Map construction method and device, electronic equipment and storage medium

Abstract

The present disclosure provides a map construction method and device, electronic equipment and storage medium, and relates to the technical field of map making. The method comprises: acquiring scene image data of a target location uploaded by a plurality of vehicles; performing road semantic feature information analysis on the scene image data to obtain driving track data of each vehicle; wherein the road semantic feature information comprises at least one of a lane line, a stop line, a guide arrow mark and a road ground mark; clustering the driving track data according to the similarity of the road semantic feature information between the driving track data to obtain a clustering result; constructing a target map of the target location according to the clustering result; and the target map is used to provide navigation or path planning for the vehicles. The difference in scene image data caused by different vehicles is reduced, and the accuracy of map construction is improved.

Description

Technical Field

[0001] This disclosure relates to the field of map production technology, and in particular to a map construction method, apparatus, electronic device and storage medium. Background Technology

[0002] With the continuous development of intelligent driving technology, high-definition maps (HDM) can be used to assist intelligent driving. HDMs are typically generated based on scene image data collected by multiple vehicles. Since multiple scene image data of the same location were collected by different vehicles, there are differences between the multiple scene image data, making it difficult to construct a map from scene image data. Summary of the Invention

[0003] This disclosure provides a map building method, apparatus, electronic device, and storage medium to solve problems in the related art.

[0004] A first aspect of this disclosure provides a map construction method, the method comprising: Acquire scene image data of the target location uploaded by multiple vehicles; The scene image data is analyzed for road semantic feature information to obtain the driving trajectory data of each vehicle; wherein, the road semantic feature information includes at least one of lane lines, stop lines, directional arrow marks, and road surface marks; Based on the similarity of road semantic feature information between the driving trajectory data, the driving trajectory data is clustered to obtain clustering results; Based on the clustering results, a target map of the target location is constructed; the target map is used to provide navigation or route planning for the vehicle.

[0005] Reduce the differences in scene image data caused by different vehicles, and improve the accuracy of map construction.

[0006] In some embodiments, the analysis of road semantic feature information on the scene image data to obtain the driving trajectory data of each vehicle includes: Perform road semantic analysis on the scene image to obtain the road semantic feature information of the scene image; The driving trajectory data is determined based on the first positioning information and the corresponding road semantic feature information in the scene image.

[0007] The association between the first location information and the road semantic features makes each driving trajectory data correspond to the actual location of the vehicle, reducing conflicts between different data.

[0008] In some embodiments, determining the driving trajectory data based on the first positioning information and the corresponding road semantic feature information in the scene image includes: Based on the speed change characteristics and angular velocity change characteristics in the first positioning information, the time interval in which the multiple vehicles are located within the target location is determined; Spatiotemporal correlation is performed on the road semantic feature information and the time interval to obtain spatiotemporal correlation data; The spatiotemporal correlation data is compressed to obtain the driving trajectory data.

[0009] By determining the time interval through speed changes and angular velocity characteristics, the trajectory deviation caused by temporary vehicle stops or detours is reduced, ensuring that the data only covers the driving process at the target location.

[0010] In some embodiments, clustering the driving trajectory data based on the similarity of road semantic feature information between the driving trajectory data to obtain clustering results includes: Based on the road configuration features, road driving direction features, and road geometric features in the road semantic feature information of the driving trajectory data, a clustering feature vector of the driving trajectory data is constructed. Based on the distribution of the driving trajectory data, the cluster similarity threshold of the cluster feature vector is determined; The driving trajectory data is clustered according to the clustering similarity threshold to obtain the clustering result.

[0011] By integrating road configuration, driving direction, and geometric features, misclassification caused by a single feature is reduced, making the clustering results more consistent with the actual road structure.

[0012] In some embodiments, clustering the driving trajectory data according to the clustering similarity threshold to obtain the clustering result includes: Driving trajectory data whose clustering feature vector is greater than the clustering similarity threshold are identified as driving trajectory data of the same type; Identify abnormal driving trajectory data among the similar driving trajectory data; The abnormal driving trajectory data is removed from the same type of driving trajectory data to obtain the clustering result.

[0013] Remove abnormal trajectories to reduce local errors in the map caused by noise.

[0014] In some embodiments, constructing a target map of the target location based on the clustering results includes: Identify matching pairs between driving trajectory data in the clustering results; Determine the pose constraints of the matching pair; the pose constraints are conditions that constrain the pose range of the matching pair. Based on the pose constraints and the matching pairs, a target map of the target location is constructed.

[0015] By limiting trajectory deviation through pose constraints, lane misalignment or map fragmentation caused by vehicle positioning errors can be reduced.

[0016] In some embodiments, the pose constraints for determining the matching pair include: Based on the geometric constraints of the matching pair, the pose transformation of the matching pair is calculated to obtain the pose transformation value of the matching pair; Using the pose transformation values, an association constraint is established between the matching pairs; the pose constraint conditions include the association constraint.

[0017] Pose transformation values ​​can reduce vehicle positioning errors.

[0018] In some embodiments, the pose constraints for determining the matching pair include: Obtain the second positioning information of the pose of the matched pair; Based on the second positioning information, positioning constraints are established between the poses of the matching pair; the pose constraints include the positioning constraints.

[0019] Positioning constraints reduce single-track positioning deviation, making the positions of similar tracks tend to be uniform.

[0020] In some embodiments, the pose constraints for determining the matching pair include: Obtain the motion data of the vehicle corresponding to the pose of the matched pair; Using the motion data, motion constraints are established between the poses of the matched pairs; the pose constraints include the motion constraints.

[0021] By using vehicle motion data to establish motion constraints, it is possible to describe the relative relationships between different matching pairs, which helps to make the spatial positions and orientations of multiple matching pairs more consistent.

[0022] In some embodiments, constructing a target map of the target location based on the pose constraints and the matching pairs includes: The pose constraints are compared with the poses of the matching pair. Based on the comparison result between the pose of the matching pair and the pose constraint, the pose of the matching pair is adjusted to obtain the adjusted pose diagram of the matching pair. The target map is constructed based on the adjusted pose graph of the matching pair.

[0023] By comparing and adjusting the poses of the matched pairs with the pose constraints, the poses of the matched pairs are optimized so that the connection relationship between multiple matched pairs conforms to the actual spatial and motion constraints, thereby improving the accuracy of the matched pairs.

[0024] A second aspect of this disclosure provides a map building apparatus, comprising: The acquisition unit is used to acquire scene image data of the target location uploaded by multiple vehicles; The analysis unit is used to analyze the road semantic feature information of the scene image data to obtain the driving trajectory data of each vehicle; wherein, the road semantic feature information includes at least one of lane lines, stop lines, directional arrow marks and road surface marks; The clustering unit is used to cluster the driving trajectory data based on the similarity of road semantic feature information between the driving trajectory data to obtain the clustering result; A construction unit is used to construct a target map of the target location based on the clustering results; the target map is used to provide navigation or route planning for the vehicle.

[0025] In some embodiments, the analysis unit is further configured to: Road semantic analysis is performed on the scene images of the target location collected by the multiple vehicles to obtain the road semantic feature information of the scene images; The driving trajectory data is determined based on the first positioning information in the scene image and the corresponding road semantic feature information.

[0026] In some embodiments, the analysis unit is further configured to: Based on the speed change characteristics and angular velocity change characteristics in the first positioning information, the time interval in which the multiple vehicles are located within the target location is determined; Spatiotemporal correlation is performed on the road semantic feature information and the time interval to obtain spatiotemporal correlation data; The spatiotemporal correlation data is compressed to obtain the driving trajectory data.

