Multi-source low-altitude inspection data cleaning and intelligent analysis processing method, system and equipment

Abstract

The invention relates to the technical field of data processing, in particular to a multi-source low-altitude inspection data cleaning and intelligent analysis processing method, system and equipment. The method comprises the following steps: responding to a low-altitude inspection task plan which is submitted by a user and comprises a task area and a task type; multi-source static spatial data related to task planning is intelligently connected and gathered, and data timeliness management is executed in the connection process; performing intelligent preprocessing on the converged multi-source static spatial data to generate a unified three-dimensional grid scene; based on the three-dimensional grid scene, automatically extracting inspection target features, and performing association analysis on an extraction result and real-time dynamic business data to generate an inspection analysis report and a risk assessment result; the inspection analysis report, the risk assessment result and the three-dimensional grid scene are integrally and visually presented, and a decision basis is provided for actual inspection in the physical world. And the intelligent level, the safety and the decision-making efficiency of the low-altitude inspection operation are improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a method, system, and equipment for cleaning and intelligent analysis of multi-source low-altitude inspection data. Background Technology

[0002] With the widespread application of drones, eVTOL aircraft, and other aircraft in low-altitude fields such as power line inspection, pipeline inspection, and urban management, the complexity of low-altitude inspection operations is increasing daily. Traditional inspection models mainly rely on the process of aircraft collecting data on-site and operators transmitting and analyzing the data afterward, which results in significant information silos and decision-making delays.

[0003] Currently, the industry generally attempts to improve efficiency by building data processing platforms. However, these solutions face the following prominent technical bottlenecks: First, at the data level, inspection tasks rely on multi-source heterogeneous spatial data such as oblique photography models, terrain, vector data, and point clouds. These data have different formats, coordinate systems, and update frequencies, making them extremely difficult to access, integrate, and manage, and hindering the formation of a unified and up-to-date data foundation. Second, at the analysis and decision-making level, existing methods mostly remain at the level of displaying two-dimensional maps or simple three-dimensional models, failing to deeply integrate static data with dynamic business data, resulting in weak risk identification capabilities and untimely early warnings. Managers find it difficult to effectively simulate and evaluate the feasibility and safety of inspection plans before task execution.

[0004] Therefore, there is an urgent need in this field for a technical solution that can integrate and intelligently fuse multi-source low-altitude data to achieve a leap from passive response to proactive prediction and intelligent decision-making. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a method, system, and equipment for cleaning and intelligent analysis of multi-source low-altitude inspection data.

[0006] In a first aspect, the present invention provides a method for cleaning and intelligent analysis of multi-source low-altitude inspection data, applied to a digital twin system, the method comprising the following steps: S1. In response to the low-altitude inspection task plan submitted by the user, which includes the task area and task type, intelligently introduce and aggregate multi-source static spatial data related to the task plan, and perform data timeliness management during the introduction process. S2. The aggregated multi-source static spatial data is intelligently preprocessed, and based on the pre-built inspection knowledge base and spatial grid engine, the preprocessed data is fused to generate a unified three-dimensional grid scene. S3. Based on the three-dimensional grid scene, automatically extract the features of the inspection target, and perform correlation analysis between the extraction results and real-time dynamic business data to generate an inspection analysis report and risk assessment results; the real-time dynamic business data includes at least real-time aircraft status data, electronic fence data and meteorological data. S4. Integrate the inspection analysis report and risk assessment results with the three-dimensional grid scene for unified visualization, providing a basis for decision-making in actual inspections in the physical world.

[0007] As a further limitation of the technical solution of the present invention, step S1 includes: In response to a user-submitted low-altitude patrol mission plan that includes the mission area and mission type, the required set of multi-source static spatial data types is determined from a pre-set mission-data mapping table based on the mission type. Based on the task area and the set of data types, automatically import multi-source static spatial data related to the task planning from the local database and external data service interface; When importing multi-source static spatial data from the local database, a data timeliness check based on a preset validity period is performed. If the data has expired, the system automatically redirects to importing real-time or updated data of the same type from an external data service interface. The data timeliness check includes: Retrieve timestamps from data in the local database; Calculate the difference between the obtained timestamp and the current time; If the difference is greater than the preset validity period for the corresponding data type, then the data is determined to have expired.

[0008] It can automatically and accurately acquire the data required for tasks, avoiding the tediousness and omissions of manual searching and improving the efficiency of task preparation. Through timeliness management, it ensures that the data base used for analysis and decision-making is always valid and fresh, fundamentally avoiding decision-making errors caused by using outdated data and ensuring the consistency between the digital twin and the physical world.

[0009] As a further limitation of the technical solution of the present invention, the preset task-data mapping table is generated and maintained in the following manner: Initialize the definition of the correspondence between different inspection task types and the required data types and attributes; Collect historical inspection task data usage records, and optimize and update the corresponding relationships using a machine learning model; specifically including: Collect the task characteristics of historical inspection tasks and the corresponding actual usage data combinations to construct a training sample set; Based on the training sample set, the multi-label classification model is trained to obtain the data recommendation model; For new inspection tasks, the data recommendation model is used to generate recommended data combinations. Recommended data combinations that meet the preset reliability criteria are updated to the preset task-data mapping table after being approved by the administrator.

[0010] By continuously learning from past successes, data recommendations become increasingly accurate, gradually solidifying and surpassing the individual experience of domain experts. This leads to a continuous improvement in the intelligence level of the entire low-altitude inspection data governance system over time.

