Many-source GPS trajectory data aggregation method

Abstract

The invention discloses a crowd-source GPS track data aggregation method, and relates to the technical field of intelligent driving, and the method comprises the steps: obtaining crowd-source GPS track data, the crowd-source GPS track data comprising a plurality of pieces of GPS track data; track preprocessing is carried out on each piece of GPS track data of the public source GPS track data, a preprocessed track data set is generated, and each piece of preprocessed track data comprises a plurality of target sampling points; obtaining target sensing semantic information of each target sampling point in the preprocessed track data set; determining a road boundary of each target sampling point based on the target perception semantic information; and performing iterative aggregation processing on all the target sampling points based on the road boundaries of all the target sampling points to generate a road skeleton road network.

Description

Technical Field

[0001] This application relates to the field of intelligent driving technology, and in particular to a method for aggregating crowdsourced GPS trajectory data. Background Technology

[0002] In the fields of intelligent transportation and autonomous driving, constructing road network skeletons using crowdsourced GPS trajectory data has become an important technical approach. Current mainstream trajectory aggregation methods primarily rely on the spatial distribution characteristics of trajectory points. For example, density-based clustering algorithms are used to fit high-density trajectory point regions to road centerlines, or physical gravitational field models are introduced to simulate the interactions between trajectory points, causing them to converge towards the center. However, these methods only use pure GPS trajectory coordinates as input, and their aggregation process typically depends on preset empirical road width parameters. Due to the complex and varied nature of actual road widths, these methods are prone to errors when dealing with typical scenarios such as wide urban roads or parallel main and auxiliary roads. These errors can lead to the misinterpretation of a single wide road as multiple parallel paths, or the incorrect merging of spatially adjacent but physically isolated roads, resulting in a chaotic and inaccurate road network skeleton topology. Therefore, a crowdsourced GPS trajectory data aggregation method is urgently needed to address these problems. Summary of the Invention

[0003] The summary section introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. This summary section is not intended to limit the key and essential technical features of the claimed technical solutions, nor is it intended to determine the scope of protection of the claimed technical solutions.

[0004] Firstly, this application provides a method for aggregation of crowdsourced GPS trajectory data, including: Acquire crowdsourced GPS trajectory data, which includes multiple GPS trajectory data; Each GPS trajectory data point from the crowdsourced GPS trajectory data is preprocessed to generate a preprocessed trajectory data set, wherein each preprocessed trajectory data point includes multiple target sampling points; Obtain the target perception semantic information of each target sampling point in the preprocessed trajectory data set; Based on the target-aware semantic information, the road boundary of each target sampling point is determined; Based on the road boundaries of all target sampling points, all target sampling points are iteratively aggregated to generate a road skeleton network.

[0005] In some implementations, the step of performing trajectory preprocessing on each GPS trajectory data point of the crowdsourced GPS trajectory data to generate a preprocessed trajectory data set includes: For each GPS trajectory data in the crowdsourced GPS trajectory data, based on a preset thinning error threshold, the GPS trajectory data is thinned to generate thinned trajectory data. For each thinned trajectory data, linear interpolation is performed on the thinned trajectory data based on the preset maximum segmentation length to generate preprocessed trajectory data of the thinned trajectory data. Based on all the preprocessed trajectory data, the preprocessed trajectory data set is generated.

[0006] In some implementations, for each GPS trajectory data point of the crowdsourced GPS trajectory data, based on a preset thinning error threshold, the GPS trajectory data is thinned to generate thinned trajectory data, including: For each GPS trajectory data, based on the preset thinning error threshold, the Douglas-Puk algorithm is used to extract feature points from the GPS trajectory data to determine a set of trajectory feature points that characterize the geometry of the GPS trajectory data. Based on the set of trajectory feature points, the thinned trajectory data of the GPS trajectory data is generated.

[0007] In some implementations, the step of performing linear interpolation on each thinned trajectory data based on a preset maximum segmentation length to generate preprocessed trajectory data includes: For each thinned trajectory data, based on the preset maximum segmentation length, distance detection processing is performed on adjacent sampling points in the thinned trajectory data to determine the set of line segments to be interpolated; Based on the linear interpolation algorithm, each line segment in the set of line segments to be interpolated is divided into equal parts to determine the corresponding set of new sampling points. Based on the sampled points in the thinned trajectory data and the set of all newly added sampled points, the preprocessed trajectory data of the thinned trajectory data is generated.

[0008] In some implementations, obtaining the target-aware semantic information of each target sampling point in the preprocessed trajectory data set includes: For each target sampling point in the preprocessed trajectory data set, based on the target trajectory identifier and target acquisition time of the target sampling point, the target perception semantic information of the target sampling point is obtained from the perception database; The target perception semantic information includes lane line semantic information and road edge line semantic information; The perception database includes multiple sampling points and perception semantic information corresponding to the trajectory identifier and acquisition time of each sampling point.

[0009] In some implementations, the road edge semantic information includes the lateral distance of the road edge, and the lane line semantic information includes the lateral distance of the lane lines. Determining the road boundary of each target sampling point based on the target perception semantic information includes: For each target sampling point, based on the target perception semantic information, it is determined whether the target sampling point has road edge semantic information; If the aforementioned road edge semantic information exists, the road boundary of the target sampling point is determined based on the lateral distance of the road edge; or, If the road edge semantic information does not exist, the road boundary of the target sampling point is determined based on the lane line lateral distance; or, If neither the road edge semantic information nor the lane line semantic information exists, the road boundary of the target sampling point is determined based on the preset default width.