[0027] In some embodiments, the clustering unit is further configured to: Based on the road configuration features, road driving direction features, and road geometric features in the road semantic feature information of the driving trajectory data, a clustering feature vector of the driving trajectory data is constructed. Based on the distribution of the driving trajectory data, the cluster similarity threshold of the cluster feature vector is determined; The driving trajectory data is clustered according to the clustering similarity threshold to obtain the clustering result.

[0028] In some embodiments, the clustering unit is further configured to: Driving trajectory data whose clustering feature vector is greater than the clustering similarity threshold are identified as driving trajectory data of the same type; Identify abnormal driving trajectory data among the similar driving trajectory data; The abnormal driving trajectory data is removed from the same type of driving trajectory data to obtain the clustering result.

[0029] In some embodiments, the building unit includes: The first determining module is used to determine matching pairs between driving trajectory data in the clustering results; The second determining module is used to determine the pose constraints of the matching pair; the pose constraints are conditions that constrain the pose range of the matching pair. A construction module is used to construct a target map of the target location based on the pose constraints and the matching pairs.

[0030] In some embodiments, the second determining module is further configured to: Based on the geometric constraints of the matching pair, the pose transformation of the matching pair is calculated to obtain the pose transformation value of the matching pair; Using the pose transformation values, an association constraint is established between the matching pairs; the pose constraint conditions include the association constraint.

[0031] In some embodiments, the second determining module is further configured to: Obtain the second positioning information of the pose of the matched pair; Based on the second positioning information, positioning constraints are established between the poses of the matching pair; the pose constraints include the positioning constraints.

[0032] In some embodiments, the second determining module is further configured to: Obtain the motion data of the vehicle corresponding to the pose of the matched pair; Using the motion data, motion constraints are established between the poses of the matched pairs; the pose constraints include the motion constraints.

[0033] In some embodiments, the building module is further configured to: The pose constraints are compared with the poses of the matching pair. Based on the comparison result between the pose of the matching pair and the pose constraint, the pose of the matching pair is adjusted to obtain the adjusted pose diagram of the matching pair. The target map is constructed based on the adjusted pose graph of the matching pair.

[0034] A third aspect of this disclosure provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the methods described in the first aspect of this disclosure.

[0035] A fourth aspect of this disclosure provides a non-transitory computer-readable storage medium that, when instructions in the storage medium are executed by a processor of a mobile terminal, enables the mobile terminal to perform the methods described in the first aspect of this disclosure.

[0036] According to a fifth aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the methods described in the embodiments of the first aspect of this disclosure.

[0037] A sixth aspect of this disclosure provides a chip including one or more interfaces and one or more processors; the interfaces are configured to receive signals from the memory of an electronic device and send signals to the processors, the signals including computer instructions stored in the memory, which, when executed by the processors, cause the electronic device to perform the methods described in the first aspect of this disclosure.

[0038] In summary, the map construction method proposed in this disclosure includes acquiring scene image data of a target location uploaded by multiple vehicles; analyzing road semantic feature information of the scene image data to obtain driving trajectory data for each vehicle; wherein the road semantic feature information includes at least one of lane lines, stop lines, directional arrow markers, and road surface markers; clustering the driving trajectory data based on the similarity of road semantic feature information between the driving trajectory data to obtain clustering results; and constructing a target map of the target location based on the clustering results; the target map is used to provide navigation or route planning for the vehicles. This method reduces the differences in scene image data caused by different vehicles and improves the accuracy of map construction.

[0039] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0040] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0041] Figure 1 A flowchart of a map construction method provided in this disclosure embodiment; Figure 2 This is a diagram illustrating a pose diagram provided in an embodiment of the present disclosure. Figure 3 This is a schematic diagram of the structure of a map building device provided in an embodiment of the present disclosure; Figure 4 This is a schematic diagram of another map building apparatus provided in an embodiment of the present disclosure; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure; Figure 6 This is a schematic diagram of the structure of a chip provided in an embodiment of the present disclosure. Detailed Implementation

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

[0043] With the continuous development of intelligent driving technology, high-definition maps (HDM) can be used to assist intelligent driving. HDMs are typically generated based on scene image data collected by multiple vehicles. Since multiple scene image data of the same location were collected by different vehicles, there are differences between the multiple scene image data, making it difficult to construct a map from scene image data.

[0044] Therefore, to address the problems existing in related technologies, this disclosure proposes a map construction method. This method includes acquiring scene image data of a target location uploaded by multiple vehicles; analyzing the scene image data for road semantic features to obtain the driving trajectory data of each vehicle; wherein the road semantic features include at least one of lane lines, stop lines, directional arrow markers, and road surface markers; clustering the driving trajectory data based on the similarity of the road semantic features between the driving trajectory data to obtain clustering results; and constructing a target map of the target location based on the clustering results; the target map is used to provide navigation or route planning for the vehicles. This method reduces the differences in scene image data caused by different vehicles and improves the accuracy of map construction.

[0045] This disclosure is not exhaustive, but merely illustrative of some embodiments, and is not intended to limit the scope of protection of this disclosure. Unless otherwise specified, each step in a particular embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment can be arbitrarily interchanged. Furthermore, the optional implementation methods in a particular embodiment can be arbitrarily combined; moreover, the embodiments can be arbitrarily combined, for example, some or all steps of different embodiments can be arbitrarily combined, and a particular embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0046] In each of the disclosed embodiments, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of the embodiments are consistent and can be referenced by each other. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0047] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure.

[0048] In the embodiments disclosed herein, unless otherwise stated, elements expressed in the singular, such as “a,” “an,” “the,” “the,” “the,” “the,” “the,” “the,” “this,” etc., may mean “one and only one,” or “one or more,” “at least one,” etc.

[0049] In some embodiments, the terms “in response to…”, “in response to determining…”, “in the case of…”, “when…”, “if…”, “if…”, etc., can be used interchangeably.

[0050] In some embodiments, the terms “greater than,” “greater than or equal to,” “not less than,” “more than,” “more than or equal to,” “not less than,” “higher than,” “higher than or equal to,” “not lower than,” and “above” can be used interchangeably, as can the terms “less than,” “less than or equal to,” “not greater than,” “less than,” “less than or equal to,” “not more than,” “lower than,” “lower than or equal to,” “not higher than,” and “below”.

[0051] The prefixes "first," "second," etc., used in the embodiments of this disclosure are merely for distinguishing different descriptive objects and do not constitute limitations on the position, order, priority, number, or content of the descriptive objects. The description of the descriptive objects is given in the context of the claims or embodiments, and the use of prefixes should not constitute unnecessary limitations. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The implementation methods described in the following exemplary embodiments do not represent all implementation methods consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0052] In the embodiments disclosed herein, "multiple" refers to two or more.

[0053] In the embodiments disclosed herein, terms such as “import”, “input”, and “read in” can be used interchangeably.

[0054] In some embodiments, devices, etc., can be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. Terms such as “device”, “equipment”, “circuit”, “network element”, “node”, “function”, “unit”, “section”, “system”, “network”, “chip”, “chip system”, “entity”, and “subject” can be used interchangeably.

[0055] Figure 1 This is a flowchart illustrating a map building method provided in an embodiment of this disclosure. This method can be applied to application scenarios such as smart terminals, for example, executed by a terminal with integrated map building functionality or a data processor within a terminal, or executed by other devices suitable for map building (such as a server). This disclosure does not limit its application. Figure 1 As shown, the map construction method includes steps S101-S104.

[0056] Step S101: Obtain scene image data of the target location uploaded by multiple vehicles.

