Method for constructing multi-source remote sensing passable map for off-road navigation of unmanned ground vehicle

Abstract

This invention discloses a method for constructing a multi-source remote sensing access map for off-road navigation of unmanned ground vehicles. The method acquires remote sensing images, digital elevation models, soil attribute data, and prior road data of the target off-road area; establishes a hierarchical hexagonal spatial index grid and maps the multi-source data to the grid; constructs a geometric measurement layer based on the mapped data; constructs a semantic expression layer based on the geometric measurement layer and remote sensing image interpretation results, recording the access semantic category and confidence level of each grid unit; obtains the final access semantic through confidence level-based resolution; calculates the comprehensive access cost and constructs a decision cost layer; and outputs a access map containing the geometric measurement layer, semantic expression layer, and decision cost layer. This invention unifies multi-source heterogeneous environmental information into access costs that can be directly invoked for path planning, improving the interpretability and planning applicability of off-road navigation maps.

Description

Technical Field

[0001] This invention relates to the fields of unmanned ground vehicle navigation map construction and remote sensing accessibility analysis, and in particular to a method for constructing multi-source remote sensing accessibility maps for off-road navigation of unmanned ground vehicles. Background Technology

[0002] In missions such as disaster relief, field inspections, border patrols, agricultural operations, and mining exploration, unmanned ground vehicles need to navigate autonomously in unstructured environments such as forests, farmland, mountains, riverbanks, and post-disaster ruins. These environments typically lack continuous, stable, and regular road networks, and a vehicle's safe passage depends not only on the presence of obstacles but also on a variety of factors, including slope, terrain roughness, surface cover, soil moisture, water distribution, and road damage.

[0003] Currently, conventional occupancy raster maps primarily represent spatial occupancy or vacancy, making it difficult to further explain factors affecting vehicle passability such as steep slopes, wet and soft soil, rough gravel, and soft surfaces. Ordinary remote sensing feature classification maps can provide surface categories, but these categories are not directly equivalent to the traffic status of unmanned ground vehicles and cannot be directly used as a cost for path planning. High-precision urban maps are mainly geared towards structured road traffic, representing lane lines, road boundaries, and road topology, and are also difficult to apply to off-road navigation.

[0004] Satellite imagery, digital elevation models (DEMs), soil thematic data, and open-source road vector data can provide a wide range of environmental prior information. Satellite imagery can be used to identify land cover, water bodies, and road signs; DEMs can derive slope and terrain roughness; soil properties can provide prior information on subsidence and slippage risks; and open road data can provide auxiliary constraints for candidate travel corridors. However, different data sources vary in spatial resolution, acquisition time, semantic granularity, positioning accuracy, and reliability. Conflicts may arise between road priors, image classification results, terrain constraints, and soil risks within the same area.

[0005] In summary, the existing technology has at least the following technical problems: I. Existing map representation methods are insufficient to uniformly convert multi-source information such as terrain, soil, water bodies, and road priors into a passage cost that can be directly accessed by unmanned ground vehicles; Second, there are scale, temporal and semantic conflicts among multi-source remote sensing data, and simple overlay or weighted fusion methods are difficult to meet the safety priority requirements in off-road navigation. Third, the lack of a clear mapping relationship between ordinary remote sensing classification results and vehicle traffic capacity and route planning costs leads to insufficient map interpretability and planning applicability. Summary of the Invention

[0006] The purpose of this invention is to provide a method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles. By constructing a three-layer progressive access map structure of "geometric measurement layer - semantic expression layer - decision cost layer", multi-source remote sensing observations, terrain constraints, soil conditions and road priors are uniformly expressed as a comprehensive access cost that can be directly called upon for path planning. This solves the problem in existing technologies that it is difficult to uniformly express multi-source remote sensing observations, terrain constraints, soil conditions and road priors as a usable access cost for path planning.