[0010] In some implementations, determining the road boundary of the target sampling point based on the lateral distance of the lane lines includes: Based on the lateral distance of the lane lines, determine the lateral distance of the outermost lane line; The road boundary of the target sampling point is determined based on the lateral distance of the outermost lane line and the preset lane width.

[0011] In some implementations, the iterative aggregation process of all target sampling points based on the road boundaries of all target sampling points to generate a road skeleton network includes: Based on the preprocessed trajectory data of all target sampling points, the heading information of each target sampling point is determined; Based on the road boundary and the heading information, multiple rounds of iterative aggregation processing are performed on all target sampling points, and the current position of each target sampling point is updated step by step in each round of iterative aggregation processing; When the update status or iteration count of all target sampling points meets the preset convergence condition during the iteration process, a road skeleton network is generated based on the updated positions of all target sampling points.

[0012] In some implementations, the step of each round of iterative aggregation processing includes: For each target sampling point across all target sampling points, perform the following steps: Based on the heading information of the target sampling point and the heading information of other target sampling points, direction consistency filtering is performed to determine the first candidate point set; Based on the road boundary of the target sampling point and the road boundary of each candidate point in the first candidate point set, spatial range filtering is performed to determine the projection point set; Based on the set of projection points and the target sampling point, determine the aggregation center point; The moving distance is determined based on the current position of the aggregation center point and the target sampling point; Based on the movement distance and the preset movement limit, the target update position of the target sampling point is determined; Based on the target update location, update the current position of the target sampling point; Based on the updated current location, update the heading information and road boundary of the target sampling point.

[0013] In some implementations, spatial range filtering is performed based on the road boundary of the target sampling point and the road boundary of each candidate point in the first candidate point set to determine the projection point set, including: For each candidate point in the first candidate point set, perform the following steps: The target sampling point is vertically projected onto the trajectory line segment where the candidate point is located to determine the coordinates of the projected point. The trajectory line segment is a subset of the preprocessed trajectory data of the candidate point. If the coordinates of the projection point are within the road boundary of the target sampling point, and / or the road boundary of the candidate point overlaps with the road boundary of the target sampling point in an area with a width greater than or equal to the preset lane width, the candidate point is added to the projection point set.

[0014] Secondly, this application proposes a crowdsourced GPS trajectory data aggregation device, comprising: A trajectory data acquisition unit is used to acquire crowdsourced GPS trajectory data, which includes multiple GPS trajectory data. The trajectory data processing unit is used to perform trajectory preprocessing on each GPS trajectory data of the crowdsourced GPS trajectory data to generate a preprocessed trajectory data set, wherein each preprocessed trajectory data includes multiple target sampling points; A semantic perception acquisition unit is used to acquire the target semantic perception information of each target sampling point in the preprocessed trajectory data set. The road boundary determination unit is used to determine the road boundary of each target sampling point based on the target perception semantic information. The skeleton road network generation unit is used to iteratively aggregate all target sampling points based on the road boundaries of all target sampling points to generate a road skeleton road network.

[0015] Thirdly, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program stored in the memory to implement the steps of the crowdsourced GPS trajectory data aggregation method of any of the first aspects described above.

[0016] Fourthly, this application also proposes a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of the crowdsourced GPS trajectory data aggregation method of any one of the first aspects.

[0017] In summary, this application provides a method for aggregating crowdsourced GPS trajectory data. By acquiring crowdsourced GPS trajectory data and performing trajectory preprocessing, the data density is optimized while preserving trajectory morphological features. Then, by combining the perceptual semantic information corresponding to each sampling point, its precise road boundary is determined. All sampling points are iteratively aggregated to ultimately generate an accurate road network skeleton. This method introduces objective perceptual semantic information to define road boundaries, enabling the trajectory aggregation process to effectively distinguish between discrete trajectories within the same road and adjacent trajectories from different physical roads, thereby improving the accuracy of road network extraction in complex scenarios. Simultaneously, the iterative aggregation mechanism progressively optimizes the trajectory point positions under road boundary constraints, ensuring that the generated road network skeleton accurately reflects the actual geometric shape of the road while maintaining the original trajectory direction, thus enhancing the method's adaptability and reliability. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating a crowdsourced GPS trajectory data aggregation method provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a single target sampling point in one round of iterative aggregation processing provided in an embodiment of this application; Figure 3 A schematic diagram of a crowdsourced GPS trajectory data aggregation device provided in this application embodiment; Figure 4 This is a schematic diagram of a crowdsourced GPS trajectory data aggregation device provided in an embodiment of this application. Detailed Implementation

[0019] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0020] Please see Figure 1 This is a flowchart illustrating a crowdsourced GPS trajectory data aggregation method provided in an embodiment of this application, which may specifically include: S110. Obtain crowdsourced GPS trajectory data, which includes multiple GPS trajectory data. For example, in step S110, crowdsourced GPS trajectory data covering a wide geographical area is constructed by accessing GPS positioning information continuously collected from multiple vehicle terminals. This crowdsourced GPS trajectory data contains a large number of discrete trajectory point sequences with spatiotemporal attributes, with each GPS trajectory recording the continuous position changes of a single vehicle over a period of time. This trajectory data reflects the driving patterns of vehicles in the actual road network; however, due to differences in acquisition equipment, signal drift, and the complexity of traffic behavior, the raw data contains a large number of redundant and noisy points, requiring subsequent processing to support the accurate generation of the road framework.

