Unmanned area vehicle parking navigation method and system based on satellite remote sensing

Abstract

The application discloses a satellite remote sensing-based unmanned area vehicle parking navigation method and system, and relates to the technical field of vehicle parking navigation.The method comprises the following steps: performing multi-level buffer zone space modeling to obtain a parking buffer zone; synchronously acquiring DEM grid data and multispectral remote sensing image data; constructing an elevation gradient distribution map; performing ground feature spectral feature interpretation to generate a no-entry area mask; performing passable area extraction to obtain a candidate parking area set; extracting associated road vector data of the parking buffer zone; performing topological reachability modeling to filter and output a target parking area and a target navigation route, and performing offline path guidance.The application solves the technical problem that, in the prior art, it is difficult to effectively identify no-entry risk areas in an unmanned area scenario, which leads to great blindness in vehicle parking site selection and insufficient safety of navigation routes due to the influence of the navigation routes, and achieves the technical effect of improving the accuracy of vehicle parking site selection in an unmanned area and the safety of navigation routes.

Description

Technical Field

[0001] This invention relates to the field of vehicle parking navigation technology, specifically to a method and system for vehicle parking navigation in uninhabited areas based on satellite remote sensing. Background Technology

[0002] When carrying out construction tasks such as power line inspection, geological exploration, and emergency rescue in uninhabited areas, vehicles need to rely on navigation systems to complete route planning and parking location selection. However, traditional navigation technology faces significant bottlenecks in this scenario: On the one hand, uninhabited areas generally lack complete ground transportation infrastructure and real-time communication networks. Conventional navigation, which relies on satellite signals, can only provide rough route guidance and cannot accurately identify terrain obstacles, dangerous features, and other restricted areas, which can easily lead to vehicles encountering traffic risks. On the other hand, construction in uninhabited areas often requires the operation of drones. Traditional navigation does not take into account the special requirements such as the minimum turning space of vehicles and the flatness of drone take-off and landing sites. It only selects parking points based on road accessibility, which can easily lead to problems such as unsuitable terrain in parking areas, which cannot meet the safe take-off and landing or operational preparation of drones, resulting in reduced construction efficiency or even equipment damage and personnel safety hazards.

[0003] Existing technologies struggle to effectively identify restricted areas in uninhabited environments, leading to significant risks associated with vehicle parking and navigation routes being easily affected, resulting in insufficient safety. Summary of the Invention

[0004] This application provides a method and system for vehicle parking navigation in uninhabited areas based on satellite remote sensing, which is used to address the technical problem in existing technologies that it is difficult to effectively identify restricted areas in uninhabited scenarios, resulting in blind selection of vehicle parking locations, susceptibility to navigation routes, and insufficient safety.

[0005] In view of the above problems, this application provides a method and system for vehicle parking navigation in uninhabited areas based on satellite remote sensing.

[0006] The first aspect of this application provides a method for vehicle parking navigation in uninhabited areas based on satellite remote sensing, the method comprising:

[0007] After receiving the construction task work point, a multi-level buffer space model is performed with the work point as the center to obtain a parking buffer. Using the geographical boundary of the parking buffer as a constraint, DEM raster data and multispectral remote sensing image data are acquired simultaneously. Terrain gradient quantization analysis is performed on the DEM raster data to construct an elevation gradient distribution map. Spectral feature interpretation of ground features is performed on the multispectral remote sensing image data to generate a restricted area mask. After merging the elevation gradient distribution map and the restricted area mask through spatial overlay analysis, passable areas are extracted using the minimum turning space constraint for vehicles to obtain a candidate parking area set. The associated road vector data of the parking buffer is extracted based on the spatial data service interface. Topology accessibility modeling is performed based on the candidate parking area set and the associated road vector data to filter and output target parking areas and target navigation routes for offline path guidance.

[0008] A second aspect of this application provides a satellite remote sensing-based vehicle parking navigation system for uninhabited areas, the system comprising:

[0009] The system comprises the following modules: a parking buffer zone acquisition module, which receives the construction task work point and performs multi-level buffer zone spatial modeling centered on the work point to obtain the parking buffer zone; an image data acquisition module, which simultaneously acquires DEM raster data and multispectral remote sensing image data, constrained by the geographical boundaries of the parking buffer zone; an elevation gradient distribution map construction module, which performs terrain gradient quantization analysis on the DEM raster data to construct an elevation gradient distribution map; a restricted area mask generation module, which performs ground feature spectral feature interpretation on the multispectral remote sensing image data to generate a restricted area mask; a parking area set acquisition module, which fuses the elevation gradient distribution map and the restricted area mask through spatial overlay analysis, and extracts passable areas based on the minimum turning space constraint for vehicles to obtain a candidate parking area set; a vector data extraction module, which extracts the associated road vector data of the parking buffer zone based on the spatial data service interface; and an offline path guidance module, which performs topology accessibility modeling based on the candidate parking area set and associated road vector data, filters and outputs target parking areas and target navigation routes, and provides offline path guidance.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] After receiving the construction task work points, multi-level buffer space modeling is performed to obtain parking buffer zones. Using the geographical boundaries of the parking buffer zones as constraints, DEM raster data and multispectral remote sensing image data are acquired simultaneously. Terrain gradient quantization analysis is performed on the DEM raster data to construct an elevation gradient distribution map. Spectral feature interpretation of ground features is performed on the multispectral remote sensing image data to generate a restricted area mask. After spatial overlay analysis to fuse the elevation gradient distribution map and the restricted area mask, passable areas are extracted using the minimum turning space constraint for vehicles, resulting in a candidate parking area set. Associated road vector data of the parking buffer zones is extracted based on the spatial data service interface. Topology accessibility modeling is performed based on the candidate parking area set and associated road vector data to filter and output target parking areas and target navigation routes for offline path guidance. This achieves the technical effect of improving the accuracy of vehicle parking location selection and the safety of navigation routes in uninhabited areas. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A schematic flowchart of a vehicle parking navigation method based on satellite remote sensing in uninhabited areas provided in this application embodiment;

[0014] Figure 2 A schematic diagram of the structure of a vehicle parking navigation system based on satellite remote sensing in uninhabited areas provided in this application embodiment.

[0015] Figure labeling: Parking buffer zone acquisition module 10, image data acquisition module 20, elevation gradient distribution map construction module 30, restricted area mask generation module 40, parking area set acquisition module 50, vector data extraction module 60, offline path guidance module 70. Detailed Implementation

[0016] This application provides a vehicle parking navigation method and system based on satellite remote sensing in uninhabited areas, which addresses the technical problem in existing technologies that makes it difficult to effectively identify restricted areas in uninhabited scenarios, resulting in blind selection of vehicle parking locations, susceptibility to navigation routes, and insufficient safety.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] Example 1, as Figure 1 As shown, this application provides a vehicle parking navigation method in uninhabited areas based on satellite remote sensing, the method comprising:

[0019] Step S100: After receiving the construction task operation point, perform multi-level buffer space modeling with the construction task operation point as the center to obtain the parking buffer.