[0057] In the embodiments of this disclosure, a vehicle refers to a motor vehicle equipped with an image acquisition device and a communication device. Each vehicle is equipped with an image acquisition device (such as a camera or other image-capturing equipment), a positioning device, and a wireless communication module, enabling it to continuously capture road scene information during operation. Scene image data uploaded by multiple vehicles is combined into crowdsourced data, which refers to scene image data collected by a large number of users driving vehicles. The target location refers to a specific geographical area where a high-precision map needs to be constructed, such as an intersection of a main urban road, a specific section of a highway, or a traffic node near a school. Scene image data refers to real-time image information collected by vehicles at the target location, including visual content of the road environment, such as lane lines, traffic lights, road markings (such as stop lines and directional arrows), and surrounding obstacles.

[0058] In the embodiments of this disclosure, when the vehicle is in motion, the vehicle positioning system detects that the vehicle has entered the target location range, triggering scene image data acquisition. The vehicle's onboard camera captures scene images of the target location in real time, and the vehicle simultaneously records the vehicle's position information at the time of image acquisition. After the acquisition is completed, the scene image data is uploaded, and the scene image data uploaded by multiple vehicles is combined into crowdsourced data.

[0059] It should be noted that the scene image data involved in the embodiments of this disclosure is de-identified data, does not involve the disclosure of user privacy, and can be uploaded after obtaining user authorization.

[0060] By collecting data from multiple vehicles at different times and from different perspectives, image deviations caused by weather, lighting, or shooting angles of a single vehicle are avoided, ensuring that the scene information of the target location is more comprehensive and more in line with the actual scene.

[0061] Step S102: Analyze the road semantic feature information of the scene image data to obtain the driving trajectory data of each vehicle; wherein, the road semantic feature information includes at least one of lane lines, stop lines, directional arrow marks and road surface marks.

[0062] In the embodiments of this disclosure, road semantic feature information refers to semantic content related to road structure extracted from scene image data, such as the direction of lane lines, the position of stop lines, the direction of directional arrows, and road surface markings (such as zebra crossings). Driving trajectory data refers to the actual path information of a vehicle at a target location, including but not limited to the vehicle's movement path, direction changes, and location points (such as intersection turning points) on the road. Lane lines are longitudinal markings drawn on the road surface to regulate vehicle driving trajectories. Stop lines are transverse markings drawn at intersections, before pedestrian crossings, or other locations where vehicles need to stop, and are substantially perpendicular to the road direction. Directional arrow markings are arrow-shaped ground graphics drawn inside lanes to indicate the permitted driving directions (such as straight, left turn, right turn, straight plus left turn, etc.). Road surface markings refer to textual and graphic data applied to the road surface to convey traffic rules or prompts.

[0063] In the embodiments of this disclosure, the scene image data is preprocessed to remove noise and improve the clarity of the scene image data; road semantic feature information in the road image data is extracted by feature recognition, and the position and attributes of each road semantic feature are marked; the road semantic feature sequence of the same vehicle is associated in chronological order by combining the positioning information corresponding to each frame of road image data; and the driving trajectory data of the vehicle driving path is generated based on the road semantic feature sequence.

[0064] The extracted road feature information can provide a basis for driving decisions, enabling intelligent driving to cope with complex traffic situations and improve driving safety.

[0065] Step S103: Based on the similarity of road semantic feature information between the driving trajectory data, the driving trajectory data is clustered to obtain the clustering result.

[0066] In the embodiments of this disclosure, similarity refers to the degree of matching between driving trajectory data in terms of road semantic features; the higher the similarity, the closer the driving trajectory data are. Clustering refers to the processing method of classifying the driving trajectory data of multiple vehicles according to the similarity of road semantic features, grouping driving trajectories with high road semantic feature matching into the same category. The clustering result refers to the set of driving trajectory data of various types obtained after clustering processing, and each set of driving trajectory data contains multiple driving trajectory data with similar road semantic features.

[0067] In the embodiments of this disclosure, road semantic features (such as the number of lanes and turning direction) are extracted for each driving trajectory data; the similarity between different driving trajectory data is calculated (e.g., comparing the matching degree of lane configuration and driving direction); and similar driving trajectory data are grouped into the same group according to a preset similarity threshold (e.g., 80%) to form a clustering result. The preset similarity threshold can be adjusted according to actual needs, such as to 70% or 90%, and this disclosure does not limit this adjustment.

[0068] Integrating similar trajectories from different vehicles reduces interference caused by differences in vehicle data.

[0069] Step S104: Based on the clustering results, construct a target map of the target location; the target map is used to provide navigation or route planning for the vehicle.

[0070] In the embodiments of this disclosure, the target map refers to a map constructed based on clustering results that fits the actual road scene of the target location. It includes road information such as the geometry of the roads at the target location, lane line positions, and traffic signs (such as stop lines and directional arrows). The target map can be used for decision support in intelligent driving technology, such as path planning based on the target map to achieve vehicle-assisted driving control. The target map can be a high-precision map.

[0071] In the embodiments of this disclosure, matching pairs of driving trajectory data are determined from various trajectory sets of clustering results, and pose constraints are established based on the road semantic feature information of the matching pairs; the poses of the matching pairs are compared with the constraints, and the poses of the matching pairs are adjusted to make the trajectory conform to the constraint requirements; similar driving trajectory data are integrated to form a unified road path; and road semantic feature information corresponding to various trajectories is superimposed to generate a target map of the target location.

[0072] Maps are generated based on multiple driving trajectory data, reducing local distortions caused by errors in a single driving trajectory data (such as lane misalignment), making the maps more stable and more suitable for intelligent driving scenarios.

[0073] According to the map construction method proposed in this disclosure, the method includes acquiring scene image data of a target location uploaded by multiple vehicles; performing road semantic feature information analysis on the scene image data to obtain driving trajectory data of each vehicle; wherein the road semantic feature information includes at least one of lane lines, stop lines, directional arrow markers, and road surface markers; clustering the driving trajectory data according to the similarity of road semantic feature information between the driving trajectory data to obtain clustering results; constructing a target map of the target location based on the clustering results; the target map is used to provide navigation or route planning for the vehicles. This method reduces the differences in scene image data caused by different vehicles and improves the accuracy of map construction.

[0074] For example, when performing the road semantic feature information analysis on the scene image data to obtain the driving trajectory data of each vehicle, it can be implemented in the following ways, but is not limited to: performing road semantic analysis on the scene image to obtain the road semantic feature information of the scene image; and determining the driving trajectory data based on the first positioning information in the scene image and the corresponding road semantic feature information.

[0075] In the embodiments of this disclosure, road semantic feature information refers to the visual features in the scene image that reflect road traffic attributes and markings, including but not limited to lane lines, stop lines, directional arrow markings, and road surface markings. The first positioning information refers to the location and time information synchronously recorded when the vehicle collects scene image data, including the latitude and longitude of the collection point and the collection time, which can accurately anchor the geographical location corresponding to each road scene image data.

[0076] In the embodiments of this disclosure, road semantic analysis is performed on scene image data to identify road elements (such as lane lines and stop lines) in the image; first positioning information is read when the road image data is collected; the identified road semantic feature information is associated with the first positioning information to calculate the vehicle's movement path at the target location and generate driving trajectory data.

[0077] The association between the first location information and the road semantic features makes each driving trajectory data correspond to the actual location of the vehicle, reducing conflicts between different data.

[0078] For example, when performing the step of determining the driving trajectory data based on the first positioning information and corresponding road semantic feature information in the scene image, it can be implemented in the following ways, but is not limited to: determining the time interval in which the multiple vehicles are located within the target location based on the driving speed change features and angular velocity change features in the first positioning information; performing spatiotemporal correlation on the road semantic feature information and the time interval to obtain spatiotemporal correlation data; and performing compression processing on the spatiotemporal correlation data to obtain the driving trajectory data.