[0007] According to one aspect of the present invention, a method for constructing a multi-source remote sensing access map for off-road navigation of unmanned ground vehicles is provided, comprising: acquiring multi-source environmental data of a target off-road area, including remote sensing imagery, digital elevation model, soil property data, and road prior data; establishing a hierarchical hexagonal spatial index grid covering the target off-road area, and mapping the multi-source environmental data to the hierarchical hexagonal spatial index grid; constructing a geometric measurement layer based on the mapped multi-source environmental data, the geometric measurement layer including at least one grid attribute among elevation, slope, terrain roughness, humidity proxy index, soil clay content, and road prior data; and constructing a geometric measurement layer based on the geometric measurement layer and remote sensing imagery. A semantic representation layer is constructed from the image interpretation results. This layer records the traffic semantic category and corresponding confidence level of each grid cell. Based on the geometric measurement layer and the semantic representation layer, semantic conflicts from different data sources in the same grid cell are resolved by confidence level classification to obtain the final traffic semantics. The comprehensive traffic cost of each grid cell is calculated based on the final traffic semantics and the geometric measurement attributes in the geometric measurement layer, and a decision cost layer is constructed. This decision cost layer records the comprehensive traffic cost of each grid cell. A multi-source remote sensing traffic map containing the geometric measurement layer, the semantic representation layer, and the decision cost layer is output for off-road navigation path planning of unmanned ground vehicles.

[0008] As a further technical solution, the hierarchical hexagonal spatial index grid is an H3-level grid. The H3-level grid is used to store grid identifiers, parent-child hierarchical relationships, neighborhood relationships, and grid center locations. The multi-source remote sensing access map uses neighborhood relationships as the spatial topological basis for path search.

[0009] As a further technical solution, the traffic semantic categories include: C0 roads or hard-surface traffic areas, C1 bare land or low vegetation areas, C2 forests or tall vegetation areas, C3 buildings or obstacle areas, and C4 water bodies or waterlogged areas; the semantic expression layer records candidate semantics, final semantics, semantic distribution, source identifiers, and confidence levels for each grid cell.

[0010] As a further technical solution, the confidence level resolution includes: when the slope derived from the digital elevation model exceeds the upper limit of the safe slope or the humidity proxy index indicates the risk of water bodies or water accumulation, a high-risk semantic is adopted; when there is a conflict between the road prior and the interpretation results of remote sensing images and there are no hard constraints of terrain or water bodies, the final traffic semantic is determined based on the road prior confidence, image classification confidence and neighborhood consistency.

[0011] As a further technical solution, the comprehensive passage cost Calculate according to the following formula: ,in, Indicates final pass semantics The corresponding fundamental impedance, Indicates slope The corresponding slope penalty factor, This indicates a soil wet and soft risk correction factor. This represents the surface roughness correction factor. Indicates normalized clay content. Indicates the humidity proxy index. This represents the terrain roughness index.

[0012] As a further technical solution, the slope penalty factor Determined according to the following piecewise function: hour, =1.0; in hour, = ;exist hour, = ;in Indicates the threshold for a gentle slope. Indicates the upper limit of the safe slope. This represents the slope penalty coefficient. This indicates the cost of making the route impassable.

[0013] As a further technical solution, the soil wet and soft risk correction factor is based on Calculation, where This indicates the risk coefficient for soft, wet soil.

[0014] As a further technical solution, the surface roughness correction factor is based on Calculation, where This represents the roughness correction factor.

[0015] According to one aspect of the present invention, a multi-source remote sensing access map construction device for off-road navigation of unmanned ground vehicles is provided, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the multi-source remote sensing access map construction method for off-road navigation of unmanned ground vehicles.