[0021] S120. Perform trajectory preprocessing on each GPS trajectory data from the crowdsourced GPS trajectory data to generate a preprocessed trajectory data set, wherein each preprocessed trajectory data includes multiple target sampling points; For example, in step S120, each GPS trajectory data is preprocessed to optimize the data structure and density to meet subsequent aggregation requirements while preserving the key geometric shape of the original trajectory. This process first removes redundant points from the trajectory using a thinning algorithm, retaining feature points that characterize the road direction to reduce the amount of data. Then, interpolation is performed on adjacent points with excessively large spacing after thinning to increase the number of sampling points, ensuring the trajectory has sufficient density to describe road details. Finally, preprocessed trajectory data consisting of multiple target sampling points is formed.

[0022] S130. Obtain the target perception semantic information of each target sampling point in the preprocessed trajectory data set; For example, in step S130, by associating the trajectory identifier and acquisition time of each target sampling point in the preprocessed trajectory data set, the corresponding target perception semantic information is retrieved and obtained from the pre-constructed perception database. This perception semantic information originates from the environmental data synchronously perceived and recorded by the on-board sensor system when the autonomous vehicle collects GPS trajectory data, mainly including the geometric description of key road elements such as lane lines and road edge lines around the vehicle in the vehicle coordinate system.

[0023] S140. Based on target-aware semantic information, determine the road boundary of each target sampling point; For example, in step S140, the road boundary at the location of each target sampling point is determined based on the target perception semantic information corresponding to that point. Specifically, firstly, it is determined whether there is directly usable road edge semantic information at the sampling point; if so, its lateral distance value is directly used as the boundary reference; if there is no road edge but there is lane line semantic information, the boundary is estimated by extending a preset lane width outward based on the position of the outermost lane line; for scenarios where semantic information is completely missing, a preset default width is used as the boundary reference. Through this process, each sampling point is given boundary constraints that conform to its actual road environment, providing precise spatial range limits for subsequent trajectory aggregation.

[0024] S150. Based on the road boundaries of all target sampling points, perform iterative aggregation processing on all target sampling points to generate a road skeleton network.

[0025] For example, in step S150, based on the road boundaries of all target sampling points, an iterative aggregation process is performed on all target sampling points to generate a road skeleton network. This process gradually optimizes the spatial position of the sampling points through multiple iterations. In each iteration, each target sampling point selects candidate points that meet the requirements of directional consistency and spatial range from other sampling points based on its heading information and road boundary constraints, and then calculates the aggregation center and updates its own position with a limited movement range accordingly. By repeatedly executing the above process and updating the heading and boundary information synchronously, all sampling points gradually converge to a position that can represent the road centerline. Based on the converged sampling point positions, a road skeleton network reflecting the real road shape is generated.

[0026] In summary, this embodiment acquires crowdsourced GPS trajectory data and performs trajectory preprocessing, optimizing data density while preserving trajectory morphological features. Then, it combines the perceptual semantic information corresponding to each sampling point to determine road boundaries, and iteratively aggregates all sampling points based on boundary constraints to ultimately generate an accurate road network skeleton. This method dynamically defines road boundaries by introducing objective perceptual semantic information, enabling the trajectory aggregation process to effectively distinguish between discrete trajectories within the same road and adjacent trajectories on different physical roads, thereby improving the accuracy of road network extraction in complex scenarios such as wide roads and parallel roads. Simultaneously, the multi-round iterative aggregation mechanism based on road boundaries and heading information gradually optimizes the spatial position of trajectory points while maintaining the original trajectory direction. This avoids trajectory deformation caused by excessive single movement and ensures that the generated road network skeleton accurately reflects the actual geometric shape of the road, overcoming the limitations of traditional methods that rely on preset road widths.

[0027] In some instances, each GPS trajectory data point from the crowdsourced GPS trajectory data undergoes trajectory preprocessing to generate a preprocessed trajectory data set, including: For each GPS trajectory data from the crowdsourced GPS trajectory data, based on a preset thinning error threshold, the GPS trajectory data is thinned to generate thinned trajectory data. For each thinned trajectory data, linear interpolation is performed on the thinned trajectory data based on the preset maximum segmentation length to generate preprocessed trajectory data of the thinned trajectory data. A set of preprocessed trajectory data is generated based on all the preprocessed trajectory data.

[0028] For example, in the trajectory preprocessing stage, each GPS trajectory data in the crowdsourced GPS trajectory data is first thinned based on a preset thinning error threshold. This process uses the Douglas-Puk algorithm, which recursively compares the perpendicular distance between trajectory points and the lines connecting adjacent feature points, retains feature points whose distance exceeds the preset thinning error threshold, and removes redundant collinear points. This reduces the amount of trajectory data while preserving the original geometry of the trajectory to the greatest extent possible, thus generating thinned trajectory data.

[0029] Next, for each thinned trajectory data, linear interpolation is performed based on a preset maximum segmentation length. This process detects the distance between adjacent sampling points in the thinned trajectory data. When the distance exceeds the preset maximum segmentation length, new sampling points are inserted at equal intervals between adjacent points using a linear interpolation algorithm until the distance between all adjacent sampling points is less than the preset maximum segmentation length. This increases the density of trajectory sampling points while maintaining the trajectory shape, generating preprocessed trajectory data.

[0030] Finally, all the preprocessed trajectory data after thinning and interpolation are integrated to form a preprocessed trajectory data set. Each preprocessed trajectory data in this set contains multiple target sampling points, which retain the key geometric features of the original trajectory and have an appropriate spatial distribution density.

[0031] In summary, the embodiments of this application, through the above-described trajectory preprocessing, can effectively optimize the trajectory data structure while preserving the key geometric features of the crowdsourced GPS trajectories. This not only reduces data redundancy and computational burden through thinning, but also ensures that the sampling point density meets the requirements of subsequent processing through interpolation, providing high-quality data input for trajectory aggregation methods, while balancing computational efficiency and processing accuracy.