[0020] Specifically, the system receives the latitude and longitude coordinates of the construction task operation point uploaded by the construction unit or supervisor via mobile device, using the WGS84 World Geodetic System coordinate system. This coordinate system provides the core benchmark for subsequent buffer zone modeling. Then, a multi-level buffer zone spatial modeling process is initiated centered on this operation point. Prioritizing the core requirements of unmanned aerial vehicle (UAV) operations in uninhabited areas, a first-level buffer zone with a preset radius (e.g., 3 kilometers) is generated. This range corresponds to the conventional safe operating radius of UAVs and serves as the primary search area to ensure the efficiency and signal stability of subsequent UAV inspections. If, after initial parking area identification within the first-level buffer zone, no candidate parking points meeting the basic conditions of flat terrain and road access are found (i.e., the candidate parking area is empty), the radius of concentric circles is expanded based on the boundary of the first-level buffer zone, up to a maximum of 5 kilometers, which conforms to the system's set limit for UAV operation maneuverability. By gradually expanding the search range and repeatedly identifying and verifying, a parking buffer zone with valid parking candidate points is obtained. Finally, the geographical boundaries for subsequent remote sensing data acquisition, terrain analysis, and path planning are determined, ensuring that the search range meets the needs of UAV operations while covering sufficient potential parking areas.

[0021] Step S200: Using the geographical boundary of the parking buffer zone as a constraint, simultaneously acquire DEM raster data and multispectral remote sensing image data.

[0022] Specifically, after determining the geographical boundary of the parking buffer zone, the system uses this boundary as a spatial constraint to simultaneously initiate the acquisition process for two types of key data. The DEM (Digital Elevation Model) raster data is primarily acquired from a geospatial data cloud platform, prioritizing 30-meter resolution GDEM data. For some high-precision scenarios, higher resolutions can be used. This data accurately reflects the elevation information within the buffer zone, providing a foundation for subsequent calculations of terrain slope and undulation. Multispectral remote sensing imagery data is acquired primarily through commercial high-resolution satellite imagery with a resolution of 0.3–0.5 meters, supplemented by publicly available data sources such as Tianditu and Google Earth. Data is also updated regularly based on project requirements, ensuring that the imagery clearly presents details of features within the buffer zone, such as roads, water bodies, and vegetation distribution, especially in areas with rapidly changing field environments. During the data acquisition process, the system automatically verifies the coordinate system of both types of data, unifying them to the WGS84 coordinate system and geographical boundaries. This ensures that the data coverage area fully matches the parking buffer zone, avoiding the impact of data misalignment or range deviation on the accuracy of subsequent analysis. It provides complete and accurate data source support for subsequent terrain gradient quantification analysis and restricted area identification.

[0023] Step S300: Perform terrain gradient quantization analysis on the DEM raster data to construct an elevation gradient distribution map.

[0024] Specifically, based on the acquired DEM raster data, the surface slope within the buffer zone is calculated using the Spatial Analyst advanced spatial analysis extension module of the ArcGIS professional geographic information system platform. A slope raster map in degrees is output. Then, a preset slope threshold, such as ≤3°, is used to meet the terrain safety requirements for drone take-off and landing and vehicle parking. The slope raster map is traversed and binarized, retaining areas with acceptable slope, generating a raster mask for suitable slope areas. Next, neighborhood elevation range analysis is performed on the DEM raster data. The difference between the maximum and minimum elevation values ​​within a fixed window is calculated using a focus statistics tool, outputting a terrain relief raster map reflecting micro-topographic undulations. Again, a preset relief threshold is applied, such as ≤3°, to ensure suitability for drone take-off and landing and vehicle parking. For slopes ≤0.5 meters, binarization segmentation is performed to filter out areas with gentle undulations, resulting in a raster mask for suitable undulation areas. Subsequently, the raster masks for suitable slope areas and suitable undulation areas are registered in spatial coordinate systems to ensure they are both in the WGS84 coordinate system. Based on the influence weight of terrain on parking (e.g., slope weight 0.6, undulation weight 0.4), raster algebra operations are performed to merge the suitability characteristics of the two types of masks into a unified quantitative index. Finally, an elevation gradient distribution map that can intuitively reflect the terrain suitability of different areas within the buffer zone is generated, providing a quantitative basis for terrain for subsequent selection of parking areas.

[0025] Step S400: Perform ground object spectral feature interpretation on the multispectral remote sensing image data to generate a no-entry area mask.

[0026] Specifically, the process begins by interpreting multispectral remote sensing image data based on the differences in spectral reflectance of different land features. For water bodies such as rivers, lakes, and swamps, predefined water-related bands, such as low reflectance in the near-infrared band and moderate reflectance in the visible light band, are used to identify regions with reflectance differences in the image. Water body area masks are then segmented and output to exclude areas where vehicles are likely to get stuck or are impassable. Simultaneously, neighborhood elevation change analysis is performed using acquired DEM raster data to locate dangerous terrain areas with abrupt elevation changes, such as cliffs and steep slopes, generating dangerous terrain area masks to prevent vehicle accidents caused by steep terrain. Subsequently, the water body area masks and dangerous terrain area masks are merged. Considering safety redundancy for vehicle passage, a preset safety distance is used, such as extending outwards by 5-10 meters around the water body and dangerous terrain, to buffer the merged area and ensure sufficient safety distance between vehicles and dangerous areas. Finally, a no-entry area mask covering all unsuitable areas is generated, clearly defining the exclusion range for subsequent extraction of passable areas.

[0027] Step S500: After fusing the elevation gradient distribution map and the restricted area mask through spatial overlay analysis, the passable area is extracted using the minimum turning space constraint of the vehicle to obtain a set of candidate parking areas.

[0028] Specifically, spatial overlay analysis is first used to analyze the elevation gradient distribution map, reflecting terrain suitability and restricted area masks (such as marking water bodies, dangerous terrain, and other impassable areas). Raster algebra operations are then performed to fuse the terrain suitability score with the impassability attribute of restricted areas, generating a accessibility score raster. Restricted areas are directly marked as "0 points," meaning completely impassable, while non-restricted areas retain their original terrain suitability score. Next, according to a preset terrain safety threshold (e.g., a score ≥ 60 points corresponding to a slope ≤ 3° and undulation ≤ 0.5 meters), the accessibility score raster is traversed and binarized to segment the data, selecting areas that meet the safety standards for passage. The system first generates a grid mask for suitable traffic areas. Then, using a preset minimum parking area (e.g., no less than 20 square meters) to ensure vehicles can park and turn around normally as a spatial constraint, it marks the grid mask for connected components, eliminating small areas with insufficient area to obtain potential areas that meet the size requirements. Finally, combining the minimum turning space constraint for vehicles (e.g., based on the turning radius of common supervised vehicles) to ensure vehicles can complete turning operations within the area, it optimizes the geometry of the areas that meet the size requirements, eliminating irregularly shaped areas that cannot meet the turning requirements of vehicles, and finally extracting a set of candidate parking areas that combine terrain suitability, safety, and operational convenience.

[0029] Step S600: Extract the associated road vector data of the parking buffer zone based on the spatial data service interface.

[0030] Specifically, by calling standardized spatial data service interfaces, prioritizing interfaces conforming to the OGC WFS (Open Geospatial Consortium) network element service standard specifications, such as the road element service interface of the Tianditu Geographic Information Public Service Platform, the system accurately extracts associated road vector data within the buffer zone using the established geographic boundaries of the parking buffer as the spatial filtering condition. The extracted data covers various road types commonly found in uninhabited areas, including unpaved dirt roads, simple rutted roads, forest trails, and connecting rural roads. It also includes topological relationships such as intersections, road segment connectivity, and road classifications such as arterial roads and branch roads. During data extraction, the system automatically verifies the coordinate system of the road vector data to ensure it is consistent with the parking buffer and DEM data, using the WGS84 coordinate system. Data integrity is also checked; if missing road data is found in some areas, such as incomplete coverage of remote uninhabited areas, previously acquired high-resolution multispectral remote sensing imagery is used for supplementary completion. This ensures that the extracted associated road vector data fully reflects the road network distribution within the parking buffer, laying a data foundation for subsequent road network map construction and parking area accessibility analysis.