[0079] In the embodiments of this disclosure, the driving speed change feature refers to the dynamic changes in the vehicle's speed during driving, such as speed fluctuations during acceleration, deceleration, or stopping. The angular velocity change feature refers to the changes in the vehicle's rotational speed when turning, such as the steering rate during a turn (e.g., a sharp turn or a gentle turn). The time interval refers to the actual time period during which the vehicle travels within the target location, such as a continuous time period from entering an intersection to leaving it. Spatiotemporal correlation data refers to data formed by combining road semantic feature information with the time interval, representing the spatial location of the feature at a specific point in time.

[0080] In embodiments of this disclosure, speed changes (e.g., detecting a vehicle decelerating from 40 km / h to 0 km / h) and angular velocity changes (e.g., detecting the steering rate when turning left) in the first positioning information are analyzed to determine the time range of the vehicle within the target location (e.g., within 10 seconds from the start of deceleration to stopping); road semantic feature information (e.g., the position of the stop line) is associated with the time interval to form spatiotemporal associated data (e.g., recording that the stop line is located at the center point of the intersection at 10:00:15); the spatiotemporal associated data is compressed (e.g., only key location points are retained, and repeated coordinates during continuous driving are removed) to generate driving trajectory data.

[0081] By determining the time interval through speed changes and angular velocity characteristics, the trajectory deviation caused by temporary vehicle stops or detours is reduced, ensuring that the data only covers the driving process at the target location.

[0082] For example, when performing the clustering of the driving trajectory data based on the similarity of road semantic feature information between the driving trajectory data to obtain the clustering result, it can be implemented in the following ways, but is not limited to: constructing a clustering feature vector of the driving trajectory data based on the road configuration features, road driving direction features, and road geometric features in the road semantic feature information of the driving trajectory data; determining a clustering similarity threshold of the clustering feature vector based on the distribution between the driving trajectory data; and clustering the driving trajectory data based on the clustering similarity threshold to obtain the clustering result.

[0083] In the embodiments of this disclosure, road configuration features refer to attributes representing road layout in road semantic feature information, including but not limited to the number of lanes, lane type (straight lane / left turn lane / right turn lane), and lane boundary markers. Road driving direction features represent the direction a vehicle travels along the road, determined based on trajectory data and road markings, including but not limited to straight, left turn, and right turn. Road geometric features represent attributes representing the physical form of the road, including but not limited to road orientation (straight / curved), curve curvature, and lane width. Clustering feature vectors refer to comprehensive representation data formed by integrating road configuration, driving direction, and geometric features of a single driving trajectory. Clustering similarity thresholds are judgment thresholds set based on the distribution density of all driving trajectory data, used to define whether two trajectories belong to the same category; the clustering similarity threshold is adapted to the actual distribution of trajectories at the target location.

[0084] In the embodiments of this disclosure, for each driving trajectory data, all road semantic feature information contained therein is traversed, road configuration features are extracted and quantified; the road configuration features are arranged in order to form a clustering feature vector of fixed length. For example, a trajectory observing the eastern entrance of an intersection has a vector that presents a numerical sequence of three lanes, one stop line, a straight arrow, a direction angle of zero degrees, no large turns, and curvature close to zero. The feature vector difference values ​​between all pairs of trajectories are calculated to generate a list of difference value distributions. If most difference values ​​are concentrated in a small range, it indicates that the data is naturally grouped; if a few difference values ​​are significantly larger, it indicates that there are obvious differences between different classes. The difference values ​​are sorted from smallest to largest to find the inflection point of the distribution curve—that is, the critical point where the rate of increase of difference values ​​changes abruptly. The value corresponding to this inflection point is selected as the clustering similarity threshold, which can classify compactly clustered trajectories into the same class without forcibly merging obviously dispersed trajectories. The clustering feature vector of each driving trajectory data is compared with all other vectors one by one to calculate the difference value. If the difference value between two trajectories is less than the determined clustering similarity threshold, then a class label is established. If trajectories A and B belong to the same category, and B and C belong to the same category, then A, B, and C are grouped into the same cluster. This process generates several initial clusters. Subsequently, within each cluster, the deviation of each trajectory vector from the cluster's average vector is calculated. Trajectories deviating by more than twice the threshold are considered abnormal and removed. The final set of remaining clusters is the clustering result.

[0085] By integrating road configuration, driving direction, and geometric features, misclassification caused by a single feature is reduced, making the clustering results more consistent with the actual road structure.

[0086] For example, when performing the clustering of the driving trajectory data according to the clustering similarity threshold to obtain the clustering result, it can be implemented in the following ways, but not limited to: determining the driving trajectory data whose clustering feature vector is greater than the clustering similarity threshold as driving trajectory data of the same type; identifying abnormal driving trajectory data in the driving trajectory data of the same type; removing the abnormal driving trajectory data in the driving trajectory data of the same type to obtain the clustering result.

[0087] In the embodiments of this disclosure, similar driving trajectory data refers to a set of multiple driving trajectories with a clustering feature vector similarity higher than a threshold, reflecting normal driving paths under the same road scenario. Abnormal driving trajectory data refers to trajectory data in a similar trajectory set that deviates from road semantic features and violates normal driving patterns, and cannot truly reflect the road scenario at the target location.

[0088] In the embodiments of this disclosure, the clustering feature vectors of each driving trajectory are compared pairwise, and trajectories with similarity higher than a threshold are selected as driving trajectory data of the same category; combined with road semantic features (lane lines, guide arrows) and driving direction features, abnormal trajectories that deviate from the normal path and violate driving rules in the same category are identified; after removing all abnormal trajectories, the remaining valid trajectories of the same category are integrated to obtain the final clustering result.

[0089] Remove abnormal trajectories to reduce local errors in the map caused by noise.

[0090] For example, when performing the step of constructing a target map of the target location based on the clustering results, it can be implemented in the following ways, but is not limited to: determining matching pairs between driving trajectory data in the clustering results; determining pose constraints of the matching pairs; the pose constraints being conditions that constrain the pose range of the matching pairs; and constructing a target map of the target location based on the pose constraints and the matching pairs.

[0091] In the embodiments of this disclosure, a matching pair refers to a correspondence established between the driving trajectory data of different vehicles, representing the same physical point on the actual road. A matching pair typically contains two or more points, which come from different trajectories but are determined to describe the same location in the road environment based on their semantics (e.g., both located at the starting point of the same lane line) or spatiotemporal proximity. Pose refers to the spatial position (latitude and longitude) and driving posture (driving direction, vehicle orientation) of the vehicle corresponding to the trajectory at the time of acquisition, and is a core parameter characterizing the spatial state of the trajectory. Pose constraints refer to the conditions that constrain the spatial position and posture range of the two trajectories in the matching pair. They are set based on the road semantic features (lane lines, stop lines, etc.) of the target location, so that the trajectory pose fits the actual road scene and does not deviate from a reasonable range.

