Road topology construction method and device, vehicle, medium, program product and chip

Abstract

This disclosure relates to a road topology construction method, apparatus, vehicle, medium, program product, and chip, and pertains to the field of intelligent driving technology. The method includes: in response to a vehicle being in a parking scenario, acquiring surrounding environmental data corresponding to the vehicle; acquiring a first parking area map corresponding to the vehicle, the first parking area map including first parking map elements; performing element recognition processing on the surrounding environmental data to obtain second parking map elements; updating the first parking area map based on the second parking map elements to obtain a second parking area map; and constructing a target road topology based on the second parking area map. This disclosure continuously updates the parking area map based on multi-frame fusion perception results and outputs a stable road topology based on the parking area map, enabling the vehicle to have a clearer road understanding ability in parking scenarios and to control the vehicle to drive autonomously within the parking area, thereby improving vehicle parking efficiency.

Description

Technical Field

[0001] This disclosure relates to the field of intelligent driving technology, and more specifically, to a road topology construction method, a road topology construction device, a vehicle, a storage medium, a computer program product, and a chip. Background Technology

[0002] Road topology constructs a structured model of the road network by analyzing the geometric positions and connections of elements such as lane lines, traffic lights, and intersections. In intelligent driving scenarios, road topology can be used to describe the connections, adjacencies, and intersections between lanes, and also to describe the permitted travel paths for vehicles.

[0003] In relevant solutions, road topology mostly exists within the vehicle's driving scenario. Some common solutions combine perceived road information such as vehicle trajectory data, perception detection, or laser point clouds to construct the topology map. With the continuous advancement of intelligent driving technology, intelligent parking functions are gradually developing towards higher levels (such as autonomous parking), which also require topology mapping based on perceived road information. Summary of the Invention

[0004] To overcome the problems existing in related technologies, this disclosure provides a road topology construction method, a road topology construction device, a vehicle, a storage medium, a computer program product, and a chip.

[0005] According to a first aspect of the present disclosure, a road topology construction method is provided. The method includes: in response to a vehicle being in a parking scenario, acquiring surrounding environmental data corresponding to the vehicle; acquiring a first parking area map corresponding to the vehicle, the first parking area map including first parking map elements; performing element recognition processing on the surrounding environmental data to obtain second parking map elements; updating the first parking area map based on the second parking map elements to obtain a second parking area map; and constructing a target road topology based on the second parking area map. This disclosure continuously updates the parking area map based on multi-frame fusion perception results, updates the parking area map through surrounding environmental data, and outputs a stable road topology, enabling vehicles to have a clearer and more accurate road understanding ability in parking scenarios; and can control the vehicle to drive autonomously in the parking area based on the target road topology, thereby improving vehicle parking efficiency.

[0006] In some exemplary embodiments of this disclosure, the step of performing element recognition processing on the surrounding environmental data to obtain a second parking map element includes: inputting the surrounding environmental data into a pre-trained element recognition model, the element recognition model including a centerline recognition sub-model, an intersection recognition sub-model, and a parking space recognition sub-model; performing centerline recognition processing on the surrounding environmental data by the centerline recognition sub-model to obtain a centerline recognition result; performing intersection recognition processing on the surrounding environmental data by the intersection recognition sub-model to obtain an intersection recognition result; performing parking space recognition processing on the surrounding environmental data by the parking space recognition sub-model to obtain a parking space recognition result; and determining the second parking map element based on the centerline recognition result, the intersection recognition result, and the parking space recognition result. The second parking map element determined through the above steps can be used as the data basis for map updates to update the first parking map element in the first parking area map.

[0007] In some exemplary embodiments of this disclosure, the first parking map element includes one or more of a first centerline, a first intersection, and a first parking space; the second parking map element includes one or more of a second centerline, a second intersection, and a second parking space. These parking map elements can be used to construct a parking area map, enabling vehicles to construct continuous roads, thereby providing a data foundation for subsequently outputting a stable road topology.

[0008] In some exemplary embodiments of this disclosure, the first parking area map includes at least a first centerline, and the second parking map element includes at least a second centerline. Updating the first parking area map based on the second parking map element to obtain a second parking area map includes: performing centerline matching between the second centerline and the first centerline to obtain a centerline matching result; responding to the centerline matching result indicating the existence of a first centerline matching the second centerline, fusing and updating the first centerline matching the second centerline based on the second centerline to obtain an updated matching centerline; and responding to the centerline matching result indicating the absence of a first centerline matching the second centerline, using the second centerline as a newly added centerline to obtain the second parking area map. Based on the centerline information in the multi-frame fusion perception results, the stability of road detection results can be improved.

[0009] In some exemplary embodiments of this disclosure, the step of fusing and updating the first centerline matching the second centerline based on the second centerline to obtain an updated matching centerline includes: determining the number of centerline data corresponding to the first centerline matching the second centerline based on the second centerline; in response to the number of centerline data reaching a first threshold, acquiring road point cloud data corresponding to the first centerline matching the second centerline; generating a road point cloud set based on the road point cloud data; and updating the first centerline matching the second centerline in the first parking area map based on the road point cloud set and road generation conditions to obtain the updated matching centerline. Through the above steps, a smooth, continuous, and high-precision matching centerline can be obtained for subsequent generation of the target road topology.

[0010] In some exemplary embodiments of this disclosure, the first parking area map includes at least a first intersection, and the second parking map element includes at least a second intersection. Updating the first parking area map based on the second parking map element to obtain a second parking area map includes: performing intersection matching between the second intersection and the first intersection based on the intersection center point to obtain an intersection matching result; responding to the intersection matching result indicating the existence of a first intersection matching the second intersection, performing a fusion update on the first intersection matching the second intersection based on the second intersection to obtain an updated matching intersection; and responding to the intersection matching result indicating the absence of a first intersection matching the second intersection, adding the second intersection as a new intersection to obtain the second parking area map. Constructing the second parking area map based on intersection information from multi-frame fusion perception results can improve the stability of road detection results.

[0011] In some exemplary embodiments of this disclosure, the step of fusing and updating the first intersection matched with the second intersection based on the second intersection to obtain an updated matched intersection includes: determining the number of intersection data corresponding to the first intersection matched with the second intersection based on the second intersection; in response to the number of intersection data reaching a second threshold, determining at least one intersection center point and intersection boundary information corresponding to the intersection center point based on historical data of the first intersection matched with the second intersection; and updating the first intersection matched with the second intersection on the first parking area map based on the intersection center point and the intersection boundary information to obtain an updated matched intersection. Updating the intersection information based on the intersection center point and boundary information determined from the observation data can yield more accurate intersection modeling results.

[0012] In some exemplary embodiments of this disclosure, the first parking area map includes at least a first parking space, and the second parking map element includes at least a second parking space. Updating the first parking area map based on the second parking map element to obtain a second parking area map includes: determining the intersection-union ratio (IUU) between the second parking space and the first parking space to obtain a parking space matching result; responding to the parking space matching result indicating the existence of a first parking space matching the second parking space, fusing and updating the first parking space matching the second parking space based on the second parking space to obtain an updated matching parking space; and responding to the parking space matching result indicating the absence of a first parking space matching the second parking space, adding the second parking space as a new parking space to obtain the second parking area map. Updating the second parking area map based on parking space information from multi-frame fusion perception results can improve the stability of road detection results.

[0013] In some exemplary embodiments of this disclosure, the step of fusing and updating the first storage location matching the second storage location based on the second storage location to obtain an updated matching storage location includes: determining the number of storage location data points for the first storage location matching the second storage location based on the second storage location; in response to the number of storage location data points reaching a third threshold, determining the coordinates of at least one storage location corner point based on historical data of the first storage location matching the second storage location; and updating the first storage location matching the second storage location on the first parking area map based on the coordinates of the storage location corner point corresponding to the storage location, thereby obtaining an updated matching storage location. Updating the storage location information based on the storage location corner point coordinates determined from the observation data can yield more accurate storage location modeling results.

[0014] In some exemplary embodiments of this disclosure, constructing a target road topology based on the second parking area map includes: constructing topology nodes based on the second parking area map, the topology nodes including intersection nodes and parking space nodes; obtaining the centerline of the second parking area map, and constructing topology edges based on the connection relationship between the centerline and the intersection nodes or parking space nodes; and constructing the target road topology based on the intersection nodes, the parking space nodes, and the topology edges. The constructed target road topology can be used in subsequent vehicle path planning and other decisions, enabling vehicles to understand roads in parking functions.

[0015] In some exemplary embodiments of this disclosure, the step of constructing a target road topology based on the second parking area map further includes: constructing a first road topology based on the second parking area map; performing anomaly detection on the first road topology to obtain anomaly detection results; and performing anomaly repair on the first road topology based on the anomaly detection results to obtain the target road topology. Through the anomaly detection and anomaly repair steps, an incremental, online approach can be provided to stably fuse discontinuous, noisy real-time sensing data into a globally consistent, geometrically smooth road topology.

[0016] In some exemplary embodiments of this disclosure, the step of performing anomaly repair on the first road topology based on the anomaly detection result to obtain the target road topology includes: determining a group of intersections to be reviewed based on the first road topology, wherein the group of intersections to be reviewed is any two intersections whose distance is less than a first distance threshold; in response to the anomaly detection result indicating that a centerline is missing between the group of intersections to be reviewed, adding a virtual centerline between the group of intersections to be reviewed to obtain the target road topology, which can enhance the continuity of the constructed road topology.

