Map data generation method and device, electronic equipment and computer readable storage medium

Abstract

The invention provides a map data generation method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: acquiring point cloud data and motion state data, and fusing the point cloud data and the motion state data to obtain pose estimation data; performing ground elevation modeling based on the pose estimation data and the point cloud data to obtain an elevation map corresponding to the point cloud data; performing incremental mapping processing based on the elevation map corresponding to the multi-frame point cloud data to obtain a global elevation map; performing feature extraction on the global elevation map to obtain topographic feature data of each grid in the global elevation map, and determining a traffic risk score of each grid based on the topographic feature data of each grid; and based on the traffic risk score of each grid, updating the global elevation map to obtain a target map. According to the invention, the integrity of terrain modeling and the accuracy of traffic risk detection can be improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a map data generation method, apparatus, electronic device, and computer-readable storage medium. Background Technology

[0002] Terrain modeling and access risk detection are core technologies for autonomous navigation of outdoor mobile robots. Their research significance lies in improving the robot's safety, adaptability, and intelligence in complex environments. Outdoor environments (such as mountains, forests, and urban ruins) are dynamic, unstructured, and uncertain. By modeling different terrain types (such as soft soil, rocks, and steep slopes) and risk factors (such as landslides, collapses, and obstacles), robots can adapt to diverse task requirements, such as disaster relief, agricultural inspection, and border patrol. Terrain modeling and access risk detection technologies can perceive environmental characteristics in real time, assisting robots in dynamically adjusting their paths and improving their autonomous navigation capabilities.

[0003] In related technologies, outdoor mobile robots are usually equipped with sensors such as LiDAR to expand their perception range. However, the sparsity of sensor data leads to a decrease in the accuracy of ground perception, which affects the integrity of terrain modeling and, consequently, the accuracy of surface passage risk detection. Summary of the Invention

[0004] This application provides a map data generation method, apparatus, electronic device, and computer-readable storage medium, which can improve the integrity of terrain modeling and the accuracy of traffic risk detection.

[0005] The technical solution of this application embodiment is implemented as follows:

[0006] This application provides a map data generation method, the method comprising:

[0007] Acquire point cloud data and motion state data, and fuse the point cloud data and motion state data to obtain pose estimation data;

[0008] Based on the pose estimation data and the point cloud data, ground elevation modeling is performed to obtain the elevation map corresponding to the point cloud data.

[0009] Incremental mapping is performed on the elevation maps corresponding to the point cloud data from multiple frames to obtain a global elevation map.

[0010] Feature extraction is performed on the global elevation map to obtain the terrain feature data of each grid in the global elevation map, and the access risk score of each grid is determined based on the terrain feature data of each grid.

[0011] Based on the access risk score of each grid, the global elevation map is updated to obtain the target map.

[0012] This application provides a map data generation apparatus, including:

[0013] The data fusion module is used to acquire point cloud data and motion state data, and fuse the point cloud data and motion state data to obtain pose estimation data;

[0014] The map modeling module is used to perform ground elevation modeling based on the pose estimation data and the point cloud data to obtain an elevation map corresponding to the point cloud data.

[0015] The incremental mapping module is used to perform incremental mapping processing based on the elevation map corresponding to multiple frames of point cloud data to obtain a global elevation map.

[0016] The feature extraction module is used to extract features from the global elevation map to obtain the terrain feature data of each grid in the global elevation map, and to determine the passage risk score of each grid based on the terrain feature data of each grid.

[0017] The map update module is used to update the global elevation map based on the access risk score of each grid to obtain the target map.

[0018] This application provides an electronic device, the electronic device comprising:

[0019] Memory is used to store executable instructions or computer programs.

[0020] The processor, when executing computer-executable instructions or computer programs stored in the memory, implements the map data generation method provided in the embodiments of this application.

[0021] This application provides a computer-readable storage medium storing computer-executable instructions or computer programs, which, when executed by a processor, implement the map data generation method provided in this application.

[0022] This application provides a computer program product, including computer-executable instructions or a computer program, which, when executed by a processor, implements the map data generation method provided in this application.

[0023] The embodiments of this application have the following beneficial effects:

[0024] By applying the embodiments of this application, point cloud data and motion state data are acquired and fused to obtain pose estimation data. Ground elevation modeling is then performed based on the pose estimation data and point cloud data to obtain an elevation map corresponding to the point cloud data. This method can generate an elevation map corresponding to a single frame of point cloud data using sparse sensor data such as point cloud data and motion state data. Incremental mapping is then performed based on the elevation maps corresponding to multiple frames of point cloud data to obtain a global elevation map. This enables global terrain modeling using elevation maps corresponding to multiple frames of point cloud data, improving the completeness of terrain modeling. Feature extraction is then performed on the global elevation map to obtain terrain feature data for each grid in the global elevation map. Based on the terrain feature data of each grid, a passage risk score for each grid is determined. Finally, based on the passage risk score of each grid, the global elevation map is updated to obtain the target map. In this way, by using sparse sensor data such as point cloud data and motion state data, a dense elevation map is incrementally constructed, which improves the accuracy and completeness of terrain modeling. Furthermore, by utilizing the access risk score of each grid in the dense elevation map, a high-precision access risk map is generated, thereby improving the accuracy of access risk detection. Attached Figure Description

[0025] Figure 1 This is a schematic diagram illustrating the application mode of the map data generation method provided in the embodiments of this application;

[0026] Figure 2 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application;

[0027] Figure 3A This is a first flowchart illustrating the map data generation method provided in this application embodiment;

[0028] Figure 3B This is a schematic diagram of the second process of the map data generation method provided in the embodiments of this application;

[0029] Figure 3C This is a schematic diagram of the third process of the map data generation method provided in the embodiments of this application;

[0030] Figure 4 This is a schematic diagram of the system architecture for generating map data provided in an embodiment of this application.

[0031] It should be noted that the terms "first" and "second" mentioned above are only used to distinguish between different options and do not represent the degree of superiority or inferiority of the options or their priority in the implementation process. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0034] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0035] In the implementation of this application, the collection and processing of relevant data should strictly comply with the requirements of relevant laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject.

[0036] In this application embodiment, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0037] Unless otherwise defined, all technical and scientific terms used in the embodiments of this application have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in the embodiments of this application is for the purpose of describing the embodiments of this application only and is not intended to limit this application.

