OSM road network parking space updating method based on public source trajectory data

Abstract

The invention discloses an OSM road network parking space updating method based on public source trajectory data, and relates to the technical field of trajectory positioning. The method comprises the following steps: based on public source trajectory data, performing topology check on road network data and performing quality cleaning on the trajectory data, and performing map matching on the processed road network data and trajectory data to obtain preprocessed data; setting a limiting range and a filtering condition based on the preprocessed data, extracting parking track point information, and then identifying a parking track point cluster by using a DBSCAN algorithm; and parking space features are extracted based on the parking track point clusters, a parking space recognition model is constructed according to the parking space features, classification recognition of roadside parking spaces, parking lot parking spaces and temporary parking spaces is performed on to-be-detected parking space information by using the trained parking space recognition model, and updating of the parking space information is completed on an OSM road network. According to the invention, better, more complete and more accurate parking space information can be provided for navigation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of trajectory positioning, in particular to an OSM road network parking space updating method based on crowd-sourced trajectory data. BACKGROUND

[0002] With the acceleration of urbanization, the number of motor vehicles has increased rapidly, and the parking pressure in large and medium-sized cities has gradually increased. People's demand for obtaining parking space information is increasing, and they hope to obtain parking space information near the current location or the destination of the trip through mobile phones or vehicle navigation devices and other daily devices. At present, the methods for obtaining parking space information include manual investigation, on-site collection of hardware such as vehicle-mounted cameras / sensors, laser radar measurement, photogrammetry, and unmanned aerial vehicle remote sensing. Due to the frequent construction and reconstruction of parking lots, the position changes quickly. The above-mentioned methods relying on manual adjustment or special equipment measurement have long cycle, high cost, and great difficulty in popularization, which cannot meet the actual demand. Therefore, the rapid collection method of parking space information has gradually become a research hotspot at home and abroad.

[0003] OpenStreetMap (OSM) is a global map collaboration project that not only allows users to freely edit and use map data, but also provides data upload and download to the public free of charge. It is a good data platform for geographic spatial data production and updating worldwide, and has been widely adopted by many enterprises, government agencies and non-profit organizations. As the core and infrastructure of OSM, road network data shows significant application value in the fields of navigation service optimization, multi-modal path planning, and urban spatial analysis. However, compared with the dynamically updated road topology structure, the static traffic information such as parking spaces attached to the road network has long been lagging behind in updating, with serious lack and insufficient updating, which restricts OSM from providing more complete and reliable parking information support for intelligent navigation, travel planning, urban parking guidance and traffic management applications.

[0004] Crowd-sourced trajectory data is a collection of spatio-temporal location information data contributed by ubiquitous mobile terminal users actively or passively. It has the characteristics of large data volume, strong real-time performance, diverse data sources, and low data acquisition cost. This data is currently growing explosively, and its implicit parking behavior characteristics provide a new way to solve the high-cost and low-coverage dilemma of traditional parking space detection methods, which can effectively make up for the lack of OSM road network parking data and help to alleviate the urban parking problem and optimize the efficiency of urban traffic operation. However, due to the large noise and irregular sampling interval of crowd-sourced trajectory data, it is quite challenging to use crowd-sourced trajectory data to achieve accurate identification and dynamic updating of OSM road network parking spaces, which has become a topic that needs to be broken through in the field of intelligent transportation. SUMMARY

[0005] The application aims to provide an OSM road network parking space updating method based on crowd-sourced trajectory data, aiming to solve or improve at least one of the above technical problems.