[0017] In some exemplary embodiments of this disclosure, the step of performing anomaly repair on the first road topology based on the anomaly detection result to obtain the target road topology includes: determining a group of centerlines to be verified based on the first road topology, wherein the group of centerlines to be verified is any two centerlines whose distance between their beginning and end is less than a second distance threshold; responding to the anomaly detection result that there is a lack of intersections between the group of centerlines to be verified, determining the intersection positions and intersection boundary information based on the centerlines to be verified; adding virtual intersections based on the intersection positions and intersection boundary information to obtain the target road topology, thereby obtaining a continuous and stable target road topology.

[0018] In some exemplary embodiments of this disclosure, the step of performing anomaly repair on the first road topology based on the anomaly detection result to obtain the target road topology further includes: in response to the anomaly detection result indicating a conflict between centerline information and storage location information, determining the centerline confidence and storage location confidence corresponding to the centerline information and the storage location information respectively; in response to the centerline confidence being greater than or equal to the storage location confidence, deleting conflicting storage locations; and in response to the centerline confidence being less than the storage location confidence, deleting conflicting centerlines. Through the above anomaly repair steps, a more accurate road topology can be obtained.

[0019] In some exemplary embodiments of this disclosure, the method further includes controlling the vehicle to perform parking operations based on the target road topology. By constructing the target road topology, the vehicle gains a clearer and more accurate road understanding in parking scenarios, enabling it to have parking roaming capabilities and explore parking spaces autonomously to perform parking operations.

[0020] According to a second aspect of the present disclosure, a road topology construction apparatus is provided, comprising: an environmental data acquisition module, configured to acquire surrounding environmental data corresponding to a vehicle in response to a parking scenario; a first map acquisition module, configured to acquire a first parking area map corresponding to the vehicle, the first parking area map including first parking map elements; an element recognition module, configured to perform element recognition processing on the surrounding environmental data to obtain second parking map elements; a map update module, configured to update the first parking area map based on the second parking map elements to obtain a second parking area map; and a topology construction module, configured to construct a target road topology based on the second parking area map.

[0021] In some exemplary embodiments of this disclosure, the first parking area map includes at least a first centerline, and the second parking map element includes at least a second centerline; the map update module includes a first map update unit, which is configured to: perform centerline matching between the second centerline and the first centerline to obtain a centerline matching result; in response to the centerline matching result indicating the existence of a first centerline matching the second centerline, perform fusion update on the first centerline matching the second centerline based on the second centerline to obtain an updated matching centerline; and in response to the centerline matching result indicating the absence of a first centerline matching the second centerline, use the second centerline as a newly added centerline to obtain the second parking area map.

[0022] In some exemplary embodiments of this disclosure, the first parking area map includes at least a first intersection, and the second parking map element includes at least a second intersection; the map update module includes a second map update unit, which is configured to: perform intersection matching between the second intersection and the first intersection based on the intersection center point to obtain an intersection matching result; in response to the intersection matching result indicating the existence of a first intersection matching the second intersection, perform a fusion update on the first intersection matching the second intersection based on the second intersection to obtain an updated matching intersection; and in response to the intersection matching result indicating the absence of a first intersection matching the second intersection, treat the second intersection as a newly added intersection to obtain the second parking area map.

[0023] In some exemplary embodiments of this disclosure, the first parking area map includes at least a first parking space, and the second parking map element includes at least a second parking space; the map update module includes a third map update unit, which is configured to: determine the intersection-union ratio (IUU) between the second parking space and the first parking space to obtain a parking space matching result; in response to the parking space matching result indicating the existence of a first parking space matching the second parking space, perform a fusion update on the first parking space matching the second parking space based on the second parking space to obtain an updated matching parking space; and in response to the parking space matching result indicating the absence of a first parking space matching the second parking space, treat the second parking space as a newly added parking space to obtain the second parking area map.

[0024] According to a third aspect of the present disclosure, a vehicle is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to: implement the steps of any of the road topology construction methods.

[0025] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium is provided, wherein when instructions in the storage medium are executed by a processor of a mobile terminal, the mobile terminal is enabled to perform any of the road topology construction methods described herein.

[0026] According to a fifth aspect of the present disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements any of the road topology construction methods described herein.

[0027] According to a sixth aspect of the present disclosure, a chip is provided, the chip including a processor and an interface; wherein the processor is configured to implement the steps of any of the road topology construction methods described in the present disclosure.

[0028] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects: The vehicle continuously updates the parking area map by combining surrounding environmental data. It can continuously update the parking area map based on multi-frame fusion perception results, and output a stable road topology by updating the parking area map with surrounding environmental data. This gives the vehicle a clearer and more accurate road understanding ability in parking scenarios. It can also control the vehicle to drive autonomously in the parking area based on the target road topology, which can improve the vehicle's parking efficiency.

[0029] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0030] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0031] Figure 1 This is a flowchart illustrating a road topology construction method according to an exemplary embodiment of the present disclosure.

[0032] Figure 2 This is a flowchart illustrating a method for performing feature identification processing on surrounding environmental data to obtain second parking map features according to an exemplary embodiment of the present disclosure.

[0033] Figure 3 This is a schematic diagram illustrating parking map elements included in a parking area map according to an exemplary embodiment of the present disclosure.

[0034] Figure 4 This is a flowchart illustrating an exemplary embodiment of the present disclosure of updating a first parking area map based on a second centerline to obtain a second parking area map.

[0035] Figure 5 This is a flowchart illustrating an exemplary embodiment of the present disclosure of updating a first parking area map based on a second intersection to obtain a second parking area map.

[0036] Figure 6 This is a flowchart illustrating an exemplary embodiment of the present disclosure of updating a first parking area map based on a second parking location to obtain a second parking area map.

[0037] Figure 7 This is a flowchart illustrating an exemplary embodiment of the present disclosure of adding a virtual centerline in a first road topology.

[0038] Figure 8 This is a flowchart illustrating an exemplary embodiment of the present disclosure of adding a virtual intersection in a first road topology.

[0039] Figure 9 This is a block diagram illustrating a road topology construction device according to an exemplary embodiment of the present disclosure.

[0040] Figure 10 This is a functional block diagram of a vehicle according to an exemplary embodiment of the present disclosure.

[0041] Figure 11 This is a schematic diagram of the structure of a chip according to an exemplary embodiment of the present disclosure. Detailed Implementation

[0042] Some embodiments of this disclosure will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. Various changes, modifications, and equivalents of the methods, apparatus, and / or systems described herein will become apparent upon understanding this disclosure. For example, the order of operations described herein is merely illustrative and is not limited to those orders set forth herein, but can be changed as will become apparent upon understanding this disclosure, except for operations that must be performed in a particular order. Furthermore, for clarity and brevity, descriptions of features known in the art may be omitted.

[0043] The embodiments described in the following examples of this disclosure are not representative of all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0044] In relevant solutions, road topology is mostly found in driving scenarios. Parking scenarios, however, often lack clear lane lines, have a weaker road concept, and suffer from various obstructions such as walls and access roads, resulting in poor road continuity, poor perception and detection performance, and difficulty in providing stable and consistent results. Furthermore, while the concept of a road is weak in parking scenarios, and there is usually no clear road centerline, road topology is still crucial. It plays a vital role in new features such as crowdsourced mapping and parking roaming. In short, parking scenarios cannot function without road topology.

[0045] The steps of the method in the exemplary embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings and examples.

[0046] Figure 1 This is a flowchart of a road topology construction method according to an exemplary embodiment of the present disclosure. Figure 1 The method described in this embodiment can be applied to vehicles, as well as to servers including local servers and cloud servers, which can be deployed on a single computer or a computer cluster consisting of multiple computers.

[0047] like Figure 1 As shown, in some embodiments, the road topology construction method of this disclosure includes: In step S110, in response to the vehicle being in a parking scenario, the surrounding environment data corresponding to the vehicle is acquired.

[0048] In this embodiment of the disclosure, when a vehicle is detected to be in a parking scenario, surrounding environmental data corresponding to the vehicle is acquired. The surrounding environmental data can be external environmental data of the vehicle collected by sensors and other devices, used to support intelligent driving, intelligent assisted driving, and vehicle operation decisions in parking scenarios. The surrounding environmental data may include, but is not limited to, visual sensor data, LiDAR data, millimeter-wave radar data, ultrasonic radar data, and positioning and pose assistance data.

[0049] For example, visual sensor data can come from front-view / rear-view / surround-view cameras, fisheye cameras, etc., and can include, but are not limited to, RGB images / video streams and depth images (acquired using binocular / time-of-flight (TOF) cameras); the environmental information included includes ground markings (parking lines, road center lines), obstacle outlines (vehicles, pedestrians, pillars), scene semantics (garage entrances, intersections, walls), and lighting and shadows (low light in underground garages, strong light in open-air scenes).

[0050] LiDAR data can come from solid-state / mechanical LiDAR (mostly deployed on the roof or around the vehicle body), and can include, but is not limited to, 3D point cloud data (including the 3D coordinates of points and reflection intensity); the environmental information that can be identified includes the precise outline of static obstacles (pillars, curbs, walls), the 3D boundaries of parking spaces, and the position and movement trend of dynamic targets (pedestrians, moving vehicles), especially suitable for textureless and low-light scenes in underground parking garages.