[0011] As a further limitation of the technical solution of the present invention, the steps in S2 include: S21. The aggregated multi-source static spatial data is uniformly registered to a preset spatial coordinate system; and based on the rules corresponding to the task type in the pre-built inspection knowledge base, the multi-source static spatial data is dynamically cleaned and enhanced in a way that matches the data type; the multi-source static spatial data includes at least oblique photogrammetry model data, terrain data, vector data and point cloud data; S22. Based on the task type, obtain the corresponding grid configuration strategy from the inspection knowledge base; based on the grid configuration strategy and the task area, use the spatial grid engine to perform multi-scale grid division of the three-dimensional space, and assign a unique integer code to each grid unit; S23. The oblique photogrammetry model data, terrain data, vector data and point cloud data preprocessed in S21 are mapped to the corresponding grid cells according to the gridding configuration strategy, and attribute association is performed. S24. Generate a unified 3D mesh scene based on the mesh cells with completed attribute associations.

[0012] The steps for dynamic cleaning and enhancement of multi-source static spatial data, tailored to the data type, include: Based on the task type, retrieve the preprocessing rule set corresponding to the data type from the inspection knowledge base; Based on the preprocessing rule set, geometric correction and texture enhancement are performed on the oblique photogrammetry model data, elevation outlier repair and data hole filling are performed on the terrain data, topological relationship repair and attribute normalization are performed on the vector data, and outlier removal, ground point classification and extraction, and non-ground point semantic segmentation are performed on the point cloud data.

[0013] Targeted cleaning and enhancement improved the quality and usability of the original data; unified grid fusion placed different types of data under the same spatial computing framework, breaking down data barriers and generating a three-dimensional grid scene with precise location, rich attributes, and computability, providing a unique and reliable data foundation for subsequent in-depth analysis.

[0014] As a further limitation of the technical solution of the present invention, step S3 includes: S31. Based on the semantic attributes of the grid cells in the three-dimensional grid scene, perform spatial query and calculation to extract target features related to the inspection task type; S32. The extracted target features are correlated with electronic fence data, real-time aircraft status data and meteorological data to identify potential risk events. S33. Based on the identified risk events, assign a comprehensive risk level to each event and automatically generate an inspection analysis report that includes a description of the risk event, its spatial location, risk level, and handling recommendations.

[0015] It can proactively and automatically identify potential risks that are difficult for humans to detect in a timely manner, and quantify and classify the risks, thereby providing managers with accurate and quantitative decision-making basis, rather than just raw data or simple alarm information.

[0016] As a further limitation of the technical solution of the present invention, step S4 includes: S41. Based on the spatial location and attributes in the risk assessment results, render the corresponding three-dimensional visualization alarm element at the corresponding grid cell in the three-dimensional grid scene, and visually highlight the associated inspection target. S42. In the visualization interface, the content of the inspection analysis report and the visualization data dashboard of real-time dynamic business data are displayed synchronously; and the two-way interactive linkage of the three components of the three-dimensional grid scene, inspection analysis report and real-time data dashboard is realized. The selection operation in any component can enable the other components to perform corresponding view positioning and content highlighting.

[0017] The analysis results are presented to decision-makers in a clear and three-dimensional way, and the interactive linkage greatly reduces the difficulty of information acquisition and understanding, making complex spatial analysis and risk assessment results clear at a glance, thus improving the efficiency and accuracy of command and decision-making.

[0018] Secondly, the present invention also provides a multi-source low-altitude inspection data cleaning and intelligent analysis processing system, applied to a digital twin system, the system comprising: The data intelligent aggregation module is used to respond to the low-altitude inspection task plan submitted by the user, which includes the task area and task type, intelligently introduce and aggregate multi-source static spatial data related to the task plan, and perform data timeliness management during the introduction process. The data fusion and scene construction module is used to intelligently preprocess the aggregated multi-source static spatial data, and based on the pre-built inspection knowledge base and spatial grid engine, fuse the preprocessed data to generate a unified three-dimensional grid scene. The intelligent analysis and risk assessment module is used to automatically extract the features of the inspection targets based on the three-dimensional grid scene, and perform correlation analysis between the extraction results and real-time dynamic business data to generate inspection analysis reports and risk assessment results; the real-time dynamic business data includes at least real-time aircraft status data, electronic fence data and meteorological data; The integrated visualization and decision support module is used to present the inspection analysis report, risk assessment results and the three-dimensional mesh scene in an integrated visualization, providing a basis for decision-making for actual inspections in the physical world.

[0019] Thirdly, the present invention also provides an electronic device, the electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory storing computer program instructions executable by the at least one processor, the computer program instructions being executed by the at least one processor to enable the at least one processor to perform the multi-source low-altitude inspection data cleaning and intelligent analysis processing method as described in the first aspect.

[0020] As can be seen from the above technical solutions, this application has the following advantages: it improves the intelligence level, safety, and decision-making efficiency of low-altitude inspection operations. By completing the simulation and evaluation of the entire process in advance in the digital space, it can reduce the blind spots in actual physical inspections, effectively avoid flight conflicts and violation risks, optimize inspection paths, and save operating costs. Attached Figure Description

[0021] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description 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.

[0022] Figure 1 This is a flowchart illustrating the method provided in an embodiment of the present invention.

[0023] Figure 2 A block diagram of a system provided in an embodiment of the present invention. Detailed Implementation

[0024] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0026] like Figure 1 As shown in the figure, this invention provides a method for cleaning and intelligent analysis of multi-source low-altitude inspection data, applied to a digital twin system. The method includes the following steps: S1. In response to the low-altitude inspection task plan submitted by the user, which includes the task area and task type, intelligently introduce and aggregate multi-source static spatial data related to the task plan, and perform data timeliness management during the introduction process. S2. The aggregated multi-source static spatial data is intelligently preprocessed, and based on the pre-built inspection knowledge base and spatial grid engine, the preprocessed data is fused to generate a unified three-dimensional grid scene. S3. Based on the three-dimensional grid scene, automatically extract the features of the inspection target, and perform correlation analysis between the extraction results and real-time dynamic business data to generate an inspection analysis report and risk assessment results; the real-time dynamic business data includes at least real-time aircraft status data, electronic fence data and meteorological data. S4. Integrate the inspection analysis report and risk assessment results with the three-dimensional grid scene for unified visualization, providing a basis for decision-making in actual inspections in the physical world.