[0016] According to one aspect of the present invention, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps of the multi-source remote sensing access map construction method for off-road navigation of unmanned ground vehicles.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: I. This invention organizes multi-source environmental data through a geometric measurement layer, a semantic expression layer, and a decision cost layer, enabling remote sensing observations, terrain constraints, soil conditions, and road priors to have traceable map representations. Second, this invention improves the reliability of passage judgment in multi-source conflict scenarios by arbitrating physical hard constraints, road priors, and remote sensing observations according to the principle of safety priority through a confidence-level conflict resolution mechanism. Third, this invention calculates the comprehensive passage cost by combining basic impedance, slope penalty, soil wet and soft risk correction and surface roughness correction, so that the output map can directly serve the path planning of unmanned ground vehicles. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles provided in an embodiment of the present invention. Detailed Implementation

[0020] It should be noted that: The terms “comprising” and “having”, and any variations thereof, in the specification, claims, and drawings of this invention are intended to cover a non-exclusive inclusion, such as a process, method, system, product, or apparatus that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined to form new technical solutions. Such combinations are not bound by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0022] like Figure 1 As shown, this embodiment of the invention provides a method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles, including the following steps: S1. Acquire and preprocess multi-source environmental data of the target area. This step involves acquiring multi-source environmental data of the target area and performing spatial registration and preprocessing. The multi-source environmental data includes at least remote sensing imagery, digital elevation models, soil property data, and prior road data. Specific steps are as follows: S1.1 Read the boundary of the mission area of ​​the unmanned ground vehicle, and acquire remote sensing images, digital elevation models, soil property data and road prior data covering the target area based on the boundary of the mission area. Among them, remote sensing images are used to provide land cover, water bodies and road signs, digital elevation models are used to derive slope and terrain roughness, soil property data are used to provide prior risk of wet and soft subsidence, and road prior data are used to provide candidate passage corridors. S1.2 Perform spatial registration, coordinate unification, cropping, resampling, and outlier checking on the multi-source environmental data to ensure that different data sources are under the same spatial reference. S1.3 Extract candidate semantics of land cover, water index and vegetation index from remote sensing images, calculate elevation, slope and terrain roughness from digital elevation model, extract clay content and moisture surrogate from soil property data, and generate road buffer and road continuity identifiers from road prior data. S1.4 Record the source identifier, acquisition time, spatial resolution, and data quality identifier for each data source to form the source information required for subsequent conflict resolution.

[0023] S2. Map the multi-source environmental data to a hierarchical hexagonal grid. This step establishes a hierarchical hexagonal spatial index grid covering the target off-road area and maps the multi-source environmental data to the hierarchical hexagonal spatial index grid.

[0024] Specifically, a hierarchical hexagonal grid is generated based on the target area, and the grid code is used as a unique index to write continuous environmental attributes and discrete semantic attributes into the corresponding grid. For slope, roughness, humidity proxy, and soil clay content, mean, maximum, or quantile statistics are used; for surface semantics and road priors, area proportion or pixel voting method is used to form semantic distribution.

[0025] Specifically, the hierarchical hexagonal spatial index grid is an H3-level grid. The H3-level grid is used to uniformly organize satellite remote sensing images, digital elevation models, soil attribute data, and road prior data, and is used to store grid identifiers, parent-child hierarchical relationships, neighborhood relationships, and grid center locations.

[0026] S3. Construct a geometric measurement layer, a semantic expression layer, and a decision cost layer.

[0027] The geometric measurement layer records at least the grid code, grid level, elevation, slope, terrain roughness, humidity proxy index, soil clay content, and road prior identifier; the semantic expression layer records candidate semantics, final semantics, semantic distribution, source identifier, and confidence level; the decision cost layer records the basic impedance, slope penalty, soil correction, roughness correction, and comprehensive traffic cost.

[0028] In the geometric measurement layer, the slope is calculated by the digital elevation model, the terrain roughness is calculated by the elevation difference of the grid neighborhood, the humidity proxy index is calculated by the spectral information of the remote sensing image, the soil property data includes clay content, and the road prior data is obtained by rasterizing or spatially intersecting the open road vector data.

[0029] In the semantic expression layer, the source identifier is used to indicate that the corresponding semantics originate from satellite remote sensing images, digital elevation models, soil attribute data, or road prior data.

[0030] S4. Group environmental observations from different sources into C0-C4 common semantics.