[0038] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0039] 1) Elevation map: a two-dimensional raster map in which each cell stores the ground elevation value corresponding to that location.

[0040] 2) Point cloud data: is a data set consisting of a large number of three-dimensional points, each containing three-dimensional coordinates.

[0041] This application provides a map data generation method, apparatus, electronic device, and computer-readable storage medium, which can improve the integrity of terrain modeling and the accuracy of traffic risk detection.

[0042] The following describes exemplary applications of the electronic devices provided in the embodiments of this application. These electronic devices can be implemented as various types of terminals such as laptops, tablets, desktop computers, set-top boxes, smartphones, smart speakers, smartwatches, smart TVs, and in-vehicle terminals, or as servers. The following will describe exemplary applications when the device is implemented as a server.

[0043] See Figure 1 , Figure 1 This is a schematic diagram illustrating the application mode of the map data generation method provided in the embodiments of this application, for example. Figure 1 The system involves server 200, network 300, and terminal 400. Terminal 400 connects to server 200 through network 300, which can be a wide area network (WAN), a local area network (LAN), or a combination of both.

[0044] During map data generation, terminal 400 collects point cloud data and motion state data through a data acquisition device and sends the point cloud data and motion state data to server 200. Server 200 acquires the point cloud data and motion state data and fuses them to obtain pose estimation data. Based on the pose estimation data and point cloud data, ground elevation modeling is performed to obtain an elevation map corresponding to the point cloud data. Incremental mapping processing is performed on the elevation map corresponding to multiple frames of point cloud data to obtain a global elevation map. Feature extraction is performed on the global elevation map to obtain terrain feature data for each grid in the global elevation map, and a passage risk score for each grid is determined based on the terrain feature data of each grid. Based on the passage risk score of each grid, the global elevation map is updated to obtain the target map. When server 200 receives a path planning request from the terminal, it parses the start and end point information from the request and performs path planning based on the target map, the start and end point information to obtain the target path. Server 200 returns the target path to terminal 400, which then moves based on the received target path. Since the target map is obtained by updating the global elevation map based on the passage risk score of each grid, the safety of passage when terminal 400 moves based on the target path determined by the target map can be guaranteed. Terminal 400 can be an intelligent robot used in outdoor scenarios.

[0045] In some embodiments, after acquiring point cloud data and motion state data through the data acquisition device, the terminal 400 can also fuse the point cloud data and motion state data to obtain pose estimation data; perform ground elevation modeling based on the pose estimation data and point cloud data to obtain an elevation map corresponding to the point cloud data; perform incremental mapping processing based on the elevation map corresponding to multiple frames of point cloud data to obtain a global elevation map; extract features from the global elevation map to obtain terrain feature data for each grid in the global elevation map, and determine the passage risk score for each grid based on the terrain feature data of each grid; update the global elevation map based on the passage risk score of each grid to obtain a target map; and then, when path planning is required, perform path planning based on the target map, starting point information, and ending point information to obtain a target path, and move based on the target path.

[0046] In some embodiments, the server (e.g., server 200) can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal 400 can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, in-vehicle terminal, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited in this embodiment.

[0047] See Figure 2 , Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may be a terminal or a server. Figure 2 The illustrated electronic device includes at least one processor 410, a memory 450, and at least one network interface 420. The various components of the electronic device are coupled together via a bus system 440. It is understood that the bus system 440 is used to implement communication between these components. In addition to a data bus, the bus system 440 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in… Figure 2 The general labeled all buses as Bus System 440.

[0048] The processor 410 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0049] The memory 450 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 450 may optionally include one or more storage devices physically located away from the processor 410.

[0050] The memory 450 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 450 described in this application embodiment is intended to include any suitable type of memory.

[0051] In some embodiments, memory 450 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0052] Operating system 451 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, and driver layer, for implementing various basic business functions and handling hardware-based tasks.

[0053] The network communication module 452 is used to reach other electronic devices via one or more (wired or wireless) network interfaces 420, exemplary network interfaces 420 including Bluetooth, WiFi, and Universal Serial Bus (USB).

[0054] In some embodiments, the apparatus provided in this application can be implemented in software. Figure 2 A map data generation device 455 stored in memory 450 is shown. It can be software in the form of programs and plug-ins, including the following software modules: data fusion module 4551, map modeling module 4552, incremental mapping module 4553, feature extraction module 4554, and map update module 4555. These modules are logically related and can therefore be arbitrarily combined or further divided according to the functions they implement. The functions of each module will be described below.

[0055] The map data generation method provided in this application will be described in conjunction with exemplary applications and implementations of the server devices provided in the embodiments of this application.

[0056] The following describes the map data generation method provided in the embodiments of this application. For example, in order to facilitate understanding of the map data generation method provided in the embodiments of this application, the method is described using an application in a robot navigation scenario.

[0057] As mentioned above, the electronic device implementing the map data generation method of this application embodiment can be a terminal, a server, or a combination of both. The following description uses an electronic device as a server as an example to illustrate the map data generation method provided in this application embodiment. See also... Figure 3A , Figure 3A This is a first flowchart illustrating the map data generation method provided in this application embodiment, which will be combined with... Figure 3A The steps shown are explained.

[0058] In step 301, point cloud data and motion state data are acquired and fused to obtain pose estimation data.

[0059] Here, point cloud data is collected using a LiDAR system mounted on the robot. Point cloud data refers to a dataset composed of a large number of three-dimensional points, each containing three-dimensional coordinates. The robot's motion state data, including high-frequency angular velocity and acceleration, is collected using an Inertial Measurement Unit (IMU) mounted on the robot. Simultaneous Localization and Mapping (SLAM) techniques such as Fast LiDAR-Inertial Odometry (Fast-LIO) and LiDAR-Inertial Odometry via Smoothing and Mapping (LIO-SAM) can be used to fuse the point cloud data and motion state data, obtaining high-precision real-time pose estimation data. This pose estimation data estimates the robot's position and orientation.

[0060] Continue to refer to Figure 3A In step 302, ground elevation modeling is performed based on pose estimation data and point cloud data to obtain an elevation map corresponding to the point cloud data.

[0061] Here, pose estimation data and point cloud data with the same timestamp are used to perform ground elevation modeling to generate an elevation map corresponding to the point cloud data. The elevation map is a two-dimensional raster map, in which each cell stores the ground elevation value corresponding to that location.

