A roadside cooperative positioning method, apparatus, device and medium

Abstract

A roadside cooperative positioning method, device, equipment and medium are disclosed, the roadside cooperative positioning method comprising: acquiring first point cloud data and second point cloud data; after two-dimensional projection mapping of the registered first point cloud data and second point cloud data, a vehicle-mounted local topological graph and a roadside candidate subgraph set are respectively constructed; single-frame graph structure and attribute matching is performed on the vehicle-mounted local topological graph and the roadside candidate subgraph set to screen out a candidate subgraph consistent with the structure of the vehicle-mounted local topological graph from the roadside candidate subgraph set; a key frame is selected and a time sequence multi-frame joint matching strategy is executed to determine a roadside target matched with the vehicle-mounted local topological graph; and a positioning result of the vehicle in a world coordinate system is solved. The present application significantly improves the robustness and accuracy of vehicle-level positioning under the perspective difference between the vehicle and the road, and can achieve stable and reliable target association and positioning in a complex dynamic scene without relying on high-precision GNSS initial values.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving perception technology, specifically to a roadside cooperative positioning method, device, equipment, and medium. Background Technology

[0002] As autonomous driving technology evolves towards full-scenario deployment, high-precision positioning has become a core foundational capability supporting path planning, collaborative perception, and decision-making control. Currently, mainstream positioning solutions primarily rely on a combination of Global Navigation Satellite Systems (GNSS) and inertial navigation systems (IMUs). However, in complex traffic environments such as urban canyons, under overpasses, or tree-lined roads, GNSS signals often face multipath interference or complete blockage, leading to a sharp decline in positioning accuracy or even failure, making it difficult to meet the stringent centimeter-level positioning requirements of autonomous driving.

[0003] To address the limitations of single-vehicle intelligent perception, vehicle-to-everything (V2X)-based assisted positioning technologies have become a research hotspot. Among these, roadside lidar, with its advantages of high installation position, wide field of view coverage, and absolute spatial reference, is widely used to assist onboard units in blind spot supplementation and positioning correction. For example, Chinese patent application CN202010457794.X discloses a road vehicle positioning and perception method based on V2X. This method acquires vehicle status information from the onboard unit and sends it to the roadside unit. The roadside unit performs spatiotemporal synchronization and correlation matching between the environmental status information and the received vehicle status information, realizing the generation and transmission of fused vehicle status information. This method improves the positioning accuracy of road vehicles to a certain extent and expands the perception range of the onboard unit.

[0004] However, existing vehicle-road cooperative positioning methods still face significant technical bottlenecks: Due to significant differences in installation height, pitch angle, and field of view coverage between vehicle-mounted and roadside LiDAR, the point cloud scanning features and occlusion states of the same target are severely inconsistent, making direct registration highly susceptible to geometric errors and prone to local optima. Simultaneously, the lack of a unified spatial reference between the vehicle's local coordinate system and the roadside global coordinate system means that traditional target matching methods relying solely on position or simple geometric dimensions are prone to target association ambiguity in complex scenarios such as dense traffic and parallel driving, making it difficult to guarantee the robustness and accuracy of cooperative positioning. Therefore, there is an urgent need for a vehicle-road cooperative positioning technology that can overcome viewpoint difference interference, does not rely on high-precision GNSS initial values, and possesses strong anti-interference capabilities to achieve stable and reliable vehicle target-level positioning in complex environments. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned technical deficiencies and propose a roadside cooperative positioning method, device, equipment and medium to solve the technical problems of insufficient robustness and accuracy of vehicle-road cooperative positioning methods in complex environments due to differences in perspective, ambiguous target matching, and lack of a unified spatial reference.

[0006] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a roadside cooperative positioning method, comprising: Acquire first point cloud data and second point cloud data, and register the first point cloud data and second point cloud data, wherein the first point cloud data and second point cloud data are point cloud data collected by roadside lidar and vehicle-mounted lidar, respectively. After performing two-dimensional projection mapping on the registered first point cloud data and second point cloud data, a local topology map of the vehicle end and a set of candidate sub-maps of the roadside end are constructed respectively. Single-frame graph structure and attribute matching is performed on the local topology map of the vehicle terminal and the candidate sub-graph set of the roadside terminal to filter out candidate sub-graphs with the same structure as the local topology map of the vehicle terminal from the candidate sub-graph set of the roadside terminal. A temporal topology map sequence is constructed based on historical frame data. Key frames are selected and a temporal multi-frame joint matching strategy is executed to determine roadside targets that match the local topology map of the vehicle terminal. Based on the local coordinates of the matched roadside target in the roadside lidar coordinate system and the preset external parameters, the positioning result of the vehicle in the world coordinate system is calculated.