[0006] To achieve the above-mentioned purpose, the application provides the following solutions: An OSM road network parking space updating method based on crowd-sourced trajectory data, comprising: Based on the crowd-sourced trajectory data, topological inspection is performed on the road network data, and quality cleaning is performed on the trajectory data, and the processed road network data and trajectory data are subjected to map matching to obtain preprocessed data including the connection relationship between trajectory points and the belonging road segments; Based on the preprocessed data, a limit range and a filtering condition are set, parking trajectory point information is extracted, and then DBSCAN algorithm is used to identify parking trajectory point clusters; Parking space features are extracted based on the parking trajectory point clusters, a parking space recognition model is constructed according to the parking space features, the trained parking space recognition model is used to classify and identify the parking space information to be tested, and the parking space information is updated on the OSM road network; the parking space information includes roadside parking spaces, parking lot parking spaces and temporary parking spaces.

[0007] Optionally, based on the preprocessed data, a limit range and a filtering condition are set, parking trajectory point information is extracted, and then DBSCAN algorithm is used to identify parking trajectory point clusters, specifically comprising: The limit range and the filtering condition are set according to the parking behavior mode at the end of the trajectory and the related information, the parking trajectory is filtered, the speed zero value points are selected from the parking trajectory as parking trajectory points, and then DBSCAN algorithm is used to cluster the parking trajectory points to obtain clustering clusters composed of dense parking trajectory points.

[0008] Optionally, the process of filtering the parking trajectory is specifically: The parking trajectory segment is defined as a continuous trajectory point sequence in which a vehicle meets the condition of keeping a set stationary state for a period of time; According to the definition, the cleaned trajectory data is traversed, if the average speed of any trajectory is lower than a preset speed threshold, it is judged whether there is a sub-segment in the trajectory that meets the following conditions: if the speed of each trajectory point in the sub-segment is lower than the preset speed threshold, and the duration of the sub-segment is greater than or equal to a preset time threshold, the sub-segment is extracted as a candidate parking trajectory segment, and the candidate parking trajectory segment is subjected to secondary screening.

[0009] Optionally, the process of secondary screening the candidate parking trajectory segment is specifically: The straight-line distance between the start point and the end point of the candidate parking trajectory segment is calculated, if the distance is greater than a preset distance threshold, the candidate parking trajectory segment calculated this time is removed. Calculate the change of direction angle between consecutive trajectory points in the candidate parking trajectory segment, and calculate the sum and standard deviation. If the sum of the change of direction angle is greater than the first angle threshold or the standard deviation is greater than the second angle threshold, the candidate parking trajectory segment is removed.

[0010] Optionally, the parking space features include first type features designed based on the geometric shape of the parking trajectory point cluster and second type features designed based on the trajectory behavior information in the cluster. The first type features at least include: the area of the minimum circumscribed rectangle of the cluster, the aspect ratio of the minimum circumscribed rectangle of the cluster, the ratio of the cluster contour area to its convex hull area, the included angle between the cluster main line direction and its nearest road direction, the number of first-order adjacent clusters of the cluster in the main line direction and the short line direction, and the shortest distance from the center point of the minimum circumscribed rectangle of the cluster to its nearest road center line. The second type features at least include: the average number of trajectory points contained in each trajectory in the cluster, the average speed of trajectory points in the cluster, the average duration of trajectories in the cluster, the average number of parking points of each trajectory in the cluster, the average duration of trajectory parking in the cluster, the proportion of U-turn or turning trajectories, the trajectory point direction diversity index calculated based on the direction distribution, the trajectory time overlap rate in the cluster, and the trajectory point time diversity index calculated based on the time distribution.

[0011] Optionally, the calculation of the trajectory point direction diversity index is specifically: Divide 0°~360° into nodes with 30° as the interval, calculate the proportion of trajectory points in each direction interval in the cluster to the total number of trajectory points, and use the Shannon diversity index to calculate the trajectory point direction diversity of the cluster.

[0012] Optionally, the calculation formula of the Shannon diversity index is: ; Wherein, N i The number of trajectory points in each direction interval in the cluster is N, and N is the total number of trajectory points, n The number of direction intervals is N.