[0031] Wherein, C0 represents roads or hard-surfaced access areas, C1 represents bare ground or low-vegetation areas, C2 represents forest or tall-vegetation areas, C3 represents building or obstacle areas, and C4 represents water bodies or waterlogged areas. For grids with a single semantic meaning, the dominant access semantic is recorded; for mixed grids such as road edges, shorelines, and forest boundaries, the semantic distribution vectors of C0-C4 are recorded. As shown in Table 1.

[0032] Table 1. Traffic Semantics and Planning Implications of C0-C4 .

[0033] S5. The final common semantics are obtained based on the confidence-based hierarchical conflict resolution mechanism.

[0034] The specific steps are as follows: S5.1 Determine whether there are physical hard constraints such as excessive slope, water cover, or obvious obstacles in the same grid. If they exist (i.e., the slope derived from the digital elevation model exceeds the upper limit of the safe slope or the humidity proxy index indicates the risk of water or water accumulation), then prioritize outputting high-risk or impassable semantics. S5.2 Determine whether the road prior is consistent with the remote sensing image observation. If the road prior is consistent with the observation of low slope, low water index and low vegetation risk, then increase the semantic confidence of road traffic. S5.3 When there is a conflict between the road prior and the interpretation results of remote sensing images and there are no hard constraints of terrain or water bodies, the final traffic semantics shall be determined based on the road prior confidence, image classification confidence and neighborhood consistency. S5.4 Output the final pass semantics, final confidence level, conflict type, accepted information source, and conflict resolution reason for subsequent map status tracing.

[0035] Furthermore, the confidence level conflict resolution mechanism includes several typical conflict handling scenarios. As shown in Table 2, firstly, when there is a conflict between road priors and water body observations, if the water body index is high and the grid is located in low-lying terrain or near the boundary of a pond, the water body risk is given priority, and the grid is classified as C4 or a high-risk waterlogging state. Secondly, when there is a conflict between road priors and vegetation observations, if the road vector is continuous and the slope and roughness are low, it can be classified as a forest road; if the road vector is isolated and the surrounding vegetation is continuous, it is classified as a C2 forest barrier. Thirdly, when there is a conflict between bare land semantics and slope constraints, if the slope exceeds the vehicle safety threshold, the hard slope constraint is given priority.

[0036] Table 2 Typical Multi-Source Conflicts and Resolution Rules .

[0037] Furthermore, when remote sensing imagery is obscured by cloud shadows, digital elevation models contain holes, road vectors are offset, or soil attribute data has low resolution, this method does not directly delete the data source. Instead, it reduces the corresponding source weight and records the data quality label. For conflicting grids that cannot be resolved, they are marked as grids to be checked, so that subsequent UAV supplementary observations or manual verification can be performed.

[0038] In this embodiment, if the final confidence level of a grid is lower than a preset threshold, the grid is marked as a grid to be verified. Grids to be verified can serve as candidate areas for supplementary observation tasks by UAVs, or they can temporarily increase the conservative cost in path planning.

[0039] S6. Calculate the comprehensive passage cost based on the final passage semantics and geometric measurement attributes. This step is based on the final passage semantics. The comprehensive passage cost is calculated by including corrections for slope, soil softness risk, and terrain roughness. .

[0040] The overall passage cost is calculated based on the product of base impedance, slope penalty, soil softness risk correction, and surface roughness correction. Base impedance is determined by the final passage semantics from C0 to C4; slope penalty is determined segmented according to the vehicle safe slope threshold; soil softness risk correction is determined based on normalized clay content and moisture proxy index; and surface roughness correction is determined based on the terrain roughness index.

[0041] The formula for calculating the comprehensive passage cost is as follows: , in, Indicates grid position The overall cost of passage; This indicates the final, acceptable semantics obtained after conflict resolution. This represents the underlying impedance corresponding to the final passage semantics; Indicates slope The corresponding slope penalty factor; Indicated by normalized clay content Humidity proxy index Commonly determined soil softening risk correction factors; Indicated by the terrain roughness index The determined surface roughness correction factors are shown in Table 3.