[0007] In some embodiments, acquiring the first point cloud data and the second point cloud data, and registering the first point cloud data and the second point cloud data, wherein the first point cloud data and the second point cloud data are point cloud data collected by roadside lidar and vehicle-mounted lidar, respectively, include: Acquire the first point cloud data and the second point cloud data; Ground plane features were extracted from the first point cloud data and the second point cloud data respectively, and the roadside ground plane normal vector and the vehicle-mounted ground plane normal vector were obtained by fitting. Construct a rotation matrix to calibrate the Z-axis of the vehicle coordinate system to be parallel to the Z-axis of the roadside coordinate system, so that the vehicle-road coordinate system is geometrically coarsely aligned on the XOY plane.

[0008] In some embodiments, after performing two-dimensional projection mapping on the registered first point cloud data and second point cloud data, constructing a local topology map for the vehicle-mounted end and a candidate sub-map set for the roadside end respectively includes: Constructing a local topology graph for the vehicle: with the vehicle itself as the central node. Neighboring vehicles within the vehicle's perception range are considered as neighboring nodes. , construct graph ;in, Represents a set of nodes. This represents the set of edges constructed based on the relative pose relationships between vehicles; Constructing a candidate sub-map set for the roadside: based on data detected by roadside lidar. Each vehicle target is analyzed sequentially, with each detected vehicle being... As the central node, construct candidate subgraphs .

[0009] In some embodiments, after performing two-dimensional projection mapping on the registered first point cloud data and second point cloud data, the process of constructing a local topology map for the vehicle-mounted end and a candidate sub-map set for the roadside end, respectively, further includes: The bounding box is modified based on the prior vehicle dimensions; The geometric center point of the target is recalculated using the corrected bounding box; Update the coordinate positions of the corresponding graph nodes based on the geometric center point.

[0010] In some embodiments, the step of performing single-frame graph structure and attribute matching on the vehicle-mounted local topology map and the roadside candidate sub-graph set to filter out candidate sub-graphs in the roadside candidate sub-graph set that have the same structure as the vehicle-mounted local topology map includes: Select candidate subgraphs from the roadside candidate subgraph set that are consistent with the local topology structure of the vehicle-mounted terminal; The subgraph isomorphism algorithm is used to determine whether there is an isomorphic mapping relationship between the local topology graph of the vehicle end and the candidate subgraph of the roadside end; In the presence of isomorphic mappings, the similarity score of edge attributes is calculated using a Gaussian weighted function; Select roadside target candidates whose matching degree reaches a preset threshold.

[0011] In some embodiments, the step of constructing a temporal topology map sequence based on historical frame data, selecting key frames, and executing a temporal multi-frame joint matching strategy to determine roadside targets that match the local topology map of the vehicle includes: Looking back from the current time Frames are used to construct the timing sequence of the vehicle-mounted terminal and the timing sequence of the roadside terminal; Calculate the similarity between any two frames in the vehicle-mounted time sequence and construct a similarity matrix; Summing is performed on each row of the similarity matrix, and the set of times when the summation value is lower than a set value is selected as the keyframe set; The cumulative matching score between the vehicle-mounted image and the roadside image is calculated in the set of keyframes, and the roadside target with the highest cumulative score is selected as the final matching result.

[0012] In some embodiments, the calculation of the vehicle's positioning result in the world coordinate system based on the local coordinates of the matched roadside target in the roadside lidar coordinate system and preset external parameters includes: Based on the local coordinates of the successfully matched roadside target in the roadside lidar coordinate system; Using the pre-calibrated external parameters of the roadside lidar, the local coordinates are transformed to the world coordinate system to obtain the vehicle's positioning result.

[0013] In a second aspect, the present invention provides a roadside cooperative positioning device, comprising: The data acquisition and registration module acquires first point cloud data and second point cloud data, and registers the first point cloud data and second point cloud data, wherein the first point cloud data and second point cloud data are point cloud data collected by roadside lidar and vehicle-mounted lidar, respectively. The topology graph construction module is used to perform two-dimensional projection mapping on the registered first point cloud data and second point cloud data, and then construct a local topology graph of the vehicle end and a set of candidate subgraphs of the roadside end, respectively. A single-frame matching module is used to perform single-frame graph structure and attribute matching between the local topology map of the vehicle terminal and the candidate sub-graph set of the roadside terminal, so as to filter out candidate sub-graphs in the roadside candidate sub-graph set that are consistent with the structure of the local topology map of the vehicle terminal. The temporal multi-frame joint matching module is used to construct a temporal topology map sequence based on historical frame data, select key frames and execute the temporal multi-frame joint matching strategy to determine the roadside targets that match the local topology map of the vehicle terminal. The positioning result calculation module is used to calculate the vehicle's positioning result in the world coordinate system based on the local coordinates of the matched roadside target in the roadside lidar coordinate system and preset external parameters.

[0014] Thirdly, the present invention provides an electronic device, comprising: a processor and a memory; The memory stores computer programs that can be executed by the processor; When the processor executes the computer program, it implements the steps in the roadside cooperative positioning method as described in any of the above.

[0015] Fourthly, the present invention provides a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the roadside cooperative positioning method as described in any of the preceding claims.