[0032] In some instances, for each GPS trajectory data point from the crowdsourced GPS trajectory data, based on a preset thinning error threshold, the GPS trajectory data is thinned to generate thinned trajectory data, including: For each GPS trajectory data, based on a preset thinning error threshold, the Douglas-Puk algorithm is used to extract feature points from the GPS trajectory data to determine the set of trajectory feature points used to characterize the geometry of the GPS trajectory data. Based on the set of trajectory feature points, thinned trajectory data of the GPS trajectory data is generated.

[0033] For example, in the thinning process, the first and last points of the trajectory curve are virtually connected to form a reference straight line. Then, the vertical distance from all intermediate points in the trajectory curve to this reference straight line is calculated, and the maximum distance value is determined. This maximum distance value is compared with a preset thinning error threshold. If the maximum distance value is less than the preset thinning error threshold, it indicates that the overall deviation of the current trajectory curve segment from the reference straight line is within the allowable range, and all intermediate points within this segment can be discarded. If the maximum distance value is greater than or equal to the preset thinning error threshold, the coordinate point corresponding to the maximum distance value is retained as a key feature point, and this point is used as a dividing line to divide the original trajectory curve into two sub-segments. The above process of distance calculation, comparison, and segmentation is recursively repeated for each sub-segment until all trajectory segments have been processed. By controlling the size of the preset thinning error threshold, the granularity of trajectory thinning can be adjusted. Finally, all the selected key feature points constitute a set of trajectory feature points that can characterize the geometry of the GPS trajectory data. Based on this set, corresponding thinned trajectory data is generated, achieving the goal of compressing the data volume while maintaining the macroscopic shape of the trajectory.

[0034] In some instances, for each thinned trajectory data, linear interpolation is performed on the thinned trajectory data based on a preset maximum segmentation length to generate preprocessed trajectory data for that thinned trajectory data, including: For each thinned trajectory data, based on the preset maximum segmentation length, distance detection processing is performed on adjacent sampling points in the thinned trajectory data to determine the set of line segments to be interpolated; Based on the linear interpolation algorithm, each line segment in the set of line segments to be interpolated is divided into equal parts to determine the corresponding set of new sampling points. Based on the sampled points in the thinned trajectory data and the set of all newly added sampled points, preprocessed trajectory data of the thinned trajectory data is generated.

[0035] For example, during the linear interpolation process for each thinned trajectory data, distance detection is first performed on adjacent sampling points in the thinned trajectory data based on a preset maximum segmentation length. The specific steps of this distance detection process are as follows: The Euclidean distance between each pair of adjacent sampling points in the thinned trajectory data is calculated sequentially, and this distance value is compared with the preset maximum segmentation length. If the distance between a pair of adjacent sampling points is greater than the preset maximum segmentation length, the trajectory segment formed by that pair of sampling points is selected. After traversing all adjacent sampling point pairs, the set of interpolation segments composed of all such trajectory segments can be determined. The preset maximum segmentation length takes into account the scale characteristics of the actual road structure and is usually set to a value greater than a preset lane width to ensure that the interpolation operation can provide a sufficient density of sampling points for road boundary definition and trajectory aggregation.

[0036] Based on a linear interpolation algorithm, each line segment in the set of line segments to be interpolated is equally divided to determine the corresponding set of new sampling points. Specifically, for each line segment to be interpolated, the number of new sampling points to be inserted is determined according to the ratio of its length to the preset maximum segment length. Then, new sampling points are evenly inserted between adjacent sampling points according to the linear interpolation formula. This equal division process ensures that the new sampling points are evenly distributed on the line segment, which not only maintains the geometric features of the original trajectory but also effectively improves the spatial resolution of the trajectory data.

[0037] Based on the original sampling points and the set of all newly added sampling points in the thinned trajectory data, preprocessed trajectory data is generated. By integrating and recombining the original and newly added sampling points in chronological order, a continuous trajectory containing more sampling points while maintaining the original direction is formed. This preprocessed trajectory data not only preserves the key geometric features of the thinned trajectory but also compensates for the problem of excessively sparse local sampling points that may have been caused by thinning through interpolation.

[0038] In summary, the embodiments of this application, through the above-described linear interpolation process, can effectively increase the density of trajectory sampling points while maintaining the original shape of the trajectory, ensuring that each road segment has a sufficient number of sampling points to describe its spatial characteristics.

[0039] In some instances, target-aware semantic information is obtained for each target sampling point in the preprocessed trajectory dataset, including: For each target sampling point in the preprocessed trajectory data set, based on the target trajectory identifier and target acquisition time of the target sampling point, the target perception semantic information of the target sampling point is obtained from the perception database. The target perception semantic information includes lane line semantic information and road edge line semantic information. The perception database includes multiple sampling points and perception semantic information corresponding to the trajectory identifier and acquisition time of each sampling point.

[0040] For example, in the process of acquiring the target perception semantic information of each target sampling point in the preprocessed trajectory data set, for each target sampling point in the set, a retrieval and matching operation is performed from the pre-built perception database based on the target trajectory identifier associated with the target sampling point and its target acquisition time in the original data stream. This perception database, as a structured data storage system, internally records a large amount of sampling point information reported by numerous vehicle perception systems, and associates and stores the corresponding trajectory identifier, acquisition time, and environmental semantic information perceived and parsed by the vehicle sensor system at that time for each sampling point entry. Specifically, the retrieval logic uses the target trajectory identifier and target acquisition time of the current target sampling point as the query key, performing exact matching or nearest neighbor matching in the perception database to obtain the environmental semantic data perceived by the vehicle at that specific time and on that specific trajectory. This data constitutes the target perception semantic information, the core content of which includes, but is not limited to, lane line semantic information and road edge line semantic information of the vehicle's location. This semantic information typically exists in the form of geometric parameters relative to the vehicle's body coordinate system.