[0013] Optionally, the calculation formula of the trajectory time overlap rate in the cluster is: ; Wherein, k is the number of trajectories in the cluster, The duration of the i-th trajectory in the cluster is k, and k is the duration of the i-th trajectory in the cluster. The duration of the i-th trajectory in the cluster is k, and k is the duration of the i-th trajectory in the cluster.

[0014] Optionally, the trajectory point time diversity index calculated based on the time distribution is calculated by the formula: ; in, P i N represents the number of trajectory points within each hourly interval from 0 to 24 hours within the cluster, and N is the total number of trajectory points.

[0015] Optionally, the process of classifying and identifying the parking space information to be tested using the trained parking space recognition model includes: The trained parking space recognition model is used to perform preliminary classification of daily parking trajectory point clusters. Weighted voting is then applied to the classification results of the same parking trajectory point cluster over multiple days, and the type with the most votes is taken as the preliminary classification type for that cluster. Let +1 represent "street parking space" and -1 represent "parking lot parking space". The formula for calculating the preliminary classification result Y is as follows: Y = ; Where sign(x) is the sign function, which returns +1 when x > 0, indicating a roadside parking space, -1 when x < 0, indicating a parking lot space, and x = 0, indicating a flat ticket, and randomly outputs the classification result; For the first i The classification result is given by day d, where d represents the number of days for classification. For the first i The weight of the classification results ; Then, if the total number of first-order neighbor clusters in the main and short directions of the cluster exceeds a preset number, a second vote is conducted based on the preliminary judgment results of the first-order neighbor clusters in the main and short directions of the cluster. The result with the most votes is taken as the final parking space type. The formula for calculating the final classification result is as follows: Y_final = ; Where n is the total number of first-order neighbor clusters in the main line and short line directions; Finally, based on the identification of roadside parking spaces and parking lot spaces, the average daily dwell time of parking spaces over multiple days is analyzed. If the median of the average dwell time over multiple days is less than 20 minutes, it is judged as a temporary parking space.

[0016] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: This invention discloses an OSM road network parking space update method based on crowdsourced trajectory data. The method includes: performing topology checks and quality cleaning on the road network data based on crowdsourced trajectory data; performing map matching on the processed road network data and trajectory data to obtain preprocessed data including the connection relationships between trajectory points and their respective road segments; extracting parking trajectory point information based on the preprocessed data by setting limitations and filtering conditions; identifying parking trajectory point clusters using the DBSCAN algorithm; extracting parking space features based on the parking trajectory point clusters; constructing a parking space recognition model based on these features; classifying and identifying roadside parking spaces, parking lot parking spaces, and temporary parking spaces using the trained parking space recognition model; and updating the parking space information on the OSM road network. This invention, while identifying parking space information, avoids the shortcomings of traditional methods, providing better, more complete, and more accurate parking space information for navigation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the OSM road network parking space update method in this embodiment. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] The purpose of this invention is to provide an OSM road network parking space update method based on crowdsourced trajectory data, aiming to solve or improve at least one of the above-mentioned technical problems.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] like Figure 1 As shown, this invention provides a method for updating parking spaces in an OSM road network based on crowdsourced trajectory data, including: Step 100: Based on the crowdsourced trajectory data, perform topology checks on the road network data and quality cleaning on the trajectory data. Then, perform map matching on the processed road network data and trajectory data to obtain preprocessed data including the connection relationship between trajectory points and their respective road segments.

[0023] Step 200: Based on the preprocessed data, set the limiting range and filtering conditions, extract the parking trajectory point information, and then use the DBSCAN algorithm to identify the parking trajectory point clusters.

[0024] As one specific implementation method, this step includes: Based on the parking behavior pattern and related information at the end of the trajectory, the restriction range and filtering conditions are set, and the parking trajectory is filtered to select the zero speed point as the parking trajectory point. Then, the DBSCAN algorithm is used to cluster the parking trajectory points to obtain a cluster composed of dense parking trajectory points.