[0031] Step S700: Perform topology accessibility modeling based on the candidate parking area set and associated road vector data, filter and output target parking areas and target navigation routes, and provide offline path guidance.

[0032] Specifically, spatial proximity analysis is performed on the candidate parking area set and associated road vector data. The associated road vector data is then reconstructed into a road network topology map including intersections and road segments. Each candidate parking area is then projected onto this topology map, and the Euclidean distance between each area and the nearest road is calculated. Accessible parking areas with a distance ≤ 50 meters that meet the basic requirements for vehicle accessibility are selected, and the geometric center points of these areas are used as candidate parking anchor points. Next, starting from the construction task, candidate parking anchor points uploaded by the mobile terminal are used as start and end nodes. Bidirectional spatial extension and topology pruning are performed on the road network topology map to remove impassable road segments, such as those that are too narrow or damaged, and an accessible road subnet is constructed. Subsequently, the shortest travel path from the starting point to each anchor point is selected within the subnet. Multiple parking navigation routes were initially screened. A multi-factor weighted scoring mechanism was then activated, quantifying the routes based on factors such as path complexity, road grade coefficient, length, environmental risk factors, and the accessibility of the corresponding parking areas, considering terrain suitability, site size, and drone take-off and landing safety index. These scores were then weighted and merged according to preset weights, such as 50% for route scores and 50% for area scores, to obtain a comprehensive score for each solution. Finally, the solutions were sorted in descending order of their comprehensive scores, and the highest-scoring solution was selected as the target parking area and target navigation route. Its coordinates, route details, and other data were converted to a mobile-friendly format and sent. This supports real-time voice broadcasting, turning guidance, and deviation alerts via offline maps on mobile devices, even in uninhabited areas without network access, thus completing offline route guidance.

[0033] In one possible implementation, step S100 further includes:

[0034] Step S110: Generate a first-level buffer zone with a preset first-level radius scale centered on the construction task operation point.

[0035] Step S120: If parking area identification is performed with the first-level buffer as the core search area and the obtained candidate parking areas are empty sets, then the concentric circle radius is expanded based on the first-level buffer until the parking candidate points are obtained as the parking buffer with a non-empty set.

[0036] Specifically, upon receiving the precise latitude and longitude coordinates of the construction task's work point, the WGS84 coordinate system is used. Based on the core requirements of drone inspection in unmanned power grid construction supervision, it is necessary to ensure that the drone can efficiently cover the work point after taking off from the parking point, while also guaranteeing signal stability and emergency operating space. Using the work point as the center, a preset primary radius scale is automatically generated, such as a circular primary buffer zone of 3 kilometers. This radius corresponds to the conventional safe operating radius of most industrial-grade rotary-wing drones on the market, minimizing unnecessary long-distance vehicle travel while reserving sufficient safety space for subsequent drone take-off, landing, and inspection operations. This circular area serves as the primary parking search core area, defining the initial geographical scope for the accurate acquisition of DEM data and remote sensing image data, and for parking area identification.

[0037] First, using the first-level buffer zone, with a radius of 3 kilometers as the core search area, preliminary parking area identification is performed. Based on simplified terrain screening, such as quickly determining whether there are flat areas with a slope of ≤3° and basic features to exclude, such as water bodies, cliffs, and other obviously prohibited areas, a preliminary assessment is made as to whether there are candidate areas within this area that meet the basic conditions of "parking is permissible". If the identification results show that the candidate parking area is an empty set, that is, there is no area in the core area that meets the basic requirements for vehicle parking and drone take-off and landing, then the circular boundary of the first-level buffer zone is used as a reference, and the zone is expanded outward according to preset expansion rules, such as expanding outward in steps of 500 meters each time, up to a maximum radius of 5 kilometers, matching the extreme operational maneuver range of drones, and concentric circle radius expansion is performed to form a new, larger buffer zone. After each expansion, the parking area identification process is repeated on the new buffer zone to continuously verify whether there are valid parking candidate points, until a buffer zone with a non-empty set of parking candidate points is identified. Finally, this buffer zone is determined as the official parking buffer zone. This avoids the situation where there are no usable parking points due to the area being too small, and also prevents excessive expansion from increasing the burden of subsequent data processing, ensuring that the search range is accurately matched with actual parking needs.

[0038] In one possible implementation, step S300 further includes:

[0039] Step S310: Calculate the surface slope based on the DEM raster data and output a slope raster map.

[0040] Step S320: Use a preset slope threshold to traverse the slope raster map for binarization segmentation to generate a raster mask for the appropriate slope area.

[0041] Step S330: Perform neighborhood elevation range analysis on the DEM raster data and output a terrain relief raster map.

[0042] Step S340: Perform binarization segmentation on the terrain undulation raster map using a preset undulation threshold to generate a raster mask for a suitable undulation area.

[0043] Step S350: After registering the grid mask of the slope suitable area and the grid mask of the undulation suitable area with spatial coordinate system, perform weighted grid algebra operation to generate the elevation gradient distribution map.

[0044] Specifically, using ArcGIS's Spatial Analyst tool module, the surface slope within the parking buffer zone is calculated based on the acquired DEM raster data, such as 30-meter resolution GDEM data. The slope raster map is output in degrees, and the value of each raster cell in the map directly reflects the terrain tilt of the corresponding area, providing basic data for subsequent slope selection.

[0045] The system retrieves a preset slope safety threshold for vehicle parking and drone take-off and landing in uninhabited areas, typically set to ≤3°. This threshold can be flexibly adjusted based on vehicle type and drone model, with the core objective of ensuring no risk of vehicle slippage and stable drone take-off and landing attitude. The resulting slope raster map is then traversed using this threshold as a standard, with each raster cell labeled with the slope value of its corresponding area. During the traversal, a binarization segmentation operation is performed. Raster cells with slope values ​​≤3° are assigned a value of "1," indicating that the slope meets safety requirements and is considered a suitable slope zone. Raster cells with slope values ​​>3° are assigned a value of "0," indicating that the slope is too steep and excluded from the suitable zone. After segmentation, a slope-suitable zone raster mask is generated, retaining only areas with acceptable slopes. This mask clearly presents the suitable slope range within the buffer zone in black and white binary form, laying the foundation for further filtering of parking areas by integrating undulation conditions.

[0046] The window size for neighborhood analysis is determined, using 3×3 or 5×5 raster cells, which can be adjusted according to the complexity of the terrain in the uninhabited area. The window size needs to balance local details with regional representativeness. Then, using this window as the unit, a window-by-window neighborhood elevation range analysis is performed on the DEM raster data used to calculate the slope. For all raster cells within each window, the corresponding elevation values ​​are extracted, and the difference between the maximum and minimum elevation values ​​within the window is calculated, i.e., the elevation range. This difference directly reflects the severity of terrain undulation in the area covered by the window; the smaller the difference, the gentler the terrain. After all windows have been analyzed, the elevation range result of each window is assigned to the central raster cell of the window, forming a terrain undulation raster map covering the entire parking buffer zone. The value of each raster in the map is the local terrain undulation at the corresponding location, providing accurate data for subsequent selection of suitable parking areas based on undulation.