[0027] In some embodiments, step S1 includes: S11. In response to the low-altitude inspection task plan submitted by the user, which includes the task area and task type, determine the required set of multi-source static spatial data types from the preset task-data mapping table according to the task type. S12. Based on the task area and the data type set, automatically import multi-source static spatial data related to the task planning from the local database and external data service interface; When importing multi-source static spatial data from the local database, a data timeliness check based on a preset validity period is performed. If the data has expired, the system automatically redirects to importing real-time or updated data of the same type from an external data service interface. The data timeliness check includes: Retrieve timestamps from data in the local database; Calculate the difference between the obtained timestamp and the current time; If the difference is greater than the preset validity period for the corresponding data type, then the data is determined to have expired.

[0028] In this embodiment of the invention, the preset task-data mapping table is generated and maintained in the following manner: Initialize the definition of the correspondence between different inspection task types and the required data types and attributes; Collect historical inspection task data usage records, and optimize and update the corresponding relationships using a machine learning model; specifically, this includes: collecting task features of historical inspection tasks and corresponding actual usage data combinations to construct a training sample set; training a multi-label classification model based on the training sample set to obtain a data recommendation model; for new inspection tasks, generating recommended data combinations using the data recommendation model; and updating the recommended data combinations that meet preset reliability conditions to the preset task-data mapping table after administrator approval.

[0029] During operation, the system will automatically collect two types of core data: historical task data and negative samples and user feedback data, forming the basis for model training. Historical mission data includes: Task characteristics: task type, geographical and geomorphological features of the task area (e.g., urban areas, farmland, mountains), and task objectives (e.g., routine patrols, post-disaster assessment).

[0030] Data Usage Log: A complete list of the data types that were ultimately retrieved and successfully used in this task.

[0031] Task performance metrics: the accuracy of the analysis, the efficiency of task execution, the quality score of the generated report, and user satisfaction feedback.

[0032] Negative samples and user feedback data include: records of invalid data manually removed by users during task execution, and records of data manually added by users that was not recommended by the mapping table.

[0033] The collected raw data is transformed into samples that can be processed by machine learning models, specifically including: Feature vectorization: Converting task features (such as task type, regional topography) into numerical feature vectors through encoding techniques.

[0034] Sample construction: Each historical task and its final data combination constitute a training sample. The feature vector is the input, and the final data combination is the output label that the model needs to learn.

[0035] Model training and intelligent recommendation, including: Model selection: Collaborative filtering recommendation algorithm or multi-label classification model is adopted. This embodiment of the invention adopts a multi-label classification model.

[0036] Collaborative filtering: The logic is to find similar tasks. If task A and task B are highly similar in features, and task A successfully uses data combination X, then X is recommended to task B.

[0037] Multi-label classification: The problem is defined as predicting all the data labels needed for a given task feature. The model outputs a probability for each data type, representing its importance to the task.

[0038] Recommendation generation: When a user creates a new task, the system inputs the features of the new task into the pre-trained model. The model then outputs a list of recommended data and their confidence scores.

[0039] The system submits high-confidence recommendations from the model (e.g., confidence > 90%) as optimization suggestions to the system administrator for review. The administrator (usually a domain expert) uses their professional experience to determine whether the recommendation is reasonable, effective, and universally applicable.

[0040] Once approved, the system will formally and persistently write this new task type-data type mapping into the pre-defined task-data mapping table.

[0041] In some embodiments, the steps in S2 include: S21. The aggregated multi-source static spatial data is uniformly registered to a preset spatial coordinate system; and based on the rules corresponding to the task type in the pre-built inspection knowledge base, the multi-source static spatial data is dynamically cleaned and enhanced in a way that matches the data type; the multi-source static spatial data includes at least oblique photogrammetry model data, terrain data, vector data and point cloud data; S22. Based on the task type, obtain the corresponding grid configuration strategy from the inspection knowledge base; based on the grid configuration strategy and the task area, use the spatial grid engine to perform multi-scale grid division of the three-dimensional space, and assign a unique integer code to each grid unit; In this step, the goal of unified spatial registration is to incorporate all data into the same spatial reference frame. Specifically, this includes: The system reads the embedded coordinate system parameters of each data file to accurately identify the original spatial reference frame for each data type. Based on the relationship between the source coordinate system and the target preset coordinate system, a coordinate transformation parameter model (including translation, rotation, scaling, and ellipsoid transformation parameters) is established. Subsequently, using this model, batch coordinate transformation calculations are performed on all geometric elements in the data (image corner points, point cloud coordinates, vector vertices). For raster data (such as imagery and terrain), after coordinate transformation, an appropriate resampling algorithm (such as cubic convolution for imagery and bilinear interpolation for terrain) is used to recalculate the value of each pixel in the new coordinate system to ensure geometric accuracy and texture continuity.

[0042] The steps for dynamic cleaning and enhancement of multi-source static spatial data, tailored to the data type, include: Based on the task type, retrieve the preprocessing rule set corresponding to the data type from the inspection knowledge base; Based on the preprocessing rule set, geometric correction and texture enhancement are performed on the oblique photogrammetry model data, elevation outlier repair and data hole filling are performed on the terrain data, topological relationship repair and attribute normalization are performed on the vector data, and outlier removal, ground point classification and extraction, and non-ground point semantic segmentation are performed on the point cloud data.

[0043] S23. The oblique photogrammetry model data, terrain data, vector data, and point cloud data preprocessed in S21 are mapped to corresponding grid cells according to the aforementioned gridding configuration strategy, and attribute associations are performed. It should be noted that the oblique photogrammetry model is a high-precision 3D reality model generated by acquiring multi-angle images of the mission area from a flight platform equipped with an oblique photogrammetry camera, followed by aerial triangulation, dense matching, and 3D reconstruction processing. The oblique photogrammetry model data provides a high-precision, realistic 3D environment of the inspection area. It is the foundation for UAV path planning and perspective simulation, and also serves as the base map for accurately locating inspection findings in 3D space.