[0025] The process of filtering parking trajectories is as follows: A parking trajectory segment is defined as a continuous sequence of trajectory points that maintain a set stationary state for a vehicle over a period of time. Based on this definition, the cleaned trajectory data is iterated. If the average speed of any trajectory is lower than a preset speed threshold, it is determined whether a sub-segment exists that meets the following conditions: if the speed of each trajectory point in the sub-segment is lower than the preset speed threshold, and the duration of the sub-segment is greater than or equal to a preset time threshold, then the sub-segment is extracted as a candidate parking trajectory segment, and a secondary screening is performed on the candidate parking trajectory segments. Calculate the straight-line distance between the start and end points of the candidate parking trajectory segment. If the distance is greater than a preset distance threshold, the candidate parking trajectory segment calculated in this instance is removed.

[0026] Calculate the change in direction angle between consecutive trajectory points within a candidate parking trajectory segment, and calculate the sum and standard deviation. If the sum of the direction angle changes is greater than the first angle threshold or the standard deviation is greater than the second angle threshold, then the candidate parking trajectory segment calculated in this instance is removed.

[0027] Step 300: Extract parking space features based on the parking trajectory point cluster, construct a parking space recognition model based on the parking space features, use the trained parking space recognition model to classify and recognize the parking space information to be tested, and update the parking space information on the OSM road network; the parking space information includes roadside parking spaces, parking lot parking spaces and temporary parking spaces.

[0028] Among them, the parking space features include a first type of features designed based on the geometric shape of parking trajectory point clusters and a second type of features designed based on trajectory behavior information within the cluster.

[0029] The first type of features includes at least: the area of ​​the minimum bounding rectangle of the cluster, the aspect ratio of the minimum bounding rectangle of the cluster, the ratio of the cluster outline area to its convex hull area, the angle between the main line direction of the cluster and the direction of the nearest road, the number of first-order neighboring clusters of the cluster in the main line direction and the short line direction, and the shortest distance from the center point of the minimum bounding rectangle of the cluster to the center line of its nearest road.

[0030] The second type of features includes at least: the average number of trajectory points per trajectory within a cluster, the average speed of trajectory points within a cluster, the average duration of trajectories within a cluster, the average number of dwell points per trajectory within a cluster, the average dwell time of trajectories within a cluster, the proportion of U-turn or turning trajectories, the trajectory point direction diversity index calculated based on direction distribution, the overlap rate between trajectory time and commuting time within a cluster, and the trajectory point time diversity index calculated based on time distribution.

[0031] The calculation of the trajectory point orientation diversity index within this cluster is as follows: Divide the area from 0° to 360° using 30° as nodes, calculate the proportion of trajectory points within each directional interval to the total number of trajectory points within the cluster, and then use the Shannon diversity index to calculate the orientation diversity of the cluster's trajectory points. The specific formula is as follows: ; in, N i Let N be the number of trajectory points within each directional interval of the cluster, and N be the total number of trajectory points. n This represents the number of directional intervals.

[0032] The calculation of the overlap rate between trajectory duration and commuting time within this cluster is as follows: First, calculate the overlap rate between the duration of each trajectory and commuting time within the cluster. Then, calculate the average of the overlap rates between the duration of all trajectories and commuting time. The specific formula is as follows: ; Where k is the number of trajectories within the cluster, For the first in the cluster i The duration of the trajectory overlaps with the commuting time. This refers to commute time.

[0033] The calculation of the temporal diversity index of trajectory points within this cluster is as follows: The time frame from 0h to 24h is divided into 1-hour intervals. The proportion of trajectory points Pi within each hourly interval to the total number of trajectory points N within the cluster is calculated. The Shannon diversity index is then used to calculate the temporal distribution diversity of trajectory points within the cluster. The specific formula is as follows: ;

[0034] in, P i N represents the number of trajectory points within each hourly interval from 0 to 24 hours within the cluster, and N is the total number of trajectory points.