[0047] First, based on the stability requirements for vehicle parking and the flatness requirements for the drone take-off and landing platform in uninhabited areas, a preset undulation threshold is determined, typically set to ≤0.5 meters. This threshold can be flexibly adjusted according to the vehicle chassis height and the drone landing gear's cushioning capacity. The core purpose is to prevent vehicle instability or drone take-off and landing imbalance caused by excessive local terrain elevation differences. Then, using this threshold as the criterion, the output terrain undulation raster map is traversed. Each raster cell in the map is labeled with the local elevation range of the corresponding area, i.e., the undulation value. During the traversal process, a binarization segmentation operation is performed to separate the undulations... Grid cells with an elevation value ≤ 0.5 meters are assigned a value of "1", indicating that the micro-topography of the area is gentle and meets the undulation requirements for parking and drone operations. Grid cells with an elevation value > 0.5 meters are assigned a value of "0", indicating that the terrain of the area is undulating and is excluded from the suitable area. After segmentation, a suitable elevation area grid mask is generated, which retains only the areas with the required elevation. This mask clearly presents the suitable elevation range within the buffer zone in a black and white binary format, providing data support for subsequent fusion with the slope suitable area mask and further narrowing down the parking area.

[0048] The generated raster masks for suitable slope and suitable undulation areas were registered using a spatial coordinate system. By verifying the geographic coordinate systems of both types of masks, it was ensured that they were both unified to the WGS84 coordinate system, consistent with the previous DEM raster data and parking buffer coordinates. Any potential coordinate offsets were corrected to ensure complete spatial alignment of the two types of masks, avoiding any impact on fusion accuracy due to misalignment. Next, based on the weighting of terrain's influence on parking and drone takeoff and landing, slope directly affects the risk of vehicle rollover and drone attitude control, and its weight was set to 0.6; undulation affects vehicle parking stability and the flatness of the drone takeoff and landing platform. The elevation gradient is set to a weight of 0.4, which can be fine-tuned according to the actual operation scenario. The weighted raster algebra operation is performed, that is, the elevation gradient value of each raster unit = slope suitability value × 0.6 + undulation suitability value × 0.4, where the suitability value "1" represents compliance and "0" represents non-compliance. After the operation is completed, an elevation gradient distribution map covering the entire parking buffer zone is generated. The raster values ​​in the map are between 0 and 1, which intuitively reflect the comprehensive terrain suitability of the corresponding area. The closer the value is to 1, the more the terrain meets the requirements of parking and drone operation, providing an accurate comprehensive terrain quantification basis for subsequent selection of passable parking areas in combination with restricted areas.

[0049] In one possible implementation, step S400 further includes:

[0050] Step S410: Define predefined water body associated bands to identify reflectance differences in the multispectral remote sensing image data, and segment and output water body region masks.

[0051] Step S420: Perform neighborhood elevation change analysis based on the DEM raster data to locate dangerous terrain area masks.

[0052] Step S430: After merging the water area mask and the dangerous terrain area mask, perform a buffer expansion with a preset safety distance to generate the restricted area mask.

[0053] Specifically, based on the unique reflection patterns of water bodies in multispectral bands, a predefined combination of bands associated with water bodies is defined, such as the ratio of near-infrared bands to green bands. Water bodies have extremely low reflectivity in the near-infrared band due to strong absorption, while their reflectivity is moderate in the green band, creating a significant difference in reflectivity compared to other land features such as vegetation and soil. Subsequently, using this band combination as the analysis benchmark, a reflectivity difference identification algorithm is applied to the multispectral remote sensing image data within the parking buffer zone. The image is traversed pixel by pixel, and reflectivity feature values ​​are calculated. Pixels whose reflectivity matches the spectral characteristics of water bodies are classified as water areas, while the remaining pixels are classified as non-water areas. Image segmentation technology is used to extract the classified water areas, generating a binary raster mask that only marks the water body's extent. Water areas are assigned a value of "1", and non-water areas are assigned a value of "0", i.e., a water area mask. This mask clearly defines the boundaries of water bodies such as rivers, lakes, swamps, and permanent water pits in uninhabited areas that are impassable by vehicles, laying the foundation for subsequent integration of dangerous terrain and generation of complete restricted areas.

[0054] Based on the acquired DEM raster data, a reasonable neighborhood analysis window is set, typically a 3×3 or 5×5 raster cell, balancing local terrain details with analysis efficiency. Then, a window-by-window neighborhood elevation change analysis is performed on the DEM raster data within the parking buffer zone. For the central raster in each window, the elevation difference between it and all surrounding raster cells is calculated, and the largest elevation difference is selected. This largest elevation difference is compared with a preset dangerous terrain threshold. If the largest elevation difference exceeds the threshold, the area where the central raster is located is determined to be dangerous terrain. After all window analyses are completed, all raster cells marked as dangerous terrain are integrated to generate a binary raster mask that only covers the dangerous terrain area. Dangerous terrain areas are assigned a value of "1", and non-dangerous terrain areas are assigned a value of "0". This dangerous terrain area mask provides data support for the dangerous terrain dimension in subsequent merging of water areas and construction of complete no-entry boundaries.

[0055] The output water area mask and the output hazardous terrain area mask are spatially superimposed and merged. Through raster algebra operations, the raster cells marked "1" in both types of masks, representing water or hazardous terrain, are uniformly retained to form a merged mask covering all initial hazardous areas, clearly defining the core area in the uninhabited area where vehicles absolutely cannot enter. Subsequently, considering the safety redundancy requirements of field operations, such as preventing vehicles from getting stuck near water edges due to slippery conditions and staying away from cliff edges to prevent accidental slippage, a buffer expansion operation is performed on the boundary of the merged mask according to a preset safety distance, set according to the complexity of the uninhabited area terrain, usually 5-10 meters. That is, a continuous safety buffer zone is added outside the boundary of the initial hazardous area to ensure that vehicles maintain a sufficient safe distance from the hazardous area. After the expansion is completed, a complete restricted area mask containing water, hazardous terrain, and surrounding safety buffer zone is generated. The restricted area is assigned a value of "1", and the passable area is assigned a value of "0". This mask clearly defines the area that must be strictly avoided in subsequent path planning and parking area selection, providing a basic guarantee for the safety of vehicle passage in the uninhabited area.

[0056] In one possible implementation, step S500 further includes:

[0057] Step S510: Perform raster algebra operations on the elevation gradient distribution map and the restricted area mask to generate a passability score raster.

[0058] Step S520: Using a preset terrain safety threshold, traverse the accessibility score grid to perform binarization segmentation, and output the accessibility suitable area grid mask.

[0059] Step S530: Using the preset minimum parking area as a spatial constraint, mark the connected components of the grid mask of the suitable passage area and filter to obtain the area that meets the size standard.

[0060] Step S540: Optimize the geometry of the target area using vehicle turning radius constraints to generate the candidate parking area set.

[0061] Specifically, the rules for grid algebra operations are clearly defined. An elevation gradient distribution map reflects terrain suitability, with grid values ​​ranging from 0 to 1. Values ​​closer to 1 indicate terrain that better meets parking and drone operation requirements. A restricted area mask defines safety boundaries; a restricted area grid value is 1, and a non-restricted area value is 0. The calculation formula is set as: Accessibility Score = Elevation Gradient Value × (1 - Restricted Area Value). During the calculation, if a grid belongs to a restricted area (restricted area value = 1), then "1 - Restricted Area Value = 0," and the accessibility score is directly assigned to 0, indicating that the area is completely impassable due to safety risks. If a grid belongs to a non-restricted area (restricted area value = 0), then the accessibility score is directly equal to the grid's elevation gradient value, preserving its terrain suitability quantification result. Through grid-by-grid operations across the entire buffer zone, a accessibility score grid is finally generated. This grid reflects both the impact of terrain on parking and eliminates safety hazards, laying the foundation for subsequent selection of suitable areas based on the score.