[0044] Terrain data provides information about the undulations of the ground. This is crucial for planning safe flight altitudes, avoiding terrain obstacles, and conducting slope analysis; it is core data for flight safety.

[0045] Vector data provides structured information about geographic entities, such as roads, administrative divisions, and the locations of facilities requiring focused inspection (such as the precise coordinates of power towers, pipelines, and wind turbines). This is core data for task planning.

[0046] Point cloud data provides the most accurate three-dimensional spatial structure information and is often used to supplement and refine oblique photogrammetry models, such as accurately calculating the distance from power lines to trees. It is the core data for fine analysis.

[0047] This step involves data mapping and attribute association based on a gridded configuration strategy, specifically including: S231: Strategy resolution and mapping priority confirmation; The grid-based configuration strategy retrieved from the inspection knowledge base is read, which specifies that: Data mapping priority: For example, in the presence of conflicts, the geometric information of the oblique photogrammetry model takes precedence over the vector white model.

[0048] Attribute association rules: Define which key attributes need to be retained.

[0049] Fusion parameters: Geographic tolerance used to determine the relationship between data and grid affiliation.

[0050] S232: Mapping and correlation of oblique photogrammetry model data; For each triangle in the oblique photogrammetry model, the engine calculates its outer envelope and finds all mesh cells that intersect with it. The geometry of the triangle (vertex coordinates) and its corresponding texture image index are then associated with these intersecting mesh cells as a set of attributes. A single mesh cell may be associated with multiple triangles, allowing the complete model to be reconstructed during rendering.

[0051] S233: Mapping and correlation of terrain data; For each raster cell in the terrain data (DEM / DSM), the engine finds its unique underlying grid cell based on its center point coordinates. The elevation value of that cell is then directly assigned to that grid cell as the core attribute. For DSM, derived attributes such as the height of surface objects can also be extracted.

[0052] S234: Mapping and association of vector data; For vector data (such as building outline polygons), the engine performs a spatial query to find all grid cells covered by or intersecting with the polygon. The semantic attributes of these grid cells are then labeled as the vector type. Simultaneously, the vector's attribute table information (such as building name and height) is inherited by these grids. Depending on the strategy, the bottom elevation field is read from the attribute table, or the elevation at that location is obtained by associating with a terrain service, or a fixed value is directly used as the bottom elevation of the grid. The building height field is read from the attribute table, or a fixed value is used, combined with the bottom elevation to calculate the building's top elevation.

[0053] S235: Mapping and association of point cloud data; For point clouds that have undergone semantic segmentation, the engine aggregates all point clouds belonging to the same grid cell. It then calculates the statistical features of the point clouds within that grid cell, for example: For vegetation point clouds, calculate the average height and maximum height.

[0054] For building point clouds, the point cloud density is calculated to reflect the integrity of the structure.

[0055] These statistical characteristics are associated as attributes of the grid cell.

[0056] S24. Based on the mesh cells with completed attribute associations, generate a unified 3D mesh scene. This step essentially involves constructing a multi-resolution pyramid structure and generating a scene description file; specifically, it includes: S241: Constructing spatial indexes and multi-resolution hierarchical structures; Based on the original scale of the grid division in S22, it automatically aggregates upwards to build multiple levels of detail. For example, the bottom 16 fine 1-meter grids are aggregated into a 4-meter grid, and so on, forming a pyramid structure.

[0057] When aggregating grids, the attributes of lower-level grids are calculated comprehensively. For example, the main feature type attribute of a parent grid can be determined based on the semantic type with the largest proportion in its child grids; its maximum height attribute is taken as the maximum value in the child grids.

[0058] S242: Scene graph organization and spatial scheduling structure generation; Create a scene graph data structure that organizes all meshes in a tree structure: Root node: Represents the entire task area.

[0059] Intermediate nodes: represent grid blocks at different levels in the pyramid.

[0060] Leaf nodes: Represent the lowest level of the grid cells, containing the richest attribute information.

[0061] Each node contains its spatial extent (bounding box) information, which is used to implement view frustum clipping and level of detail scheduling.

[0062] S243: Generate a standardized scene description file; Generate one or more scene description files that conform to open standards. This file defines the global coordinate system, metadata, and attribute table structure of the entire scene, describes the tree structure of the scene graph and the link path to the actual mesh data file corresponding to each node, and includes rules for switching the level of detail for dynamic loading.

[0063] S244: Standardized Scenario Description File Serialization and Output; Each lowest-level mesh cell (or a small collection of meshes packaged for performance optimization) and all its associated properties are serialized into a compact binary tile file. During this process, the system compresses the geometry, texture, and attribute data to reduce storage and network transmission overhead; ensuring that the output tile file conforms to the selected criteria and can be directly recognized and loaded by common 3D engines.

[0064] S245: Encapsulation of scene metadata and permission information; The scene's metadata, such as creation time, coordinate reference system, data source information, and access permissions, is encapsulated into a scene description file or a separate metadata file, ultimately outputting a standardized 3D mesh scene. This scene can serve as a digital twin base, directly available for querying, analysis, and visualization in step S3. This ensures the scene's integrity and manageability.

[0065] In some embodiments, step S3 includes: S31. Based on the semantic attributes of the grid cells in the three-dimensional grid scene, perform spatial query and calculation to extract target features related to the inspection task type; This step utilizes the semantic properties of the grid for targeted search and feature quantization, specifically including: S311: Task-driven semantic query and initial target screening; Based on the current inspection task type, the corresponding target recognition rule is retrieved from the inspection knowledge base. This rule defines the semantic type of the target to be searched and its preliminary screening conditions.

[0066] Example (Power Line Inspection): The rule instructs the system to find all grid cells in the scene with [semantic attribute]='power line'.

[0067] Example (illegal building monitoring): The rule instructs the system to find all grid cells with [semantic attribute]='building' and [building height]>15 meters (in order to filter out high-rise buildings for key monitoring).