[0051] Millimeter-wave radar data can come from forward / rear / lateral millimeter-wave radars and can include, but is not limited to, the target's distance, velocity, azimuth angle, and radial acceleration. It is mainly used to detect dynamic targets at medium and long distances (such as vehicles approaching at garage entrances and exits) and to make up for the shortcomings of vision in rain, fog, and strong light.

[0052] Ultrasonic radar data can come from ultrasonic probes around the vehicle (core sensors for parking scenarios), and can include, but are not limited to, the relative distance and orientation of obstacles; it is specifically designed for short-range perception (e.g., 0.1m–5m) and can identify vehicles, curbs, and low obstacles (e.g., parking locks) in front of and behind parking spaces.

[0053] Positioning and pose assistance data can come from inertial measurement units or positioning systems, and may include, but are not limited to, the vehicle's own attitude angles (pitch, roll, yaw), driving speed, cumulative mileage, and coarse positioning coordinates; used to calibrate the spatial coordinate system of environmental data to ensure that the perceived data matches the vehicle's own position.

[0054] In step S120, a first parking area map corresponding to the vehicle is obtained, and the first parking area map includes first parking map elements.

[0055] In this embodiment of the disclosure, the first parking area map can be a parking area map that has already been constructed for the vehicle. The parking area map can be a customized area map for the parking scenario, with the parking lot / parking segment as its scope. The elements of the first parking map can be the elements that constitute the first parking area map, such as including but not limited to the first center line, the first intersection, and the first parking space.

[0056] Obtain the first parking area map corresponding to the vehicle. For example, the first parking area map can be the parking area map newly constructed by the vehicle at time t1. In this embodiment, the vehicle can continuously update the first parking area map at time t1 by acquiring surrounding environmental data. In some special cases, such as when the vehicle switches from a driving scenario to a parking scenario, the first parking area map may be empty, and can be updated subsequently based on the continuously acquired surrounding environmental data.

[0057] In step S130, the surrounding environment data is processed for element identification to obtain the second parking map elements.

[0058] In this embodiment of the disclosure, the second parking map element can be a parking map element obtained after performing element recognition processing on the surrounding environmental data. The second parking map element can be used to update the first parking area map, thereby obtaining the latest parking area map.

[0059] After acquiring the surrounding environmental data, feature recognition processing is performed to determine whether the surrounding environmental data contains parking map features such as center lines, intersections, and parking spaces. The parking map features obtained from the feature recognition processing of the surrounding environmental data are used as the second parking map features.

[0060] In step S140, the first parking area map is updated based on the second parking map elements to obtain the second parking area map.

[0061] In this embodiment of the disclosure, the second parking area map can be a parking area map obtained by updating the first parking area map based on the elements of the second parking map. The second parking area map can be a parking area map constructed at time t2.

[0062] In step S150, the target road topology is constructed based on the second parking area map.

[0063] In this embodiment of the disclosure, the target road topology can be a road topology constructed based on a second parking area map. The road topology in this disclosure can be constructed based on the road centerline and road intersections. The centerline (topology edge) connects to the intersection (topology node), and the intersection connects to the next centerline, describing the road characteristics. Parking spaces can also be used as topology nodes, and other obstacles can define some topology constraint relationships of traffic rules attached to the topology nodes or topology edges.

[0064] After generating the second parking area map, a target road topology can be constructed based on the second parking area map. For example, the target road topology can be used to describe the connection between intersection A and lane B and parking space C. Subsequent vehicles can perform path search and accessibility judgment based on the target road topology.

[0065] It is readily understood by those skilled in the art that in a parking scenario, the vehicle can iteratively execute the above steps. For example, after constructing the parking area map at time t2, it can be used as the new first parking area map. Then, based on the surrounding environmental data acquired at time t3, the parking area map at time t2 is updated to obtain the parking area map at time t3. Then, based on the parking area map at time t3 and the surrounding environmental data acquired at time t4, the parking area map at time t3 is updated to obtain the parking area map at time t4, and so on. Through these iterative steps, the vehicle continuously generates the latest parking area map, thereby constructing the target road topology.

[0066] The road topology construction method disclosed herein allows vehicles to continuously update parking area maps by combining surrounding environmental data. It can continuously update parking area maps based on multi-frame fusion perception results, update parking area maps using surrounding environmental data, and output stable road topology, enabling vehicles to have a clearer and more accurate road understanding ability in parking scenarios. It can also control vehicles to drive autonomously in parking areas based on target road topology, thereby improving vehicle parking efficiency.

[0067] Figure 2 This is a flowchart illustrating, according to an exemplary embodiment of the present disclosure, a process of performing feature identification processing on surrounding environmental data to obtain second parking map features. Figure 1 Based on the road topology construction method shown, Figure 1 Step S130 shown may include the following steps.

[0068] In step S210, the surrounding environment data is input into a pre-trained feature recognition model, which includes a centerline recognition sub-model, an intersection recognition sub-model, and a parking space recognition sub-model.

[0069] The acquired surrounding environment data may include, but is not limited to, images of the surrounding environment at different times, descriptive information about the surrounding environment, and other road observation data. For each frame of received surrounding environment data, for the centerline, redundancy and overlap are first removed, and longer, more informative road segments are preferentially retained. Furthermore, the focus is on key areas within a specified range of the vehicle's location, ignoring distant or irrelevant road information. This information is used as input to the feature recognition model to extract the centerline, thereby improving processing efficiency.

[0070] Similarly, for intersections and parking spaces, the surrounding environmental data can be processed to remove overlapping intersection detections and intersections with abnormal geometric shapes (intersections with particularly small radii); the surrounding environmental data can be processed to remove parking space detections with abnormal geometric shapes (such as non-rectangular parking spaces) and overlapping parking spaces. The focus is on intersections and parking spaces within a specified area of ​​the vehicle's location, ignoring extremely distant intersections and parking spaces, as input to the feature recognition model. The pre-processed surrounding environmental data is then input into the pre-trained feature recognition model, which can then perform feature recognition and extraction from the surrounding environmental data to obtain the second parking map features.

[0071] In step S220, the centerline recognition sub-model performs centerline recognition processing on the surrounding environment data to obtain the centerline recognition result. Road centerline extraction (line-like features) can be performed through various semantic segmentation + centerline fitting methods and end-to-end detection methods.

[0072] For the semantic segmentation + centerline fitting approach, a lane detection model, such as a lane network model, can be used, combined with a deep semantic segmentation model to segment each lane region, and then centerline fitting can be applied to generate a smooth road centerline. For the end-to-end line detection approach, a dedicated line detection network, such as an end-to-end lane detection model, can be used to directly output the parameters of the road centerline (such as start / end coordinates, slope, etc.), reducing post-processing steps.

[0073] In step S230, the intersection recognition sub-model performs intersection recognition processing on the surrounding environment data to obtain the intersection recognition result. Intersection feature extraction (key point / regional features) can employ networks such as deep learning models used for object detection, treating multiple intersections as target categories and detecting their bounding boxes and types. Intersection points can be derived from the center of the bounding box or preset anchor points. Intersection categories may include, but are not limited to, crossroads, T-junctions, L-junctions, etc.

[0074] In step S240, the storage location identification sub-model performs storage location identification processing on the surrounding environmental data to obtain the storage location identification result. This can be done in the following way: For storage location identification, instance segmentation can be used to directly output the storage location mask. For example, using network models such as fast region convolutional neural networks, each storage location is treated as an independent instance, and the complete mask of the storage location and the coordinates of its four interior corners are output. Then, the status of the storage location is determined, including adding a classification head to the segmentation branch, and judging the "idle / occupied" status of the storage location by the pixel texture (such as whether there is a vehicle) in the mask.

[0075] Alternatively, a single-stage object detection model or target detection model can be used for key point detection. First, the four inner corner points of the storage location are detected, and then the outline and orientation of the storage location are fitted by the minimum bounding rectangle / perspective transformation.

[0076] In step S250, the second parking map element is determined based on the centerline recognition result, the intersection recognition result, and the parking space recognition result. After obtaining the centerline recognition result, the intersection recognition result, and the parking space recognition result, the second parking map element can be determined based on the three recognition results, thereby serving as the data basis for map updating and updating the first parking map element in the first parking area map.

[0077] refer to Figure 3 , Figure 3 This is a schematic diagram illustrating parking map elements included in a parking area map according to an exemplary embodiment of the present disclosure. Figure 3 This includes elements such as centerline 310, intersection 320, and parking space 330. The centerline, also known as the road centerline, can be the centerline corresponding to a road where vehicles can travel; it can be a virtual connecting line. An intersection can be the point where two or more roads meet, used to connect them. A parking space, also known as a parking line, can be a location used to mark where vehicles can be parked.

[0078] The first parking map elements may include one or more of the following: a first centerline, a first intersection, and a first parking space. The second parking map elements may include one or more of the following: a second centerline, a second intersection, and a second parking space. The first parking map elements contained in the first parking area map are matched with the corresponding second parking map elements extracted through element recognition processing. The first parking area map is then updated based on these second parking map elements to obtain the second parking area map. These parking map elements can be used to construct parking area maps, enabling vehicles to build continuous roads, thus providing a data foundation for subsequently outputting a stable road topology.