[0062] Combining the comprehensive requirements of vehicle parking stability and drone take-off and landing safety in uninhabited areas, a preset terrain safety threshold is established, typically set to ≥0.6. This threshold corresponds to terrain conditions with a slope ≤3° and undulation ≤0.5 meters in non-restricted areas, simultaneously meeting the core requirements of risk-free vehicle parking and controllable drone attitude. This threshold can be flexibly adjusted according to vehicle type and drone model. Subsequently, using this threshold as the criterion, a grid-by-grid passability scoring grid is generated, with grid values ​​ranging from 0 to 1; higher values ​​indicate better passability. Binarization segmentation is performed during the traversal process. The operation assigns a value of "1" to grid cells with a accessibility score ≥ 0.6, indicating that the area simultaneously meets the requirements of safety (i.e., not restricted) and terrain suitability (i.e., the slope and undulation meet the standards). Grid cells with a score < 0.6 are assigned a value of "0", indicating that the area or terrain is unsuitable or poses a safety hazard, and are therefore excluded. After segmentation, the output retains only the accessibility grid mask of the qualified areas. This mask presents the area within the buffer zone with true parking potential in a clear binary form, providing accurate boundaries for subsequent further screening based on area and shape.

[0063] Based on the dimensions of commonly used monitoring vehicles in uninhabited areas, such as small pickup trucks and SUVs, and their parking and operation requirements, a preset minimum parking area is set, typically no less than 20 square meters. This area must simultaneously accommodate the vehicle body, passenger access space, and temporary turnover space, and can be flexibly adjusted according to the vehicle type. Subsequently, for the output suitable passage area grid mask, only "1" values ​​represent suitable areas, and a connected component labeling algorithm is applied. By identifying consecutive adjacent "1" value grid blocks in the mask, i.e., connected components, a single connected component represents a continuous suitable area. Based on the actual geographical area of ​​the grid unit, such as a 30-meter resolution DEM corresponding to a single grid area of ​​900 square meters, the actual area occupied by each connected component needs to be calculated according to the data resolution. Finally, the area of ​​each connected component is compared with the preset minimum parking area, and connected components with an area greater than or equal to the minimum parking area are selected. These areas are potential parking areas that meet the scale requirements, i.e., scale-compliant areas. At the same time, small connected components with insufficient area that cannot meet the vehicle parking requirements are eliminated, providing a candidate range that conforms to the basic space scale for subsequent morphological optimization.

[0064] Based on the types of vehicles commonly used for monitoring in uninhabited areas, such as small pickup trucks and SUVs for work, a turning radius constraint is set for each vehicle, typically 5-8 meters. This constraint must match the vehicle's actual minimum turning radius to prevent vehicles from being unable to turn due to narrow areas. Subsequently, the selected areas that meet the size requirements are subjected to geometric morphological analysis. Morphological algorithms, such as dilation and erosion operations, are used to detect the integrity of the internal space of each area, determining whether there are narrow passages, sharp corners, or protruding obstacles, such as small areas of unidentified rocks. These structures can prevent vehicles from completing a full turn. For areas with morphological defects, local segments that do not meet the turning radius requirements are removed, such as trimming narrow sections or removing sharp protrusions, retaining only the open areas that can accommodate a full turn. Finally, all the morphologically optimized areas are integrated to form a set of candidate parking areas that meet both area requirements and operational feasibility. Each area can ensure safe parking and smooth turning for vehicles, providing practical target area data for subsequent accessibility analysis.

[0065] In one possible implementation, step S700 further includes:

[0066] Step S710: Perform spatial proximity analysis on the candidate parking area set and associated road vector data to locate multiple candidate parking anchor points.

[0067] Step S720: Based on the starting point of the construction task and the multiple candidate parking anchor points, perform road network graph modeling of the associated road vector data, and initially screen and locate multiple parking navigation routes corresponding to multiple accessible parking areas.

[0068] Step S730: Perform multi-factor weighted scoring on the multiple parking navigation routes and multiple accessible parking areas, and filter and output the target parking area and target navigation route.

[0069] Step S740: Send the target parking area and target navigation route to the construction task mobile terminal for offline path guidance, wherein the construction task starting point and construction task work point are both sent through the construction task mobile terminal.

[0070] Specifically, the associated road vector data is first restored into a road network topology map containing road segments, intersection nodes, and connection relationships. Then, the geometric range of each area in the candidate parking area set is projected onto this topology map. The Euclidean distance between each area and the nearest road is calculated using a spatial proximity algorithm. Accessible parking areas with a distance ≤ 50 meters that meet the short-distance vehicle connection requirements are selected. Subsequently, the center points of these accessible parking areas are used as candidate parking anchor points to clarify the core endpoint coordinates of subsequent route planning, forming multiple candidate anchor point datasets.

[0071] Next, using the starting point coordinates uploaded by the mobile terminal of the construction task as the starting point and multiple candidate parking anchor points as the ending points, a bidirectional topology extension is performed in the road network topology map. First, excessively narrow road sections, such as those with a width of less than 3 meters or damaged and impassable sections, are eliminated from the topology map to construct a simplified accessible road subnet. Then, the shortest path algorithm, such as Dijkstra's algorithm, is used to calculate the optimal travel path from the starting point to each candidate anchor point within the subnet, and multiple parking navigation routes corresponding to multiple accessible parking areas are obtained through initial screening, ensuring that the routes are all based on actual accessible road planning.

[0072] A multi-factor weighted scoring mechanism is implemented. For each parking navigation route, the comprehensive score is quantified by combining route complexity (such as the number of turns, road surface smoothness, road grade coefficient (e.g., main road ×1.0, rural road ×0.6), route length, and environmental risk factors (e.g., whether it is near a steep slope). For the corresponding accessible parking area, the comprehensive score is quantified by combining terrain suitability (e.g., whether the slope and undulation meet the standards), site size (e.g., whether the area is ≥20㎡), and drone take-off and landing safety index (e.g., whether it is unobstructed). The total score of each solution is obtained by combining the route score (50%) and the area score (50%). The solutions are then sorted in descending order of total score, and the solutions with the highest scores are selected as the target parking area and target navigation route.

[0073] Finally, the latitude and longitude coordinates and site attributes of the target parking area, such as area and slope, as well as the node coordinates and turning prompts of the target navigation route, are converted into an offline data format supported by the mobile device and sent to the mobile device for the construction task. After receiving the data, the mobile device can load the route based on the built-in offline map without relying on the network, and provide offline route guidance services such as voice broadcast, deviation reminders, and estimated arrival time through real-time GPS global positioning system positioning, ensuring the accuracy and stability of navigation in uninhabited areas throughout the entire process.

[0074] In one possible implementation, step S710 further includes:

[0075] Step S711: Restore the associated road vector data to obtain a road network topology map.

[0076] Step S712: Project multiple candidate parking areas from the candidate parking area set onto the road network topology map to perform nearest road distance queries and obtain multiple Euclidean proximity values.

[0077] Step S713: Compare the multiple Euclidean proximity values ​​using a preset reachability distance threshold to filter out the multiple reachable parking areas from the multiple candidate parking areas.

[0078] Step S714: Use the geometric center point of the plurality of accessible parking areas as the plurality of candidate parking anchor points.

[0079] Specifically, the system reads information such as road segment coordinates, intersection nodes, and inter-segment connections from the associated road vector data. It then uses a topology reconstruction algorithm to reconstruct a road network topology map that conforms to the actual road orientation and connectivity logic. In the map, nodes represent intersections, line segments represent passable road segments, and the length, width, and other attributes of each road segment are labeled to form the basic road network model for subsequent distance analysis.

[0080] The polygonal geometry of each candidate parking area in the candidate parking area set is accurately mapped to the road network topology map using a coordinate projection algorithm to ensure that the area location and the road network are aligned in the same spatial coordinate system. Then, for each projected candidate area, the nearest road distance query algorithm is started to calculate the straight-line distance from the area boundary to the nearest road segment in the topology map, i.e., the Euclidean proximity, and the Euclidean proximity value corresponding to each candidate area is recorded to quantify the spatial correlation between the area and the road.