[0016] Compared with existing technologies, the roadside cooperative positioning method, device, equipment, and medium provided by this invention perform geometric coarse registration of the vehicle-road coordinate system by extracting ground plane features. This overcomes the viewpoint difference between vehicle-mounted and roadside lidar without relying on high-precision GNSS initial values, achieving rapid coarse alignment of heterogeneous coordinate systems. Simultaneously, it projects the three-dimensional target onto a two-dimensional plane to construct a spatial relationship graph containing topological structure and attribute features. Matching objects with the same structure as the local topological graph at the vehicle end are selected from the candidate sub-graph set at the roadside end, effectively solving the target matching ambiguity problem caused by inconsistent point cloud features in dense traffic flow and occlusion scenarios. Furthermore, by combining a temporal multi-frame joint matching strategy, a temporal topological graph sequence is constructed, and key frames are selected for temporal matching. Historical motion information is used to eliminate the symmetry multiple solutions and mismatch problems that may arise from the geometric structure of a single frame, ensuring high accuracy and robustness of the positioning results in complex dynamic scenarios such as dense traffic, high-speed driving, and limited GNSS signals, and possessing good engineering practicality. Attached Figure Description

[0017] Figure 1 This is a flowchart of the roadside cooperative positioning method provided in the embodiments of the present invention; Figure 2 This is a structural block diagram of the roadside cooperative positioning device provided in an embodiment of the present invention; Figure 3 This is a structural block diagram of the electronic device provided in an embodiment of the present invention.

[0018] Explanation of reference numerals in the attached figures: 100. Roadside cooperative positioning method; 200. Roadside cooperative positioning device; 210. Data Acquisition and Registration Module; 220. Ground plane feature extraction and coarse registration module; 211. Ground plane feature extraction unit; 212. Coordinate system coarse registration unit; 220. Topology graph construction module; 221. Vehicle-mounted topology graph construction unit; 222. Roadside candidate subgraph set construction unit; 223. Bounding box correction and coordinate update unit; 230. Single-frame matching module; 231. Subgraph isomorphism judgment unit; 232. Edge attribute similarity calculation unit; 233. Candidate set filtering unit; 240. Temporal multi-frame joint matching module; 241. Sequence graph construction unit; 242. Keyframe selection unit; 243. Multi-frame cumulative matching unit; 250. Positioning Result Calculation Module; 300. Electronic device; 301. Device bus; 302. Processor; 303. Storage medium; 3031. Operating system; 3032. Computer program; 304. Internal memory; 305. Network interface. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] To address the technical challenges of vehicle-road cooperative positioning methods in complex environments, such as difficulties in point cloud registration due to perspective differences, ambiguities in target matching, and insufficient robustness and accuracy in the absence of a unified spatial reference, this invention provides a roadside cooperative positioning method, device, equipment, and medium. By utilizing ground plane consistency to achieve geometric coarse alignment of the vehicle-road coordinate system, and introducing graph matching and temporal multi-frame optimization strategies based on vehicle topology relationships, this invention achieves high-precision and robust positioning of connected vehicles in the world coordinate system without the need for high-precision GNSS initial values.

[0021] Therefore, this implementation plan proposes a roadside cooperative positioning method 100, which is described below in conjunction with... Figure 1 This invention describes a roadside cooperative positioning method 100 according to an exemplary embodiment of the present disclosure, including the following steps S110-S160. The present invention provides a roadside cooperative positioning method, including: Step S110: Acquire first point cloud data and second point cloud data, and register the first point cloud data and second point cloud data, wherein the first point cloud data and second point cloud data are point cloud data collected by roadside lidar and vehicle-mounted lidar, respectively.

[0022] Step S120: After performing two-dimensional projection mapping on the registered first point cloud data and second point cloud data, construct the vehicle-mounted local topology map and the roadside candidate sub-map set respectively.

[0023] Step S130: Perform single-frame graph structure and attribute matching on the local topology map of the vehicle terminal and the candidate sub-graph set of the roadside terminal to filter out candidate sub-graphs with the same structure as the local topology map of the vehicle terminal from the candidate sub-graph set of the roadside terminal.

[0024] Step S140: Construct a temporal topology map sequence based on historical frame data, select key frames and execute a temporal multi-frame joint matching strategy to determine roadside targets that match the local topology map of the vehicle terminal.

[0025] Step S150: Based on the local coordinates of the matched roadside target in the roadside lidar coordinate system and the preset external parameters, calculate the vehicle's positioning result in the world coordinate system.

[0026] The implementation process of each step is illustrated below as an example.