[0041] In summary, this application's embodiments, through trajectory identification and acquisition time, effectively fuse discrete GPS trajectory points with vehicle-perceived environmental information, providing environmental context for subsequent steps. This method ensures that each trajectory sampling point not only includes its spatial location attributes but also the physical boundary description of its surrounding environment. This allows the trajectory aggregation process to overcome the limitations of relying solely on the spatial distribution of GPS coordinates, instead utilizing real, real-time perceived road structure information as aggregation constraints. This plays a crucial role in improving the accuracy of road skeleton generation at the data source level.

[0042] In some instances, road edge semantic information includes the lateral distance of the road edge, and lane line semantic information includes the lateral distance of the lane lines. Based on the target perception semantic information, the road boundary of each target sampling point is determined, including: For each target sampling point, based on the target perception semantic information, it is determined whether the target sampling point has road edge semantic information; If road edge semantic information exists, the road boundary of the target sampling point is determined based on the lateral distance of the road edge; or, if road edge semantic information does not exist, the road boundary of the target sampling point is determined based on the lateral distance of the lane line; or, if neither road edge semantic information nor lane line semantic information exists, the road boundary of the target sampling point is determined based on the preset default width.

[0043] For example, the presence of road edge semantic information at the target sampling point can be determined based on the target perception semantic information. Road edge semantic information typically includes the lateral distance of the road edge in the vehicle coordinate system. This lateral distance is defined as the Y-value of the road edge point coordinate in the vehicle coordinate system (with the vehicle's forward direction as the X-axis), with the vehicle's center of gravity as the origin, pointing to the left or right side of the vehicle along the Y-axis.

[0044] If the determination result indicates the existence of road edge semantic information, the single or double road boundary of the target sampling point is directly determined based on the value of the lateral distance of the road edge. If the determination result indicates the absence of road edge semantic information, the existence of lane line semantic information is further checked. This information also includes the lateral distance of the lane lines in the vehicle coordinate system. If it exists, the lateral distance of the outermost lane line is determined from all available lane line lateral distances. Then, based on the lateral distance of the outermost lane line, a preset fixed length is added outward. This fixed length is usually set to a preset lane width, and this is used as the estimated road boundary of the target sampling point. If neither road edge semantic information nor lane line semantic information exists at the target sampling point, such as on a suburban road without lane markings, the road boundary of the target sampling point is directly determined based on a preset default width value. This default width is usually set according to empirical values ​​for typical road scenarios and can be set to 1.5 times the preset lane width.

[0045] In summary, the embodiments of this application, through hierarchical judgment and processing logic, can assign appropriate road boundaries to each target sampling point. This method prioritizes the most reliable road edge information, uses lane line information for reasonable estimation when edge information is missing, and activates default rules when semantic information is completely lacking, ensuring the robustness and adaptability of road boundary determination. This boundary determination method based on actual perception information overcomes the limitations of fixed preset road widths in traditional methods, enabling each trajectory sampling point to obtain spatial constraints consistent with its specific spatiotemporal context, thereby improving the geometric accuracy and topological correctness of road skeleton generation.

[0046] In some instances, based on the road boundaries of all target sampling points, an iterative aggregation process is performed on all target sampling points to generate a road skeleton network, including: Based on the preprocessed trajectory data of all target sampling points, the heading information of each target sampling point is determined; Based on road boundary and heading information, multiple rounds of iterative aggregation processing are performed on all target sampling points, and the current position of each target sampling point is updated step by step in each round of iterative aggregation processing; When the update status or iteration count of all target sampling points meets the preset convergence condition during the iteration process, a road skeleton network is generated based on the updated positions of all target sampling points.

[0047] For example, in the process of iteratively aggregating road boundaries based on all target sampling points to generate a road skeleton network, the heading information of each target sampling point needs to be determined first based on the preprocessed trajectory data of all target sampling points. This heading information is obtained by calculating the direction angle of the vector formed by each target sampling point and its adjacent subsequent sampling points in the preprocessed trajectory data, thereby assigning each point an attribute representing its instantaneous driving direction.

[0048] Based on the established road boundaries and heading information, multi-round iterative aggregation processing is performed on all target sampling points. In each iteration, each target sampling point is traversed, and a series of operations are performed to progressively optimize its spatial position. Specifically, for the currently processed target sampling point, its heading information is first used to perform directional consistency filtering with the heading information of other target sampling points, selecting candidate points whose heading angle is less than a preset angle threshold to form a first candidate point set. Next, spatial range filtering is performed based on the road boundary of the current target sampling point and the road boundary of each candidate point in the first candidate point set to obtain a projection point set. Then, based on the position of this projection point set and the current target sampling point itself, an aggregation center point is determined by calculating its coordinate mean. Subsequently, the movement distance between the current target sampling point and the aggregation center point is calculated, and the movement distance is constrained according to a preset movement limit ratio, thereby determining the target update position of the current target sampling point in this iteration. Finally, the current position of the target sampling point is overwritten using the updated target position, and its heading information is recalculated based on its new position and its relationship with neighboring points in the preprocessed trajectory data. Simultaneously, its road boundaries are updated synchronously as the position moves to maintain spatial constraint consistency. This process is executed sequentially for all target sampling points in this iteration.