[0092] In embodiments of this disclosure, each trajectory in the clustering results is grouped, and road semantic feature sequences (e.g., lane lines, stop lines, signs) of all trajectories within the group are extracted. By comparing feature types, spatial distribution order, and geometric shapes, the poses of different trajectories observing the same physical element are identified. The matching rules require that the feature types observed by two poses must be the same (both are stop lines or both are left-turn arrows), the spatial distance between the observation points must be within a preset range, and the differences in feature shape parameters (e.g., length, curvature) must be within tolerance. Pose that meet the conditions are paired to form a set of matching pairs. For example, if all three trajectories observe a stop line at an intersection, then the corresponding poses of any two trajectories constitute a pair of matching pairs. For each pair of matched poses, three types of constraints are computed in parallel: geometric constraints require that the road geometry observed by the two poses must be consistent, such as the difference in lane length not exceeding ten centimeters and the same curvature direction; localization constraints read the difference in the original localization coordinates of the two poses as the prior range for pose transformation to prevent the optimization result from deviating excessively from the original observation; motion constraints retrieve the vehicle motion data (speed, steering angle) at the corresponding time of the two poses to verify whether the relative motion between the poses conforms to the laws of vehicle dynamics. These three types of constraints together constitute the pose constraints of the matched pair; the tighter the constraints, the smaller the optimization space. The pose constraints of all matched pairs are integrated into a global optimization objective. Starting from the initial pose, the system makes minor adjustments to the pose parameters of each trajectory, and successively checks the degree of constraint satisfaction for each matched pair. Each adjustment improves geometric consistency, reduces the difference from the original localization, and improves the consistency with motion data. After repeated iterations, the constraints of all matched pairs are simultaneously satisfied, at which point the poses of each trajectory converge to a globally consistent state. The road semantic features associated with these optimized poses are fused, the average coordinates of overlapping elements are taken, the topological connections are unified, and finally a target map of the target location is generated.

[0093] By limiting trajectory deviation through pose constraints, lane misalignment or map fragmentation caused by vehicle positioning errors can be reduced. Map fragmentation refers to data deviations that result in discontinuous or discontinuous areas in the map.

[0094] For example, when executing the determination of the pose constraints of the matching pair, it can be implemented in the following ways, but is not limited to: performing pose transformation calculation on the matching pair according to the geometric constraints of the matching pair to obtain the pose transformation value of the matching pair; using the pose transformation value to establish the association constraint between the matching pair; the pose constraint includes the association constraint.

[0095] In the embodiments of this disclosure, geometric constraints refer to the objectively existing spatial geometric relationship restrictions between the road structural features corresponding to the poses of the two sets of road data in the matching pair, including but not limited to fixed geometric attributes such as relative position, angular relationship, distance range, and shape fit between road structural features. Pose transformation calculation refers to the process of analyzing the spatial deviation (position offset, attitude tilt) of the two poses in the matching pair and calculating the adjustment parameters required to make the two poses tend towards spatial consistency. This process must conform to the actual logic of the road structural features and not deviate from the spatial constraints of the physical road. Pose transformation values ​​refer to the adjustment parameters obtained through pose transformation calculation, including position transformation values ​​(such as the distance of translation from one pose to another) and attitude transformation values ​​(such as the angle required to adjust the orientation of one pose to be consistent with another pose), which are the core data for quantifying the spatial differences between the two poses. Association constraints refer to the rules established based on pose transformation values ​​that limit the spatial relationship between the two poses in the matching pair, clarifying the allowable deviation range of the two associated poses (such as the position deviation not exceeding the translation distance corresponding to the pose transformation value, and the attitude deviation not exceeding the angle corresponding to the transformation value).

[0096] In the embodiments of this disclosure, for each matching pair, the spatial states of the two poses are observed: positional deviation (e.g., the distance by which the lane line corresponding to one pose is offset relative to the lane line of another pose) and attitude deviation (e.g., the angle of inclination of the stop line corresponding to one pose relative to the stop line of another pose), to determine the specific direction and degree of the deviation. Taking one pose in the matching pair as a reference (e.g., the pose whose acquisition time is closer to the actual road condition), the parameters that need to be adjusted for the other pose are calculated: the position transformation value is the translation distance required to eliminate the positional deviation (e.g., if the offset is 0.5 meters, the transformation value is 0.5 meters), and the attitude transformation value is the adjustment angle required to eliminate the attitude deviation (e.g., if the tilt is 2 degrees, the transformation value is 2 degrees), ensuring that the two poses after adjustment conform to the spatial attributes of the same physical target.

[0097] In the embodiments of this disclosure, the pose transformation value of each matching pair is converted into specific constraint rules to determine the allowable deviation range between the two poses in the matching pair. For example, when the position transformation value is 0.5 meters, the constraint rule is that the position deviation between the two poses in the matching pair does not exceed 0.5 meters; when the attitude transformation value is 2 degrees, the constraint rule is that the attitude tilt deviation between the two poses in the matching pair does not exceed 2 degrees. The constraint rules corresponding to all matching pairs are summarized, and rule conflicts are removed (if the constraint rules for the same type of road features are inconsistent, the rule that is more in line with the actual road structure shall prevail), forming a unified set of associated constraints. The pose transformation value can reduce vehicle positioning errors.

[0098] For example, when executing the determination of the pose constraints of the matching pair, it can be implemented in the following ways, but is not limited to: obtaining second positioning information of the poses of the matching pair; establishing positioning constraints between the poses of the matching pair based on the second positioning information; the pose constraints include the positioning constraints.

[0099] In the embodiments of this disclosure, the second positioning information refers to the original positioning data directly measured by vehicle sensors corresponding to each pose point of the matching pair. Positioning constraints refer to the conditions established based on the second positioning information that constrain the spatial positional relationship between the two poses of the matching pair, limiting the reasonable range of positional deviation between the two in the same acquisition scenario.

[0100] In the embodiments of this disclosure, for each pair of matched poses, the original positioning records corresponding to the two poses are retrieved. The physical distance difference between the two positioning coordinates is recorded as a priori reference value for pose transformation. The original coordinate difference between the two poses is used as a reference value, and a tolerance range is calculated in conjunction with the positioning accuracy evaluation value. The tolerance range constitutes a positioning constraint, and the optimized matched poses ensure that their positional differences fall within this range. If the range is exceeded, the optimization result is considered to deviate excessively from the original reliable observation, and the constraint is not satisfied. The positioning constraint is then integrated into the overall pose constraint condition.

[0101] Positioning constraints reduce single-track positioning deviation, making the positions of similar tracks tend to be uniform.

[0102] For example, when executing the determination of the pose constraints of the matching pair, it can be implemented in the following ways, but is not limited to: obtaining motion data of the vehicle corresponding to the pose of the matching pair; using the motion data to establish motion constraints between the poses of the matching pair; the pose constraints include the motion constraints.

[0103] In the embodiments of this disclosure, vehicle motion data refers to information related to the vehicle's motion state generated during driving, such as vehicle speed, acceleration, angular velocity, steering wheel rotation angle, and directional changes. This data is typically acquired through sensors on the vehicle (such as accelerometers, gyroscopes, and speed sensors) and is used to describe the vehicle's dynamic behavior. Motion constraints refer to the motion relationships between multiple matched pairs of poses derived from vehicle motion data. By analyzing vehicle motion, the relative relationships between matched pairs during motion can be clarified, such as the displacement and rotation angle between two matched pairs.

[0104] In the embodiments of this disclosure, motion data generated by the vehicle during driving is acquired through sensors on the vehicle (such as gyroscopes, accelerometers, etc.). This motion data can be used to describe information such as the vehicle's speed, acceleration, and steering angle during driving.

[0105] In embodiments of this disclosure, changes in vehicle motion are analyzed based on acquired vehicle motion data, and motion constraints between multiple matching pairs of poses are derived based on the vehicle's dynamic behavior. For example, suppose a vehicle travels through a section of road A and then turns to enter road B. By analyzing motion data such as vehicle acceleration and steering angle, the system can derive the relative position and angle constraints between road A and road B. Suppose the vehicle turns left at the end of road A and enters road B; the angle change between road A and road B is derived based on the vehicle's steering angle and speed data. The angle change is the motion constraint, which is used to establish the spatial relationship between two matching pairs.

[0106] In the embodiments of this disclosure, motion constraints between the poses of the matching pairs are established using vehicle motion data, and these motion constraints are transformed into pose constraints. These pose constraints are used to help ensure that the spatial relative relationships between multiple matching pairs conform to the actual vehicle motion, thereby improving the accuracy and reliability of subsequent tasks such as localization, map building, and path planning.