[0068] The system executes an efficient spatial database query to quickly filter out all grid cells that meet the rules, forming a set of candidate targets.

[0069] S312: Aggregation and segmentation of target instances; The initially selected mesh cells are discrete. This step aggregates them into complete, independent target individuals. A 3D connectivity analysis is performed on the candidate mesh cells. Spatially adjacent and semantically identical meshes are merged into the same target instance. For example, all power line meshes belonging to the same transmission tower are aggregated into a power line string target, and all meshes belonging to the same building are aggregated into a building instance. A unique ID is assigned to each aggregated independent target.

[0070] S313: Calculation of geometric and spatial relationship characteristics; For each aggregated target instance, its precise geometric features and spatial relationship with the surrounding environment are calculated. For planar targets (such as buildings), its 3D outer cuboid is calculated to extract contour coordinates, base area, building height, volume, etc. For linear targets (such as power lines), the center points of the aggregated mesh are fitted into 3D vector lines, and their length and sag (deflection) are calculated.

[0071] For power lines, the system calculates the 3D Euclidean distance from each point on the power line vector to the vegetation grid below and around it, automatically identifying the minimum clearance distance and its location. It also calculates the nearest distance between the target and other terrain features (such as buildings and roads).

[0072] S314: Change Detection and State Feature Analysis; The extracted target features are compared with historical feature data archived in the knowledge base. Newly added building outlines, expanded areas, or disappeared features are identified through vector overlay analysis or raster difference algorithms. The sag data of the same power line at different times is compared to determine if its deformation exceeds the safety threshold. Targets are then labeled with status tags such as "new," "demolished," "deformed," and "normal."

[0073] S32. The extracted target features are correlated with electronic fence data, real-time aircraft status data, and meteorological data to identify potential risk events. This step specifically realizes the fusion and conflict detection of multi-source data at the three-dimensional space and business rule level, including the following steps: S321: Correlation calculation with electronic fence data, i.e., spatial compliance analysis; Acquire task-related, published geofence data that defines the three-dimensional spatial boundaries and management rules of areas such as no-fly zones, height-restricted zones, and protected areas.

[0074] The three-dimensional geometry of the target features extracted by S31 (such as the outline of a building or the cylindrical buffer zone of a planned route) is quickly intersected with the three-dimensional boundary of the electronic fence.

[0075] If the outline of a building intersects with the fence of a restricted construction zone, it will trigger a risk event of illegal construction.

[0076] If a planned flight path intersects with the no-fly zone fence, it will trigger a flight path violation risk event.

[0077] S322: Correlation calculation with real-time aircraft situational data, i.e., flight safety analysis; The system accesses real-time aircraft situational data from sources such as air traffic control and ADS-B, including the aircraft's unique identifier, latitude and longitude, altitude, speed, and heading. A three-dimensional safety buffer zone is generated, centered on static obstacles extracted by S31, according to safety standards. The system determines whether the real-time aircraft's position falls within any safety buffer zone; if so, a static obstacle approach risk event is triggered.

[0078] Based on the aircraft's real-time status, its short-term (e.g., 30-60 seconds) flight trajectory is predicted. This predicted trajectory is then compared with the safety buffer zone of static obstacles and the predicted trajectories of other aircraft using a four-dimensional (spatial + temporal) conflict detection system. If the predictions intersect, a flight conflict risk event is triggered.

[0079] S323: Correlation calculation with meteorological data, i.e., environmental safety analysis; Acquire gridded meteorological data (such as wind speed, wind direction, visibility, and precipitation) for the task area and match this data with the spatial location of the target features.

[0080] The real-time wind speed is compared with the wind speed threshold for safe drone operation, and the visibility is compared with the minimum requirements for visual inspection. If the threshold is exceeded, it indicates that a severe weather operation risk event has been triggered in the area.

[0081] By combining wind speed and direction with the building's geometric features extracted from S31, and by consulting wind load calculation tables or performing simple fluid dynamics analysis, the risk of wind-induced vibration in the building is assessed. If the assessment results exceed the safe range, a structural safety warning event is triggered for the building.

[0082] S33. Based on the identified risk events, assign a comprehensive risk level to each event and automatically generate an inspection analysis report that includes a description of the risk event, its spatial location, risk level, and handling recommendations.

[0083] This step quantifies risk based on a risk assessment matrix and generates a structured report based on templates and a knowledge base; specifically, it includes: S331: Quantitative calculation of risk level; This step quantifies risk events into a comprehensive level using a predefined risk assessment matrix. This matrix is ​​a function of two dimensions: the probability of risk occurrence and the degree of potential impact.

[0084] Calculate a probability value for each risk event based on its type and current data status. The following formula is recommended for quantitative assessment:

[0085] In the formula, The probability of risk occurrence is normalized to the [0,1] interval to be calculated. This is a baseline probability based on risk type. It is derived from historical statistics (e.g., the historical frequency of aircraft intrusions into the buffer zone). This is a dynamic probability based on real-time data. For example, for flight conflicts, it can be calculated in real time based on aircraft speed and heading angle; for weather risks, it can be calculated based on the degree to which wind speed approaches a threshold. These are the weights for the type probability and the data probability, respectively. This allows the system to flexibly adjust the emphasis between historical experience and real-time situation.

[0086] Assess the potential severity of consequences should a risk event occur. The following formula is recommended:

[0087] In the formula, This represents the final calculated degree of influence. This is the inherent severity coefficient for the risk type. It is defined by expert knowledge (e.g., a collision risk coefficient of 1.0 and an environmental warning coefficient of 0.3). This is the risk scale coefficient, which is proportional to the target value or scope involved in the risk (for example, the coefficient is 1.0 for the main power grid and 0.6 for ordinary lines).

[0088] Calculated and The values ​​are mapped to the risk assessment matrix shown in Table 1 to determine the final comprehensive risk level.

