High-precision map construction method and device, electronic equipment and storage medium

Abstract

The invention relates to a high-precision map construction method and device, electronic equipment and a storage medium, and is applied to the technical field of automatic driving, and the method comprises the steps: collecting inertial measurement unit data, wheel speed meter data and image data around a vehicle in a driving process of the vehicle in a park; the image data, the inertial measurement unit data and the wheel speed meter data are processed by using a tight coupling synchronous positioning and mapping algorithm, and the position coordinates of the vehicle are determined; performing feature extraction on the image data by using a pre-trained deep learning model to obtain geometric features and semantic features of each element in pre-constructed N target layers; the N target layers are constructed based on the original layers, and N is smaller than the number of the original layers; and constructing high-precision map data based on the geometric features and semantic features of each element and the position coordinates of the vehicle. According to the invention, the efficiency of high-precision map construction can be improved, and the cost of high-precision map construction can be reduced.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and in particular to high-precision map construction methods, devices, electronic devices and storage media. Background Technology

[0002] High-precision maps, as a key component of autonomous driving systems, provide vehicles with prior environmental information with centimeter-level accuracy, forming the foundation for precise positioning, reliable perception, and safe planning. In open road scenarios, high-precision maps are generated based on point cloud data and motion trajectory information collected by multi-sensor fusion (LiDAR, GDS, inertial measurement units, etc.), through complex correlation calculations and parameter optimization, resulting in high-precision maps rich in semantic information. These maps typically contain more than 15 semantic layers (such as lane lines, traffic signs, and curbs) to cope with complex and ever-changing traffic environments.

[0003] As autonomous driving technology expands into specific scenarios, especially with the increasing application of low-speed driverless sanitation vehicles in industrial parks (including parks, factories, and municipal roads), the demand for high-precision maps is showing significant differences. For example, industrial park scenarios have relatively simple road structures, limited traffic element types, unstable GPS (Global Positioning System) signals, a lack of clear curb guidance, and mixed pedestrian and vehicle traffic. Directly applying high-precision map construction methods from open road scenarios to industrial park scenarios would generate a large amount of redundant data. Furthermore, it would require expensive specialized data acquisition equipment and significant time for manual editing and quality control, resulting in long map production cycles and high costs. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a high-precision map construction method, apparatus, electronic device, storage medium, and computer program product.

[0005] According to a first aspect of this application, a method for constructing a high-precision map is provided, comprising: The system collects inertial measurement unit data, wheel speed meter data, and image data of the vehicle's surroundings as it moves within the park. Using a tightly coupled synchronous positioning and mapping algorithm, the image data, inertial measurement unit data, and wheel speedometer data are processed to determine the vehicle's position coordinates; The image data is used to extract features using a pre-trained deep learning model to obtain the geometric and semantic features of each element in N pre-constructed target layers; the N target layers are constructed based on the original layers, and N is less than the number of original layers; Based on the geometric and semantic features of each element, as well as the vehicle's location coordinates, high-precision map data is constructed.

[0006] Optionally, the high-precision map construction method further includes: Based on the relationships between the various elements, a spatial topological relationship is constructed between the various elements so that the vehicle can perform trajectory planning based on the topological relationship and the high-precision map data.

[0007] Optionally, the high-precision map construction method further includes: After constructing the high-precision map data, the high-precision map data is checked using a park rule base pre-constructed based on park scene knowledge; When a defect is determined in the high-precision map data, a corresponding defect warning message is displayed.

[0008] Optionally, the high-precision map construction method further includes: Collect point cloud data around the vehicle while it is driving within the park; The process involves using a pre-trained deep learning model to extract features from the image data, obtaining the geometric and semantic features of each element in N pre-constructed target layers, including: The image data and point cloud data are used to extract features using a pre-trained deep learning model to obtain the geometric and semantic features of each element in the pre-constructed N target layers.

[0009] Optionally, N is 5, and the 5 target layers include: road boundaries, passable areas, restricted areas, speed bumps, and entrances / exits.

[0010] According to a second aspect of this application, a high-precision map building apparatus is provided, comprising: The data acquisition module is used to collect inertial measurement unit data, wheel speed meter data, and image data around the vehicle during its operation within the park. The vehicle positioning module is used to process the image data, inertial measurement unit data and wheel speedometer data using a tightly coupled synchronous positioning and mapping algorithm to determine the vehicle's position coordinates. The layer feature extraction module is used to extract features from the image data using a pre-trained deep learning model to obtain the geometric and semantic features of each element in N pre-constructed target layers; the N target layers are constructed based on the original layers, and N is less than the number of original layers; The high-precision map data generation module is used to construct high-precision map data based on the geometric and semantic features of each element and the position coordinates of the vehicle.