[0049] The iterative process continues until a preset convergence condition is met. The preset convergence condition includes one or more of the following: the number of iterations reaches a preset maximum iteration threshold, which is usually set to a fixed value, such as 20, based on experience; or the maximum lateral movement distance of all target sampling points in a single iteration is less than a first preset distance threshold, which is usually set to a small value, such as 0.1 meters; or the average lateral movement distance of all target sampling points in a single iteration is less than a second preset distance threshold, which is usually set to a smaller value, such as 0.05 meters. After any convergence condition is met, a road skeleton network is generated based on the updated positions of all target sampling points. Specifically, the updated target sampling points are fitted into continuous line segments according to their connection relationships in the preprocessed trajectory data, forming the final road skeleton network.

[0050] In summary, the iterative aggregation process in this embodiment introduces dual filtering of direction and spatial range, supplemented by a constrained movement update strategy, ensuring that trajectory points can stably and progressively converge towards the road centerline under the constraints of the actual road physical boundaries and driving directions. This method effectively avoids the path splitting problem caused by traditional aggregation algorithms in wide road scenarios and the path merging error caused in parallel road scenarios, thereby improving the geometric accuracy and topological correctness of the generated road skeleton network. Simultaneously, the updates to heading and boundaries during the iteration process maintain the stability of the trajectory morphology, enabling the final result to more realistically reflect the actual road direction and structure.

[0051] Please see Figure 2 This is a schematic diagram of a single target sampling point in an iterative aggregation processing structure provided in an embodiment of this application, including: Point A represents the target sampling point currently undergoing aggregation processing. Points B and C represent other target sampling points retained after direction filtering, and their trajectory segments meet the requirement of consistency with the heading of point A. Point B1 is the projection point obtained by vertically projecting target sampling point A onto the trajectory segment containing point B, and point C1 is the projection point obtained by vertically projecting target sampling point A onto the trajectory segment containing point C. The coordinates of points A, B1, and C1 are included in the calculation, and the aggregation center point D for this iteration is determined by calculating the average of these coordinates. The vertical projection point of point D in the Y-axis direction is D1, and the straight-line distance L between point A and point D1 is the theoretical lateral distance that the current target sampling point A needs to move towards the aggregation center. In the actual update process, to avoid trajectory deformation caused by excessive single movement, point A is not moved directly to point D1. Based on a preset movement limit ratio, such as 1 / 10 or 8 / 10 of this distance L, the movement distance of point A in this iteration is used, thereby achieving a gradual adjustment of the target sampling point's position.

[0052] In some instances, the steps of each iteration of the aggregation process include: For each target sampling point across all target sampling points, perform the following steps: Based on the heading information of the target sampling point and the heading information of other target sampling points, direction consistency filtering is performed to determine the first candidate point set; Based on the road boundary of the target sampling point and the road boundary of each candidate point in the first candidate point set, spatial range filtering is performed to determine the projection point set, including: For each candidate point in the first candidate point set, perform the following steps: The target sampling point is vertically projected onto the trajectory line segment where the candidate point is located to determine the coordinates of the projected point. The trajectory line segment is a subset of the preprocessed trajectory data of the candidate point. If the coordinates of the projected point are within the road boundary of the target sampling point, and / or the road boundary of the candidate point overlaps with the road boundary of the target sampling point in an area with a width greater than or equal to the preset lane width, the candidate point is added to the projected point set.

[0053] Based on the set of projection points and the target sampling point, determine the aggregation center point; The moving distance is determined based on the current position of the aggregation center point and the target sampling point; Based on the movement distance and preset movement restrictions, the target update location of the target sampling point is determined; Update the current position of the target sampling point based on the target update location; Based on the updated current location, update the heading information and road boundary of the target sampling point.

[0054] For example, in each iteration of the aggregation process, for each target sampling point among all target sampling points, a direction consistency filter is first performed based on the heading information of that target sampling point and the heading information of other target sampling points. This step calculates the heading angle between the trajectory segment of the current target sampling point and the trajectory segments of other target sampling points, and compares this angle with a preset angle threshold to filter out other target sampling points whose heading differences are within an acceptable range, forming a first candidate point set. The heading information is represented by the vector direction angle formed by the target sampling point and its subsequent adjacent sampling points. The preset angle threshold is set according to the actual road direction variation characteristics, for example, 30 degrees, to ensure that the selected candidate points have the same driving direction as the current target sampling point.

[0055] Based on the road boundary of the target sampling point and the road boundaries of each candidate point in the first candidate point set, spatial range filtering is performed to determine the projection point set. For each candidate point in the first candidate point set, vertical projection processing is performed, that is, the current target sampling point is vertically projected onto the trajectory line segment where the candidate point is located, and the coordinates of the projection point are calculated. Here, the trajectory line segment refers to the line segment formed by the candidate point and its subsequent adjacent sampling points in the preprocessed trajectory data. Then, it is determined whether the coordinates of the projection point meet one of the following conditions: the coordinates of the projection point are within the road boundary of the current target sampling point, or the road boundary of the candidate point and the road boundary of the current target sampling point have an overlapping area with a width greater than or equal to the preset lane width. If either condition is met, the candidate point is added to the projection point set. The preset lane width is usually set according to standard road design specifications, for example, 3.5 meters, to ensure that trajectories with sufficient overlapping areas can be correctly identified as trajectories of the same road.

[0056] Based on the set of projection points and the target sampling point, the aggregation center point is determined by calculating the average coordinates of these points. Specifically, the coordinates of the current target sampling point are included in the calculation along with the coordinates of all projection points in the projection point set. The coordinates of the aggregation center point are obtained by calculating the arithmetic mean of these points on the two-dimensional plane. This aggregation center point represents the ideal center position of the current target sampling point and its aggregateable neighboring points in the current iteration, providing a target reference for subsequent position updates.