[0062] In some embodiments, see Figure 3B , Figure 3B This is a schematic diagram of the second process of the map data generation method provided in the embodiments of this application. Figure 3A Step 302 shown can be achieved through... Figure 3B Steps 3021 to 3025 are implemented, and will be explained in detail below.

[0063] In step 3021, distortion correction is performed on the point cloud data based on the motion state data to obtain the corrected point cloud data.

[0064] Here, because the robot's movement causes motion distortion in the point cloud data, distortion correction is necessary. In step 3021, the timestamps of the motion state data and the point cloud data are synchronized to obtain the transformation matrix from point cloud data to motion state data. The transformation matrix is ​​used to transform the point cloud data to the inertial measurement unit coordinate system, and intra-frame motion compensation is performed on the transformed point cloud data to obtain compensated point cloud data. Then, the transformation matrix is ​​used to transform the compensated point cloud data to the lidar coordinate system to obtain the corrected point cloud data.

[0065] In step 3022, obstacle points and non-obstacle points are determined from each point in the corrected point cloud data.

[0066] Here, the corrected point cloud data includes multiple 3D points, from which obstacle points and non-obstacle points are determined.

[0067] In some embodiments, determining obstacle points and non-obstacle points from points in the corrected point cloud data can be achieved through the following steps: for each point in the corrected point cloud data, determining the point's neighboring points in the horizontal direction, and determining a first distance between the neighboring points and any two adjacent points in the point; determining the point's adjacent points in the vertical direction, and determining a height difference and a second distance between the point and its adjacent points; determining the point as an obstacle point when at least one first distance is greater than a first threshold; or determining the point as an obstacle point when the height difference is greater than a second threshold and the second distance is less than a third threshold; determining the point as a non-obstacle point when there is no first distance greater than the first threshold, a height difference less than or equal to the second threshold, and a second distance greater than or equal to the third threshold.

[0068] Here, for each point in the corrected point cloud data, since points are densely distributed in the horizontal direction, we determine the neighboring points in the horizontal direction for each point. These neighboring points include at least two points from the left and right neighbors of the given point. Using the coordinates of each neighboring point and the coordinates of the given point, we calculate the distance between each neighboring point and its two adjacent points, obtaining multiple first distances. Since points are sparsely distributed in the vertical direction, we determine the nearest adjacent point in the vertical direction as the neighboring point. Using the coordinates of the neighboring point and the coordinates of the given point, we calculate the height difference between the given point and the neighboring point, and then calculate the distance between the given point and the neighboring point, which is the second distance.

[0069] The first threshold is a preset horizontal distance threshold. A point is identified as an obstacle point when at least one horizontal distance is greater than the first threshold. The second threshold is a preset height threshold, and the third threshold is a preset spatial distance threshold. A point is identified as an obstacle point when the height difference between the point and its adjacent points is greater than the second threshold, and the second distance between the point and its adjacent points is less than the third threshold. A point is identified as a non-obstacle point when no horizontal distance is greater than the first threshold, the height difference is less than or equal to the second threshold, and the second distance is greater than or equal to the third threshold.

[0070] For example, for a specific point in the corrected point cloud data, determine the horizontal neighborhood points of that point. These neighborhood points include the three points in the left neighborhood and the three points in the right neighborhood, totaling six neighborhood points. Calculate the distances between each of these six neighborhood points and any two adjacent points, obtaining six first distances. Identify the closest vertical neighbor to the current point as the adjacent point, calculate the height difference between the current point and the adjacent point, and then calculate the distance between them, obtaining a second distance. Using a first threshold of 0.2 meters, a second threshold of 0.1 meters, and a third threshold of 1 meter as an example, if at least one of the six first distances is greater than the first threshold (i.e., at least one first distance is greater than 0.2 meters), the point is identified as an obstacle point. Alternatively, if the height difference is greater than the second threshold and the second distance is less than the third threshold (i.e., the height difference is greater than 0.1 meters and the second distance is less than 1 meter), the point is identified as an obstacle point. A point is determined to be a non-obstacle point when none of the six first distances is greater than 0.2 meters, the height difference is less than or equal to 0.1 meters, and the second distance is greater than or equal to 1 meter.

[0071] In this embodiment, for each point in the corrected point cloud data, a first distance is determined between the horizontal neighboring points of a point and two adjacent points of that point, and a second distance is determined between the vertical neighboring points of that point and the vertical neighboring points of that point. The first distance, the height difference, and the second distance are used to classify obstacle points and non-obstacle points, thereby achieving accurate classification of each point in the corrected point cloud data by combining horizontal distance analysis and vertical height difference analysis, thus improving the accuracy of determining obstacle points and non-obstacle points.

[0072] Continue to refer to Figure 3B In step 3023, obstacle points and non-obstacle points are projected onto a two-dimensional grid map to determine the non-obstacle grids and unknown grids in the two-dimensional grid map.

[0073] Here, a two-dimensional grid map is a method of discretizing environmental space into a regular grid representation. Each grid cell stores the occupancy probability of that location, i.e., statistical information about whether an obstacle exists at that location. Obstacle points and non-obstacle points are projected onto the two-dimensional grid map, and the non-obstacle grids and unknown grids in the two-dimensional grid map are determined based on the points projected into each grid cell.

[0074] In some embodiments, projecting obstacle points and non-obstacle points onto a two-dimensional grid map and determining non-obstacle grids and unknown grids in the two-dimensional grid map can be achieved through the following steps: projecting obstacle points and non-obstacle points onto a two-dimensional grid map and determining the number of obstacle points in each grid of the two-dimensional grid map; for each grid, determining the occupancy probability of the grid based on the number of obstacle points in the grid; when the occupancy probability is less than or equal to a fourth threshold, the grid is determined to be a non-obstacle grid; grids without obstacle points or non-obstacle points are determined to be unknown grids.

[0075] Here, after projecting obstacle points and non-obstacle points onto a 2D grid map, the number of obstacle points in each grid is determined. Each grid has a corresponding initial occupancy probability. Based on the number of obstacle points detected in a grid, a Bayesian update model is used to adjust the initial occupancy probability of that grid, resulting in the grid's occupancy probability. A fourth threshold is a preset upper limit for the occupancy probability of a grid. When the occupancy probability of a grid is less than or equal to the fourth threshold, the grid is determined to be a non-obstacle grid. For each grid in the 2D grid map, grids without any obstacle points or non-obstacle points are determined to be unknown grids; that is, unknown grids are grids without any laser points.