[0027] Step S110: Acquire first point cloud data and second point cloud data, and register the first point cloud data and second point cloud data. The first point cloud data and second point cloud data are point cloud data collected by roadside lidar and vehicle-mounted lidar, respectively. Roadside lidar is typically fixedly installed on utility poles, traffic light poles, or dedicated poles beside the road, and its scanning range covers the target road segment. The collected point cloud data contains three-dimensional spatial information of various targets such as the road surface, vehicles, pedestrians, and roadside facilities. Each point consists of its three-dimensional coordinates (x, y, z) and optional reflection intensity information. The acquired first point cloud data will serve as a reference for subsequent ground fitting and coordinate system one. The registration method is based on the ground plane features extracted by the roadside lidar and vehicle-mounted lidar, performing geometric coarse registration of the vehicle-road coordinate system. This step aims to utilize the consistency of observations of the common feature of the ground plane by the vehicle-mounted and roadside lidars to establish an initial spatial alignment relationship and reduce initial attitude errors.

[0028] Step S110 specifically includes the following steps: Acquire the first point cloud data and the second point cloud data; Ground plane features are extracted from the first and second point cloud data respectively, and the roadside ground plane normal vector and the vehicle-mounted ground plane normal vector are obtained by fitting. Specifically, this includes: Identify point cloud data collected by roadside lidar and fit the roadside ground plane equation :

[0029] in, The unit normal vector of the roadside ground plane. This is the offset. Let be the coordinates of any point on the plane.

[0030] A ground segmentation algorithm is used to extract a subset of points belonging to the ground from the second point cloud data to filter out interference from non-ground points such as vehicles, pedestrians, and road poles on the ground fitting. Then, the RANSAC algorithm is used to fit the vehicle-mounted ground plane equation. :

[0031] in, The unit normal vector of the vehicle-mounted ground plane. The offset is used. Among these, the Patchwork++ algorithm is preferred for ground segmentation, as it can adapt to different road slopes and terrain variations across multiple regions, improving the accuracy and robustness of ground point extraction.

[0032] Construct rotation matrix This makes the vehicle-mounted ground plane normal vector Normal vector to the roadside ground plane Alignment, i.e. Specifically, it includes: To convert the vehicle-mounted ground plane normal vector Normal vector to the roadside ground plane For alignment, first determine the axis of rotation, which is calculated using the following formula: ; Calculate the vehicle-mounted ground plane normal vector Normal vector to the roadside ground plane The rotation angle between them is calculated using the following formula: ; Calculate the rotation matrix using Rodrigues' rotation formula :

[0033] in It is a 3×3 identity matrix. For vectors An antisymmetric matrix.

[0034] The Z-axis of the vehicle-mounted coordinate system is calibrated to be parallel to the Z-axis of the roadside coordinate system, thus achieving a coarse geometric alignment of the vehicle-road coordinate systems in the XOY plane. Specifically, this is achieved by calculating the centroid of the vehicle-mounted ground points. centroid of the roadside ground point Calculate the translation vector .pass The second point cloud data Transform to the roadside coordinate system to complete the coplanar coarse alignment of the XOY plane.

[0035] Step S120: After performing two-dimensional projection mapping on the registered first point cloud data and second point cloud data, construct the vehicle-mounted local topology map and the roadside candidate sub-map set respectively.

[0036] Step S120 specifically includes the following steps: Constructing a local topology diagram for the vehicle-mounted terminal Using the vehicle's own location as the central node. Neighboring vehicles within the vehicle's perception range are considered as neighboring nodes. , construct graph ;in, This represents a set of nodes, including the center node and its neighboring nodes. This represents the set of edges constructed based on the relative pose relationships between vehicles; for a graph... Each edge Its attributes include distance information. With direction information .

[0037] Constructing candidate subgraphs for roadside ends Based on all vehicle targets detected by the roadside lidar, let the total number of detected targets be... Each of the detected vehicles in turn As the central node, (i=1,2,..., Referring to the mapping rules in step S310, construct... This yields the roadside candidate subgraph set. .

[0038] Size-Prior-Based Bounding Box Correction. To address the issue of L-shaped or significantly smaller bounding boxes appearing in roadside views due to occlusion of vehicle targets, leading to target center point shifts, a size-prior-based bounding box correction mechanism is introduced. This mechanism uses size threshold constraints for large and small vehicles to geometrically correct the length and width of the bounding box of roadside detected targets. Based on the corrected bounding box, the geometric center point of the target is recalculated, thereby updating the coordinate positions of the corresponding graph nodes.

[0039] Specifically, let the length of the detected bounding box be... Width is The length is adjusted according to the following logic. and width :

[0040] In the formula, and These are the length and width of the optimized bounding box, respectively. and These are the length and width of the detected bounding box, respectively. , , , , , , , These are the length and width thresholds for large and small vehicles, respectively.

[0041] Step S130: Perform single-frame graph structure and attribute matching on the local topology map of the vehicle terminal and the candidate sub-graph set of the roadside terminal to filter out candidate sub-graphs with the same structure as the local topology map of the vehicle terminal from the candidate sub-graph set of the roadside terminal.