[0079] Figure 4 This is a flowchart illustrating an exemplary embodiment of the present disclosure, showing how to update a first parking area map based on a second centerline to obtain a second parking area map. Figure 1 Based on the road topology construction method shown, Figure 1Step S140 shown may include the following steps.

[0080] In step S410, the second centerline is matched with the first centerline to obtain a centerline matching result. For the second centerline extracted from the surrounding environment data after feature recognition processing, it can be compared and associated with all lane centerlines (i.e., the first centerlines) in the pre-constructed first parking area map to obtain a centerline matching result. There can be one or more second centerlines. If there are multiple second centerlines, centerline matching is performed for each second centerline separately, and the matching operations for multiple centerlines can be performed simultaneously.

[0081] For example, by comprehensively evaluating the proximity, directional consistency, and geometric overlap between the segments of the first and second centerlines, a best match is found in the first parking area map for each identified second centerline. This best match is considered the first centerline that matches the second centerline. If no match is found in the first parking area map, it is assumed that no matching centerline exists. The centerline matching result is determined through this process.

[0082] In step S420, in response to the centerline matching result indicating the existence of a first centerline matching the second centerline, the first centerline matching the second centerline is fused and updated based on the second centerline to obtain the updated matching centerline.

[0083] If the centerline matching result shows a matching centerline corresponding to the second centerline, then the second centerline can be appended to the matching centerline as a new observation. For example, road centerlines can be labeled with continuous line segments (e.g., each centerline labeled with ≥5 key coordinate points) to distinguish centerline instances of different lanes. The second centerline extracted through feature recognition processing can then be used as a new observation for the matching centerline based on the aforementioned labeling information of the second centerline.

[0084] The annotation information of the second center line is used as a new observation data of the matching center line. After being associated with the matching center line, the matching center line can be fused and updated based on the second center line and all the historical observation data corresponding to the matching center line to obtain the updated matching center line.

[0085] In step S430, in response to the centerline matching result that there is no first centerline matching the second centerline, the second centerline is used as a new centerline to obtain the second parking area map.

[0086] If the centerline matching result shows no first centerline matching the second centerline, it is considered that no matching item corresponding to the second centerline was found in the first parking area map. Therefore, the second centerline can be identified as a new, previously unseen road and added to the parking area map. Furthermore, for newly discovered road centerlines, a completely new lane centerline data record can be created in the parking area map, using the currently identified and extracted centerline observation data as the initial observation data for the new centerline. The first parking area map is updated by identifying and extracting the second parking map elements to obtain the second parking area map. Since the second parking area map is constructed based on multi-frame fusion perception results, it can improve the stability of road detection results.

[0087] exist Figure 4 Based on the map update steps for the second parking area shown, Figure 4 Step S420 shown can be achieved through the following steps.

[0088] The number of centerline data points corresponding to the first centerline that matches the second centerline is determined. The first centerline that matches the second centerline can also be called the matching centerline. The number of centerline data points can be determined based on the observation data of the currently identified second centerline and the existing historical observation data of the first centerline that matches the second centerline. For example, if the first centerline that matches the second centerline has 3 historical observation data points, and the second centerline contains 1 observation data point, then the number of centerline data points is 4. For the second centerline obtained through feature identification processing, its corresponding observation data can be used as supplementary observation data for the matching centerline, and the matching centerline can be merged and updated. This is equivalent to continuously using new observation data to enrich the matching centerline.

[0089] In response to the centerline data quantity reaching a first quantity threshold, road point cloud data corresponding to the first centerline matching the second centerline is acquired. The first quantity threshold can be a pre-configured data value, such as 2. If the centerline data quantity is greater than or equal to the first quantity threshold (e.g., 4 is greater than 2), then road point cloud data corresponding to the first centerline matching the second centerline is acquired. Since a matching centerline is labeled by multiple coordinate data points, the data used to describe the matching centerline can be point cloud data.

[0090] For the first centerline that matches the second centerline, each observation data point has its own corresponding road point cloud data. These road point cloud data points can be aggregated, integrated, and aligned to form a complete road point cloud set about the matching centerline. The road point cloud set can be a set of point cloud data composed of multiple road point cloud data points.

[0091] After obtaining the road point cloud set, the matching centerline is updated in the first parking area map based on the road point cloud set and the road generation conditions, resulting in the updated matching centerline. The road generation conditions are the constraints that must be followed when generating the centerline to ensure that the generated road centerline is a path that vehicles can travel on. For example, road generation conditions may include, but are not limited to, continuity constraints and curvature constraints.

[0092] Curvature can be used to describe the degree of bending of a curve at a certain point. Curvature constraints are restrictions on the maximum curvature value of a road or vehicle's trajectory. For example, the larger the steering wheel angle during vehicle movement, the greater the curvature of the vehicle's trajectory and the smaller the turning radius. Therefore, curvature that cannot be achieved even when the steering wheel is turned to its maximum angle should not occur. Corresponding constraints usually need to be pre-defined for the curvature of road topology.

[0093] Continuity constraints can refer to making the road topology as smooth as possible, and avoiding topological routes that require vehicles to frequently switch driving directions (such as frequently turning left and right alternately) to follow, thereby ensuring smooth driving. Corresponding constraints can also be pre-defined for the continuity of the road topology.

[0094] During the centerline matching update process, a curve fitting algorithm can be used to mathematically model the aggregated road point cloud set, generating a lane centerline that perfectly fits all the observation data of the centerlines while maintaining high smoothness (i.e., continuous curvature change), which serves as the updated centerline. Through the above steps, a smooth, continuous, and high-precision matching centerline is obtained, which serves as the data basis for subsequently generating the target road topology.

[0095] Figure 5 This is a flowchart illustrating an exemplary embodiment of the present disclosure, showing how to update a first parking area map based on a second intersection to obtain a second parking area map. Figure 1 Based on the road topology construction method shown, Figure 1 The step S140 shown may also include the following steps.

[0096] In step S510, the second intersection and the first intersection are matched based on the intersection center point to obtain an intersection matching result. For the second intersection extracted from the surrounding environment data, it can be compared and associated with all the first intersections in the pre-constructed first parking area map to obtain an intersection matching result. The number of second intersections can be one or more. If there are multiple second intersections, intersection matching is performed for each second intersection separately, and the matching operations for multiple intersections can be performed simultaneously.

[0097] For example, an intersection can be modeled as a geometric shape, such as a circle, a rectangle, or other geometric shapes. The center point of the model can represent the intersection's location, and the boundary information of the geometric shape can represent the intersection's boundaries.

[0098] For each extracted second intersection, it is compared and correlated with all intersections in the pre-constructed first parking area map. For example, by comparing the proximity of the intersection center points, a best match is found in the first parking area map for each observed second intersection. If no match is found for the second intersection in the first parking area map, it is considered that no matching intersection exists. Through this intersection matching process, the intersection matching result is determined.

[0099] In step S520, in response to the intersection matching result indicating the existence of a first intersection matching the second intersection, the first intersection matching the second intersection is fused and updated based on the second intersection to obtain the updated matching intersection.

[0100] If the intersection matching result shows a matching intersection corresponding to the second intersection, then the second intersection can be added as a new observation data point to the matching intersection. For example, depending on the intersection modeling method, the intersection in this embodiment can be represented by the center point of a geometric shape and its corresponding boundary information. Additionally, intersection types can be labeled, such as T-shaped, cross-shaped, and L-shaped. The center point and boundary information of the geometric shape corresponding to the intersection are used as the observation data for the second intersection to distinguish different intersection instances.

[0101] The second intersection extracted through feature recognition processing can be used as a new observation data point for the matching intersection, and then associated with the matching intersection. Subsequently, the matching intersection can be fused and updated based on all historical observation data corresponding to the second intersection and the matching intersection to obtain the updated matching intersection.

[0102] In step S530, in response to the intersection matching result that there is no first intersection that matches the second intersection, the second intersection is added as a new intersection to obtain the second parking area map.

[0103] If the intersection matching result shows no matching intersection corresponding to the second intersection, it is considered that no matching item corresponding to the second intersection was found in the first parking area map. Therefore, the second intersection can be recognized as a new intersection and added to the parking area map. Furthermore, for newly discovered intersections, a completely new intersection data record can be created in the parking area map, using the currently identified and extracted intersection observation data as the initial observation data for the new intersection. The first parking area map is updated by identifying and extracting the second parking map elements to obtain the second parking area map. Since the second parking area map is constructed based on multi-frame fusion perception results, it can improve the stability of road detection results.

[0104] exist Figure 5 Based on the map update steps for the second parking area shown, Figure 5 Step S520 shown can be achieved through the following steps.

[0105] For the first intersection that matches the second intersection, also known as the matching intersection, the number of intersection data points corresponding to the first intersection that matches the second intersection is determined. The number of intersection data points can be determined based on the observation data of the currently identified second intersection and the existing historical observation data of the matching intersection. For example, if the matching intersection has two historical observation data points and the second intersection contains one observation data point, then the number of intersection data points corresponding to the matching intersection is three. For the second intersection obtained through feature recognition processing, its corresponding observation data can be used as supplementary observation data for the matching intersection, and the matching intersection can be merged and updated, thereby continuously enriching the matching intersection with new observation data.