[0076] For example, taking a fourth threshold of 70% as an example, for a certain grid in a two-dimensional grid map, the number of obstacle points in the grid is determined to be 60, and the initial occupancy probability of the grid is 30%. Based on the 60 obstacle points detected in the grid, the initial occupancy probability of the grid is adjusted using a Bayesian update model, resulting in an occupancy probability of 60%. Since the occupancy probability of the grid is less than 70%, the grid is determined to be a non-obstacle grid.

[0077] Continue to refer to Figure 3B In step 3024, based on the pose estimation data, the corrected point cloud data is converted to the world coordinate system to obtain the converted point cloud data.

[0078] Here, a coordinate system transformation matrix is ​​constructed using pose estimation data. For each point in the corrected point cloud data, the three-dimensional coordinates of the point are multiplied by the coordinate system transformation matrix to transform the corrected point cloud data to the world coordinate system, thus obtaining the transformed point cloud data.

[0079] In step 3025, the height information of each point located in the non-obstacle grid in the converted point cloud data is determined, and an elevation map corresponding to the point cloud data is generated based on the height information of each point.

[0080] Here, points located within the non-obstacle grid are identified from the transformed point cloud data. For each point within the non-obstacle grid, its height, or height information, is determined using its coordinates. Based on the height information of each point within a non-obstacle grid, the height corresponding to that non-obstacle grid is determined. Then, using the heights corresponding to multiple non-obstacle grids, an elevation map corresponding to the point cloud data is generated.

[0081] In some embodiments, generating an elevation map corresponding to point cloud data based on the height information of each point can be achieved through the following steps: summing and averaging the heights of each point in the non-obstructive grid to obtain a first height corresponding to the non-obstructive grid; for each unknown grid, determining at least one target non-obstructive grid adjacent to the unknown grid, and determining the weight corresponding to at least one target non-obstructive grid based on the distance between the unknown grid and at least one target non-obstructive grid; weighting the heights of at least one target non-obstructive grid based on the weights corresponding to at least one target non-obstructive grid to obtain a second height corresponding to the unknown grid; and generating an elevation map corresponding to the point cloud data based on the first height and the second height.

[0082] Here, for each non-obstructive grid, the height of each point in the non-obstructive grid is summed and averaged to obtain the height corresponding to the non-obstructive grid, i.e., the first height. For each unknown grid, non-obstructive grids whose distance from the unknown grid is less than or equal to a preset distance are identified as target non-obstructive grids. The Euclidean distances between the unknown grids and each target non-obstructive grid are determined, and then each Euclidean distance is used as a weight corresponding to each target non-obstructive grid. For example, the Euclidean distances can be used as weights for target non-obstructive grids, or the Euclidean distances can be normalized to obtain the weights for each target non-obstructive grid. Finally, the heights of multiple target non-obstructive grids are weighted and summed using the weights of multiple target non-obstructive grids to obtain the height corresponding to the unknown grid, i.e., the second height. The first and second heights are input into visualization tools such as heatmaps (colors represent height) and contour maps to generate and output an elevation map corresponding to the point cloud data.

[0083] In some embodiments, a Bayesian generalized kernel inference method can be used to predict the second height corresponding to the unknown grid. Further, this can be achieved through the following steps: Using the weights corresponding to multiple target non-obstructive grids, the heights of the multiple target non-obstructive grids are weighted and summed to obtain the initial height corresponding to the unknown grid. The difference between the initial height of the unknown grid and the height of a target non-obstructive grid is calculated, and this difference is determined as the first update weight corresponding to a target non-obstructive grid. The weight corresponding to the target non-obstructive grid is multiplied by the first update weight to obtain the second update weight corresponding to a target non-obstructive grid. Using the second update weights corresponding to multiple target non-obstructive grids, the heights of the multiple target non-obstructive grids are weighted to obtain the second height corresponding to the unknown grid.

[0084] In this embodiment, for each unknown grid, a weight corresponding to at least one target non-obstacle grid is determined based on the distance between the unknown grid and at least one target non-obstacle grid. Based on the weight corresponding to at least one target non-obstacle grid, the height of at least one target non-obstacle grid is weighted to obtain a second height corresponding to the unknown grid. This achieves the goal of combining the distance information between the unknown grid and the target non-obstacle grid and the height information of the target non-obstacle grid to determine the second height corresponding to the unknown grid, thereby improving the accuracy of predicting the height of the unknown grid.

[0085] Continue to refer to Figure 3A In step 303, incremental mapping is performed based on the elevation map corresponding to the multi-frame point cloud data to obtain a global elevation map.

[0086] Here, the elevation map corresponding to a point cloud data frame at the current moment is used to update the elevation map corresponding to a point cloud data frame at the previous moment, so as to realize the incremental mapping of the local elevation map and obtain the global elevation map.

[0087] In some embodiments, see Figure 3C , Figure 3C This is a schematic diagram of the third process of the map data generation method provided in the embodiments of this application. Figure 3A Step 303 shown can be achieved through... Figure 3C Steps 3031 to 3032 are implemented, and will be explained in detail below.

[0088] In step 3031, the first elevation map corresponding to the point cloud data at the first moment and the second elevation map corresponding to the point cloud data at the second moment are determined from the elevation maps corresponding to the multi-frame point cloud data.

[0089] Here, the first time point and the second time point are adjacent, and the first time point is less than the second time point, meaning the first time point is the previous time point adjacent to the second time point. Using the timestamps of the elevation maps, the elevation map corresponding to the point cloud data at the first time point, i.e., the first elevation map, is determined from the elevation maps corresponding to the point cloud data at multiple frames, and the elevation map corresponding to the point cloud data at the second time point, i.e., the second elevation map, is also determined.

[0090] In step 3032, the height of the corresponding grid in the first elevation map is updated based on the height of each grid in the second elevation map to obtain the global elevation map.

