensing technology according to claim 1, characterized in that, The step of performing weak edge response segmentation on the multimodal feature map to generate a fragmented set of terrain structure units includes: Multi-scale edge gradient calculation is performed on the multi-modal feature map to obtain a multi-scale edge response map; Adaptive edge connection with fracture detection is performed on the multi-scale edge response map to generate a continuous edge contour map; Based on the continuous edge contour map, a watershed transformation with morphological constraints is performed on the multimodal feature map to obtain the fragmented set of geomorphic structural units.

7. The terrain feature recognition method based on remote sensing technology according to claim 6, characterized in that, The expression for the continuous edge contour map is: in, Represents a continuous edge contour map. Represents a multi-scale edge response map. Represents the coordinates of the edge points. Point Edge direction angle, This represents the kernel function for directional consistency. express arrive The connection path, Indicates the maximum fracture distance threshold. It represents Euclidean distance.

8. A terrain feature recognition device based on remote sensing technology, characterized in that, The device includes: The data preprocessing module is used to acquire input multi-source remote sensing images and auxiliary terrain data, perform data preprocessing on the multi-source remote sensing images and auxiliary terrain data, and obtain fused data with radiometric correction and geometric registration. An anti-aliasing feature extraction module is used to extract anti-aliasing pixel features from the fused data and generate a multimodal feature map. The weak edge segmentation module is used to perform weak edge response segmentation on the multimodal feature map to generate a fragmented set of geomorphic structure units; The feature response analysis module is used to perform feature response analysis on the set of geomorphic structural units to obtain the semantic preliminary judgment confidence and fracture boundary identifier of each geomorphic structural unit. The knowledge graph construction module is used to construct a geomorphological knowledge graph that includes micro-topographic rules and morphological constraints based on pre-defined micro-topographic morphology rules. The fracture sensing connection module is used to perform fracture sensing topological connection on the geomorphic structural units with fracture boundary markers based on the fracture boundary markers and the geomorphic morphology knowledge graph, and generate geomorphic entity objects with complete geometric boundaries. The morphological semantic verification module is used to perform morphological-semantic consistency verification on the landform entity object based on the initial semantic confidence level, and obtain the terrain micro-landform recognition result with semantic label and confidence level.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the terrain feature recognition method based on remote sensing technology as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the terrain feature recognition method based on remote sensing technology as described in any one of claims 1 to 7.

Images