[0107] By using vehicle motion data to establish motion constraints, it is possible to describe the relative relationships between different matching pairs, which helps to make the spatial positions and orientations of multiple matching pairs more consistent.

[0108] For example, when performing the step of constructing a target map of the target location based on the pose constraints and the matching pair, it can be implemented in the following ways, but is not limited to: comparing the pose constraints with the pose of the matching pair; adjusting the pose of the matching pair based on the comparison result of the pose of the matching pair and the pose constraints to obtain an adjusted pose map of the matching pair; and constructing the target map based on the adjusted pose map of the matching pair.

[0109] In the embodiments of this disclosure, the pose constraints are broken down into specific, directly comparable indicators, ensuring that each indicator has a clear quantitative standard. For example, positioning constraints include: a straight-line distance of ≤30 meters between any two poses, an altitude deviation of ≤2 meters, and latitude and longitude within a reasonable range for the target location; motion constraints include: a gradual change in turning angle between adjacent poses of ≤40°, a speed change of ≤10 km / h, and orientation consistent with the motion state; and correlation constraints include: a road structure feature matching degree of ≥80%, the number of lane lines, and spacing consistent with the reference pose. All indicators must be determined in conjunction with the actual conditions of the target location (such as road size and driving characteristics) to ensure the rationality and operability of the constraints.

[0110] In the embodiments of this disclosure, the poses of the reference matching pair are used as the core nodes. All adjusted poses are sorted out according to spatial proximity and motion correlation to form an ordered correlation network. For example, poses of straight road segments are arranged according to driving direction, and poses of intersections are distributed according to intersection quadrants. The correlation relationships are marked: the constraint relationships between each pose and other poses are clearly defined in the figure (such as the distance relationship corresponding to the positioning constraint, and the turning gradient relationship corresponding to the motion constraint), so that the pose graph has a clear logical direction. The structured pose graph is output: the integrated poses and correlation relationships are output in a visual or data-driven form to form a complete pose graph. The pose graph is used to construct a map to obtain the target map.

[0111] By comparing and adjusting the poses of the matched pairs with the pose constraints, the poses of the matched pairs are optimized so that the connection relationship between multiple matched pairs conforms to the actual spatial and motion constraints, thereby improving the accuracy of the matched pairs.

[0112] In the embodiments of this disclosure, road scene images are acquired through a surround-view camera system mounted on a user vehicle. A semantic segmentation model can be used to analyze the images in real time, extracting key semantic elements such as lane lines, stop lines, and road signs, and predicting the trajectory of the road centerline based on semantic correlation. For a single vehicle passing through an intersection, the following steps are performed: Trajectory interval determination: Combining the speed change characteristics of the Global Positioning System (GPS) and the angular velocity change characteristics of the Inertial Measurement Unit (IMU), the time intervals for the vehicle entering and leaving the intersection are accurately identified; Semantic element integration: Semantic features continuously acquired during the entry and exit times are time-series aggregated to construct a link data structure containing spatiotemporal correlation. The entry link includes the lane centerline vector, lane line vector, road surface markings, and stop line vector positions; the exit link includes the lane centerline vector, lane line vector, and road surface markings; Lightweight data processing: Semantic vector elements are geometrically compressed, and data redundancy is reduced by using reference feature anchor points and relative offsets to generate data packets for cloud transmission.

[0113] In the embodiments of this disclosure, the intersection link clustering based on a greedy strategy involves: performing structured clustering processing on multiple data collections from the same intersection, specifically including: composite feature extraction: constructing a clustering feature vector by integrating lane configuration features, road structure direction features, and roadside geometric features; dynamic threshold setting: adaptively adjusting the clustering similarity threshold parameter according to the real-time data distribution characteristics; greedy clustering execution: adopting an incremental matching strategy to establish initial matching pairs through feature similarity measurement. When a candidate link meets the similarity conditions in all three dimensions—lane configuration mode, main driving direction, and geometric feature distribution—it is determined to be a homogeneous link; abnormal matching filtering: detecting and eliminating abnormal matching pairs in the clustering results through a random sampling consensus algorithm to ensure the geometric topological consistency of the clustering results. Initial value calculation for multi-constraint relative transformation: for the candidate matching pairs formed by clustering, performing matching alignment based on semantic elements to achieve multi-dimensional feature matching optimization: for each candidate matching pair, calculating the main direction of the matching pair based on vector elements, and calculating the initial attitude transformation of the matching pair based on the main direction. Based on the corresponding lane information, the system finds the semantic elements of the matching pairs, such as lane centerlines, lane lines, and stop lines, and constructs multimodal residual constraints using the geometric information of these elements: Lane centerline constraint: establishing a distance residual model between the point set and the reference line segment; Lane line association constraint: using joint modeling of distance constraints between parallel lines and direction consistency constraints; Landmark point matching: establishing point-to-point geometric constraints through feature descriptor matching; Stop line alignment: constructing position alignment constraints using the intersection characteristics of line segments; Based on these constraints, the optimizer is used to calculate and find the optimal relative transformation.

[0114] In the embodiments of this disclosure, a reference benchmark is selected. To avoid excessive constraints on pairwise link matching pairs under large amounts of data, leading to excessively long graph optimization time, an evaluation index including lane integrity, geometric coverage, and matching effectiveness is established to screen high-quality links as reference benchmarks, forming a unidirectional matching network centered on the reference links. The selection principles for reference links are as follows: the number of lanes in the selected link must be greater than a certain threshold, such as the average number of lanes in the same type of links; the average length of the lanes in the selected link must be greater than a certain threshold, such as the average length of the lanes in the same type of links; the outermost lane line of the selected link must be the edge line extracted by perception. Then, the candidate reference links are calculated with other links using the above relative transformation calculation method, and the number of lanes in the link and the number of successfully calculated matching pairs are comprehensively considered to select the best reference link. To facilitate a better understanding of the pose graph, such as Figure 2 As shown, Figure 2This is a diagram illustrating a pose graph provided in an embodiment of this disclosure. The pose graph optimization across multiple data passes involves: constructing a distributed optimization framework integrating multi-source constraints; node modeling: abstracting each link as a pose node in three-dimensional space, including position coordinates and attitude parameters; constraint edge construction: absolute positioning constraints: adding prior GNSS positioning constraints in stable GNSS signal areas; kinematic constraints: constructing relative pose constraints between adjacent frames based on IMU and wheel speedometer data; semantic matching constraints: establishing association constraints between links using the initial relative transformation values ​​calculated in step S3; robust optimization strategy: introducing a kernel function mechanism to suppress the influence of abnormal constraints, designing a multi-stage iterative optimization process, including a coarse matching stage and a fine optimization stage; topology filtering: removing isolated links through connection integrity detection, filtering matching results with insufficient confidence using an optimization residual threshold, and outputting an optimized structure with topological consistency. This significantly reduces data transmission pressure and storage costs: by deploying a lightweight semantic segmentation model on the vehicle side, only key vector elements such as lane lines, stop lines, and road signs are extracted. Data transmission volume is significantly reduced, effectively alleviating cloud storage pressure. Optimizing computational complexity and real-time processing capabilities: A unidirectional matching mechanism centered on reference links is introduced. Core reference links are selected through a three-dimensional evaluation system of lane integrity, geometric coverage, and matching effectiveness, reducing the computational complexity of fully connected matching. Achieving lane-level high-precision map stitching: Through joint optimization of multimodal constraints (lane centerline distance, lane line parallelism, landmark point matching, and stop line intersection alignment), a precise geometric alignment benchmark is established during the initial value calculation stage of relative transformation. Combined with an adaptive iterative optimization strategy, this meets the high-precision map requirements of intelligent driving. Supporting flexible map updates: The standardized link data structure supports incremental data fusion. The optimized topology achieves self-cleaning by removing isolated links and low-confidence matches, adapting to the continuous update characteristics of crowdsourced data.