[0081] Based on the actual needs of short-distance vehicle shuttles in uninhabited areas, and to avoid vehicles getting stuck due to long-distance travel on unpaved roads, a preset reachable distance threshold is set, usually ≤50 meters, which can be adjusted according to the complexity of the terrain. The Euclidean proximity of each candidate area is compared with this threshold, and candidate areas with Euclidean proximity ≤50 meters are selected. These areas are identified as reachable parking areas because they are close to the road and can be reached by short-distance driving. At the same time, areas that are too far from the road and are difficult to reach in practice are removed.

[0082] The center point of each accessible parking area is calculated by taking the geometric mean of the area boundary coordinates. The latitude and longitude of the center point are obtained by taking the geometric mean of the area boundary coordinates. These center points are used as the endpoint coordinates of the subsequent path planning, i.e., candidate parking anchor points. All anchor points together constitute multiple candidate parking anchor point datasets, which provide clear target coordinate support for the next step of building a navigation route from the starting point to the anchor point.

[0083] In one possible implementation, step S720 further includes:

[0084] Step S721: By using the starting point of the construction task and multiple candidate parking anchor points as start and end nodes, a bidirectional spatial extension is performed on the road network topology map to perform topology pruning and obtain a reachable road subnet.

[0085] Step S722: Filter the shortest travel path from the starting point of the construction task to the multiple candidate parking anchor points in the reachable road subnet, and generate the multiple parking navigation routes.

[0086] Specifically, using the starting point coordinates uploaded by the mobile terminal of the construction task as the starting node and the coordinates of multiple candidate parking anchor points as the ending nodes as the core, a bidirectional spatial extension is initiated in the restored road network topology map, extending from the starting node to all adjacent road segments, and simultaneously extending from each ending node to adjacent road segments in the opposite direction. During the process, a topology pruning operation is performed: isolated road segments in the road network that are not related to the starting-ending connection path are removed, such as branch roads far from the starting and ending nodes, road segments that do not meet the conditions for vehicle passage, such as narrow roads with a width of less than 3 meters, road segments marked as damaged, and road segments that do not form a connection with the starting / ending extension path during the extension process; finally, all road segments and nodes that are connected to each ending point and meet the passage requirements are retained, forming an accessible road subnet focusing on core passage needs, which greatly reduces the complexity of subsequent path calculations.

[0087] Within the reachable road subnet, the starting point of the construction task is used as a unified starting point, and each candidate parking anchor point is used as an independent endpoint. The shortest path algorithm, such as Dijkstra's algorithm, is called to calculate the path while taking into account both path length and traffic efficiency. The algorithm traverses the connected road segments within the subnet, calculates all possible routes from the starting point to each endpoint, and selects the path with the shortest total length and all road segments that meet the vehicle traffic conditions. Each endpoint corresponds to an optimal path. These paths together constitute multiple parking navigation routes that correspond one-to-one with multiple candidate parking anchor points. Each route contains specific node coordinates, such as turning points, road segment connection points, and road segment attributes, such as road surface type and length, providing detailed route data for subsequent scoring, filtering, and mobile navigation.

[0088] In one possible implementation, step S730 further includes:

[0089] Step S731: Retrieve the first traffic information of the first parking navigation route, wherein the first traffic information includes route traffic complexity, road grade coefficient, route length and environmental risk factor.

[0090] Step S732: Retrieve the first area attribute information of the first accessible parking area, wherein the first area attribute information includes terrain suitability, site size and drone take-off and landing safety index.

[0091] Step S733: Quantitatively evaluate the attribute information of the first area and the road condition information to obtain the comprehensive route score and the comprehensive area score.

[0092] Step S734: Weight and fuse the path comprehensive score and the region comprehensive score to output the first fused score.

[0093] Step S735: By analogy, perform multi-factor weighted scoring on the multiple parking navigation routes and multiple accessible parking areas to obtain multiple fusion scores.

[0094] Step S736: Based on the descending order of the multiple fusion scores, filter and output the target parking area and the target navigation route.

[0095] Specifically, for the first parking navigation route to be evaluated—that is, the planned route from the starting point of the construction task to a candidate parking anchor point—the first road condition information corresponding to the route is accurately retrieved from the pre-set road network attribute database and environmental risk database. The route complexity is generated by statistically analyzing parameters such as the number of sharp bends, the length of continuous gradient changes, and road surface smoothness fluctuations, used to quantify the difficulty of driving operations. The road grade coefficient is set according to the road paving type, such as asphalt road, gravel road, unpaved road, and design capacity, such as single lane / double lane, with a higher coefficient indicating better road conditions. The route length is the actual mileage from the start to the end of the route, usually in kilometers, directly reflecting travel time and fuel consumption. The environmental risk factor combines real-time or historical environmental data to indicate whether there are temporary or potential risks such as rockfall risk areas, seasonal water accumulation points, and soft roadbed sections along the route, using a coefficient value of 0 to 1 to quantify the risk level, with values ​​closer to 1 indicating higher risk. By integrating these four types of information, comprehensive data support is formed for the feasibility of the first parking navigation route, laying the foundation for subsequent quantitative scoring.

[0096] For the first accessible parking area matching the first parking navigation route, and for potential parking areas already filtered by proximity, the corresponding first area attribute information is precisely retrieved from the previous terrain analysis database and regional feature database. Among these, terrain suitability directly uses the quantitative value of the area from the generated elevation gradient distribution map, ranging from 0 to 1. The closer the value is to 1, the more suitable the area's slope and undulation are for stable vehicle parking and a flat drone take-off and landing platform. Site size refers to the calculated actual area of ​​the region, typically in square meters, ensuring sufficient space for complete vehicle parking, personnel access, and drone preparation and operation. The drone take-off and landing safety index comprehensively considers factors such as the distribution of surrounding obstacles (e.g., tree height, distance to utility poles), ground hardness (e.g., presence of soft soil or gravel pits), and airspace conditions (e.g., whether it is in a low-altitude flight path interference zone). A standardized algorithm generates an index value of 0 to 1, with a value closer to 1 indicating lower safety risks for drone take-off, landing, and hovering. By integrating these three types of core attribute information, the system comprehensively reflects the adaptability of the first accessible parking area to vehicle parking and drone operations, providing regional data support for subsequent quantitative evaluation.

[0097] Standardized evaluation rules and weights were set for each factor of the first road condition information and the first area attribute information. For the first road condition information, "Route Traffic Complexity" was scored from 0 to 100 based on parameters such as sharp bends and gradient changes, with lower complexity resulting in higher scores, and a weight of 30%. "Road Grade Coefficient" was scored from 0 to 100 based on paving quality and traffic capacity, with higher grades resulting in higher scores, and a weight of 30%. "Route Length" was scored from 0 to 100 based on actual mileage, with shorter lengths resulting in higher scores, and a weight of 20%. "Environmental Risk Factor" was scored from 0 to 100 based on risk level, with lower risks resulting in higher scores, and a weight of 20%. The comprehensive route score for the first parking navigation route is calculated using weighted summation, with a maximum score of 100. For the first area attribute information, "terrain suitability" is converted from a raw value of 0 to 1 to a score of 0 to 100, with values ​​closer to 1 receiving higher scores (40% weight); "site size" is calculated based on the actual area, corresponding to a score of 0 to 100, with adequacy and sufficient size resulting in higher scores (30% weight); and "drone take-off and landing safety index" is converted from a raw value of 0 to 1 to a score of 0 to 100, with values ​​closer to 1 receiving higher scores (30% weight). Similarly, the comprehensive area score for the first accessible parking area is calculated using weighted summation, with a maximum score of 100. Ultimately, this quantification process transforms the previously fragmented road condition and area information into intuitive and comparable scoring results, providing a unified standard for subsequent integrated evaluation of the two.