[0035] As a specific implementation method, the process of classifying and recognizing the parking space information to be tested using the trained parking space recognition model in this step includes: The trained parking space recognition model is used to perform preliminary classification of daily parking trajectory point clusters. Weighted voting is then applied to the classification results of the same parking trajectory point cluster over multiple days, and the type with the most votes is taken as the preliminary classification type for that cluster. Let +1 represent "street parking space" and -1 represent "parking lot parking space". The formula for calculating the preliminary classification result Y is as follows: Y = ;

[0036] Where sign(x) is the sign function, which returns +1 when x > 0, indicating a roadside parking space, -1 when x < 0, indicating a parking lot space, and x = 0, indicating a flat ticket, and randomly outputs the classification result; For the first i The classification result is given by day d, where d represents the number of days for classification. For the first i The weights of the classification results, and: ;

[0037] Then, if the total number of first-order neighbor clusters in the main and short directions of the cluster exceeds a preset number, a second vote is conducted based on the preliminary judgment results of the first-order neighbor clusters in the main and short directions of the cluster. The result with the most votes is taken as the final parking space type. The formula for calculating the final classification result is as follows: Y_final = ;

[0038] Where n is the total number of first-order neighbor clusters in the main line and short line directions; Finally, based on the identification of roadside parking spaces and parking lot spaces, the average daily dwell time of parking spaces over multiple days is analyzed. If the median of the average dwell time over multiple days is less than 20 minutes, it is judged as a temporary parking space.

[0039] Based on the above technical solution, the following specific embodiments are provided.

[0040] Step 100, Data Preprocessing: Perform topology checks on the road network data, perform quality cleaning on the trajectory data, and then perform map matching on the road network and trajectory data to connect the trajectory points with their respective road segments.

[0041] Step 200, Parking trajectory point cluster identification: Based on the preprocessing of crowdsourced trajectory data, limit range and filtering conditions are set to conduct research on parking trajectory point information extraction, and then the DBSCAN algorithm is used to complete the parking trajectory point cluster identification.

[0042] The identification of the parking trajectory point clusters described in step 200 specifically includes: Step 2.1, considering whether there are parking behavior patterns such as sudden speed changes, low-speed hovering, right-angle turns, and reverse driving in the end trajectory, setting filtering conditions based on information such as speed, direction, distance, and time to filter the parking trajectories, and screening the speed zero-value points from the parking trajectories as parking trajectory points. Step 2.2, using the DBSCAN clustering algorithm that can discover clusters of any shape and effectively filter out noise points from the data to cluster the parking trajectory points, and obtaining clustering clusters composed of dense parking trajectory points.

[0043] Further, the filtering of the parking trajectories in step 2.1 specifically includes: Step 2.1.1 Define the parking trajectory segment as a continuous sequence of trajectory points where the vehicle remains in a "stationary" or "nearly stationary" state for a period of time, and then set filtering conditions based on the definition of the parking trajectory segment to obtain candidate parking trajectories; Step 2.1.2 To avoid identifying patterns such as low-speed hovering, U-turns, and right-angle turns that do not conform to the "parking behavior pattern", set exclusion conditions for the candidate parking trajectories to obtain more accurate parking trajectories.

[0044] Further, the exclusion conditions described in step 2.1.1 are set as follows: If the average speed of a certain trajectory is lower than the low-speed condition (V_threshold), then further determine whether there is a segment of the trajectory that satisfies the definition of the parking trajectory segment in this trajectory, that is, the speed of each point in this segment satisfies the low-speed condition (vp < V_threshold), and the start-stop time difference of this segment is greater than or equal to a certain time threshold (T_duration >= T_threshold). If it exists, extract this segment of the trajectory from this trajectory as a candidate parking trajectory. After all the cleaned trajectories are traversed, the candidate parking trajectory information in the research area is obtained.