[0042] Table 3. Explanation of Variables in the Comprehensive Traffic Cost Formula .

[0043] Gradient is one of the primary factors affecting vehicle passability in off-road terrain. As the gradient... Increasing the slope will increase vehicle traction requirements, braking risks, and rollover risks; Exceeding the vehicle's safe gradient limit When this occurs, the area should be considered impassable. To balance continuous cost representation and hard safety constraints, this embodiment uses a piecewise function to construct the slope penalty factor. ,in Used to control the intensity of cost increase in medium-to-high slope areas. Refer to Table 4 for slope penalty factors. The calculation formula is: .

[0044] Considering the difficulty in consistently obtaining measured soil moisture content across large-scale scenarios, this embodiment employs a combination of dynamic spectral response and static soil properties to describe soil softening risk. Specifically, image spectral information is used to characterize short-term water bodies or wetting responses, while clay content from the SoilGrids database is used to represent relatively stable soil texture properties; a moisture proxy index is constructed using NDWI. and normalized clay content Together, they form a soil wet and soft risk correction factor, among which Used to control the amplification effect of soil softening risk on the total cost. Soil softening risk correction factor. The calculation formula is: .

[0045] Surface roughness is used to characterize the complexity of local undulations within a grid. Areas with concentrated gravel, ditch edges, and irregular potholes may not constitute absolute obstacles, but they can increase chassis impact risk, attitude fluctuations, and control difficulty. This embodiment constructs a surface roughness correction factor based on the Terrain Roughness Index (TRI). ,in Used to control the intensity of correction for the overall passage cost caused by rough surfaces. Surface roughness correction factor. The calculation formula is: .

[0046] Table 4. Parameters of the toll cost model and their determination basis .

[0047] In a preferred embodiment, the C0 hard passage area corresponds to the lowest basic impedance, the C1 soft passage area corresponds to a basic impedance higher than C0, and C2 forest, C3 obstacle, and C4 water body correspond to impassable or high-risk impedances. For mixed grids, the basic impedance is not directly determined by a single dominant category, but is calculated by weighting based on the semantic area proportions of C0-C4; if the proportion of impassable semantics exceeds a preset safety threshold, the entire grid is upgraded to a high-risk state.

[0048] Furthermore, the grid attribute table includes basic index fields, geometric measurement fields, semantic expression fields, conflict resolution fields, and planning cost fields, as shown in Table 5. The basic index fields include grid code, grid level, parent grid code, and grid center coordinates; the geometric measurement fields include average elevation, slope statistics, terrain roughness index, humidity proxy index, soil clay content, and road prior identifiers; the semantic expression fields include dominant semantics, semantic distribution, semantic confidence, and source list; the conflict resolution fields include conflict identifier, conflict type, accepted source, and resolution reason; and the planning cost fields include basic impedance, slope penalty, soil correction, roughness correction, and comprehensive traffic cost.

[0049] Table 5. Field composition of the grid attribute table .

[0050] S7. Output the prior travel cost map. This step outputs a multi-source remote sensing travel map containing the geometric measurement layer, semantic representation layer, and decision cost layer. This multi-source remote sensing travel map is used for global path planning of unmanned ground vehicles.

[0051] The prior travel cost graph uses a hierarchical hexagonal grid as its carrier, recording the travel semantics, confidence level, source information, and comprehensive travel cost of each grid. During path planning, each traversable grid is treated as a node, and edges are established between adjacent grids. The edge cost is determined based on the comprehensive travel cost and spatial distance of adjacent grids. Inaccessible grids are treated as non-expandable nodes.

[0052] Furthermore, the map storage module employs a sparse index structure, using grid encoding as the primary key to store grid attributes. Compared to dense raster arrays, the sparse index structure only stores valid grids and their attributes within the target area, reducing storage overhead in large-scale off-road scenarios. The path planning interface module generates a planning map structure based on grid adjacency relationships, using traversable grids as nodes, spatial connections between adjacent grids as edges, and determining edge costs based on comprehensive traversal costs and spatial distances.