[0011] Optionally, the high-precision map building device further includes: The spatial topology relationship construction module is used to construct the spatial topology relationship between the various elements based on the association relationship between the various elements, so that the vehicle can perform trajectory planning based on the topology relationship and the high-precision map data.

[0012] Optionally, the high-precision map building device further includes: The map detection module is used to check the high-precision map data after it has been constructed, using a park rule base pre-built based on park scene knowledge. The defect alert module is used to display corresponding defect alert information when it is determined that there is a defect in the high-precision map data.

[0013] Optionally, the data acquisition module is also used to collect point cloud data around the vehicle while it is driving in the park; The layer feature extraction module is specifically used to extract features from the image data and point cloud data using a pre-trained deep learning model, so as to obtain the geometric and semantic features of each element in the pre-constructed N target layers.

[0014] Optionally, N is 5, and the 5 target layers include: road boundaries, passable areas, restricted areas, speed bumps, and entrances / exits.

[0015] According to a third aspect of this application, an electronic device is provided, comprising: a processor configured to execute a computer program stored in a memory, wherein the computer program, when executed by the processor, implements the method described in the first aspect.

[0016] According to a fourth aspect of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect.

[0017] According to a fifth aspect of this application, a computer program product is provided that, when the computer program product is run on a computer, causes the computer to perform the method described in the first aspect.

[0018] The technical solution provided in this application has the following advantages compared with the prior art: Data is collected using a low-cost sensor combination (such as binocular cameras, inertial measurement units, and wheel speed sensors). The vehicle's position coordinates are calculated, and a pre-trained deep learning model extracts the geometric and semantic features of each element in N target layers from the image data, thus constructing high-precision map data. Even in park environments with weak GPS signals, positioning accuracy is guaranteed, providing accurate geographic coordinates for elements. The N target layers constructed from the original layers are suitable for park scenarios, and N is less than the number of original layers, simplifying the number of layers and reducing the workload of data collection and processing. Combined with automated mapping tools, the mapping cycle and cost of high-precision park maps can be reduced, thus meeting the urgent need for automated mapping due to frequent park map updates. The simplified and standardized layer structure facilitates end-to-end element recognition and annotation by AI algorithms, providing a clear data interface and process foundation for subsequent integration with fully automated recognition models, lowering the technical threshold for achieving fully automated mapping in the future. The simplified map data is smaller, and transmission and loading are faster, making it ideal for vehicle-mounted platforms with limited computing and storage resources. This enables low-speed sanitation vehicles to perform real-time positioning and planning more efficiently. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of a high-precision map construction method in an embodiment of this application; Figure 2 This is a schematic diagram of the high-precision map building device in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of an electronic device in an embodiment of this application. Detailed Implementation

[0022] To better understand the above-mentioned objectives, features, and advantages of this application, the solution of this application will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of this application, but this application may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some embodiments of this application, and not all embodiments.

[0024] Existing high-precision map construction methods for open road scenarios involve detailed layer divisions, typically including 15 or more layers such as Lane, Boundary, Road, Intersection, Signal, Gate, Pillar, Free area, Prohibited area, Parking, Clear area, Safety island, Stop line, Crosswalk, and Speedbump. Collecting and processing information from 15 layers requires expensive specialized data acquisition equipment (such as multi-line LiDAR and high-precision integrated navigation systems) and significant time spent on manual editing and quality control, resulting in long map production cycles and high costs, which does not meet the needs of rapid updates and cost-sensitive environments in park settings.

[0025] The high-precision map production process, especially in the feature extraction and semantic annotation stages, still heavily relies on the manual judgment and operation of cartographers. This not only restricts the improvement of mapping efficiency but also introduces the risk of subjective errors, making it difficult to meet the urgent need for automated mapping due to frequent updates to park maps. Overly complex map data specifications consume a large amount of storage space and computing resources. For low-speed sanitation vehicles with relatively limited computing power that require real-time route replanning, loading and processing information from 15 layers would create unnecessary computational load and affect system real-time performance.

[0026] Based on this, this application provides a high-precision map construction method that can reduce the number of layers in the high-precision map. While retaining the necessary navigation and operational information for low-speed autonomous sanitation scenarios in the park, it significantly reduces the complexity and cost of map data collection, processing, and updating, reduces reliance on specialized data collection equipment and manual intervention, and improves the production efficiency and consistency of high-precision maps for the park. It ensures that the optimized map specifications accurately support the special needs of sanitation operations, such as automatic cleaning, avoidance of restricted areas, and identification of specific facilities, and is compatible with the computing platform of low-speed sanitation vehicles, reducing the resource overhead of real-time processing while ensuring safety.