[0091] Here, based on the height of each non-obstruction grid in the second elevation map, the height of the corresponding non-obstruction grid in the first elevation map is updated to obtain the updated height of the non-obstruction grid. Similarly, based on the height of each unknown grid in the second elevation map, the height of the corresponding unknown grid in the first elevation map is updated to obtain the updated height of the unknown grid. The updated heights of the non-obstruction grids and the unknown grids are then input into visualization tools such as heatmaps and contour maps to generate and output a global elevation map.

[0092] In this embodiment, a first elevation map corresponding to the point cloud data at the first moment and a second elevation map corresponding to the point cloud data at the second moment are determined from the elevation maps corresponding to the point cloud data at multiple frames. Based on the height of each grid in the second elevation map, the height of the corresponding grid in the first elevation map is updated to obtain a global elevation map. This realizes the incremental construction of a dense elevation map using sparse sensor data, which improves the accuracy and completeness of terrain modeling.

[0093] Continue to refer to Figure 3AIn step 304, feature extraction is performed on the global elevation map to obtain the terrain feature data of each grid in the global elevation map, and the access risk score of each grid is determined based on the terrain feature data of each grid.

[0094] Here, for each grid in the global elevation map, the terrain features of that grid are extracted to obtain the terrain feature data, which includes slope, elevation difference, and roughness. Based on the slope, elevation difference, and roughness of each grid, a passage risk score is determined for each grid.

[0095] The slope can be calculated from the angle between the normal of the current grid and the Z-axis of the world coordinate system. Specifically, the heights of all grids within a preset radius of the current grid are determined, and the average height is calculated. Principal Component Analysis (PCA) is used to calculate the normal of the current grid based on the average height, and the inverse cosine function is applied to the Z-axis component of the normal to obtain the slope of the current grid. Several target grids closest to the center point of the current grid are identified, and the height differences between the current grid and each of the target grids are calculated. The maximum height difference is determined as the elevation difference of the current grid. The roughness of the grid is obtained by calculating the standard deviation of the heights of all grids.

[0096] In some embodiments, determining the access risk score for each grid based on the terrain feature data of each grid can be achieved through the following steps: for each grid, when the grid's slope is greater than a fifth threshold, or the grid's elevation difference is greater than a sixth threshold, or the grid's roughness is greater than a seventh threshold, the grid's access risk score is determined to be a first value; when the grid's slope is less than or equal to the fifth threshold, the grid's elevation difference is less than or equal to the sixth threshold, and the grid's roughness is less than or equal to the seventh threshold, the grid's access risk score is determined to be a second value.

[0097] Here, the fifth threshold is a preset slope threshold, the sixth threshold is a preset elevation difference threshold, and the seventh threshold is a preset roughness threshold. For each grid in the global elevation map, when the grid's slope is greater than the fifth threshold, or the grid's elevation difference is greater than the sixth threshold, or the grid's roughness is greater than the seventh threshold, the grid's traffic risk score is determined to be the first value, which can be 0, indicating that there is a traffic risk at the grid's location. When the grid's slope is less than or equal to the fifth threshold, the grid's elevation difference is less than or equal to the sixth threshold, and the grid's roughness is less than or equal to the seventh threshold, the grid's traffic risk score is determined to be the second value, which can be 1, indicating that there is no traffic risk at the grid's location.

[0098] For example, the fifth threshold is 30 degrees, the sixth threshold is 10 meters, and the seventh threshold is 50 micrometers. For the first grid in the global elevation map, the slope of the first grid is 10 degrees, the elevation difference of the first grid is 20 meters, and the roughness of the first grid is 40 micrometers. Therefore, the slope of the first grid is less than the fifth threshold, the elevation difference of the first grid is greater than the sixth threshold, and the roughness of the first grid is less than the seventh threshold. Thus, the passage risk score of the first grid is determined to be 0, indicating that there is a passage risk at the location of the first grid.

[0099] For the second grid in the global elevation map, the slope of the second grid is 10 degrees, the elevation difference of the second grid is 8 meters, and the roughness of the second grid is 30 micrometers. Therefore, the slope of the second grid is less than the fifth threshold, the elevation difference of the second grid is less than the sixth threshold, and the roughness of the second grid is less than the seventh threshold. The passage risk score of the second grid is determined to be 1, indicating that there is no passage risk at the location of the second grid.

[0100] In this embodiment, for each grid in the global elevation map, when the grid's slope is greater than a preset slope threshold, or the grid's elevation difference is greater than a preset elevation difference threshold, or the grid's roughness is greater than a preset roughness threshold, the grid's traffic risk score is determined as a first value. When the grid's slope is less than or equal to a preset slope threshold, the grid's elevation difference is less than or equal to a preset elevation difference threshold, and the grid's roughness is less than or equal to a preset roughness threshold, the grid's traffic risk score is determined as a second value. This achieves a comprehensive determination of the grid's traffic risk score based on terrain feature data such as grid slope, elevation difference, and roughness, thereby improving the accuracy of traffic risk detection for each grid's location.

[0101] Continue to refer to Figure 3A In step 305, the global elevation map is updated based on the access risk score of each grid to obtain the target map.

[0102] Here, the target map is a access risk map. For each grid in the global elevation map, the access risk score of the grid is input into visualization tools such as a geographic information system to draw and output the target map.

[0103] In some embodiments, the map data generation method provided in this application can be applied in the field of cloud technology. Point cloud data and motion state data are acquired on a cloud platform and fused to obtain pose estimation data. Ground elevation modeling is then performed based on the pose estimation data and point cloud data to obtain an elevation map corresponding to the point cloud data. This method can utilize sparse sensor data such as point cloud data and motion state data to generate an elevation map corresponding to a single frame of point cloud data. Incremental mapping is then performed based on multiple frames of elevation maps to obtain a global elevation map. This enables global terrain modeling using elevation maps corresponding to multiple frames of point cloud data, improving the completeness of terrain modeling. Feature extraction is then performed on the global elevation map to obtain terrain feature data for each grid. A traffic risk score for each grid is determined based on the terrain feature data, and the global elevation map is updated based on the traffic risk score for each grid to obtain the target map. In this way, by using sparse sensor data such as point cloud data and motion state data, a dense elevation map is incrementally constructed, which improves the accuracy and completeness of terrain modeling. Furthermore, by utilizing the access risk score of each grid in the dense elevation map, a high-precision access risk map is generated, thereby improving the accuracy of access risk detection.