[0053] In one embodiment, after receiving the start point, end point, and prior travel cost map, the path planner prioritizes roads, hard-surface access areas, and low-slope regions, while avoiding water bodies, forests, obstacles, and areas with excessive slope. If soil softness risk and roughness correction are introduced, the planned path will further avoid soft surfaces and undulating gravel areas, thereby improving the feasibility of the planned path.

[0054] Furthermore, this invention can be adapted to different unmanned ground vehicles. For tracked vehicles, a higher slope threshold and a lower soft surface penalty can be set; for wheeled vehicles or heavy-duty vehicles, a more conservative slope threshold and a higher wet / soft risk penalty can be set. Thus, the same geometric measurement layer and semantic expression layer can derive different decision cost layers according to different vehicle capabilities.

[0055] Furthermore, the method of the present invention can be executed on a server, an in-vehicle computing platform, or an edge computing platform. In offline mode, a priori passage cost map is generated and sent to the in-vehicle terminal before the vehicle departs; in online mode, it can be linked with the UAV observation update process to update the cost of the affected area after local observations arrive.

[0056] This invention does not limit the specific source of remote sensing data. Remote sensing images can come from satellites, aerial platforms, or drones; digital elevation models can come from publicly available DEMs, lidar point cloud reconstructions, or photogrammetric reconstructions; soil properties can come from publicly available soil databases, field sampling, or hydrological models; road priors can come from open maps, historical trajectories, or manual plotting.

[0057] This invention does not limit the implementation to a specific hierarchical hexagonal grid. As long as the spatial index has hierarchical encoding, neighborhood query, and parent-child index functions, it can be used as a spatial organization unit. The advantage of using a hexagonal grid is that the neighborhood relationship is more balanced, which helps to reduce the bias of the path search direction and facilitates the subsequent writing back of local observation results.

[0058] This invention is applicable to scenarios such as forest patrol, disaster relief, border patrol, mining exploration, farmland operations, riverbank inspection, and field scientific research. In these scenarios, road structures are incomplete or frequently changing, and vehicles must comprehensively assess the risk of passage based on terrain, soil, water bodies, and obstacles. This invention can generate a large-scale prior travel cost map before the vehicle arrives, providing a foundation for global path planning.

[0059] This invention also provides a multi-source remote sensing access map construction device for off-road navigation of unmanned ground vehicles. The device includes a memory and a processor. The memory stores a computer program, and the processor, coupled to the memory, executes the computer program to implement the steps of the multi-source remote sensing access map construction method described in any of the above embodiments. This device can be deployed on the onboard computing platform of an unmanned ground vehicle, a ground station server, or an edge computing node.

[0060] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of the multi-source remote sensing access map construction method described in any of the above method embodiments. The storage medium includes, but is not limited to, ROM, RAM, magnetic disk, optical disk, flash memory, or other non-volatile readable storage media.