[0042] Step S130 specifically includes the following steps: The VF2 subgraph isomorphism algorithm is used to determine the local topology of the vehicle terminal. Does it have an isomorphic mapping relationship with the roadside candidate subgraph? If there exists a node correspondence such that when any two nodes in the vehicle-mounted graph are connected by an edge, the two corresponding nodes in the candidate subgraph are also connected by an edge, and when no two nodes in the vehicle-mounted graph are connected by an edge, the two corresponding nodes in the candidate subgraph are also not connected by an edge, then the candidate subgraph is determined to be structurally isomorphic to the vehicle-mounted graph and is retained; otherwise, the two are determined to be structurally inconsistent and the candidate subgraph is directly eliminated, thus enabling the selection of subgraphs structurally similar to the vehicle-mounted graph. Consistent candidate roadside subplots .

[0043] In the presence of isomorphic mappings, for isomorphic subgraphs The similarity score of edge attributes is calculated using a Gaussian weighted function; the similarity function is defined. as follows:

[0044] In the formula, Let be any edge of the vehicle-mounted graph; and Represents a node and In candidate subgraph The matching node in the middle; Represents nodes in the vehicle graph and The Euclidean distance between them; Subgraph The Euclidean distance between corresponding nodes; Represents nodes in the vehicle graph and The relative angle difference between them; For subgraph The angle difference between corresponding nodes; This is the Gaussian smoothing coefficient of the distance difference, which controls the tolerance range of the distance error; The Gaussian smoothing coefficient for the angle difference controls the tolerance range of the angle error.

[0045] Based on the obtained similarity scores of each candidate sub-image, a preliminary selection of roadside target candidate sets with high matching degree is made. Specifically, a similarity threshold is set, and roadside targets corresponding to candidate sub-images with similarity scores greater than or equal to the threshold are included in the candidate set; candidate sub-images with similarity scores lower than the threshold are removed, and the final output roadside target candidate set is used as the single-frame matching result.

[0046] Step S140: Construct a temporal topology graph sequence based on historical frame data, select key frames, and execute a multi-frame temporal joint matching strategy to determine the roadside target that matches the local topology graph of the vehicle. This step addresses the geometric symmetry ambiguity problem that may exist in single-frame matching (e.g., multiple roadside candidate subgraphs have similar matching scores to the vehicle graph, making it impossible to uniquely determine the matching target). A temporal graph matching strategy is constructed. By tracing back multiple frames of historical point clouds, temporal topology graph sequences for both the vehicle and roadside ends are constructed. The similarity features of the temporal graph sequences are calculated, and the historical moment with the most significant graph structure diversity is selected as the key frame. A joint decision is then made to determine the unique roadside target that truly corresponds to the connected vehicle, improving the uniqueness and robustness of the matching.

[0047] Step S140 specifically includes the following steps: Time series diagram construction. Starting with the current time... Based on, trace back Frame historical point cloud data were used to construct time series diagrams for the vehicle-mounted terminal. And roadside end timing sequence By introducing temporal information and utilizing the continuous changes in the graph structure during vehicle motion, the ambiguity caused by spatial symmetry can be overcome, providing a multi-moment matching basis for subsequent keyframe selection and joint decision-making.

[0048] Similarity matrix calculation and keyframe optimization. Calculate the similarity between any two frames in the vehicle-mounted time sequence and construct a similarity matrix. .

[0049] Select a set of keyframes. Then, analyze the similarity matrix. Sum each row and select the set of times when the sum is lower than a set value. As a set of keyframes.

[0050] Joint judgment. Only in the keyframe set. The final matching decision is made, and the roadside target index with the highest cumulative matching score is selected.

[0051] in, This represents the local topology of the vehicle-mounted terminal at the nth keyframe. This represents the candidate sub-map of the i-th candidate target at the roadside end at the same time. The similarity function defined for the above steps. This cumulative score reflects the consistency of candidate target i with the vehicle-mounted image across all keyframes, thereby identifying the unique target vehicle corresponding to the roadside end. By extending single-frame spatial matching to temporal multi-frame joint optimization, and leveraging the diverse characteristics of graph structure evolution over time, the matching ambiguity problem in symmetrical scenes is effectively solved, significantly improving the accuracy and robustness of cross-view target association.

[0052] Step S150: Based on the local coordinates of the matched roadside target in the roadside lidar coordinate system and the preset external parameters, calculate the vehicle's positioning result in the world coordinate system.

[0053] Step S150 specifically includes the following steps: Obtain the roadside targets that were successfully matched in the above steps. Local coordinates in the roadside lidar coordinate system .

[0054] Acquire the pre-calibrated external parameters of the roadside lidar relative to the world coordinate system, including the rotation matrix. Translation vector .

[0055] Using the rotation matrix and translation vector, the local coordinate rigid body is transformed to the world coordinate system, thus obtaining the absolute position of the connected vehicle in the world coordinate system. :

[0056] in, Let be a rotation matrix. It is a translation vector. To match the local coordinates of the target in the roadside lidar coordinate system.

[0057] The calculated absolute position As the final high-precision positioning result output for connected vehicles, it is used to correct the accumulated errors of the onboard GNSS / IMU.