[0104] The following will describe an exemplary application of the map data generation method provided in this application embodiment in a robot navigation scenario.

[0105] Outdoor mobile robots are widely used in outdoor scenarios to complete various pre-defined tasks, such as power line inspection, surface mapping, and reconnaissance. As the complexity of outdoor applications increases, higher demands are placed on the robots' perception capabilities, requiring them to accurately model the surrounding terrain and detect passage risks. Terrain modeling and passage risk detection generate a surface passage risk map for subsequent robot planning and navigation. Outdoor mobile robots are typically equipped with LiDAR to extend their perception range; however, the sparsity of radar point clouds can affect the completeness of terrain modeling, thus impacting the accuracy of surface passability detection.

[0106] This application proposes a map data generation method to address the problems existing in related technologies, which includes the following improvements compared to related technologies:

[0107] By utilizing sparse sensor data (point cloud data and motion state data in the above embodiments) to perceive the local environment, a dense elevation map is constructed. Historical frame information is integrated into the elevation map to improve the accuracy and completeness of terrain modeling. This enables the generation of a two-dimensional obstacle map (global elevation map in the above embodiments) and a passage risk map (target map in the above embodiments). This solves the problems of decreased ground perception accuracy and low passage risk detection accuracy caused by the sparsity of sensor data. It can be widely used in environmental modeling and perception of outdoor mobile robots, autonomous navigation of unstructured terrain, and other fields.

[0108] The following describes the process of generating map data. The map data generation process consists of five parts: data acquisition, data preprocessing, ground elevation modeling, incremental mapping, and terrain feature extraction. For an example, refer to [link / reference]. Figure 4 , Figure 4 This is a schematic diagram of the system architecture for generating map data provided in an embodiment of this application.

[0109] Regarding the data acquisition process, the system's sensors include a LiDAR 401 and an Inertial Measurement Unit 402. The LiDAR collects laser point clouds, and the Inertial Measurement Unit (IMU) collects the robot's high-frequency angular velocity and acceleration information (motion state data in the above embodiment). Based on Simultaneous Localization and Mapping (SLAM) technology, specifically using methods such as Fast LiDAR-Inertial Odometry (Fast-LIO) and LiDAR-Inertial Odometry via Smoothing and Mapping (LIO-SAM), the laser point cloud, angular velocity, and acceleration information are fused to perform real-time robot pose estimation, thereby obtaining accurate radar inertial odometry information (pose estimation data in the above embodiment).

[0110] Regarding the data preprocessing process, since the robot's movement causes motion distortion in the point cloud, high-frequency 6-DOF angular velocity and acceleration data output by the IMU can be used to perform intra-frame motion compensation on the distorted point cloud, resulting in a corrected single-frame point cloud. The corrected single-frame point cloud consists of multiple points, and each point is divided into obstacle points and non-obstacle points based on horizontal distance difference analysis and vertical height difference analysis. For each point, if the horizontal distance between the point and its adjacent points is greater than a preset distance threshold, or if the vertical height difference between the point and its adjacent points is greater than a preset height threshold, then the point is marked as an obstacle point. Specifically, for each laser beam in the horizontal direction, the distance between a point on the laser beam and three points on each side is calculated. For example, for the fifth point of the first laser beam, the distance between this point and two adjacent points in its three neighboring areas on the left and right is calculated. If at least one distance exceeds a preset distance threshold, it indicates that the point is located at the edge of an obstacle, and the point is identified as an obstacle point, thus achieving horizontal obstacle (such as a tree trunk) recognition.

[0111] When the height difference between a point and its adjacent points in the adjacent line bundle in the vertical direction is greater than 0.1 meters, and the distance between them is less than 1 meter, it indicates that there is a significant drop near that point, and the point is identified as an obstacle point, thus achieving the identification of vertical obstacles (such as puddles). Then, the obstacle points and non-obstacle points are filtered using point cloud techniques and projected onto a 2D grid map. The grid cells are observed based on the obstacle state of each point, and the 2D grid map is updated probabilistically. Each grid stores its corresponding occupancy probability. When an obstacle point falls into a grid, the occupancy probability of that grid increases. When the occupancy probability exceeds a set threshold (e.g., 0.6), the grid is identified as an obstacle grid.

[0112] For the ground elevation modeling process, odometry information is used to rotate and transform the laser point cloud to a global coordinate system, and the height of points falling into each non-obstacle grid is statistically analyzed. For each non-obstacle grid, the mean and variance of the heights of all points in that grid are calculated to obtain the height and height variance of the non-obstacle grid. To address the sparsity of radar data, the Bayesian Generalized Kernel Inference (BGK) method is used to predict the height of unknown grids. Unknown grids are those not observed by the lidar, meaning that no obstacle or non-obstacle points fall into them; therefore, the BGK method is used to estimate the height of unknown grids.

[0113] Bilateral filtering is introduced into the BGK method. Specifically, it iterates through the non-obstacle grids (hereinafter referred to as known grids) with known elevation distributions within the Gaussian kernel radius of the current unknown grid, and determines the spatial weights based on the Euclidean distance between the current unknown grid and the known grids. The initial height of the unknown grid is obtained by weighting the heights of all known grids according to the spatial weights. The difference between the initial height of the unknown grid and the heights of the known grids is calculated, and the range weights are calculated using this difference. The spatial weights are multiplied by the range weights to obtain updated spatial weights. The heights of all known grids are then weighted according to these updated spatial weights to obtain the estimated height of the unknown grid. The process of predicting the height of the unknown grid considers both the spatial weights related to grid distances and the range weights related to the height differences between grids, thereby solving the edge blurring problem caused by the BGK method.

[0114] Regarding the incremental mapping process, after generating an elevation map corresponding to a single frame of point cloud data using the heights of non-obstructive and unknown grids, a global elevation map can be incrementally constructed based on information from consecutive radar frames using a rolling sliding window method. This improves the stability and accuracy of terrain modeling. For each frame of the elevation map, the height corresponding to each grid is traversed. For grids overlapping between two observation frames, a first-order Kalman filter can be used to update the height of the non-obstructive grids. For an example, please refer to [link to example]. Figure 4 Bayesian kernel inference, bilateral filtering, and Kalman filtering are performed on the point cloud data to generate elevation map 403.