[0061] In summary, this invention discloses a method for constructing a multi-source remote sensing access map for off-road navigation of unmanned ground vehicles. This method acquires satellite remote sensing images, digital elevation models, soil property data, and road prior data of the target off-road area; establishes a hierarchical hexagonal spatial index grid; and maps multi-source environmental data uniformly to the grid. Based on the mapping results, a geometric measurement layer is constructed, including elevation, slope, terrain roughness, humidity proxy index, soil properties, and road priors. The grid access status is divided into five categories of access semantics: C0 road priors, C1 bare ground or loose surface, C2 vegetation, C3 artificial structures or obstacles, and C4 water bodies or stagnant water, and the category confidence and source information are recorded. The final access semantics are obtained through conflict resolution based on confidence level classification. A comprehensive access cost is calculated based on the final access semantics, as well as slope, soil softness risk, and terrain roughness correction terms, forming a multi-source remote sensing access map containing a geometric measurement layer, a semantic expression layer, and a decision cost layer. This method can transform heterogeneous remote sensing and geographic prior data into a comprehensive traffic cost map that can be used for unmanned ground vehicle path planning.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles, characterized in that, include: Acquire multi-source environmental data for the target off-road area, including remote sensing imagery, digital elevation model, soil property data, and prior road data; A hierarchical hexagonal spatial index grid covering the target off-road area is established, and the multi-source environmental data is mapped to the hierarchical hexagonal spatial index grid. A geometric measurement layer is constructed based on the mapped multi-source environmental data. The geometric measurement layer includes at least one grid attribute among elevation, slope, terrain roughness, humidity proxy index, soil clay content, and road prior. A semantic expression layer is constructed based on the geometric measurement layer and the remote sensing image interpretation results. The semantic expression layer records the common semantic categories and corresponding confidence levels of each grid cell. Based on the geometric measurement layer and the semantic expression layer, the semantic conflicts from different data sources in the same grid cell are resolved by confidence level classification to obtain the final common semantics; The comprehensive passage cost of each grid cell is calculated based on the final passage semantics and the geometric measurement attributes in the geometric measurement layer, and a decision cost layer is constructed, which records the comprehensive passage cost of each grid cell. The output includes a multi-source remote sensing access map containing the geometric measurement layer, semantic expression layer, and decision cost layer, which is used for off-road navigation path planning for unmanned ground vehicles.

2. The method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles according to claim 1, characterized in that, The hierarchical hexagonal spatial index grid is an H3-level grid. The H3-level grid is used to store grid identifiers, parent-child hierarchical relationships, neighborhood relationships, and grid center locations. The multi-source remote sensing access map uses neighborhood relationships as the spatial topological basis for path search.

3. The method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles according to claim 1, characterized in that, The traffic semantic categories include: C0 roads or hard-surface traffic areas, C1 bare land or low vegetation areas, C2 forest or tall vegetation areas, C3 buildings or obstacle areas, and C4 water bodies or waterlogged areas; the semantic expression layer records candidate semantics, final semantics, semantic distribution, source identifiers, and confidence levels for each grid cell.

4. The method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles according to claim 1, characterized in that, The confidence level resolution includes: When the slope derived from the digital elevation model exceeds the upper limit of the safe slope or the humidity proxy index indicates the risk of water bodies or water accumulation, a high-risk semantic is adopted. When there is a conflict between road priors and remote sensing image interpretation results, and there are no hard constraints on terrain or water bodies, the final traffic semantics are determined based on road prior confidence, image classification confidence, and neighborhood consistency.

5. The method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles according to claim 1, characterized in that, The comprehensive passage cost Calculate according to the following formula: ， in, Indicates final pass semantics The corresponding fundamental impedance, Indicates slope The corresponding slope penalty factor, This indicates a soil wet and soft risk correction factor. Indicates the surface roughness correction factor. Indicates normalized clay content. Indicates the humidity proxy index. This represents the terrain roughness index.

6. The method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles according to claim 5, characterized in that, The slope penalty factor Determined according to the following piecewise function: exist hour, =1.0; exist hour, = ; exist hour, = ; in Indicates the threshold for a gentle slope. Indicates the upper limit of the safe slope. This represents the slope penalty coefficient. This indicates the cost of making the route impassable.

7. The method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles according to claim 5, characterized in that, The soil softening risk correction factor is according to Calculation, where This indicates the risk coefficient for soft, wet soil.

8. The method for constructing multi-source remote sensing access maps for off-road navigation of unmanned ground vehicles according to claim 5, characterized in that, The surface roughness correction factor is according to Calculation, where This represents the roughness correction factor.

9. A multi-source remote sensing access map construction device for off-road navigation of unmanned ground vehicles, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the multi-source remote sensing access map construction method for off-road navigation of unmanned ground vehicles as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The system contains a computer program that, when executed by a processor, implements the steps of the multi-source remote sensing access map construction method for off-road navigation of unmanned ground vehicles as described in any one of claims 1 to 8.

Images