[0027] See Figure 1 , Figure 1 This is a flowchart of a high-precision map construction method in an embodiment of this application, which may include the following steps: Step S102: Collect inertial measurement unit data, wheel speed meter data, and image data of the vehicle's surroundings during its operation within the park.

[0028] The vehicle is equipped with binocular cameras, an inertial measurement unit (IMU), and wheel speed sensors, which are used to collect image data of the vehicle's surroundings, IMU data, and wheel speed data, respectively. When constructing a high-precision map, the vehicle travels through various areas within the park to cover all areas, collecting the aforementioned sensor data during its movement.

[0029] Step S104: Using a tightly coupled synchronous positioning and mapping algorithm, the image data, inertial measurement unit data, and wheel speedometer data are processed to determine the vehicle's position coordinates.

[0030] Tightly coupled Simultaneous Localization and Mapping (SLAM) algorithms achieve autonomous localization and environmental mapping by deeply fusing raw data from multiple sensors. This method directly incorporates measurement data from different sensors into a unified optimization framework, rather than processing them independently and then fusing the results, thus fully leveraging the complementary advantages of the sensors. Even in park environments with weak GPS signals, it ensures positioning accuracy, providing accurate geographic coordinates for various elements in subsequent target layers.

[0031] Step S106: Use a pre-trained deep learning model to extract features from the image data to obtain the geometric and semantic features of each element in the pre-constructed N target layers.

[0032] This application, through in-depth analysis of the park scene, reconstructs the scene based on 15 original layers, resulting in N target layers. N is less than the number of original layers, meaning the original layers are simplified into N core layers suitable for the park scene. Optionally, N is 5, and the 5 target layers include: road boundaries, passable areas, restricted areas, speed bumps, and entrances / exits. A layer simplification comparison table is as follows: ; Among them, Boundary (road boundary): describes the boundary of the driveway and sidewalk within the park, and is the basis for path planning.

[0033] Free_area (accessible area): Identifies open areas where vehicles can safely pass, such as plazas and parking lots, enhancing the flexibility of route planning.

[0034] Prohibited_area: Clearly marked areas such as flower beds, pools, equipment areas, and densely populated pedestrian areas where vehicles are prohibited from entering, ensuring safety.

[0035] Speedbump: Precisely marks the location and geometry of speed bumps, allowing vehicles to decelerate smoothly in advance.

[0036] Gate (Entrance / Exit): Defines key connection points such as entrances / exits and turnstiles in the park, used for global path start and end point planning.

[0037] The optimized map specifications are suitable for sanitation scenarios in industrial parks. They can ensure basic navigation through Boundary and Gate, and ensure operational safety and precision through Prohibited_area, Speedbump, and Free_area, thereby improving the reliability and efficiency of the entire sanitation autonomous driving system.

[0038] Deep learning models, such as image segmentation networks based on DeepLabV3+, are trained as follows: Multiple sample images are acquired and used as input data. Geometric and semantic information of each element in each of the N target layers is extracted from each sample image, and this information is used as label data. An image segmentation network based on DeepLabV3+ is selected, and the neural network is trained using the input and label data to generate a deep learning model. This deep learning model is then used to extract features from image data collected from vehicles, yielding the geometric and semantic features of each element in the N target layers.

[0039] Optionally, in addition to collecting image data of the area surrounding the vehicle while it is driving within the park, point cloud data of the area surrounding the vehicle can also be collected. A pre-trained deep learning model is used to extract features from the image and point cloud data, obtaining the geometric and semantic features of each element in N pre-constructed target layers. By combining the image and point cloud data, the accuracy of the extracted geometric and semantic features of each element in the N target layers can be improved. Accordingly, the deep learning model can be an image segmentation network and a point cloud segmentation network based on DeepLabV3+.

[0040] Step S108: Based on the geometric and semantic features of each element, as well as the vehicle's location coordinates, construct high-precision map data.

[0041] The geometric and semantic features of each element identified from the image data are located in the image coordinate system. By using camera extrinsic parameters, the coordinates can be transformed into the vehicle coordinate system. Based on the vehicle's position coordinates, the coordinates can be further transformed into the map coordinate system, thereby constructing high-definition map data.

[0042] Optionally, after constructing the high-precision map data, a pre-built park rule base based on park scene knowledge can be used to inspect the high-precision map data. Specifically, it can check layer integrity (e.g., whether each boundary is closed), attribute completeness (e.g., whether each speedbump contains a height attribute), and topological consistency (e.g., Free_area should not overlap with Prohibited_area). When defects are found in the high-precision map data, corresponding defect prompts are displayed.