[0115] For the terrain feature extraction process, terrain features such as slope, elevation difference, and roughness are extracted from the global elevation map for each grid to calculate a terrain accessibility score, thus assessing the ease of terrain access. Slope can be calculated from the angle between the grid surface normal and the Z-axis of the global coordinate system. Specifically, the height of all grids within a preset radius of the current grid is determined, and the average height is calculated. Principal Component Analysis (PCA) is used to calculate the normal of the current grid based on the average height, and the inverse cosine function is applied to the Z-axis component of the normal to obtain the grid slope. The elevation difference of the grid can be calculated from the maximum elevation difference between the grid center point and its k nearest neighboring grids. Specifically, multiple target grids closest to the current grid center point are identified, and the elevation differences between the current grid and each of the target grids are calculated. The maximum elevation difference is determined as the elevation difference of the current grid. The roughness of the grid is calculated from the standard deviation of the heights of all grids.

[0116] To ensure the reliability of grid observations and reduce the impact of dynamic obstacles, a passage score can be calculated when the number of observations for each grid cell exceeds the observation threshold. When any of the slope, elevation difference, or roughness exceeds the corresponding threshold, the passage score for the grid cell is set to 0, indicating that it is impassable. When the slope, elevation difference, or roughness is below the corresponding threshold, the passage score for the grid cell is set to 1. For other cases, the passage score can be determined using the following formula (1):

[0117]

[0118] Among them, T geo s represents the pass score. cri h cri r cri α1, α2, and α3 represent the maximum critical thresholds for dangerous situations such as the robot slipping, tipping over, or getting stuck. α1, α2, and α3 represent the weights of each feature, and the sum of the weights is 1.

[0119] For example, please continue to refer to [the example]. Figure 4 A passage risk map 404 is generated based on the passage score of each grid cell for the robot's autonomous navigation in unstructured terrain.

[0120] In the aforementioned robot navigation scenarios, generating two-dimensional obstacle maps and terrain traversal risk maps can improve the completeness and accuracy of terrain modeling and autonomous navigation in unstructured terrain environments. Simultaneously, it enhances the detection accuracy and real-time performance of traversal risks, reduces computational resource consumption, and is suitable for deployment on low-computing-power onboard devices.

[0121] The following description continues to illustrate the exemplary structure of the map data generation device 455 provided in the embodiments of this application as a software module. In some embodiments, such as Figure 2 As shown, the software modules stored in the map data generation device 455 in the memory 450 may include: a data fusion module 4551, used to acquire point cloud data and motion state data, and fuse the point cloud data and motion state data to obtain pose estimation data; a map modeling module 4552, used to perform ground elevation modeling based on the pose estimation data and point cloud data to obtain an elevation map corresponding to the point cloud data; an incremental mapping module 4553, used to perform incremental mapping processing based on the elevation map corresponding to multiple frames of point cloud data to obtain a global elevation map; a feature extraction module 4554, used to extract features from the global elevation map to obtain terrain feature data of each grid in the global elevation map, and determine the access risk score of each grid based on the terrain feature data of each grid; and a map update module 4555, used to update the global elevation map based on the access risk score of each grid to obtain a target map.

[0122] In some embodiments, the map modeling module 4552 is further configured to: correct the distortion of point cloud data based on motion state data to obtain corrected point cloud data; determine obstacle points and non-obstacle points from each point in the corrected point cloud data; project the obstacle points and non-obstacle points onto a two-dimensional grid map to determine the non-obstacle grids and unknown grids in the two-dimensional grid map; convert the corrected point cloud data to the world coordinate system based on pose estimation data to obtain converted point cloud data; determine the height information of each point located in the non-obstacle grids in the converted point cloud data; and generate an elevation map corresponding to the point cloud data based on the height information of each point.

[0123] In some embodiments, the map modeling module 4552 is further configured to, for each point in the corrected point cloud data, determine the neighboring points of the point in the horizontal direction and determine the first distance between the neighboring points and two adjacent points of the point; determine the adjacent points of the point in the vertical direction and determine the height difference and second distance between the point and the adjacent points; determine the point as an obstacle point when there is at least one first distance greater than a first threshold; or determine the point as an obstacle point when the height difference is greater than the second threshold and the second distance is less than a third threshold; determine the point as a non-obstacle point when there is no first distance greater than the first threshold, the height difference is less than or equal to the second threshold, and the second distance is greater than or equal to the third threshold.

[0124] In some embodiments, the map modeling module 4552 is further configured to project obstacle points and non-obstacle points onto a two-dimensional grid map, determine the number of obstacle points in each grid of the two-dimensional grid map; for each grid, determine the occupancy probability of the grid based on the number of obstacle points in the grid; when the occupancy probability is less than or equal to a fourth threshold, determine the grid as a non-obstacle grid; and determine grids without obstacle points or non-obstacle points as unknown grids.

[0125] In some embodiments, the map modeling module 4552 is further configured to sum and average the heights of each point in the non-obstruction grid to obtain a first height corresponding to the non-obstruction grid; for each unknown grid, determine at least one target non-obstruction grid adjacent to the unknown grid, and determine the weight corresponding to at least one target non-obstruction grid based on the distance between the unknown grid and at least one target non-obstruction grid; perform weighted processing on the heights of at least one target non-obstruction grid based on the weights corresponding to at least one target non-obstruction grid to obtain a second height corresponding to the unknown grid; and generate an elevation map corresponding to the point cloud data based on the first height and the second height.

[0126] In some embodiments, the incremental mapping module 4553 is further configured to determine, from the elevation maps corresponding to the point cloud data at the first moment and the point cloud data at the second moment, the first moment and the second moment are adjacent and the first moment is less than the second moment; based on the height of each grid in the second elevation map, the height of the corresponding grid in the first elevation map is updated to obtain a global elevation map.

[0127] In some embodiments, the terrain feature data includes slope, elevation difference, and roughness. The feature extraction module 4554 is further configured to, for each grid, determine the grid's access risk score as a first value when the grid's slope is greater than a fifth threshold, or the grid's elevation difference is greater than a sixth threshold, or the grid's roughness is greater than a seventh threshold; and determine the grid's access risk score as a second value when the grid's slope is less than or equal to the fifth threshold, the grid's elevation difference is less than or equal to the sixth threshold, and the grid's roughness is less than or equal to the seventh threshold.