[0058] Figure 2 A schematic diagram of a roadside cooperative positioning device 200 according to another embodiment of the present disclosure is shown, such as... Figure 2 As shown, the roadside cooperative positioning device 200 includes: The data acquisition and registration module 210 acquires first point cloud data and second point cloud data, and registers the first point cloud data and second point cloud data, wherein the first point cloud data and second point cloud data are point cloud data collected by roadside lidar and vehicle-mounted lidar, respectively. The topology map construction module 220 is used to construct a local topology map of the vehicle end and a set of candidate sub-maps of the roadside end respectively after performing two-dimensional projection mapping on the registered first point cloud data and second point cloud data. The single-frame matching module 230 is used to filter candidate subgraphs that are consistent with the local topology map structure of the vehicle end from the candidate subgraph set at the roadside end, and to perform single-frame map structure and attribute matching based on the attribute features of the candidate subgraphs; The temporal multi-frame joint matching module 240 is used to construct a temporal topology map sequence based on historical frame data, select key frames and execute a temporal multi-frame joint matching strategy to determine roadside targets that match the local topology map of the vehicle terminal. The positioning result calculation module 250 is used to calculate the positioning result of the vehicle in the world coordinate system based on the local coordinates of the matched roadside target in the roadside lidar coordinate system and the preset external parameters.

[0059] Preferably, in this embodiment, the data acquisition and registration module 210 includes a ground plane feature extraction unit 211 and a coordinate system coarse registration unit 212; wherein, the ground plane feature extraction unit 211 is used to acquire first point cloud data and second point cloud data, and extract ground plane features from the first point cloud data and the second point cloud data respectively, and fit to obtain the roadside ground plane normal vector and the vehicle-mounted ground plane normal vector; the coordinate system coarse registration unit 212 is used to construct a rotation matrix based on the roadside ground plane normal vector and the vehicle-mounted ground plane normal vector obtained by the ground plane feature extraction unit 211, and adjust the vehicle-mounted coordinate system through the rotation matrix so that the vehicle-mounted ground plane normal vector is aligned with the roadside ground plane normal vector, thereby calibrating the Z-axis of the vehicle-mounted coordinate system to be parallel to the Z-axis of the roadside coordinate system, and realizing the geometric coarse alignment of the vehicle-road coordinate system on the XOY plane.

[0060] Preferably, in this embodiment, the topology graph construction module 220 includes an on-board topology graph construction unit 221, a roadside candidate subgraph set construction unit 222, and a bounding box correction and coordinate update unit 223; wherein, the on-board topology graph construction unit 221 is used to take the vehicle itself as the center node. Neighboring vehicles within the vehicle's perception range are considered as neighboring nodes. , construct graph ;in, Represents a set of nodes. The edge set is constructed based on the relative pose relationships between vehicles; the roadside candidate subgraph set construction unit 222 is used for constructing the edge set based on the roadside lidar detection. Each vehicle target is analyzed sequentially, with each detected vehicle being... As the central node, construct candidate subgraphs The bounding box correction and coordinate update unit 223 is used to correct the bounding box based on the prior vehicle size, recalculate the geometric center point of the target using the corrected bounding box, and update the coordinate position of the corresponding graph node based on the geometric center point.

[0061] Preferably, in this embodiment, the single-frame matching module 230 includes a subgraph isomorphism judgment unit 231, an edge attribute similarity calculation unit 232, and a candidate set filtering unit 233; wherein, the subgraph isomorphism judgment unit 231 is used to use a subgraph isomorphism algorithm to determine whether there is an isomorphic mapping relationship between the local topology graph of the vehicle end and the candidate subgraph of the roadside end; the edge attribute similarity calculation unit 232 is used to calculate the similarity score of the edge attributes using a Gaussian weighted function when there is an isomorphic mapping; the candidate set filtering unit 233 is used to filter out the roadside target candidate set whose matching degree reaches a preset threshold.

[0062] Preferably, in this embodiment, the temporal multi-frame joint matching module 240 includes a temporal graph sequence construction unit 241, a key frame selection unit 242, and a multi-frame cumulative matching unit 243; wherein, the temporal graph sequence construction unit 241 is used to trace back Nt frames based on the current time to construct the vehicle-mounted temporal graph sequence and the roadside temporal graph sequence respectively; the key frame selection unit 242 is used to calculate the similarity between any two frames in the vehicle-mounted temporal graph sequence and construct a similarity matrix, summing each row of the similarity matrix, and selecting the set of times with a summation value lower than a set value as the key frame set; the multi-frame cumulative matching unit 243 is used to calculate the cumulative matching score between the vehicle-mounted map and the roadside map in the key frame set, and finally select the roadside target with the highest cumulative score as the final result of matching with the vehicle-mounted local topology map.

[0063] The aforementioned roadside cooperative positioning device can be implemented as a computer program 3032, which can, for example, Figure 3 The electronic device 300 shown is running on it.