[0057] Based on the current positions of the aggregation center point and the target sampling point, the relative positional relationship between the two in space is calculated. This method mainly focuses on the lateral position adjustment, which has a more significant impact on the road skeleton morphology. This lateral movement distance is specifically represented by the projection component of the aggregation center point along the Y-axis in the local coordinate system of the current sampling point. Figure 2 As shown, this lateral movement distance is the straight-line distance L between point A and the perpendicular projection point D1 of point D, and its direction is perpendicular to the direction of the current trajectory segment. This distance vector clearly indicates the theoretical magnitude and direction of the lateral movement that the current target sampling point A needs to make towards the aggregation center point D.

[0058] Based on the movement distance and a preset movement limit, the target update position of the target sampling point is determined. The preset movement limit is typically set as a proportional coefficient of the movement distance, such as 1 / 10 or 8 / 10. This coefficient is chosen to balance trajectory shape preservation and convergence speed. Multiplying the movement distance by this coefficient yields the actual allowed movement amount. Then, the coordinates of the current target sampling point are adjusted according to the movement direction to determine its target update position in this iteration. This mechanism of limiting the movement amplitude ensures gradual adjustment of the trajectory point position, avoiding trajectory deformation caused by excessive single movement.

[0059] Based on the target update location, update the current position of the target sampling point. This step uses the calculated target update coordinates to overwrite the original coordinates of the current target sampling point, completing the position adjustment in this iteration. It is important to note that during the iteration process, the position updates of all target sampling points are based on the state after the previous iteration, i.e., a synchronous update strategy is adopted to ensure that the iteration order does not affect the final aggregation result.

[0060] Based on the updated current position, the heading information and road boundary of the target sampling point are updated. The heading information is updated by recalculating the vector direction angle formed by the target sampling point and its subsequent adjacent sampling points in the preprocessed trajectory data. The road boundary is updated by maintaining the original spatial relationship between the road boundary and the sampling point, based on the new position of the target sampling point. That is, as the sampling point moves, its road boundary also shifts accordingly, thus maintaining accurate spatial constraints. This update mechanism ensures that in subsequent iterations, direction selection and range constraints can be based on the latest trajectory shape and spatial relationships.

[0061] In summary, the embodiments of this application, through the aforementioned iterative aggregation process, gradually converge scattered trajectory points towards the road centerline while maintaining the original trajectory shape. This process, by introducing a dual filtering mechanism of directional consistency and spatial range, combined with a constrained position update strategy, avoids the path splitting problem that occurs in wide-road scenarios and the path merging error that occurs in parallel-road scenarios, thereby improving the accuracy and topological correctness of the generated road skeleton network. Simultaneously, the progressive update method ensures the stability of the trajectory aggregation process, enabling the generated road network skeleton to accurately reflect the geometric characteristics of the actual roads.

[0062] Please see Figure 3 The diagram below illustrates the structure of a crowdsourced GPS trajectory data aggregation device provided in this application embodiment, comprising: The trajectory data acquisition unit 21 is used to acquire crowdsourced GPS trajectory data, which includes multiple GPS trajectory data. The trajectory data processing unit 22 is used to perform trajectory preprocessing on each GPS trajectory data of the crowdsourced GPS trajectory data to generate a preprocessed trajectory data set, wherein each preprocessed trajectory data includes multiple target sampling points; The semantic perception acquisition unit 23 is used to acquire the target semantic perception information of each target sampling point in the preprocessed trajectory data set; The road boundary determination unit 24 is used to determine the road boundary of each target sampling point based on target perception semantic information; The skeleton road network generation unit 25 is used to perform iterative aggregation processing on all target sampling points based on the road boundaries of all target sampling points to generate a road skeleton road network.

[0063] Please see Figure 4 This application also provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, it implements the steps of the crowdsourced GPS trajectory data aggregation method.

[0064] Since the electronic device described in this embodiment is the device used to implement the crowdsourced GPS trajectory data aggregation device in the embodiments of this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method described in the embodiments of this application. Therefore, how the electronic device implements the method in the embodiments of this application will not be described in detail here. Any device used by those skilled in the art to implement the method in the embodiments of this application falls within the scope of protection of this application.

[0065] In practice, when the computer program 311 is executed by the processor, it can implement any of the embodiments corresponding to the first aspect.

[0066] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0067] Those skilled in the art will understand that embodiments of this application can provide methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media containing computer-readable program code.

[0068] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0069] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0070] These computer program instructions can also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0071] This application also provides a computer program product, which includes computer software instructions that, when executed on a processing device, cause the processing device to perform... Figure 1 The flowchart of a crowdsourced GPS trajectory data aggregation method in the corresponding embodiment.

[0072] A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, computer instructions may be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium may be any usable medium that a computer can store or a data storage device such as a server or data center that integrates one or more usable media. The usable medium may be a magnetic medium, an optical medium, or a semiconductor medium, etc.

[0073] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0074] In the several embodiments provided in this application, 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; multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Furthermore, the mutual couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[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 this embodiment according to actual needs.

[0076] Furthermore, the functional units in the various embodiments of this application 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 units described above can be implemented in the form of hardware and / or software functional units.

[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 computer-readable storage medium. Based on this understanding, the technical solution of this application, 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 and includes several instructions to cause a computer device to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, magnetic disks, or optical disks.

[0078] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

[0079] Although preferred embodiments have been described in this specification, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications that fall outside the scope of this specification.

[0080] Obviously, those skilled in the art can make various modifications to this specification without departing from its spirit and scope. Therefore, this specification also intends to include any modifications that fall within the scope of the claims and their equivalents.