[0115] Corresponding to the map construction method described above, this invention also proposes a map construction apparatus. Since the apparatus embodiments of this invention correspond to the method embodiments described above, details not disclosed in the apparatus embodiments can be referred to in the method embodiments described above, and will not be repeated here.

[0116] Figure 3 This is a schematic diagram of a map building apparatus provided in an embodiment of the present disclosure. The map building apparatus includes: Acquisition unit 41 is used to acquire scene image data of target locations uploaded by multiple vehicles; The analysis unit 42 is used to analyze the road semantic feature information of the scene image data to obtain the driving trajectory data of each vehicle; wherein, the road semantic feature information includes at least one of lane lines, stop lines, directional arrow marks and road surface marks; Clustering unit 43 is used to cluster the driving trajectory data according to the similarity of road semantic feature information between the driving trajectory data to obtain clustering results; The construction unit 44 is used to construct a target map of the target location based on the clustering results; the target map is used to provide navigation or route planning for the vehicle.

[0117] According to the map building apparatus disclosed herein, the apparatus includes acquiring scene image data of a target location uploaded by multiple vehicles; analyzing road semantic feature information of the scene image data to obtain driving trajectory data of each vehicle; wherein the road semantic feature information includes at least one of lane lines, stop lines, directional arrow markers, and road surface markers; clustering the driving trajectory data based on the similarity of road semantic feature information between the driving trajectory data to obtain clustering results; and constructing a target map of the target location based on the clustering results; the target map is used to provide navigation or route planning for the vehicles. This reduces the differences in scene image data caused by different vehicles and improves the accuracy of map building.

[0118] In one possible implementation of the embodiments of this disclosure, such as Figure 4 As shown, the analysis unit 42 is further used for: Road semantic analysis is performed on the scene images of the target location collected by the multiple vehicles to obtain the road semantic feature information of the scene images; The driving trajectory data is determined based on the first positioning information in the scene image and the corresponding road semantic feature information.

[0119] In one possible implementation of the embodiments of this disclosure, such as Figure 4 As shown, the analysis unit 42 is further used for: Based on the speed change characteristics and angular velocity change characteristics in the first positioning information, the time interval in which the multiple vehicles are located within the target location is determined; Spatiotemporal correlation is performed on the road semantic feature information and the time interval to obtain spatiotemporal correlation data; The spatiotemporal correlation data is compressed to obtain the driving trajectory data.

[0120] In one possible implementation of the embodiments of this disclosure, such as Figure 4 As shown, the clustering unit 43 is further used for: Based on the road configuration features, road driving direction features, and road geometric features in the road semantic feature information of the driving trajectory data, a clustering feature vector of the driving trajectory data is constructed. Based on the distribution of the driving trajectory data, the cluster similarity threshold of the cluster feature vector is determined; The driving trajectory data is clustered according to the clustering similarity threshold to obtain the clustering result.

[0121] In one possible implementation of the embodiments of this disclosure, such as Figure 4 As shown, the clustering unit 43 is further used for: Driving trajectory data whose clustering feature vector is greater than the clustering similarity threshold are identified as driving trajectory data of the same type; Identify abnormal driving trajectory data among the similar driving trajectory data; The abnormal driving trajectory data is removed from the same type of driving trajectory data to obtain the clustering result.

[0122] In one possible implementation of the embodiments of this disclosure, such as Figure 4 As shown, the building unit 44 includes: The first determining module 441 is used to determine matching pairs between driving trajectory data in the clustering results; The second determining module 442 is used to determine the pose constraint conditions of the matching pair; the pose constraint conditions are conditions that constrain the pose range of the matching pair. The construction module 443 is used to construct a target map of the target location based on the pose constraints and the matching pairs.

[0123] In one possible implementation of the embodiments of this disclosure, such as Figure 4 As shown, the second determining module 442 is further configured to: Based on the geometric constraints of the matching pair, the pose transformation of the matching pair is calculated to obtain the pose transformation value of the matching pair; Using the pose transformation values, an association constraint is established between the matching pairs; the pose constraint conditions include the association constraint.

[0124] In one possible implementation of the embodiments of this disclosure, such as Figure 4 As shown, the second determining module 442 is further configured to: Obtain the second positioning information of the pose of the matched pair; Based on the second positioning information, positioning constraints are established between the poses of the matching pair; the pose constraints include the positioning constraints.

[0125] In one possible implementation of the embodiments of this disclosure, such as Figure 4 As shown, the second determining module 442 is further configured to: Obtain the motion data of the vehicle corresponding to the pose of the matched pair; Using the motion data, motion constraints are established between the poses of the matched pairs; the pose constraints include the motion constraints.

[0126] In one possible implementation of the embodiments of this disclosure, such as Figure 4 As shown, the building module 443 is further configured to: The pose constraints are compared with the poses of the matching pair. Based on the comparison result between the pose of the matching pair and the pose constraint, the pose of the matching pair is adjusted to obtain the adjusted pose diagram of the matching pair. The target map is constructed based on the adjusted pose graph of the matching pair.

[0127] Since the apparatus provided in this embodiment corresponds to the methods provided in the above embodiments, the implementation of the methods is also applicable to the apparatus provided in this embodiment, and will not be described in detail in this embodiment.

[0128] The methods and apparatus provided in the embodiments of this application have been described above. To implement the functions of the methods provided in the embodiments of this application, the electronic device may include a hardware structure and software modules, and may implement the above functions in the form of a hardware structure, software modules, or a hardware structure plus software modules. One of the above functions may be executed in the form of a hardware structure, software modules, or a hardware structure plus software modules.

[0129] Figure 5 This is a block diagram illustrating an electronic device 500 for implementing the above-described map construction method, according to an exemplary embodiment. For example, the electronic device 500 may be applied to servers, cloud environments, vehicle terminals, operation service platforms, various computer platforms, terminal systems, and web page systems.

[0130] Reference Figure 5 The electronic device 500 may include one or more of the following components: processing component 502, memory 504, power supply component 506, multimedia component 508, audio component 510, input / output (I / O) interface 512, sensor component 514, and communication component 516.

[0131] Processing component 502 typically controls the overall operation of electronic device 500, such as operations associated with display, telephone calls, data communication, camera operation, and recording. Processing component 502 may include one or more processors 520 to execute instructions to perform all or part of the steps of the methods described above. Furthermore, processing component 502 may include one or more modules to facilitate interaction between processing component 502 and other components. For example, processing component 502 may include a multimedia module to facilitate interaction between multimedia component 508 and processing component 502.

[0132] Memory 504 is configured to store various types of data to support the operation of electronic device 500. Examples of this data include instructions for any application or method operating on electronic device 500, including messages, pictures, videos, etc. Memory 504 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0133] Power supply component 506 provides power to various components of electronic device 500. Power supply component 506 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 500.

[0134] Multimedia component 508 includes a screen that provides an output interface between electronic device 500 and user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of touch or swipe actions but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 508 includes a front-facing camera and / or a rear-facing camera. When electronic device 500 is in an operating mode, such as a shooting mode or video mode, the front-facing camera and / or rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0135] Audio component 510 is configured to output and / or input audio signals. For example, audio component 510 includes a microphone (MIC) configured to receive external audio signals when electronic device 500 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 504 or transmitted via communication component 516. In some embodiments, audio component 510 also includes a speaker for outputting audio signals.