[0064] On the other hand, embodiments of this application also provide an electronic device 300, which is a host computer or a server. Please refer to [link / reference]. Figure 3 , Figure 3 This is a schematic block diagram of an electronic device provided in an embodiment of the present invention. The electronic device 300 includes a processor 302, a memory, and a network interface 305 connected via a device bus 301. The memory may include a storage medium 303 and an internal memory 304.

[0065] The storage medium 303 can store the operating system 3031 and the computer program 3032. When the computer program 3032 is executed, it enables the processor 302 to execute the intelligent power-saving control method.

[0066] The processor 302 provides computing and control capabilities to support the operation of the entire electronic device 300.

[0067] The internal memory 304 provides an environment for the computer program 3032 in the storage medium 303 to run. When the computer program 3032 is executed by the processor 302, the processor 302 can execute the intelligent power-saving control method.

[0068] The network interface 305 is used for network communication, such as providing data transmission. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the electronic device 300 to which the present invention is applied. The specific electronic device 300 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0069] The processor 302 is used to run the computer program 3032 stored in the memory to implement the roadside cooperative positioning method disclosed in the embodiments of the present invention.

[0070] Those skilled in the art will understand that Figure 3 The embodiments of the computer device shown do not constitute a limitation on the specific configuration of the computer device. In other embodiments, the computer device may include more or fewer components than shown, or combine certain components, or have different component arrangements. For example, in some embodiments, the computer device may include only a memory and processor 302. In such embodiments, the structure and function of the memory and processor 302 are different from those shown. Figure 3 The embodiments shown are consistent and will not be described again here.

[0071] It should be understood that, in this embodiment of the invention, the processor 302 may be a Central Processing Unit (CPU), or it may be another general-purpose processor 302, a digital signal processor 302 (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor 302 may be a microprocessor 302, or it may be any conventional processor 302, etc.

[0072] In another embodiment of the present invention, a computer-readable storage medium 303 is provided. This computer-readable storage medium 303 may be a non-volatile computer-readable storage medium 303 or a volatile computer-readable storage medium 303. The computer-readable storage medium 303 stores a computer program 3032, which, when executed by a processor 302, implements the roadside cooperative positioning method disclosed in this embodiment of the present invention.

[0073] Those skilled in the art will readily understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0074] In the embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Units with the same function may be grouped into one unit. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, or may be electrical, mechanical, or other forms of connection.

[0075] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0076] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0077] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium 303. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium 303 and includes several instructions to cause an electronic device 300 (which may be a personal computer, a backend server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium 303 includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a magnetic disk, or an optical disk.

[0078] This invention performs geometric coarse registration of the vehicle-road coordinate system by extracting ground plane features. It overcomes the viewpoint difference between vehicle-mounted and roadside lidar without relying on high-precision GNSS initial values, achieving rapid coarse alignment of heterogeneous coordinate systems. Simultaneously, it projects the 3D target onto a 2D plane to construct a spatial relationship graph containing topological structure and attribute features. Matching objects with the same structure as the local topological graph at the vehicle end are selected from the candidate sub-graph set at the roadside end, effectively solving the target matching ambiguity problem caused by inconsistent point cloud features in dense traffic flow and occlusion scenarios. In addition, by combining a temporal multi-frame joint matching strategy, a temporal topological graph sequence is constructed and key frames are selected to perform temporal matching. Historical motion information is used to eliminate the symmetry multiple solutions and mismatch problems that may be caused by the geometric structure of a single frame, ensuring high accuracy and strong robustness of the positioning results in complex dynamic scenarios such as dense traffic, high-speed driving, and limited GNSS signals, and has good engineering practicality.

[0079] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A roadside cooperative positioning method, characterized in that, include: Acquire first point cloud data and second point cloud data, and register the first point cloud data and second point cloud data, wherein the first point cloud data and second point cloud data are point cloud data collected by roadside lidar and vehicle-mounted lidar, respectively. After performing two-dimensional projection mapping on the registered first point cloud data and second point cloud data, a local topology map of the vehicle end and a set of candidate sub-maps of the roadside end are constructed respectively. Single-frame graph structure and attribute matching is performed on the local topology map of the vehicle terminal and the candidate sub-graph set of the roadside terminal to filter out candidate sub-graphs with the same structure as the local topology map of the vehicle terminal from the candidate sub-graph set of the roadside terminal. A temporal topology map sequence is constructed based on historical frame data. Key frames are selected and a temporal multi-frame joint matching strategy is executed to determine roadside targets that match the local topology map of the vehicle terminal. Based on the local coordinates of the matched roadside target in the roadside lidar coordinate system and the preset external parameters, the positioning result of the vehicle in the world coordinate system is calculated.

2. The roadside cooperative positioning method according to claim 1, characterized in that, Acquire first point cloud data and second point cloud data, and register the first point cloud data and second point cloud data, wherein the first point cloud data and second point cloud data are point cloud data collected by roadside lidar and vehicle-mounted lidar, respectively, including: Acquire the first point cloud data and the second point cloud data; Ground plane features were extracted from the first point cloud data and the second point cloud data respectively, and the roadside ground plane normal vector and the vehicle-mounted ground plane normal vector were obtained by fitting. Construct a rotation matrix so that the vehicle-mounted ground plane normal vector is aligned with the roadside ground plane normal vector; The Z-axis of the vehicle coordinate system is calibrated to be parallel to the Z-axis of the roadside coordinate system, so that the vehicle-road coordinate system is geometrically coarsely aligned on the XOY plane.