[0045] Further, the exclusion conditions described in step 2.1.2 are set as follows: (1) Considering that during the entire low-speed duration of parking, the total displacement distance of the vehicle is very small, which can effectively distinguish low-speed hovering. Calculate the straight-line distance D_total between the first point and the last point of the candidate parking trajectory segment. If D_total > D_threshold, then eliminate this candidate parking trajectory; (2) Considering that the direction (heading angle) of the vehicle changes violently or continuously within the low-speed segment, it is very likely to be hovering, turning around, or making a right-angle turn. Therefore, calculate the direction angle changes between consecutive points of the candidate parking trajectory segment, and sum and calculate the standard deviation: If the total direction change Heading_total or the standard deviation of the direction change Heading_std is very large ( Heading_total>Ht_threshold or If Heading_std>Hs_threshold), then the candidate parking trajectory is removed.

[0046] Step 300, Parking Space Type Identification: After identifying the parking trajectory point clusters, a training model is constructed based on the parking space characteristics. The clusters are analyzed, identified, and classified to distinguish information such as roadside parking spaces, parking lot parking spaces, and temporary parking spaces. The parking space information is then updated on the OSM road network.

[0047] The parking space type determination in step 300 specifically includes: Step 3.1 involves analyzing the trajectory behavior of vehicles passing through parking spaces and designing features based on clusters and trajectory information within clusters to ensure a more comprehensive representation of parking space information.

[0048] Step 3.2: To ensure data integrity and usability and eliminate feature redundancy, missing value processing and feature selection are performed based on feature standardization.

[0049] Step 3.3: Label the clusters as roadside parking spaces, parking spaces in parking lots, and non-parking spaces. 70% are used as training samples and 30% are used as test samples. Use a random forest classification model as the basic classifier, taking into account the seven-day behavior pattern and adjacency information, and use a voting method to further accurately identify parking spaces in the clusters.

[0050] Furthermore, the cluster-based design features include: (1) Area: The area of ​​the smallest bounding rectangle of the cluster; (2) Shape: The aspect ratio of the minimum bounding rectangle of the cluster; (3) Convex hull: The ratio of the area of ​​the cluster outline to the area of ​​its convex hull; (4) Direction: The angle between the direction of the main line of the cluster and the direction of its nearest road; (5) Horizontal adjacency information: the number of first-order neighboring clusters along the main direction of the cluster; (6) Vertical adjacency information: the number of first-order neighbor clusters in the short-line direction of the cluster; (7) Distance to road segment: The shortest distance from the center point of the smallest bounding rectangle of the cluster to the center line of the nearest road.

[0051] Furthermore, the design features based on trajectory information within clusters include: (1) Number of trajectory points in a cluster: the number of trajectory points in a cluster divided by the number of trajectories; (2) Cluster trajectory velocity: the average velocity of trajectory points in a cluster; (3) Cluster trajectory duration: The average duration of trajectories within a cluster; (4) Cluster trajectory dwell time: the number of trajectory dwell points within a cluster divided by the number of trajectories; (5) Trajectory dwell time within cluster: average dwell time of trajectories within clusters; (6) U-turn rate: the number of U-turn trajectories within a cluster divided by the total number of trajectories; (7) Cluster trajectory point orientation diversity: Divide 0°~360° every 30°, calculate the proportion of trajectory point Ni in each directional interval within the cluster to the total number of trajectory points N, and use the Shannon diversity index to calculate the cluster trajectory point orientation diversity. (8) Cluster trajectory time overlap rate with commuting time: the average overlap rate of trajectory time and commuting time within a cluster; (9) Temporal diversity of cluster trajectory points: Divide 0h~24h every 1h, calculate the proportion of trajectory point Pi in each hour interval within the cluster to the total number of trajectory points N, and use the Shannon diversity index to calculate the temporal distribution diversity of cluster trajectory points.