[0136] I / O interface 512 provides an interface between processing component 502 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0137] Sensor assembly 514 includes one or more sensors for providing state assessments of various aspects of electronic device 500. For example, sensor assembly 514 may detect the on / off state of electronic device 500, the relative positioning of components such as the display and keypad of electronic device 500, changes in position of electronic device 500 or a component of electronic device 500, the presence or absence of user contact with electronic device 500, orientation or acceleration / deceleration of electronic device 500, and temperature changes of electronic device 500. Sensor assembly 514 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 514 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 514 may also include an accelerometer, gyroscope, magnetometer, pressure sensor, or temperature sensor.

[0138] Communication component 516 is configured to facilitate wired or wireless communication between electronic device 500 and other devices. Electronic device 500 can access wireless networks based on communication standards, such as WiFi, 2G or 3G, 4G LTE, 5G NR (NewRadio), or combinations thereof. In one exemplary embodiment, communication component 516 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 516 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0139] In an exemplary embodiment, the electronic device 500 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0140] Embodiments of this disclosure also provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the methods described in the above embodiments of this disclosure, including one or a combination of the steps of the method embodiments.

[0141] Embodiments of this disclosure also provide a computer program product comprising a computer program executable by a programmable device, the computer program having, when executed by the programmable device, the method described in the above embodiments of this disclosure.

[0142] For cases where electronic devices can be chips or chip systems, see [link to relevant documentation]. Figure 6 The diagram shows the structure of the chip. Figure 6 The chip shown includes a processor 601 and an interface 602. There can be one or more processors 601, and multiple interfaces 602.

[0143] Optionally, the chip also includes a memory 603 for storing necessary computer programs and data.

[0144] Those skilled in the art will also understand that the various illustrative logical blocks and steps listed in the embodiments of this application can be implemented by electronic hardware, computer software, or a combination of both. Whether such functionality is implemented through hardware or software depends on the specific application and the overall system design requirements. Those skilled in the art can implement the functionality using various methods for each specific application, but such implementation should not be construed as exceeding the scope of protection of the embodiments of this application.

[0145] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0146] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0147] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a system including a processing module or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0148] It should be understood that various parts of the embodiments of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0149] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc.

[0150] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A map construction method, characterized in that, The method includes: Acquire scene image data of the target location uploaded by multiple vehicles; The scene image data is analyzed for road semantic feature information to obtain the driving trajectory data of each vehicle; wherein, the road semantic feature information includes at least one of lane lines, stop lines, directional arrow marks, and road surface marks; Based on the similarity of road semantic feature information between the driving trajectory data, the driving trajectory data is clustered to obtain clustering results; Based on the clustering results, a target map of the target location is constructed; the target map is used to provide navigation or route planning for the vehicle.

2. The method according to claim 1, characterized in that, The analysis of road semantic feature information on the scene image data to obtain the driving trajectory data of each vehicle includes: Perform road semantic analysis on the scene image to obtain the road semantic feature information of the scene image; The driving trajectory data is determined based on the first positioning information and the corresponding road semantic feature information in the scene image.

3. The method according to claim 2, characterized in that, The step of determining the driving trajectory data based on the first positioning information and the corresponding road semantic feature information in the scene image includes: Based on the speed change characteristics and angular velocity change characteristics in the first positioning information, the time interval in which the multiple vehicles are located within the target location is determined; Spatiotemporal correlation is performed on the road semantic feature information and the time interval to obtain spatiotemporal correlation data; The spatiotemporal correlation data is compressed to obtain the driving trajectory data.

4. The method according to claim 1, characterized in that, The step of clustering the driving trajectory data based on the similarity of road semantic feature information between the driving trajectory data to obtain the clustering results includes: Based on the road configuration features, road driving direction features, and road geometric features in the road semantic feature information of the driving trajectory data, a clustering feature vector of the driving trajectory data is constructed. Based on the distribution of the driving trajectory data, the cluster similarity threshold of the cluster feature vector is determined; The driving trajectory data is clustered according to the clustering similarity threshold to obtain the clustering result.

5. The method according to claim 4, characterized in that, The step of clustering the driving trajectory data according to the clustering similarity threshold to obtain the clustering result includes: Driving trajectory data whose clustering feature vector is greater than the clustering similarity threshold are identified as driving trajectory data of the same type; Identify abnormal driving trajectory data among the similar driving trajectory data; The abnormal driving trajectory data is removed from the same type of driving trajectory data to obtain the clustering result.

6. The method according to claim 1, characterized in that, The step of constructing a target map of the target location based on the clustering results includes: Identify matching pairs between driving trajectory data in the clustering results; Determine the pose constraints of the matching pair; the pose constraints are conditions that constrain the pose range of the matching pair. Based on the pose constraints and the matching pairs, a target map of the target location is constructed.

7. The method according to claim 6, characterized in that, The pose constraints for determining the matching pair include: Based on the geometric constraints of the matching pair, the pose transformation of the matching pair is calculated to obtain the pose transformation value of the matching pair; Using the pose transformation values, an association constraint is established between the matching pairs; the pose constraint conditions include the association constraint.

8. The method according to claim 6, characterized in that, The pose constraints for determining the matching pair include: Obtain the second positioning information of the pose of the matched pair; Based on the second positioning information, positioning constraints are established between the poses of the matching pair; the pose constraints include the positioning constraints.

9. The method according to claim 6, characterized in that, The pose constraints for determining the matching pair include: Obtain the motion data of the vehicle corresponding to the pose of the matched pair; Using the motion data, motion constraints are established between the poses of the matched pairs; the pose constraints include the motion constraints.

10. The method according to claim 6, characterized in that, The step of constructing the target map of the target location based on the pose constraints and the matching pairs includes: The pose constraints are compared with the poses of the matching pair. Based on the comparison result between the pose of the matching pair and the pose constraint, the pose of the matching pair is adjusted to obtain the adjusted pose diagram of the matching pair. The target map is constructed based on the adjusted pose graph of the matching pair.

11. A map building device, characterized in that, The device includes: The acquisition unit is used to acquire scene image data of the target location uploaded by multiple vehicles; The analysis unit is used to analyze the road semantic feature information of the scene image data to obtain the driving trajectory data of each vehicle; wherein, the road semantic feature information includes at least one of lane lines, stop lines, directional arrow marks and road surface marks; The clustering unit is used to cluster the driving trajectory data based on the similarity of road semantic feature information between the driving trajectory data to obtain the clustering result; A construction unit is used to construct a target map of the target location based on the clustering results; the target map is used to provide navigation or route planning for the vehicle.

12. The apparatus according to claim 11, characterized in that, The analysis unit is also used for: Road semantic analysis is performed on the scene images of the target location collected by the multiple vehicles to obtain the road semantic feature information of the scene images; The driving trajectory data is determined based on the first positioning information in the scene image and the corresponding road semantic feature information.

13. The apparatus according to claim 11, characterized in that, The clustering unit is also used for: Based on the road configuration features, road driving direction features, and road geometric features in the road semantic feature information of the driving trajectory data, a clustering feature vector of the driving trajectory data is constructed. Based on the distribution of the driving trajectory data, the cluster similarity threshold of the cluster feature vector is determined; The driving trajectory data is clustered according to the clustering similarity threshold to obtain the clustering result.

14. The apparatus according to claim 11, characterized in that, The building unit includes: The first determining module is used to determine matching pairs between driving trajectory data in the clustering results; The second determining module is used to determine the pose constraints of the matching pair; the pose constraints are conditions that constrain the pose range of the matching pair. A construction module is used to construct a target map of the target location based on the pose constraints and the matching pairs.

15. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method of any one of claims 1-10.

16. A computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-10.

17. A program product, characterized in that, Includes computer instructions for causing a computer to perform the method of any one of claims 1-10.

Images