3. The roadside cooperative positioning method according to claim 1, characterized in that, After performing two-dimensional projection mapping on the registered first and second point cloud data, a local topology map for the vehicle-mounted end and a set of candidate sub-maps for the roadside end are constructed, including: Constructing a local topology graph for the vehicle: with the vehicle itself as the central node. Neighboring vehicles within the vehicle's perception range are considered as neighboring nodes. , construct graph ;in, Represents a set of nodes. This represents the set of edges constructed based on the relative pose relationships between vehicles; Constructing a candidate sub-map set for the roadside: based on data detected by roadside lidar. Each vehicle target is analyzed sequentially, with each detected vehicle being... As the central node, construct candidate subgraphs .

4. The roadside cooperative positioning method according to claim 1 or 3, characterized in that, After performing two-dimensional projection mapping on the registered first and second point cloud data, the process of constructing a local topology map for the vehicle-mounted end and a candidate sub-map set for the roadside end, respectively, further includes: The bounding box is modified based on the prior vehicle dimensions; The geometric center point of the target is recalculated using the corrected bounding box; Update the coordinate positions of the corresponding graph nodes based on the geometric center point.

5. The roadside cooperative positioning method according to claim 1, characterized in that, The step of performing single-frame graph structure and attribute matching between the vehicle-mounted local topology map and the roadside candidate sub-graph set, in order to filter out candidate sub-graphs in the roadside candidate sub-graph set that have the same structure as the vehicle-mounted local topology map, includes: Select candidate subgraphs from the roadside candidate subgraph set that are consistent with the local topology structure of the vehicle-mounted terminal; The subgraph isomorphism algorithm is used to determine whether there is an isomorphic mapping relationship between the local topology graph of the vehicle end and the candidate subgraph of the roadside end; In the presence of isomorphic mappings, the similarity score of edge attributes is calculated using a Gaussian weighted function; Select roadside target candidates whose matching degree reaches a preset threshold.

6. The roadside cooperative positioning method according to claim 1, characterized in that, The process of constructing a temporal topology map sequence based on historical frame data, selecting key frames, and executing a multi-frame temporal joint matching strategy to determine roadside targets that match the local topology map on the vehicle end includes: Looking back from the current time Frames are used to construct the timing sequence of the vehicle-mounted terminal and the timing sequence of the roadside terminal; Calculate the similarity between any two frames in the vehicle-mounted time sequence and construct a similarity matrix; Summing is performed on each row of the similarity matrix, and the set of times when the summation value is lower than a set value is selected as the keyframe set; The cumulative matching score between the vehicle-mounted image and the roadside image is calculated in the set of keyframes, and the roadside target with the highest cumulative score is selected as the final matching result.

7. The roadside cooperative positioning method according to claim 1, characterized in that, The local coordinates of the roadside target in the roadside lidar coordinate system, based on the matched local coordinates and preset external parameters, are used to calculate the vehicle's positioning result in the world coordinate system, including: Based on the local coordinates of the successfully matched roadside target in the roadside lidar coordinate system; Using the pre-calibrated external parameters of the roadside lidar, the local coordinates are transformed to the world coordinate system to obtain the vehicle's positioning result.

8. A roadside cooperative positioning device, characterized in that, include: The data acquisition and registration module acquires first point cloud data and second point cloud data, and registers the first point cloud data and second point cloud data, wherein the first point cloud data and second point cloud data are point cloud data collected by roadside lidar and vehicle-mounted lidar, respectively. The topology graph construction module is used to perform two-dimensional projection mapping on the registered first point cloud data and second point cloud data, and then construct a local topology graph of the vehicle end and a set of candidate subgraphs of the roadside end, respectively. A single-frame matching module is used to perform single-frame graph structure and attribute matching between the local topology map of the vehicle terminal and the candidate sub-graph set of the roadside terminal, so as to filter out candidate sub-graphs in the roadside candidate sub-graph set that are consistent with the structure of the local topology map of the vehicle terminal. The temporal multi-frame joint matching module is used to construct a temporal topology map sequence based on historical frame data, select key frames and execute the temporal multi-frame joint matching strategy to determine the roadside targets that match the local topology map of the vehicle terminal. The positioning result calculation module is used to calculate the vehicle's positioning result in the world coordinate system based on the local coordinates of the matched roadside target in the roadside lidar coordinate system and preset external parameters.

9. An electronic device, characterized in that, include: Processor and memory; The memory stores computer programs that can be executed by the processor; When the processor executes the computer program, it implements the steps in the roadside cooperative positioning method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the roadside cooperative positioning method as described in any one of claims 1-7.

Images