Claims

1. A method for aggregating crowdsourced GPS trajectory data, characterized in that, The method comprises the following steps: obtain crowd-sourced GPS trajectory data, the crowd-sourced GPS trajectory data comprising a plurality of GPS trajectory data; perform trajectory preprocessing on each GPS trajectory data of the crowd-sourced GPS trajectory data to generate a set of preprocessed trajectory data, wherein each preprocessed trajectory data comprises a plurality of target sampling points; obtain target perception semantic information of each target sampling point in the set of preprocessed trajectory data; determine road boundaries of each target sampling point based on the target perception semantic information; perform iterative aggregation processing on all target sampling points based on the road boundaries of all target sampling points to generate a road skeleton network.

2. The method of claim 1, wherein, The method of performing trajectory preprocessing on each GPS trajectory data of the crowd-sourced GPS trajectory data to generate a set of preprocessed trajectory data comprises the following steps: for each GPS trajectory data of the crowd-sourced GPS trajectory data, perform thinning processing on the GPS trajectory data based on a preset thinning error threshold to generate thinned trajectory data of the GPS trajectory data; for each thinned trajectory data, perform linear interpolation processing on the thinned trajectory data based on a preset maximum segmentation length to generate preprocessed trajectory data of the thinned trajectory data; generate the set of preprocessed trajectory data based on all the preprocessed trajectory data.

3. The method of claim 2, wherein, The method of performing thinning processing on each GPS trajectory data of the crowd-sourced GPS trajectory data based on a preset thinning error threshold to generate thinned trajectory data of the GPS trajectory data comprises the following steps: for each GPS trajectory data, perform feature point extraction processing on the GPS trajectory data based on the preset thinning error threshold using a Douglas-Peucker algorithm to determine a set of trajectory feature points for representing the geometric shape of the GPS trajectory data, and generate the thinned trajectory data of the GPS trajectory data based on the set of trajectory feature points.

4. The method of claim 2, wherein, The method of performing linear interpolation processing on each thinned trajectory data based on a preset maximum segmentation length to generate preprocessed trajectory data of the thinned trajectory data comprises the following steps: for each thinned trajectory data, perform distance detection processing on adjacent sampling points in the thinned trajectory data based on the preset maximum segmentation length to determine a set of to-be-interpolated line segments; based on a linear interpolation algorithm, perform equal division cutting processing on each line segment in the set of to-be-interpolated line segments to determine a corresponding set of newly added sampling points; generate the preprocessed trajectory data of the thinned trajectory data based on the sampling points in the thinned trajectory data and all the sets of newly added sampling points.

5. The method of claim 1, wherein, The method of obtaining target perception semantic information of each target sampling point in the set of preprocessed trajectory data comprises the following steps: for each target sampling point in the set of preprocessed trajectory data, obtain target perception semantic information of the target sampling point from a perception database based on a target trajectory identifier and a target collection time of the target sampling point; wherein the target perception semantic information comprises lane line semantic information and road edge line semantic information; the perception database comprises a plurality of sampling points and perception semantic information corresponding to the trajectory identifier and collection time of each sampling point.

6. The method of claim 5, wherein, The road edge semantic information comprises a road edge lateral distance, and the lane line semantic information comprises a lane line lateral distance. For each target sampling point, based on the target perception semantic information, it is determined whether the target sampling point has road edge semantic information. If the target sampling point has the road edge semantic information, a road boundary of the target sampling point is determined based on the road edge lateral distance. If the target sampling point does not have the road edge semantic information, a road boundary of the target sampling point is determined based on the lane line lateral distance. If the target sampling point does not have the road edge semantic information and does not have the lane line semantic information, a road boundary of the target sampling point is determined based on a preset default width.

7. The method of claim 6, wherein, The determination of the road boundary of the target sampling point based on the lane line lateral distance comprises: determining an outermost lane line lateral distance based on the lane line lateral distance; and determining the road boundary of the target sampling point based on the outermost lane line lateral distance and a preset lane width.

8. The method of claim 1, wherein, The iterative aggregation processing of all target sampling points comprises: determining heading information of each target sampling point based on preprocessed trajectory data of all target sampling points; performing multi-round iterative aggregation processing of all target sampling points based on the road boundary and the heading information, and updating a current position of each target sampling point in each round of iterative aggregation processing; generating a road skeleton network based on updated positions of all target sampling points when an update state or an iteration number of all target sampling points in the iterative process meets a preset convergence condition.

9. The method of claim 8, wherein, The steps of each round of iterative aggregation processing comprise: for each target sampling point of all target sampling points, the following steps are performed: performing direction consistency filtering based on heading information of the target sampling point and heading information of other target sampling points to determine a first candidate point set; performing spatial range filtering based on the road boundary of the target sampling point and a road boundary of each candidate point in the first candidate point set to determine a projection point set; determining an aggregation center point based on the projection point set and the target sampling point; determining a movement distance based on the aggregation center point and the current position of the target sampling point; determining a target update position of the target sampling point based on the movement distance and a preset movement limit; updating the current position of the target sampling point based on the target update position; updating the heading information and the road boundary of the target sampling point based on the updated current position.

10. The method of claim 9, wherein, The spatial range filtering based on the road boundary of the target sampling point and the road boundary of each candidate point in the first candidate point set to determine the projection point set comprises: for each candidate point in the first candidate point set, the following steps are performed: performing vertical projection processing of the target sampling point to a trajectory line segment in which the candidate point is located to determine a projection point coordinate, wherein the trajectory line segment is a subset of preprocessed trajectory data of the candidate point; and If the projection point coordinate is located within the road boundary range of the target sampling point, and / or the road boundary of the candidate point and the road boundary of the target sampling point have an overlapping region with a width greater than or equal to a preset lane width, the candidate point is added to the projection point set.

Images