[0089] Table 1: Risk Assessment Matrix

[0090] S332: Inspection analysis report is automatically generated; Based on the task type, retrieve the corresponding structured template (such as HTML, Markdown, or Word template) from the report template library. The following elements will be automatically populated into the corresponding positions in the template: Risk Event Description: Generates a natural language description from a pre-built event description library, based on the risk type and key parameters. Example: At coordinates (E116.XXX, N39.XXX), an aircraft B-1234 is identified as having a potential collision risk with the 500kV high-voltage line safety buffer zone in 120 seconds.

[0091] Spatial location: Directly enter the code or latitude and longitude of the grid where the risk event is located.

[0092] Risk level: Enter the overall risk level calculated in step S331 (e.g., "high risk").

[0093] On-site snapshot: Automatically captures 3D scene screenshots of risk points and embeds them into the report.

[0094] Handling Recommendations: The system queries the inspection knowledge base and automatically matches and outputs predefined standardized handling recommendations based on the risk type and level. Examples: 1. Immediately contact aircraft B-1234 via the communication system and request it to change its course or altitude; 2. Notify power grid maintenance personnel to closely monitor the status of the line.

[0095] The completed content is then combined into a final report file and pushed to the designated user or system.

[0096] In some embodiments, step S4 includes: S41. Based on the spatial location and attributes in the risk assessment results, render the corresponding 3D visual alarm elements at the corresponding grid cells in the 3D mesh scene, and visually highlight the associated inspection targets; this step maps risk attributes to visual variables and renders them in real time in the 3D scene; specifically, it includes the following steps: S411: Visual encoding strategy matching; The risk assessment results are read, and based on the two core attributes of risk type and risk level, a corresponding rendering strategy is matched from a predefined visual encoding library. This strategy defines: Alarm element icons: Different risk types use different 3D models or sprite icons (e.g., collision risk uses a flashing red obstacle cone model, and illegal construction uses a yellow exclamation mark model).

[0097] Color mapping: Risk levels are coded using colors, typically following the traffic light principle. High risk: Red (RGB:255,0,0) Medium risk: Yellow (RGB:255,255,0) Low risk: Blue (RGB:0,0,255) Dynamic behavior: such as flicker frequency, intensity and period of pulse halo.

[0098] S412: Positioning and rendering of 3D alarm elements; Based on the spatial location (grid encoding or latitude and longitude) in the risk assessment results, the spatial grid engine decoder converts the location into specific coordinates (X, Y, Z) in the 3D scene.

[0099] At this coordinate, the 3D alarm element matched in S411 is instantiated and rendered. To enhance the alert effect, its visual attributes change dynamically over time t.

[0100] S413: Visual highlighting of associated inspection targets; This step not only marks risk points but also highlights the affected entities or associated targets. Based on the correlation between the risk event and the target features extracted from S31, identify one or a group of mesh cells that need to be highlighted. Dynamically modify the rendering material or shader parameters of these target meshes. A commonly used highlighting method is the outer glow effect, whose halo intensity I can be modulated based on the risk level.

[0101] In the formula, The intensity of the halo used in the final rendering. Based on the intensity of the halo. The risk level is quantified (e.g., low=1, medium=2, high=3). This is the intensity adjustment coefficient, used to control the degree of visual difference between different risk levels.

[0102] S42. In the visualization interface, the content of the inspection analysis report and the real-time dynamic business data visualization dashboard are displayed synchronously; and the two-way interactive linkage of the three components—the 3D grid scene, the inspection analysis report, and the real-time data dashboard—is realized. Selection operations in any component will cause the other components to perform corresponding view positioning and content highlighting. This step is based on an event-driven publish-subscribe pattern to achieve state synchronization across the three ends, specifically including: S421: Registration and listening of the unified event manager; During system initialization, a unified event manager is created. The three client components—the 3D mesh scene, the analysis report component, and the real-time data dashboard—will register the events they are interested in with this manager and subscribe to events from other components.

[0103] S422: Integration from scenarios to reports and dashboards; A user clicks a flashing alert icon representing high risk in a 3D scene. The scene component captures this click event and obtains the unique ID of the clicked target (e.g., risk_id: 508). The scene component publishes an 'objectSelected' event to the event manager, carrying this ID as a parameter. The event manager broadcasts this event to the subscribed reporting and data dashboard components. Upon receiving the event, the reporting component scrolls to and highlights the report entry with risk_id 508 in its HTML or UI structure (e.g., by changing the background color of that entry to yellow). Upon receiving the event, the data dashboard component locates and highlights real-time data related to risk ID 508 in its list (e.g., highlighting the involved aircraft ID in the "Aircraft List").

[0104] S423: The linkage from reports to scenarios and dashboards; The user selected the record "Aircraft B-1234 Intruding into the No-Fly Zone" in the analysis report list. The reporting component published a 'reportItemSelected' event, carrying the spatial location and target ID associated with this record. The 3D scene component subscribed to this event. Upon receiving it, it immediately executed a smooth camera animation, flying and zooming to the spatial location carried by the event, and maximizing the display of the alarm element and the associated no-fly zone fence at that location in the center of the view. The data dashboard component subscribed to this event. Upon receiving it, it highlighted B-1234 in the "Aircraft List" and simultaneously updated other charts, highlighting the data series related to B-1234.

[0105] S424: From Kanban boards to the integration of scenarios and reports; A user clicks on a severe convective weather area on the "Weather Radar Chart" of the real-time data dashboard. The dashboard component publishes a 'dataPointSelected' event, carrying the geographic boundary coordinates of the weather area. The 3D scene component subscribes to this event. Upon receiving it, it immediately draws a three-dimensional transparent overlay (such as a red, semi-transparent dome) of the weather area in the 3D scene based on the coordinates, and adjusts the viewpoint to make it fully visible. The reporting component subscribes to this event. Upon receiving it, it automatically filters and highlights all inspection plans or risk events located in or predicted to pass through this weather area.