[0106] In response to the number of intersection data points reaching a second threshold, at least one intersection center point and its corresponding intersection boundary information are determined based on the historical data of the matched intersection. The second threshold can be a pre-configured data value, such as 2. If the number of intersection data points for the matched intersection is greater than or equal to the second threshold (e.g., 3 is greater than 2), then at least one intersection center point and its corresponding intersection boundary information are determined based on the historical data of the matched intersection.

[0107] Based on the intersection center point and boundary information, the matching intersection is updated on the map of the first parking area, resulting in the updated matching intersection. After determining the intersection center point and boundary information based on the observation data, the matching intersection can be updated according to the above information. For example, for the matched intersection, mean filtering can be used to update the intersection center point and boundary information determined in the observation data, thereby obtaining a more accurate intersection modeling result.

[0108] Figure 6This is a flowchart illustrating an exemplary embodiment of the present disclosure, showing how to update a first parking area map based on a second parking location to obtain a second parking area map. Figure 1 Based on the road topology construction method shown, Figure 1 Step S140 shown may include the following steps.

[0109] In step S610, the intersection-union ratio between the second storage location and the first storage location is determined to obtain the storage location matching result. For the extracted second storage location, it can be compared and associated with all the first storage locations in the pre-constructed first parking area map to obtain the storage location matching result. The number of second storage locations can be one or more. If there are multiple second storage locations, storage location matching is performed for each second storage location separately, and the matching operations for multiple storage locations can be performed simultaneously.

[0110] For example, parking spaces can be modeled based on corner information. For instance, a parking space can be determined by marking the coordinates of its four interior corners. In addition, semantic annotation can be used to mark the type of parking space (such as perpendicular / parallel / diagonal) and its vacancy status (occupied / vacant).

[0111] For the extracted second storage location, it is compared and associated with all storage locations in the already constructed first parking area map. For example, based on the coordinates of the four corner points of each storage location, its position and detection box in the map can be determined. Based on the above storage location labeling information, the overlap ratio (CLO) between the second storage location and the existing first storage locations in the first parking area map is calculated. This allows for finding an optimal match in the first parking area map as the corresponding storage location for the second storage location. The CLO can be the ratio of the intersection area to the union area between the detection boxes of two storage locations. If no match is found for the second storage location in the first parking area map, it is considered that no matching storage location exists for the second storage location. Through this storage location matching process, the storage location matching result is determined.

[0112] In step S620, in response to the storage location matching result indicating the existence of a first storage location matching the second storage location, the first storage location matching the second storage location is merged and updated based on the second storage location to obtain the updated matching storage location.

[0113] The first storage location that matches the second storage location can be called the matching storage location. If the storage location matching result shows that there is a first storage location that matches the second storage location, then the second storage location can be appended as a new observation data point to the first storage location that matches the second storage location. For example, depending on the storage location modeling method, the storage location in this embodiment can be represented by the coordinates of the four interior corner points of the storage location. In addition, the storage location type and storage location vacancy status can also be labeled. The coordinates of the four interior corner points corresponding to the storage location are used as the observation data of the second storage location to distinguish different storage location instances in the parking area map.

[0114] The above-mentioned annotation information of the second storage location is used as a new observation data of the matching storage location, and it is associated with the matching storage location. Subsequently, the matching storage location can be fused and updated based on all historical observation data corresponding to the second storage location and the matching storage location to obtain the updated matching storage location.

[0115] In step S630, in response to the fact that the storage location matching result does not contain a first storage location that matches the second storage location, the second storage location is added as a new storage location to obtain the second parking area map.

[0116] If the parking space matching result shows no matching first parking space for the second parking space, it is considered that no matching item corresponding to the second parking space was found in the first parking area map. Therefore, the second parking space can be recognized as a new parking space and added to the parking area map. Furthermore, for the newly discovered parking space, a completely new parking space data record can be created for it in the parking area map, using the currently identified and extracted parking space observation data as the initial observation data for the new parking space. Through these steps, the parking area map can be updated, ensuring that the second parking area map is constructed based on multi-frame fusion perception results, thus improving the stability of road detection results.

[0117] exist Figure 6 Based on the map update steps for the second parking area shown, Figure 6 Step S620 shown can be achieved through the following steps.

[0118] For a matching storage location, the number of storage location data corresponding to the matching storage location is determined based on the second storage location. The number of storage location data can be determined based on the observation data of the currently identified second storage location and the existing historical observation data of the matching storage location. For example, if the matching storage location already has 4 historical observation data entries, and the second storage location contains 1 observation data entry, then the number of storage location data corresponding to the matching storage location is 5. For the second storage location obtained through feature identification processing, its corresponding observation data can be used as supplementary observation data for the matching storage location, and the matching storage location can be merged and updated, thereby continuously enriching the matching storage location with new observation data.

[0119] In response to the number of storage location data reaching a third threshold, the coordinates of at least one storage location corner point are determined based on the historical data of the matching storage locations. The third threshold can be a pre-configured data value; the first, second, and third thresholds can be the same or different. In this disclosure, these thresholds can be set according to map update requirements. If the number of storage location data for a matching storage location is greater than or equal to the third threshold, the coordinates of the four inner corner points used to describe the storage location detection box are determined based on the historical data of the matching storage locations.

[0120] Based on the corner coordinates of the corresponding storage location, the matching storage location is updated on the map of the first parking area to obtain the updated matching storage location. After determining the corner coordinates of the storage location, the matching storage location can be updated according to the above information. For example, for the matched storage location, a mean filter can be used to update the corner coordinates of the storage location determined in the observation data, thereby obtaining a more accurate storage location modeling result.

[0121] exist Figure 1 Based on the road topology construction method shown, Figure 1 Step S150 shown may include the following steps.

[0122] A topology node is constructed based on the second parking area map, and the topology node includes intersection nodes and parking space nodes.

[0123] In this embodiment, the foundation for constructing the road topology is nodes (intersections) and edges (road centerlines), which can be accurately extracted and classified from the second parking area map. For topological nodes, they can be located based on map data features; topological nodes can include intersection nodes and parking space nodes. As seen in the above-described construction process of the second parking area map, the second parking area map contains intersections, and intersection nodes are constructed based on the extracted intersections. For parking spaces, from the corner points detected in the second parking area map, corner points facing the drivable area can be selected, and the midpoint of the parking space entrance can be fitted as the access node for the parking space. The parking space node can be connected to the road centerline.

[0124] The centerline of the second parking area map is obtained, and a topological edge is constructed based on the connection relationship between the centerline and the intersection node or parking space node. In this embodiment, the centerline in the second parking area map can be used as the connecting edge for constructing the target road topology. For example, by extracting the point set of the road centerline, segmenting it according to the connectivity range of the nodes, a channel edge connecting two nodes is formed. The start and end node IDs of the channel edge and the basic parameters of the centerline (such as length and approximate direction) are recorded to construct the topological edge.

[0125] The target road topology is constructed based on the intersection node, the storage location node, and the topological edge.

[0126] In this embodiment, when the topological elements such as intersections, storage locations, and road connection edges are determined, a unique ID can be assigned to each element. For example, the ID of a topological node can be Node_001, Node_002, etc., the ID of a topological edge can be Edge_001, Edge_002, etc., and a storage location can also be assigned a corresponding unique ID as a node, such as Node1_001, Node1_002, etc.

[0127] After element encoding is completed, the topology graph structure is constructed. Each node is indexed, and each node is associated with its directly reachable neighboring nodes and corresponding edge IDs. Connectivity is then verified to obtain the basic topology graph. Since the basic topology graph only contains connectivity relationships, access constraints need to be added to adapt to the rules of actual parking scenarios. Node constraints and edge constraints can be added separately.

[0128] Edge constraints can include, but are not limited to, directional constraints, size constraints, and priority constraints. For example, directional constraints can indicate the direction of travel on an edge (one-way / two-way), such as allowing only one-way travel ("downward → exit") on the edge of a garage ramp, and prohibiting reverse travel. Size constraints can record the minimum travel width and maximum allowable turning radius of an edge, filtering out passages that cannot accommodate target vehicles (e.g., small cars can pass, but large SUVs cannot). Priority constraints can indicate the travel priority of an edge (e.g., the priority of the main passage edge is higher than that of the side passage edge), and the higher priority passage is selected first during path planning.

[0129] Node constraints primarily target intersection nodes and can include, but are not limited to, intersection priorities and turning constraints. For example, intersection priorities can specify pedestrian / vehicle priority rules, and safety waiting logic can be added to intersection nodes during path planning. Turning constraints can specify the minimum turning radius for T-junctions / crossroads to ensure vehicles do not scrape against surrounding obstacles when turning. Constraint information can be bound to corresponding topological nodes and edges in the form of attribute key-value pairs. For road topologies with added constraints, topology verification, optimization, and dynamic updates can be performed to obtain the target road topology. This target road topology can be used in subsequent vehicle path planning and other decisions, enabling vehicles to understand roads in parking functions and initially establishing parking roaming capabilities, allowing vehicles to explore routes autonomously in parking scenarios.

[0130] Furthermore, in Figure 1 Based on the road topology construction method shown, Figure 1 Step S150 shown can also be achieved through the following steps: A first road topology is constructed based on the second parking area map; anomaly detection is performed on the first road topology to obtain anomaly detection results. Due to the large blind spots, winding roads, and poor perception stability in parking scenarios, additional steps can be used to verify the accuracy of the topology. After constructing the first road topology, anomaly detection is performed on it to obtain anomaly detection results. The anomaly detection disclosed in this invention can be performed from multiple dimensions: for example, parking space information can be integrated, and when a road centerline incorrectly extends to a parking space, this is considered an anomaly. Since some blind spots in the map may contain anomalies in road centerlines and intersections, anomaly detection is also performed on centerlines and intersections in certain scenarios to obtain corresponding anomaly detection results.