[0043] When using high-precision map data for autonomous driving planning algorithms, it's necessary to understand the relationships between map layers to plan vehicle detours. Therefore, spatial topological relationships between various elements can be constructed based on these relationships, enabling vehicles to plan routes based on these topological connections and the high-precision map data. For example, algorithms can be designed to automatically establish spatial topological relationships between the boundary area and the free and prohibited areas. By automatically identifying the relationship where a free area is surrounded by a boundary area, a connected, drivable road network can be generated. Sanitation vehicles equipped with this high-precision map data can then operate based on these spatial topological relationships between various elements.

[0044] In this embodiment, a dedicated compiler can also be developed to employ hierarchical encoding and instantiation techniques for the data of N target layers. For example, a geometrically simple Speedbump can be parameterized, while only one copy of the data is stored for recurring Gate models. The output format is compatible with mainstream autonomous driving frameworks such as Autoware and Apollo, and the data size is significantly reduced.

[0045] The high-precision map construction method of this application uses a low-cost sensor combination (e.g., binocular cameras, inertial measurement units, and wheel speed sensors) to collect data. It calculates the vehicle's position coordinates and uses a pre-trained deep learning model to extract the geometric and semantic features of each element in N target layers from the image data, thereby constructing high-precision map data. Even in park environments with weak GPS signals, positioning accuracy is guaranteed, providing accurate geographic coordinates for elements. The N target layers constructed based on the original layers are suitable for park scenarios, and N is less than the number of original layers, simplifying the number of layers and reducing the workload of data collection and processing. Combined with automated mapping tools, the mapping cycle and cost of high-precision park maps can be reduced, thus meeting the urgent need for automated mapping due to frequent park map updates. The simplified and standardized layer structure is more conducive to end-to-end element recognition and annotation by AI algorithms, providing a clear data interface and process foundation for subsequent integration with fully automated recognition models, lowering the technical threshold for achieving fully automated mapping in the future. The simplified map data is smaller, and transmission and loading are faster, making it very suitable for in-vehicle platforms with limited computing and storage resources. This enables low-speed sanitation vehicles to perform real-time positioning and planning more efficiently.

[0046] It should be noted that although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0047] Corresponding to the above method embodiments, this application also provides a high-precision map construction device, see [link to relevant documentation]. Figure 2 The high-precision map building device 200 includes: The data acquisition module 202 is used to collect inertial measurement unit data, wheel speed meter data, and image data around the vehicle during its operation within the park. The vehicle positioning module 204 is used to process image data, inertial measurement unit data and wheel speedometer data using a tightly coupled synchronous positioning and mapping algorithm to determine the vehicle's position coordinates. The layer feature extraction module 206 is used to extract features from image data using a pre-trained deep learning model to obtain the geometric and semantic features of each feature in N pre-constructed target layers; the N target layers are constructed based on the original layers, and N is less than the number of original layers; The high-precision map data generation module 208 is used to construct high-precision map data based on the geometric and semantic features of each element and the location coordinates of the vehicle.

[0048] Optionally, the high-precision map building device 200 also includes: The spatial topology relationship construction module is used to construct spatial topology relationships between various elements based on their associations, so that vehicles can perform trajectory planning based on topology relationships and high-precision map data.

[0049] Optionally, the high-precision map building device 200 also includes: The map inspection module is used to inspect the high-precision map data after it has been constructed, using a park rule base that has been pre-built based on park scene knowledge. The defect alert module is used to display corresponding defect alert information when defects are found in high-precision map data.

[0050] Optionally, the data acquisition module 202 is also used to collect point cloud data around the vehicle while it is driving in the park. The layer feature extraction module is specifically used to extract features from image data and point cloud data using a pre-trained deep learning model, so as to obtain the geometric and semantic features of each element in N pre-constructed target layers.

[0051] Optionally, N is 5, and the 5 target layers include: road boundaries, passable areas, restricted areas, speed bumps, and entrances / exits.

[0052] The specific details of each module or unit in the above-mentioned device have been described in detail in the corresponding methods, so they will not be repeated here.

[0053] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0054] This application also provides an electronic device, including: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the high-precision map construction method described in this example embodiment.

[0055] Reference Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device in an embodiment of this application. The specific embodiments of this application do not limit the specific implementation of the electronic device.

[0056] like Figure 3As shown, the electronic device may include: a processor 302, a communication interface 304, a memory 306, and a communication bus 308.

[0057] The processor 302, communication interface 304, and memory 306 communicate with each other via communication bus 308.