[0106] In some embodiments, step S4 further includes: In the three-dimensional mesh scene, the pre-planned inspection route is visualized; Based on scenario data and real-time dynamic business data, the air routes are simulated and conflict detected, and air route segments with risks are identified with differentiated visual styles. It provides the function of visually comparing multiple inspection plans on the same screen.

[0107] like Figure 2 As shown in the figure, this invention also provides a multi-source low-altitude inspection data cleaning and intelligent analysis processing system, applied to a digital twin system. The system includes: The data intelligent aggregation module is used to respond to the low-altitude inspection task plan submitted by the user, which includes the task area and task type, intelligently introduce and aggregate multi-source static spatial data related to the task plan, and perform data timeliness management during the introduction process. The data fusion and scene construction module is used to intelligently preprocess the aggregated multi-source static spatial data, and based on the pre-built inspection knowledge base and spatial grid engine, fuse the preprocessed data to generate a unified three-dimensional grid scene. The intelligent analysis and risk assessment module is used to automatically extract the features of the inspection targets based on the three-dimensional grid scene, and perform correlation analysis between the extraction results and real-time dynamic business data to generate inspection analysis reports and risk assessment results; the real-time dynamic business data includes at least real-time aircraft status data, electronic fence data and meteorological data; The integrated visualization and decision support module is used to present the inspection analysis report, risk assessment results and the three-dimensional mesh scene in an integrated visualization, providing a basis for decision-making for actual inspections in the physical world.

[0108] In some embodiments, the data intelligent aggregation module is specifically configured to: In response to a user-submitted low-altitude patrol mission plan that includes the mission area and mission type, the required set of multi-source static spatial data types is determined from a pre-set mission-data mapping table based on the mission type. Based on the task area and the set of data types, automatically import multi-source static spatial data related to the task planning from the local database and external data service interface; When importing multi-source static spatial data from the local database, a data timeliness check based on a preset validity period is performed. If the data has expired, the system automatically switches to importing real-time or updated data of the same type from an external data service interface.

[0109] In some embodiments, the data fusion and scene construction module specifically includes: The preprocessing unit is used to uniformly register the aggregated multi-source static spatial data to a preset spatial coordinate system; and based on the rules corresponding to the task type in the pre-built inspection knowledge base, to perform dynamic cleaning and enhancement processing on the multi-source static spatial data that matches the data type; the multi-source static spatial data includes at least oblique photogrammetry model data, terrain data, vector data and point cloud data; The grid division unit is used to obtain the corresponding grid configuration strategy from the inspection knowledge base according to the task type; based on the grid configuration strategy and the task area, the spatial grid engine is used to perform multi-scale grid division of the three-dimensional space, and each grid unit is assigned a unique integer code; The data mapping unit is used to map the preprocessed oblique photogrammetry model data, terrain data, vector data, and point cloud data to the corresponding grid cells according to the gridding configuration strategy, and to perform attribute association. The scene generation unit is used to generate a unified 3D mesh scene based on mesh units with completed attribute associations.

[0110] In some embodiments, the intelligent analysis and risk assessment module specifically includes: The feature extraction unit is used to perform spatial query and calculation based on the semantic attributes of the grid cells in the three-dimensional grid scene, and extract target features related to the inspection task type. The correlation analysis unit is used to perform correlation calculations with the extracted target features and electronic fence data, real-time aircraft status data and meteorological data respectively to identify potential risk events; The report generation unit is used to assign a comprehensive risk level to each identified risk event and automatically generate an inspection analysis report that includes a description of the risk event, its spatial location, risk level, and handling recommendations.

[0111] In some embodiments, the integrated visualization and decision support module is specifically configured to: Based on the spatial location and attributes in the risk assessment results, the corresponding 3D visualization alarm elements are rendered at the corresponding grid cells in the 3D grid scene, and the associated inspection targets are visually highlighted. The visualization interface synchronously displays the content of the inspection analysis report and the visualization data dashboard of real-time dynamic business data; and realizes bidirectional interactive linkage between the three components of the three-dimensional grid scene, inspection analysis report and real-time data dashboard. The selection operation in any component can enable the other components to perform corresponding view positioning and content highlighting.

[0112] This invention also provides an electronic device, comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus. The communication bus can be used for information transmission between the electronic device and sensors. The processor can call logical instructions in memory to execute the following methods: S1. In response to a user-submitted low-altitude inspection task plan containing the task area and task type, intelligently introduce and aggregate multi-source static spatial data related to the task plan, and perform data timeliness management during the introduction process; S2. Perform intelligent preprocessing on the aggregated multi-source static spatial data, and based on a pre-built inspection knowledge base and airspace grid engine, fuse the preprocessed data to generate a unified three-dimensional grid scene; S3. Based on the three-dimensional grid scene, automatically extract the inspection target features, and perform correlation analysis with real-time dynamic business data to generate an inspection analysis report and risk assessment results; the real-time dynamic business data includes at least real-time aircraft status data, electronic fence data, and meteorological data; S4. Integrate the inspection analysis report, risk assessment results, and the three-dimensional grid scene for visual presentation, providing a decision-making basis for actual inspections in the physical world.

[0113] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0114] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for cleaning and intelligent analysis of multi-source low-altitude inspection data, characterized in that, Applied to digital twin systems, the method includes the following steps: S1. Responding to the user-submitted low-altitude inspection task plan, which includes the task area and task type, intelligently introduces and aggregates multi-source static spatial data related to the task plan, and performs data timeliness management during the introduction process. S2. Intelligent preprocessing of the aggregated multi-source static spatial data, and fusion of the preprocessed data into a unified three-dimensional grid scene based on the pre-built inspection knowledge base and spatial grid engine. S3. Based on a 3D grid scene, automatically extract the features of the inspection target, and perform correlation analysis with real-time dynamic business data to generate an inspection analysis report and risk assessment results; real-time dynamic business data includes at least real-time aircraft status data, electronic fence data, and meteorological data. S4. Integrate inspection analysis reports, risk assessment results, and 3D mesh scenes into a unified visual presentation to provide decision-making basis for actual inspections in the physical world.