[0131] After obtaining the anomaly detection results, the first road topology is repaired based on these results to obtain the target road topology. For example, if the road centerline incorrectly extends into a parking space, the centerline extending into the parking space can be erased to ensure the correctness of the topology. For other anomaly detection results of centerlines and intersections, anomaly repair is used to obtain a smooth, continuous, and high-precision centerline. This centerline is then associated with intersection elements in the second parking area map to ensure that the centerline connects to intersections, and intersections connect to other centerlines, constructing a complete topology network and ensuring the consistency and continuity of the road topology. Through the above anomaly detection and repair steps, an incremental, online method can be provided to stably fuse discontinuous and noisy real-time sensing data into a globally consistent and geometrically smooth road topology.

[0132] Figure 7 This is a flowchart illustrating an exemplary embodiment of the present disclosure of adding a virtual centerline to a first road topology. Based on the above steps, when the anomaly detection result indicates a missing centerline, anomaly repair is performed on the first road topology to obtain the target road topology, which may include the following steps.

[0133] Step S710: Based on the first road topology, determine the intersection group to be reviewed. The intersection group to be reviewed consists of any two intersections whose distance is less than a first distance threshold. Virtual center lines are typically added between two relatively close intersections. Therefore, when performing center line anomaly detection, all intersections in the second parking area map can be traversed, and any two intersections whose distance is less than the first distance threshold can be considered as the intersection group to be reviewed. The intersection group to be reviewed can be two intersections to be anomaly detected, and the distance between these two intersections is less than the first distance threshold. The intersection distance can be the distance between the center points of the two intersections.

[0134] For example, by iterating through all intersections, a group of intersections to be reviewed is first selected. If the distance between the center points of two intersections minus the radii of the two intersections is less than a certain empirical value, a first distance threshold A, then it is considered a group of intersections to be reviewed, and other intersections are not considered. Using this method, all groups of intersections to be reviewed within the second parking area map are determined.

[0135] Step S720: In response to the anomaly detection result indicating a lack of centerlines between the intersections to be reviewed, virtual centerlines are added between the intersections to be reviewed to obtain the target road topology. The virtual centerline can be a supplementary road centerline added between two intersections in the intersection group to be reviewed, in order to ensure the continuity of the road topology when a centerline is missing between them.

[0136] For the identified groups of intersections to be reviewed, determine whether there is a connection between two intersections within the group. This is done by traversing all existing centerlines and checking if there is a centerline whose starting and ending points are less than a specified distance threshold from these two intersections. If there is no connection between two intersections in the group, it can be considered a case of missing centerlines. In this case, a virtual centerline can be added between the two intersections to ensure the continuity of the road topology.

[0137] Figure 8 This is a flowchart illustrating an exemplary embodiment of the present disclosure of adding a virtual intersection in a first road topology.

[0138] Step S810: Based on the first road topology, determine a group of center lines to be verified. The group of center lines to be verified consists of any two center lines whose distance between their beginning and end is less than a second distance threshold. The group of center lines to be verified can be multiple road center lines to be detected for anomalies, and the distance between the beginning and end of these two center lines is less than the second distance threshold.

[0139] For example, traverse all centerlines in the first road topology, find all centerlines whose distance between their beginning and end is less than the second distance threshold, extend the two centerlines to obtain the intersection, search for whether there is already an intersection near the intersection, if there is, skip it, so that the intersection has been bound to these endpoints; if there is no intersection at the intersection, it is considered that there is no intersection between the two centerlines.

[0140] Step S820: In response to the anomaly detection result indicating a lack of intersections between the centerline groups to be reviewed, the intersection locations and boundary information are determined based on the centerlines to be reviewed. When there is a lack of intersections between two centerlines in the centerline group to be reviewed, an intersection is generated centered on the intersection of the extensions of the two centerlines, and the intersection boundary information is obtained by referring to the distance from the intersection to the endpoints of the two centerlines.

[0141] Step S830: Add virtual intersections based on the intersection locations and intersection boundary information to obtain the target road topology. For centerlines without any bound intersections, virtual intersections can be added based on the intersection locations and intersection boundary information determined through the above steps, thereby obtaining a continuous and stable target road topology.

[0142] Based on the anomaly detection results, the first road topology is anomaly repaired to obtain the target road topology. This process may further include the following steps: In response to the anomaly detection result indicating a conflict between centerline information and storage location information, the centerline confidence and storage location confidence are determined respectively for the centerline information and the storage location information. The centerline confidence can be the degree of certainty that the model's prediction of the centerline identification is correct. The storage location confidence can be the degree of certainty that the model's prediction of the storage location identification is correct. When a conflict is detected between a centerline and a storage location, such as when a centerline overlaps with a storage location or when some topological lines extend into the storage location, the centerline confidence and storage location confidence corresponding to the aforementioned centerline and storage location can be determined respectively.

[0143] In response to the centerline confidence score being greater than or equal to the parking space location confidence score, conflicting parking spaces are deleted. Conflicting parking spaces can be incorrectly located spaces on the parking area map. By comparing the centerline confidence score and the parking space location confidence score, if the centerline confidence score is greater than or equal to the parking space location confidence score, the centerline is considered to have a higher confidence score, and the centerline detection result may be more accurate. In this case, conflicting parking spaces that conflict with the centerline can be deleted.

[0144] In response to the centerline confidence being less than the parking space location confidence, conflicting centerlines are deleted. Conflicting centerlines can be erroneously existing road centerlines in the parking area map. If the centerline confidence is less than the parking space location confidence, the parking space is considered to have a higher confidence level, and the parking space detection result may be more accurate. In this case, conflicting centerlines that conflict with that parking space can be deleted. Through the above anomaly repair steps, a more accurate road topology can be obtained. Furthermore, this disclosure can also empower crowdsourced mapping based on the above road topology construction method. After obtaining road topology information, multiple crowdsourced mapping trips can construct the road topology information into the map, giving the map complete driving routes.

[0145] The target road topology is constructed through the above steps, and the vehicle can then be controlled to perform parking operations based on this topology. By constructing the target road topology, the vehicle gains a clearer understanding of roads in parking scenarios, enabling it to navigate and explore parking areas autonomously. Furthermore, it empowers the vehicle with crowdsourced mapping capabilities, allowing road topology information to be incorporated into the map, resulting in a complete driving route.

[0146] The following are embodiments of the apparatus disclosed herein, which can be used to execute embodiments of the method disclosed herein. For details not disclosed in the apparatus embodiments of this disclosure, please refer to the embodiments of the method disclosed herein.

[0147] Figure 9 This is a block diagram illustrating a road topology construction device according to an exemplary embodiment of the present disclosure. The device of this embodiment can be applied to vehicles or electronic devices, including local servers and cloud servers, which can be deployed in a computer or a computer cluster consisting of multiple computers, and may also include smartphones and smart tablet terminal devices.

[0148] like Figure 9 As shown, the road topology construction device 900 may include: an environmental data acquisition module 910, a first map acquisition module 920, an element recognition module 930, a map update module 940, and a topology construction module 950.

[0149] The system includes: an environmental data acquisition module 910, used to acquire surrounding environmental data corresponding to the vehicle in response to the vehicle being in a parking scenario; a first map acquisition module 920, used to acquire a first parking area map corresponding to the vehicle, the first parking area map including first parking map elements; an element recognition module 930, used to perform element recognition processing on the surrounding environmental data to obtain second parking map elements; a map update module 940, used to update the first parking area map based on the second parking map elements to obtain a second parking area map; and a topology construction module 950, used to construct a target road topology based on the second parking area map.

[0150] In some exemplary embodiments of this disclosure, the feature recognition module 930 includes a feature recognition unit for inputting the surrounding environment data into a pre-trained feature recognition model. The feature recognition model includes a centerline recognition sub-model, an intersection recognition sub-model, and a parking space recognition sub-model. The centerline recognition sub-model performs centerline recognition processing on the surrounding environment data to obtain a centerline recognition result. The intersection recognition sub-model performs intersection recognition processing on the surrounding environment data to obtain an intersection recognition result. The parking space recognition sub-model performs parking space recognition processing on the surrounding environment data to obtain a parking space recognition result. Based on the centerline recognition result, the intersection recognition result, and the parking space recognition result, the second parking map feature is determined.

[0151] In some exemplary embodiments of this disclosure, the first parking area map includes at least a first centerline, and the second parking map element includes at least a second centerline; the map update module 940 includes a first map update unit, which is configured to: perform centerline matching between the second centerline and the first centerline to obtain a centerline matching result; in response to the centerline matching result indicating the existence of a first centerline matching the second centerline, perform fusion update on the first centerline matching the second centerline based on the second centerline to obtain an updated matching centerline; and in response to the centerline matching result indicating the absence of a first centerline matching the second centerline, use the second centerline as a newly added centerline to obtain the second parking area map.