[0098] Based on the core requirements of safe vehicle passage and efficient drone operation in uninhabited area construction tasks, a fusion weight of the route comprehensive score and the area comprehensive score is preset, usually set at 1:1. If the task focuses more on drone operation, the weight of the area score can be appropriately increased; if the focus is on passage efficiency, the weight of the route score can be adjusted, providing flexibility and adaptability. Then, according to the calculation rule of first fusion score = route comprehensive score × route weight + area comprehensive score × area weight, the route comprehensive score of the first parking navigation route and the area comprehensive score of the corresponding first reachable parking area are weighted and summed to obtain the overall evaluation score of the route-area combination scheme, i.e., the first fusion score. This score integrates the originally independent route and area evaluation dimensions into a unified and comparable quantitative result, intuitively reflecting the comprehensive performance of the combination scheme in both passage safety and operational adaptability, providing a core basis for the subsequent batch sorting of schemes and selection of the optimal scheme.

[0099] Following the evaluation logic of steps S731-S734, all remaining parking navigation routes and their matching accessible parking area combinations are processed one by one. First, the road condition information and area attribute information of each combination are retrieved. Then, each factor is quantified through standardized rules and the corresponding path comprehensive score and area comprehensive score are calculated. Finally, the combined scores of each combination are obtained by weighting and fusing according to preset weights. In the end, multiple combined scores covering all candidate combinations are formed, realizing a unified quantitative evaluation of all schemes.

[0100] The obtained multiple fusion scores are sorted in descending order of value. The higher the fusion score, the better the overall performance of the corresponding combination scheme in terms of route feasibility and regional operation adaptability. Then, the reachable parking area and parking navigation route corresponding to the fusion score ranked first in the ranking results are extracted and determined as the target parking area and target navigation route, respectively. This ensures that the output scheme can not only meet the needs of safe and efficient vehicle passage, but also provide a suitable, safe and stable operating site for UAV inspection operations. Finally, the optimal execution scheme adapted to UAV inspection tasks in uninhabited areas is formed.

[0101] Example 2, based on the same inventive concept as the satellite remote sensing-based vehicle parking navigation method in the aforementioned examples, such as... Figure 2 As shown, this application provides a vehicle parking navigation system for uninhabited areas based on satellite remote sensing. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0102] The parking buffer acquisition module 10 is used to receive the construction task operation point and then perform multi-level buffer space modeling with the construction task operation point as the center to obtain the parking buffer.

[0103] The image data acquisition module 20 is used to simultaneously acquire DEM raster data and multispectral remote sensing image data, constrained by the geographical boundaries of the parking buffer zone.

[0104] The elevation gradient distribution map construction module 30 is used to perform terrain gradient quantization analysis on the DEM raster data and construct an elevation gradient distribution map.

[0105] The restricted area mask generation module 40 is used to perform ground object spectral feature interpretation on the multispectral remote sensing image data to generate a restricted area mask.

[0106] The parking area set acquisition module 50 is used to extract the passable area by spatially overlaying analysis and fusing the elevation gradient distribution map and the restricted area mask, and then extracting the passable area with the minimum turning space constraint of the vehicle to obtain a candidate parking area set.

[0107] The vector data extraction module 60 is used to extract the associated road vector data of the parking buffer based on the spatial data service interface.

[0108] The offline path guidance module 70 performs topology accessibility modeling based on the candidate parking area set and associated road vector data, filters and outputs the target parking area and target navigation route, and provides offline path guidance.

[0109] Furthermore, the system is also used to implement the following functions:

[0110] Spatial proximity analysis is performed on the candidate parking area set and associated road vector data to locate multiple candidate parking anchor points. Based on the construction task starting point and the multiple candidate parking anchor points, a road network graph model of the associated road vector data is constructed to initially screen and locate multiple parking navigation routes corresponding to multiple reachable parking areas. Multi-factor weighted scoring is performed on the multiple parking navigation routes and multiple reachable parking areas to filter and output target parking areas and target navigation routes. The target parking area and target navigation routes are sent to the construction task mobile terminal for offline path guidance, wherein the construction task starting point and construction task work point are both sent through the construction task mobile terminal.

[0111] Furthermore, the system is also used to implement the following functions:

[0112] A first-level buffer zone with a preset first-level radius is generated centered on the construction task operation point; if parking area identification is performed with the first-level buffer zone as the core search area and the obtained candidate parking areas are empty sets, then the concentric circle radius is expanded based on the first-level buffer zone until the parking candidate points are obtained as the parking buffer zone with a non-empty set.

[0113] Furthermore, the system is also used to implement the following functions:

[0114] Surface slope is calculated based on the DEM raster data, and a slope raster map is output. The slope raster map is then traversed using a preset slope threshold for binarization segmentation, generating a raster mask for suitable slope areas. Neighborhood elevation range analysis is performed on the DEM raster data, and a terrain relief raster map is output. The terrain relief raster map is then binarized using a preset relief threshold, generating a raster mask for suitable relief areas. After spatial coordinate system registration of the suitable slope area raster mask and the suitable relief area raster mask, weighted raster algebra operations are performed to generate the elevation gradient distribution map.

[0115] Furthermore, the system is also used to implement the following functions:

[0116] The reflectance difference of the multispectral remote sensing image data is identified by predefined water body associated bands, and the water body area mask is segmented and output. Neighborhood elevation change analysis is performed based on the DEM raster data to locate the dangerous terrain area mask. After merging the water body area mask and the dangerous terrain area mask, a buffer expansion with a preset safety distance is performed to generate the restricted area mask.

[0117] Furthermore, the system is also used to implement the following functions:

[0118] Raster algebra operations are performed on the elevation gradient distribution map and the restricted area mask to generate a accessibility score raster. A preset terrain safety threshold is used to traverse the accessibility score raster and perform binarization segmentation to output a suitable access area raster mask. With a preset minimum parking area as a spatial constraint, the suitable access area raster mask is labeled with connected components to filter out areas that meet the size standard. The geometric shape of the areas that meet the size standard is optimized using a vehicle turning radius constraint to generate the candidate parking area set.

[0119] Furthermore, the system is also used to implement the following functions:

[0120] The associated road vector data is restored to obtain a road network topology map; multiple candidate parking areas in the candidate parking area set are projected onto the road network topology map to perform nearest neighbor distance queries to obtain multiple Euclidean proximity values; the multiple Euclidean proximity values ​​are compared with a preset reachable distance threshold to filter out the multiple reachable parking areas from the multiple candidate parking areas; the geometric center point of the multiple reachable parking areas is used as the multiple candidate parking anchor points.

[0121] Furthermore, the system is also used to implement the following functions:

[0122] By using the starting point of the construction task and multiple candidate parking anchor points as start and end nodes, a bidirectional spatial extension is performed on the road network topology map to perform topology pruning, resulting in a reachable road subnet; the shortest travel path from the starting point of the construction task to the multiple candidate parking anchor points is filtered in the reachable road subnet to generate the multiple parking navigation routes.