[0058] Communication interface 304 is used to communicate with other electronic devices or servers.

[0059] The processor 302 is used to execute program 310, specifically the relevant steps in the above method embodiments.

[0060] Specifically, program 310 may include program code that includes computer operation instructions.

[0061] Processor 302 may be a central processing unit, a specific integrated circuit, or one or more integrated circuits configured to implement the embodiments of this application. The one or more processors included in the smart device may be processors of the same type, such as one or more CPUs; or they may be processors of different types, such as one or more CPUs and one or more ASICs.

[0062] Memory 306 is used to store program 310. Memory 306 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0063] Specifically, program 310 can be used to cause processor 302 to execute the steps in the above-described high-precision map construction method embodiment.

[0064] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the devices and modules described above can be referred to the corresponding process descriptions in the foregoing method embodiments, and will not be repeated here.

[0065] In this embodiment of the application, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the above-described high-precision map construction method.

[0066] It should be noted that the computer-readable storage medium shown in this application can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable storage medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, radio frequency, etc., or any suitable combination thereof.

[0067] In this embodiment of the application, a computer program product is also provided, which, when run on a computer, causes the computer to execute the above-described high-precision map construction method.

[0068] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0069] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for constructing high-precision maps, characterized in that, include: The system collects inertial measurement unit data, wheel speed meter data, and image data of the vehicle's surroundings as it moves within the park. Using a tightly coupled synchronous positioning and mapping algorithm, the image data, inertial measurement unit data, and wheel speedometer data are processed to determine the vehicle's position coordinates; The image data is used to extract features using a pre-trained deep learning model to obtain the geometric and semantic features of each element in N pre-constructed target layers; the N target layers are constructed based on the original layers, and N is less than the number of original layers; Based on the geometric and semantic features of each element, as well as the vehicle's location coordinates, high-precision map data is constructed.

2. The method according to claim 1, characterized in that, The method further includes: Based on the relationships between the various elements, a spatial topological relationship is constructed between the various elements so that the vehicle can perform trajectory planning based on the topological relationship and the high-precision map data.

3. The method according to claim 1, characterized in that, The method further includes: After constructing the high-precision map data, the high-precision map data is checked using a park rule base pre-constructed based on park scene knowledge; When a defect is determined in the high-precision map data, a corresponding defect warning message is displayed.

4. The method according to claim 1, characterized in that, The method further includes: Collect point cloud data around the vehicle while it is driving within the park; The process involves using a pre-trained deep learning model to extract features from the image data, obtaining the geometric and semantic features of each element in N pre-constructed target layers, including: The image data and point cloud data are used to extract features using a pre-trained deep learning model to obtain the geometric and semantic features of each element in the pre-constructed N target layers.

5. The method according to claim 1, characterized in that, N is 5, and the 5 target layers include: road boundaries, passable areas, restricted areas, speed bumps, and entrances / exits.

6. A high-precision map building device, characterized in that, include: The data acquisition module is used to collect inertial measurement unit data, wheel speed meter data, and image data around the vehicle during its operation within the park. The vehicle positioning module is used to process the image data, inertial measurement unit data and wheel speedometer data using a tightly coupled synchronous positioning and mapping algorithm to determine the vehicle's position coordinates. The layer feature extraction module is used to extract features from the image data using a pre-trained deep learning model to obtain the geometric and semantic features of each element in N pre-constructed target layers; the N target layers are constructed based on the original layers, and N is less than the number of original layers; The high-precision map data generation module is used to construct high-precision map data based on the geometric and semantic features of each element and the position coordinates of the vehicle.

7. The apparatus according to claim 6, characterized in that, The device further includes: The spatial topology relationship construction module is used to construct the spatial topology relationship between the various elements based on the association relationship between the various elements, so that the vehicle can perform trajectory planning based on the topology relationship and the high-precision map data.

8. The apparatus according to claim 6, characterized in that, The device further includes: The map detection module is used to check the high-precision map data after it has been constructed, using a park rule base pre-built based on park scene knowledge. The defect alert module is used to display corresponding defect alert information when it is determined that there is a defect in the high-precision map data.

9. The apparatus according to claim 6, characterized in that, The data acquisition module is also used to collect point cloud data around the vehicle while it is driving in the park. The layer feature extraction module is specifically used to extract features from the image data and point cloud data using a pre-trained deep learning model, so as to obtain the geometric and semantic features of each element in the pre-constructed N target layers.

10. The apparatus according to claim 6, characterized in that, N is 5, and the 5 target layers include: road boundaries, passable areas, restricted areas, speed bumps, and entrances / exits.

Images