[0128] This application provides a computer program product, which includes computer-executable instructions or a computer program stored in a computer-readable storage medium. The processor of an electronic device reads the computer-executable instructions or computer program from the computer-readable storage medium and executes the computer-executable instructions or computer program, causing the electronic device to perform the map data generation method provided in this application.

[0129] This application provides a computer-readable storage medium storing computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the processor will execute the map data generation method provided in this application. For example, ... Figure 3A The map data generation method is shown.

[0130] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0131] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0132] As an example, computer-executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0133] As an example, computer-executable instructions can be deployed to execute on a single electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0134] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. A method for generating map data, characterized in that, The method includes: Acquire point cloud data and motion state data, and fuse the point cloud data and motion state data to obtain pose estimation data; Based on the pose estimation data and the point cloud data, ground elevation modeling is performed to obtain the elevation map corresponding to the point cloud data. Incremental mapping is performed on the elevation maps corresponding to the point cloud data from multiple frames to obtain a global elevation map. Feature extraction is performed on the global elevation map to obtain the terrain feature data of each grid in the global elevation map, and the access risk score of each grid is determined based on the terrain feature data of each grid. Based on the access risk score of each grid, the global elevation map is updated to obtain the target map.

2. The method according to claim 1, characterized in that, The step of performing ground elevation modeling based on the pose estimation data and the point cloud data to obtain an elevation map corresponding to the point cloud data includes: Based on the motion state data, the point cloud data is distorted to obtain distorted point cloud data; From each point in the corrected point cloud data, identify obstacle points and non-obstacle points; Project the obstacle points and non-obstacle points onto a two-dimensional grid map to determine the non-obstacle grids and unknown grids in the two-dimensional grid map; Based on the pose estimation data, the corrected point cloud data is converted to the world coordinate system to obtain the converted point cloud data. Determine the height information of each point located in the non-obstacle grid in the transformed point cloud data, and generate an elevation map corresponding to the point cloud data based on the height information of each point.

3. The method according to claim 2, characterized in that, The step of determining obstacle points and non-obstacle points from each point in the corrected point cloud data includes: For each point in the corrected point cloud data, determine the neighboring points of the point in the horizontal direction, and determine the first distance between the neighboring points and any two adjacent points of the point. Determine the adjacent points that are vertically adjacent to the point, and determine the height difference and second distance between the point and the adjacent points; When at least one of the first distances is greater than the first threshold, the point is determined to be an obstacle point; Alternatively, if the height difference is greater than the second threshold and the second distance is less than the third threshold, the point is determined to be an obstacle point. When there is no point where the first distance is greater than the first threshold, the height difference is less than or equal to the second threshold, and the second distance is greater than or equal to the third threshold, the point is determined to be a non-obstacle point.

4. The method according to claim 2, characterized in that, The step of projecting the obstacle points and non-obstacle points onto a two-dimensional grid map, and determining the non-obstacle grids and unknown grids in the two-dimensional grid map, includes: The obstacle points and the non-obstacle points are projected onto the two-dimensional grid map to determine the number of obstacle points in each grid of the two-dimensional grid map; For each grid, the occupancy probability of the grid is determined based on the number of obstacle points in the grid; When the occupancy probability is less than or equal to the fourth threshold, the grid is determined to be a non-obstacle grid; Mesh grids lacking both the stated obstacle points and the stated non-obstacle points are defined as unknown meshes.

5. The method according to claim 2, characterized in that, The step of generating an elevation map corresponding to the point cloud data based on the height information of each point includes: The first height corresponding to the non-obstacle grid is obtained by summing and averaging the heights of each point in the non-obstacle grid. For each unknown grid, at least one target non-obstacle grid adjacent to the unknown grid is determined, and based on the distance between the unknown grid and at least one target non-obstacle grid, the weight corresponding to at least one target non-obstacle grid is determined; Based on the weights corresponding to at least one of the target non-obstacle grids, the heights of at least one of the target non-obstacle grids are weighted to obtain the second height corresponding to the unknown grid. Based on the first altitude and the second altitude, an elevation map corresponding to the point cloud data is generated.

6. The method according to any one of claims 1 to 5, characterized in that, The incremental mapping process based on the elevation maps corresponding to multiple frames of the point cloud data to obtain a global elevation map includes: From the elevation maps corresponding to the point cloud data in multiple frames, determine the first elevation map corresponding to the point cloud data at the first time and the second elevation map corresponding to the point cloud data at the second time, wherein the first time and the second time are adjacent and the first time is less than the second time; Based on the height of each grid in the second elevation map, the height of the corresponding grid in the first elevation map is updated to obtain a global elevation map.

7. The method according to any one of claims 1 to 5, characterized in that, The terrain feature data includes slope, elevation difference, and roughness. The determination of the accessibility risk score for each grid based on the terrain feature data for each grid includes: For each of the grids, when the slope of the grid is greater than a fifth threshold, or the elevation difference of the grid is greater than a sixth threshold, or the roughness of the grid is greater than a seventh threshold, the passage risk score of the grid is determined to be a first value. When the slope of the grid is less than or equal to the fifth threshold, the elevation difference of the grid is less than or equal to the sixth threshold, and the roughness of the grid is less than or equal to the seventh threshold, the passage risk score of the grid is determined to be the second value.

8. A map data generation device, characterized in that, The device includes: The data fusion module is used to acquire point cloud data and motion state data, and fuse the point cloud data and motion state data to obtain pose estimation data; The map modeling module is used to perform ground elevation modeling based on the pose estimation data and the point cloud data to obtain an elevation map corresponding to the point cloud data. The incremental mapping module is used to perform incremental mapping processing based on the elevation map corresponding to multiple frames of point cloud data to obtain a global elevation map. The feature extraction module is used to extract features from the global elevation map to obtain the terrain feature data of each grid in the global elevation map, and to determine the passage risk score of each grid based on the terrain feature data of each grid. The map update module is used to update the global elevation map based on the access risk score of each grid to obtain the target map.

9. An electronic device, characterized in that, The electronic device includes: Memory is used to store executable instructions or computer programs. A processor, when executing computer-executable instructions or computer programs stored in the memory, implements the map data generation method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions or a computer program, characterized in that, When the computer-executable instructions or computer program are executed by a processor, the map data generation method according to any one of claims 1 to 7 is implemented.

Images