[0123] Furthermore, the system is also used to implement the following functions:

[0124] The system retrieves first road condition information for the first parking navigation route, including route complexity, road grade coefficient, route length, and environmental risk factors. It also retrieves first regional attribute information for the first accessible parking area, including terrain suitability, site size, and UAV take-off and landing safety index. The system quantitatively evaluates the first regional attribute information and the first road condition information to obtain a comprehensive route score and a comprehensive regional score. It then weights and fuses the comprehensive route score and the comprehensive regional score to output a first fused score. Similarly, it performs multi-factor weighted scoring on multiple parking navigation routes and multiple accessible parking areas to obtain multiple fused scores. Based on the descending order of the multiple fused scores, it filters and outputs the target parking area and the target navigation route.

[0125] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0126] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0127] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A vehicle parking navigation method in uninhabited areas based on satellite remote sensing, characterized in that, The method includes: After receiving the construction task work point, a multi-level buffer space model is performed with the construction task work point as the center to obtain the parking buffer. Using the geographical boundaries of the parking buffer zone as constraints, DEM raster data and multispectral remote sensing image data are acquired simultaneously. Perform terrain gradient quantization analysis on the DEM raster data to construct an elevation gradient distribution map; Perform ground feature interpretation on the multispectral remote sensing image data to generate a no-entry area mask; After fusing the elevation gradient distribution map and the restricted area mask through spatial overlay analysis, the passable area is extracted using the minimum turning space constraint of the vehicle to obtain a set of candidate parking areas. Extract the associated road vector data of the parking buffer zone based on the spatial data service interface; Based on the candidate parking area set and associated road vector data, topology accessibility modeling is performed to filter and output the target parking area and target navigation route for offline path guidance; The method includes performing topology reachability modeling based on the candidate parking area set and associated road vector data, filtering and outputting target parking areas and target navigation routes, and providing offline path guidance. Spatial proximity analysis is performed on the candidate parking area set and associated road vector data to locate multiple candidate parking anchor points; Based on the starting point of the construction task and the multiple candidate parking anchor points, a road network diagram model of the associated road vector data is constructed to initially screen and locate multiple parking navigation routes corresponding to multiple accessible parking areas; The multiple parking navigation routes and multiple accessible parking areas are scored using a multi-factor weighted method, and the target parking area and target navigation route are then selected and output. The target parking area and target navigation route are sent to the construction task mobile terminal for offline path guidance, wherein the construction task starting point and construction task work point are both sent through the construction task mobile terminal; After fusing the elevation gradient distribution map and the restricted area mask through spatial overlay analysis, the passable area is extracted using the minimum turning space constraint for vehicles to obtain a set of candidate parking areas. The method includes: Perform raster algebra operations on the elevation gradient distribution map and the restricted area mask to generate a passability score raster; The preset terrain safety threshold is used to traverse the accessibility scoring grid and perform binarization segmentation to output a suitable access area grid mask; Using a preset minimum parking area as a spatial constraint, the grid mask of the suitable passage area is marked with connected components to filter out areas that meet the size requirements. The geometric shape of the qualified area is optimized by using vehicle turning radius constraints to generate the candidate parking area set.

2. The vehicle parking navigation method in uninhabited areas based on satellite remote sensing as described in claim 1, characterized in that, After receiving the construction task work point, a multi-level buffer space model is performed with the construction task work point as the center to obtain the parking buffer. The method includes: A first-level buffer zone with a preset first-level radius is generated centered on the construction task operation point; If parking area identification is performed with the first-level buffer as the core search area and the obtained candidate parking areas are empty sets, then the concentric circle radius is expanded based on the first-level buffer until the parking buffer where the parking candidate points are non-empty sets is obtained.

3. The vehicle parking navigation method in uninhabited areas based on satellite remote sensing as described in claim 1, characterized in that, Perform terrain gradient quantization analysis on the DEM raster data to construct an elevation gradient distribution map. The method includes: Surface slope is calculated based on the DEM raster data, and a slope raster map is output. The slope raster image is traversed using a preset slope threshold to perform binarization segmentation, generating a raster mask for the appropriate slope area; Perform neighborhood elevation range analysis on the DEM raster data and output a terrain relief raster map; The terrain relief raster map is binarized using a preset relief threshold to generate a raster mask for a suitable relief area. After registering the grid masks for the slope-suitable area and the undulation-suitable area with spatial coordinate systems, weighted grid algebra operations are performed to generate the elevation gradient distribution map.

4. The vehicle parking navigation method in uninhabited areas based on satellite remote sensing as described in claim 3, characterized in that, The method involves interpreting the spectral features of ground features in the multispectral remote sensing image data to generate a no-entry area mask, the method comprising: Predefined water body associated bands are used to identify reflectance differences in the multispectral remote sensing image data, and water body region masks are segmented and output. Based on the DEM raster data, perform neighborhood elevation change analysis to locate dangerous terrain mask areas; After merging the water area mask and the dangerous terrain area mask, a buffer expansion with a preset safety distance is performed to generate the restricted area mask.

5. The vehicle parking navigation method in uninhabited areas based on satellite remote sensing as described in claim 1, characterized in that, Based on the starting point of the construction task and the multiple candidate parking anchor points, a road network graph model of the associated road vector data is constructed to initially screen and locate multiple parking navigation routes corresponding to multiple reachable parking areas. The method includes: By using the starting point of the construction task and multiple candidate parking anchor points as start and end nodes, a bidirectional spatial extension is performed on the road network topology map to perform topology pruning and obtain a reachable road subnet. The road network topology map is obtained by restoring the associated road vector data. In the reachable road subnet, the shortest travel path from the starting point of the construction task to the multiple candidate parking anchor points is selected to generate the multiple parking navigation routes.

6. The vehicle parking navigation method in uninhabited areas based on satellite remote sensing as described in claim 1, characterized in that, The method involves performing multi-factor weighted scoring on the multiple parking navigation routes and multiple accessible parking areas, and then filtering and outputting the target parking area and target navigation route. Retrieve the first traffic information of the first parking navigation route, wherein the first traffic information includes route traffic complexity, road grade coefficient, route length and environmental risk factor; Retrieve the first area attribute information of the first accessible parking area, wherein the first area attribute information includes terrain suitability, site size and drone take-off and landing safety index; The first area attribute information and the first road condition information are quantitatively evaluated to obtain the comprehensive route score and the comprehensive area score. The path comprehensive score and the region comprehensive score are weighted and merged to output the first fused score; By analogy, multiple factors are weighted and scored for the multiple parking navigation routes and multiple accessible parking areas to obtain multiple fusion scores; Based on the descending order of the multiple fusion scores, the target parking area and target navigation route are filtered and output.

7. A vehicle parking navigation system for uninhabited areas based on satellite remote sensing, characterized in that, The system is used to implement the satellite remote sensing-based vehicle parking navigation method in uninhabited areas as described in any one of claims 1-6, and the system comprises: The parking buffer zone acquisition module is used to receive the construction task operation point and then perform multi-level buffer space modeling with the construction task operation point as the center to obtain the parking buffer zone. The image data acquisition module is used to simultaneously acquire DEM raster data and multispectral remote sensing image data, constrained by the geographical boundaries of the parking buffer zone. The elevation gradient distribution map construction module is used to perform terrain gradient quantization analysis on the DEM raster data and construct an elevation gradient distribution map. The restricted area mask generation module is used to perform ground spectral feature interpretation on the multispectral remote sensing image data and generate a restricted area mask. The parking area set acquisition module is used to extract the passable area by spatial overlay analysis and fusion of the elevation gradient distribution map and the restricted area mask, and then extract the candidate parking area set by the minimum turning space constraint of the vehicle. The vector data extraction module is used to extract the associated road vector data of the parking buffer based on the spatial data service interface; The offline route guidance module is used to perform topology accessibility modeling based on the candidate parking area set and associated road vector data, filter and output the target parking area and target navigation route, and provide offline route guidance.

Images