[0152] In some exemplary embodiments of this disclosure, the first map update unit includes a centerline update subunit, which is configured to: determine the number of centerline data corresponding to a first centerline matching the second centerline based on the second centerline; in response to the number of centerline data reaching a first quantity threshold, acquire road point cloud data corresponding to the first centerline matching the second centerline; generate a road point cloud set based on the road point cloud data; and update the first centerline matching the second centerline in the first parking area map based on the road point cloud set and road generation conditions to obtain an updated matching centerline.

[0153] In some exemplary embodiments of this disclosure, the first parking area map includes at least a first intersection, and the second parking map element includes at least a second intersection; the map update module 940 includes a second map update unit, which is configured to: perform intersection matching between the second intersection and the first intersection based on the intersection center point to obtain an intersection matching result; in response to the intersection matching result indicating the existence of a first intersection matching the second intersection, perform a fusion update on the first intersection matching the second intersection based on the second intersection to obtain an updated matching intersection; and in response to the intersection matching result indicating the absence of a first intersection matching the second intersection, treat the second intersection as a newly added intersection to obtain the second parking area map.

[0154] In some exemplary embodiments of this disclosure, the second map update unit includes an intersection update subunit, which is used to determine, based on the second intersection, the number of intersection data corresponding to a first intersection matching the second intersection; in response to the number of intersection data reaching a second quantity threshold, determine at least one intersection center point and intersection boundary information corresponding to the intersection center point based on historical data of the first intersection matching the second intersection; and update the first intersection matching the second intersection in the first parking area map based on the intersection center point and the intersection boundary information to obtain the updated matching intersection.

[0155] In some exemplary embodiments of this disclosure, the first parking area map includes at least a first parking space, and the second parking map element includes at least a second parking space; the map update module 940 includes a third map update unit, which is configured to: determine the intersection-union ratio (IUU) between the second parking space and the first parking space to obtain a parking space matching result; in response to the parking space matching result indicating the existence of a first parking space matching the second parking space, perform a fusion update on the first parking space matching the second parking space based on the second parking space to obtain an updated matching parking space; and in response to the parking space matching result indicating the absence of a first parking space matching the second parking space, add the second parking space as a new parking space to obtain the second parking area map.

[0156] In some exemplary embodiments of this disclosure, the third map update unit includes a storage location update subunit, which is used to determine, based on the second storage location, the number of storage location data corresponding to a first storage location matching the second storage location; in response to the number of storage location data reaching a third quantity threshold, determine the storage location corner coordinates corresponding to at least one storage location based on historical data of the first storage location matching the second storage location; and update the first storage location matching the second storage location in the first parking area map based on the storage location corner coordinates corresponding to the storage location, thereby obtaining the updated matching storage location.

[0157] In some exemplary embodiments of this disclosure, the topology construction module 950 includes a first topology construction unit, which is used to construct topology nodes based on the second parking area map, the topology nodes including intersection nodes and parking space nodes; obtain the centerline of the second parking area map, construct topology edges based on the connection relationship between the centerline and the intersection nodes or parking space nodes; and construct the target road topology according to the intersection nodes, the parking space nodes and the topology edges.

[0158] In some exemplary embodiments of this disclosure, the topology construction module 950 further includes a second topology construction unit, which is used to construct a first road topology based on the second parking area map; perform anomaly detection on the first road topology to obtain anomaly detection results; and perform anomaly repair on the first road topology according to the anomaly detection results to obtain the target road topology.

[0159] In some exemplary embodiments of this disclosure, the second topology construction unit includes a first anomaly repair subunit, which is configured to: determine a group of intersections to be reviewed based on the first road topology, wherein the group of intersections to be reviewed is any two intersections whose distance is less than a first distance threshold; and, in response to the anomaly detection result that the group of intersections to be reviewed lacks a centerline, add a virtual centerline between the group of intersections to be reviewed to obtain the target road topology.

[0160] In some exemplary embodiments of this disclosure, the second topology construction unit includes a second anomaly repair subunit, which is configured to: determine a group of centerlines to be verified based on the first road topology, wherein the group of centerlines to be verified is any two centerlines whose distance between their beginning and end is less than a second distance threshold; in response to the anomaly detection result indicating that there is a lack of intersections between the group of centerlines to be verified, determine the intersection location and intersection boundary information based on the centerlines to be verified; and add virtual intersections based on the intersection location and the intersection boundary information to obtain the target road topology.

[0161] In some exemplary embodiments of this disclosure, the second topology construction unit includes a third anomaly repair subunit, which is configured to: in response to the anomaly detection result indicating a conflict between centerline information and storage location information, determine the centerline confidence and storage location confidence corresponding to the centerline information and the storage location information, respectively; in response to the centerline confidence being greater than or equal to the storage location confidence, delete the conflicting storage location; and in response to the centerline confidence being less than the storage location confidence, delete the conflicting centerline.

[0162] In some exemplary embodiments of this disclosure, the vehicle further includes a vehicle parking module, which is used to control the vehicle to perform parking operations based on the target road topology.

[0163] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0164] Figure 10 This is a functional block diagram of a vehicle according to an exemplary embodiment of the present disclosure. For example, vehicle 1000 can be a hybrid vehicle, a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or other types of vehicles. Vehicle 1000 can be an intelligent driving vehicle, a semi-intelligent driving vehicle, or a non-intelligent driving vehicle.

[0165] Reference Figure 10 The vehicle 1000 may include various subsystems, such as an infotainment system 1010, a perception system 1020, a decision control system 1030, a drive system 1040, and a computing platform 1050. The vehicle 1000 may also include more or fewer subsystems, and each subsystem may include multiple components. Furthermore, each subsystem and component of the vehicle 1000 can be interconnected via wired or wireless means.

[0166] In some embodiments, the infotainment system 1010 may include a communication system, an entertainment system, and a navigation system, etc.

[0167] The perception system 1020 may include several types of sensors for sensing information about the environment surrounding the vehicle 1000. For example, the perception system 1020 may include a global positioning system, an inertial measurement unit, a lidar, a millimeter-wave radar, an ultrasonic radar, and a camera device.

[0168] The decision control system 1030 may include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

[0169] The drive system 1040 may include components that provide powered motion to the vehicle 1000. In one embodiment, the drive system 1040 may include an engine, an energy source, a transmission system, and wheels. The engine may be one or a combination of internal combustion engines, electric motors, and compressed air engines. The engine is capable of converting energy provided by the energy source into mechanical energy.

[0170] Some or all of the functions of the vehicle 1000 are controlled by a computing platform 1050. The computing platform 1050 may include at least one processor 1051 and a memory 1052, the processor 1051 being able to execute instructions 1053 stored in the memory 1052.

[0171] The processor 1051 can be any conventional processor, such as a commercially available CPU. The processor may also include graphics processing units (GPUs), field-programmable gate arrays (FPGAs), systems on chips (SoCs), application-specific integrated circuits (ASICs), or combinations thereof.

[0172] The memory 1052 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0173] In addition to instruction 1053, memory 1052 can also store data, such as road maps, route information, vehicle location, direction, and speed. The data stored in memory 1052 has been processed using data anonymization technology to remove personally identifiable sensitive information. The collection, storage, and use of the aforementioned data have all been subject to the explicit consent of the data subject. The processed data does not contain any user privacy information, meeting compliance and security requirements for data use. The data stored in memory 1052 can be used by computing platform 1050.

[0174] In this embodiment of the disclosure, processor 1051 may execute instructions 1053 to complete all or part of the steps of the road topology construction method described above.

[0175] In some embodiments of this disclosure, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory including instructions that can be executed by a processor of a device to perform the methods described above. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0176] In some embodiments of this disclosure, a non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by a processor of a mobile terminal, enables the mobile terminal to execute a road topology construction method. The method includes: in response to a vehicle being in a parking scenario, acquiring surrounding environmental data corresponding to the vehicle; acquiring a first parking area map corresponding to the vehicle, the first parking area map including first parking map elements; performing element recognition processing on the surrounding environmental data to obtain second parking map elements; updating the first parking area map based on the second parking map elements to obtain a second parking area map; and constructing a target road topology based on the second parking area map.

[0177] In some embodiments of this disclosure, a computer program product is also provided, including a computer program that, when executed by a processor, implements a road topology construction method. The method includes: in response to a vehicle being in a parking scenario, acquiring surrounding environmental data corresponding to the vehicle; acquiring a first parking area map corresponding to the vehicle, the first parking area map including first parking map elements; performing element recognition processing on the surrounding environmental data to obtain second parking map elements; updating the first parking area map based on the second parking map elements to obtain a second parking area map; and constructing a target road topology based on the second parking area map.

[0178] Figure 11 This is a block diagram illustrating a chip according to an exemplary embodiment of the present disclosure. Figure 11 As shown, the chip 1100 may include a processor 1101 and an interface 1102. The interface 1102 may be connected to one or more memories 1103, and the interface 1102 may be used by the memories 1103 to receive or acquire data or instructions. Additionally, all or part of the memories 1103 may be located outside the chip 1100. The processor 1101 is configured to execute processing steps stored in the aforementioned road topology construction method. The specific methods executed by the processor have been described in detail in the embodiments relating to the method, and will not be elaborated upon here.

[0179] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the appended claims.

[0180] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A road topology construction method, characterized in that, include: In response to a vehicle being in a parking scenario, acquire the surrounding environment data corresponding to the vehicle; Obtain the first parking area map corresponding to the vehicle, wherein the first parking area map includes first parking map elements; The surrounding environment data is processed for element identification to obtain the second parking map elements; The first parking area map is updated based on the second parking map elements to obtain the second parking area map; Construct the target road topology based on the second parking area map.