2. The method for cleaning and intelligent analysis of multi-source low-altitude inspection data according to claim 1, characterized in that, The steps in S1 include: S11. In response to the user-submitted low-altitude inspection task plan containing the task area and task type, determine the required set of multi-source static spatial data types from the preset task-data mapping table according to the task type. S12. Based on the task area and data type set, automatically import multi-source static spatial data related to task planning from the local database and external data service interface; When importing multi-source static spatial data from the local database, a data timeliness check based on a preset validity period is performed. If the data has expired, it will automatically switch to importing real-time or updated similar data from an external data service interface.

3. The method for cleaning and intelligent analysis of multi-source low-altitude inspection data according to claim 2, characterized in that, Data timeliness verification includes: Retrieve timestamps from data in the local database; Calculate the difference between the obtained timestamp and the current time; If the difference is greater than the preset validity period for the corresponding data type, then the data is determined to have expired.

4. The method for cleaning and intelligent analysis of multi-source low-altitude inspection data according to claim 3, characterized in that, The pre-defined task-data mapping table is generated and maintained in the following ways: Initialize the definition of the correspondence between different inspection task types and the required data types and attributes; Collect historical inspection task data usage records and optimize and update the corresponding relationships using machine learning models; specifically including: Collect the task characteristics of historical inspection tasks and the corresponding actual usage data combinations to construct a training sample set; Based on the training sample set, the multi-label classification model is trained to obtain the data recommendation model; For new inspection tasks, a data recommendation model is used to generate recommended data combinations. Recommended data combinations that meet the preset reliability criteria will be updated to the preset task-data mapping table after being approved by the administrator.

5. The method for cleaning and intelligent analysis of multi-source low-altitude inspection data according to claim 4, characterized in that, The steps in S2 include: S21. Register the aggregated multi-source static spatial data to a preset spatial coordinate system; and based on the rules corresponding to the task type in the pre-built inspection knowledge base, perform dynamic cleaning and enhancement processing on the multi-source static spatial data that matches the data type; the multi-source static spatial data includes at least oblique photogrammetry model data, terrain data, vector data and point cloud data; S22. Based on the task type, obtain the corresponding grid configuration strategy from the inspection knowledge base; based on the grid configuration strategy and the task area, use the spatial grid engine to perform multi-scale grid division of the three-dimensional space, and assign a unique integer code to each grid unit; S23. The oblique photogrammetry model data, terrain data, vector data and point cloud data preprocessed in S21 are mapped to the corresponding grid cells according to the grid configuration strategy, and attribute association is performed. S24. Generate a unified 3D mesh scene based on the mesh cells with completed attribute associations.

6. The method for cleaning and intelligent analysis of multi-source low-altitude inspection data according to claim 5, characterized in that, The steps for dynamic cleaning and enhancement of multi-source static spatial data, tailored to the data type, include: Based on the task type, retrieve the preprocessing rule set corresponding to the data type from the inspection knowledge base; Based on the preprocessing rule set, geometric correction and texture enhancement are performed on the oblique photogrammetry model data, elevation outlier repair and data hole filling are performed on the terrain data, topological relationship repair and attribute normalization are performed on the vector data, and outlier removal, ground point classification and extraction, and non-ground point semantic segmentation are performed on the point cloud data.

7. The method for cleaning and intelligent analysis of multi-source low-altitude inspection data according to claim 6, characterized in that, The steps in S3 include: S31. Based on the semantic attributes of grid cells in the 3D grid scene, perform spatial query and calculation to extract target features related to the inspection task type; S32. The extracted target features are correlated with electronic fence data, real-time aircraft status data and meteorological data to identify potential risk events. S33. Based on the identified risk events, assign a comprehensive risk level to each event and automatically generate an inspection analysis report that includes a description of the risk event, its spatial location, risk level, and handling recommendations.

8. The method for cleaning and intelligent analysis of multi-source low-altitude inspection data according to claim 7, characterized in that, The steps in S4 include: S41. Based on the spatial location and attributes in the risk assessment results, render the corresponding 3D visualization alarm elements at the corresponding grid cells in the 3D grid scene, and visually highlight the associated inspection targets. S42. In the visualization interface, the content of the inspection analysis report and the visualization data dashboard of real-time dynamic business data are displayed synchronously; and the two-way interactive linkage of the three components of the three-dimensional grid scene, inspection analysis report and real-time data dashboard is realized. The selection operation in any component can enable the other components to perform corresponding view positioning and content highlighting.

9. A multi-source low-altitude inspection data cleaning and intelligent analysis processing system, characterized in that, Applied to a digital twin system, the system includes: The data intelligent aggregation module is used to respond to the low-altitude inspection task plan submitted by the user, which includes the task area and task type, intelligently introduce and aggregate multi-source static spatial data related to the task plan, and perform data timeliness management during the introduction process. The data fusion and scene construction module is used to intelligently preprocess the aggregated multi-source static spatial data, and based on the pre-built inspection knowledge base and spatial grid engine, fuse the preprocessed data to generate a unified three-dimensional grid scene. The intelligent analysis and risk assessment module is used to automatically extract the features of the inspection targets based on the three-dimensional grid scene, and perform correlation analysis between the extraction results and real-time dynamic business data to generate inspection analysis reports and risk assessment results; the real-time dynamic business data includes at least real-time aircraft status data, electronic fence data and meteorological data; The integrated visualization and decision support module is used to present the inspection analysis report, risk assessment results and the three-dimensional mesh scene in an integrated visualization, providing a basis for decision-making for actual inspections in the physical world.

10. An electronic device, characterized in that, The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores computer program instructions executable by the at least one processor, the computer program instructions being executed by the at least one processor to enable the at least one processor to perform the multi-source low-altitude inspection data cleaning and intelligent analysis processing method as described in any one of claims 1 to 8.

Images