2. The method according to claim 1, characterized in that, The step of performing element recognition processing on the surrounding environmental data to obtain the second parking map elements includes: The surrounding environment data is input into a pre-trained feature recognition model, which includes a centerline recognition sub-model, an intersection recognition sub-model, and a parking space recognition sub-model. The centerline recognition sub-model performs centerline recognition processing on the surrounding environment data to obtain the centerline recognition result; The intersection recognition sub-model performs intersection recognition processing on the surrounding environment data to obtain the intersection recognition result; The surrounding environmental data is processed by the storage location identification sub-model to obtain the storage location identification result. Based on the centerline recognition result, the intersection recognition result, and the parking space recognition result, the second parking map element is determined.

3. The method according to claim 1, characterized in that, The first parking map element includes one or more of the first centerline, the first intersection, and the first parking space; the second parking map element includes one or more of the second centerline, the second intersection, and the second parking space.

4. The method according to claim 1, characterized in that, The first parking area map includes at least a first center line, and the second parking map element includes at least a second center line; updating the first parking area map based on the second parking map element to obtain the second parking area map includes: The second centerline is matched with the first centerline to obtain the centerline matching result. In response to the centerline matching result indicating the existence of a first centerline matching the second centerline, the first centerline matching the second centerline is fused and updated based on the second centerline to obtain the updated matching centerline; In response to the centerline matching result that there is no first centerline that matches the second centerline, the second centerline is used as a new centerline to obtain the second parking area map.

5. The method according to claim 4, characterized in that, The step of fusing and updating the first centerline that matches the second centerline based on the second centerline to obtain the updated matching centerline includes: Based on the second centerline, determine the number of centerline data corresponding to the first centerline that matches the second centerline; In response to the centerline data quantity reaching a first quantity threshold, road point cloud data corresponding to the first centerline that matches the second centerline is acquired; A road point cloud set is generated based on the road point cloud data; Based on the road point cloud set and road generation conditions, the first centerline matching the second centerline is updated in the first parking area map to obtain the updated matching centerline.

6. The method according to claim 1, characterized in that, The first parking area map includes at least a first intersection, and the second parking map features include at least a second intersection; updating the first parking area map based on the second parking map features to obtain the second parking area map includes: Based on the center point of the intersection, the second intersection and the first intersection are matched to obtain the intersection matching result; In response to the intersection matching result indicating the existence of a first intersection that matches the second intersection, the first intersection that matches the second intersection is merged and updated based on the second intersection to obtain the updated matching intersection; In response to the intersection matching result that there is no first intersection that matches the second intersection, the second intersection is added as a new intersection to obtain the second parking area map.

7. The method according to claim 6, characterized in that, The step of fusing and updating the first intersection that matches the second intersection based on the second intersection to obtain the updated matching intersection includes: Based on the second intersection, determine the number of intersection data corresponding to the first intersection that matches the second intersection; In response to the intersection data quantity reaching a second quantity threshold, at least one intersection center point and the intersection boundary information corresponding to the intersection center point are determined based on the historical data of the first intersection that matches the second intersection; Based on the intersection center point and the intersection boundary information, the first intersection that matches the second intersection is updated in the first parking area map to obtain the updated matching intersection.

8. The method according to claim 1, characterized in that, The first parking area map includes at least a first parking space, and the second parking map element includes at least a second parking space; updating the first parking area map based on the second parking map element to obtain the second parking area map includes: Determine the intersection-union ratio of the second storage location and the first storage location to obtain the storage location matching result; In response to the storage location matching result indicating the existence of a first storage location that matches the second storage location, the first storage location that matches the second storage location is merged and updated based on the second storage location to obtain the updated matching storage location; In response to the fact that there is no first storage location that matches the second storage location, the second storage location is added as a new storage location, and the second parking area map is obtained.

9. The method according to claim 8, characterized in that, The step of fusing and updating the first storage location that matches the second storage location based on the second storage location to obtain the updated matching storage location includes: Based on the second storage location, determine the number of storage location data for the first storage location that matches the second storage location; In response to the number of storage location data reaching a third quantity threshold, the coordinates of the corner point of at least one storage location are determined based on the historical data of the first storage location that matches the second storage location; Based on the corner coordinates of the storage location, the first storage location that matches the second storage location is updated on the map of the first parking area to obtain the updated matching storage location.

10. The method according to claim 1, characterized in that, The step of constructing the target road topology based on the second parking area map includes: A topology node is constructed based on the second parking area map, and the topology node includes intersection nodes and parking space nodes; Obtain the centerline of the second parking area map, and construct topological edges based on the connection relationship between the centerline and the intersection node or parking space node; The target road topology is constructed based on the intersection node, the storage location node, and the topological edge.

11. The method according to claim 1, characterized in that, The step of constructing the target road topology based on the second parking area map also includes: Construct a first road topology based on the second parking area map; Anomaly detection is performed on the first road topology to obtain anomaly detection results; Based on the anomaly detection results, the first road topology is repaired to obtain the target road topology.

12. The method according to claim 11, characterized in that, The step of repairing the first road topology based on the anomaly detection results to obtain the target road topology includes: Based on the first road topology, a group of intersections to be reviewed is determined, wherein the group of intersections to be reviewed is any two intersections whose distance is less than a first distance threshold; In response to the anomaly detection result that the intersections to be reviewed lack centerlines, virtual centerlines are added between the intersections to be reviewed to obtain the target road topology.

13. The method according to claim 11, characterized in that, The step of repairing the first road topology based on the anomaly detection results to obtain the target road topology includes: Based on the first road topology, a group of center lines to be verified is determined. The group of center lines to be verified consists of any two center lines whose distance between their beginning and end is less than a second distance threshold. In response to the anomaly detection result that there is a lack of intersections between the centerline groups to be reviewed, the intersection locations and intersection boundary information are determined based on the centerlines to be reviewed; Virtual intersections are added based on the intersection location and the intersection boundary information to obtain the target road topology.

14. The method according to claim 11, characterized in that, The step of repairing the first road topology based on the anomaly detection results to obtain the target road topology further includes: In response to the anomaly detection result indicating a conflict between the centerline information and the storage location information, the centerline confidence and storage location confidence corresponding to the centerline information and the storage location information are determined respectively. In response to the centerline confidence being greater than or equal to the library location confidence, conflicting library locations with conflicts are deleted; If the confidence level of the centerline is less than the confidence level of the library location, the conflicting centerlines are deleted.

15. The method according to any one of claims 1-14, characterized in that, The method further includes: The vehicle is controlled to perform parking operations based on the target road topology.

16. A road topology construction device, characterized in that, include: The environmental data acquisition module is used to acquire the surrounding environmental data of the vehicle in response to the vehicle being in a parking scenario. The first map acquisition module is used to acquire the first parking area map corresponding to the vehicle, and the first parking area map includes first parking map elements. The feature recognition module is used to perform feature recognition processing on the surrounding environment data to obtain the second parking map features; The map update module is used to update the first parking area map based on the second parking map elements to obtain the second parking area map; The topology building module is used to build the target road topology based on the second parking area map.

17. The apparatus according to claim 16, characterized in that, The first parking area map includes at least a first center line, and the second parking map element includes at least a second center line; the map update module includes a first map update unit, which is used to: perform center line matching between the second center line and the first center line to obtain a center line matching result; In response to the centerline matching result indicating the existence of a first centerline matching the second centerline, the first centerline matching the second centerline is fused and updated based on the second centerline to obtain the updated matching centerline; In response to the centerline matching result that there is no first centerline that matches the second centerline, the second centerline is used as a new centerline to obtain the second parking area map.

18. The apparatus according to claim 16, characterized in that, The first parking area map includes at least a first intersection, and the second parking map features include at least a second intersection; the map update module includes a second map update unit, which is used to: perform intersection matching between the second intersection and the first intersection based on the intersection center point to obtain an intersection matching result; In response to the intersection matching result indicating the existence of a first intersection that matches the second intersection, the first intersection that matches the second intersection is merged and updated based on the second intersection to obtain the updated matching intersection; In response to the intersection matching result that there is no first intersection that matches the second intersection, the second intersection is added as a new intersection to obtain the second parking area map.

19. The apparatus according to claim 16, characterized in that, The first parking area map includes at least a first parking space, and the second parking map element includes at least a second parking space; the map update module includes a third map update unit, which is used to: determine the intersection-union ratio of the second parking space and the first parking space to obtain the parking space matching result; In response to the storage location matching result indicating the existence of a first storage location that matches the second storage location, the first storage location that matches the second storage location is merged and updated based on the second storage location to obtain the updated matching storage location; In response to the fact that there is no first storage location that matches the second storage location, the second storage location is added as a new storage location, and the second parking area map is obtained.

20. A vehicle, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to implement the road topology construction method according to any one of claims 1 to 15.

21. A non-transitory computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the mobile terminal, the mobile terminal is able to perform the road topology construction method according to any one of claims 1 to 15.

22. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the road topology construction method as described in any one of claims 1 to 15.

23. A chip, characterized in that, The chip includes a processor and an interface; The processor is configured to implement the road topology construction method according to any one of claims 1 to 15.

Images