[0052] Furthermore, step 3.3, which involves considering seven-day behavioral patterns and adjacency information and using a voting method to further accurately identify parking spaces within clusters, includes: (1) Parking trajectory point clusters are identified for the daily trajectory data, the above-mentioned design features are calculated, and a random forest classifier is used to identify parking spaces. Then, the parking space results identified over seven days are voted on, and the parking space result with the most votes is assigned to that cluster.

[0053] (2) If the total number of first-order neighbor clusters in the main and short-line directions of a cluster is greater than 2, then the results of the first-order neighbor clusters in the main and short-line directions are further considered, and a vote is taken to identify the parking space type with the most votes. In other cases, the seven-day voting results of the cluster shall prevail.

[0054] (3) Then, based on the identification of roadside parking spaces and parking lots, analyze the average daily parking space trajectory dwell time over seven days. If the median of the average dwell time over seven days is less than 20 minutes, it is judged as a temporary parking space.

[0055] In summary, the technical solution described in this embodiment, on the one hand, extracts and identifies parking space information based on crowdsourced trajectory data, offering advantages such as low cost and short cycle time. While accurately identifying parking space information, it also avoids the shortcomings of traditional methods and provides better, more socially acceptable parking space information updates. On the other hand, this invention can not only identify and extract parking space information but also classify specific categories of parking spaces, such as roadside parking spaces, parking lot parking spaces, and temporary parking spaces, thus better supplementing parking space information and providing better, more complete, and more accurate parking space information for navigation. Furthermore, this technical solution is scalable; it can be further optimized by integrating multi-dimensional data from remote sensing, POI, and regional functions to improve the accuracy of the judgment.

[0056] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0057] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for updating parking spaces in an OSM road network based on crowdsourced trajectory data, characterized in that, include: Based on crowdsourced trajectory data, topology checks are performed on road network data and quality cleaning is performed on trajectory data. The processed road network data and trajectory data are then matched with a map to obtain preprocessed data including the connection relationship between trajectory points and their respective road segments. Based on the preprocessed data, the range of restrictions and filtering conditions are set to extract the parking trajectory point information, and then the DBSCAN algorithm is used to identify the parking trajectory point clusters. Parking space features are extracted based on the parking trajectory point clusters, and a parking space recognition model is constructed based on the parking space features. The trained parking space recognition model is used to classify and recognize the parking space information to be tested, and the parking space information is updated on the OSM road network. The parking space information includes roadside parking spaces, parking lot parking spaces, and temporary parking spaces.

2. The OSM road network parking space update method based on crowdsourced trajectory data according to claim 1, characterized in that, The process involves setting limits and filtering conditions based on the preprocessed data, extracting parking trajectory point information, and then using the DBSCAN algorithm to identify parking trajectory point clusters. Specifically, this includes: Based on the parking behavior pattern and related information at the end of the trajectory, the restriction range and filtering conditions are set, and the parking trajectory is filtered to select the zero speed point as the parking trajectory point. Then, the DBSCAN algorithm is used to cluster the parking trajectory points to obtain a cluster composed of dense parking trajectory points.

3. The OSM road network parking space update method based on crowdsourced trajectory data according to claim 1, characterized in that, The process of filtering parking trajectories is as follows: A parking trajectory segment is defined as a continuous sequence of trajectory points that satisfy a set condition for a vehicle to remain stationary over a period of time. According to the definition, after traversing the cleaned trajectory data, if the average speed of any trajectory is lower than the preset speed threshold, it is determined whether there is a sub-segment in the trajectory that meets the following conditions: if the speed of each trajectory point in the sub-segment is lower than the preset speed threshold, and the duration of the sub-segment is greater than or equal to the preset time threshold, then the sub-segment is extracted as a candidate parking trajectory segment, and the candidate parking trajectory segment is further filtered.

4. The OSM road network parking space update method based on crowdsourced trajectory data according to claim 3, characterized in that, The process of performing a secondary screening of candidate parking trajectory segments is as follows: Calculate the straight-line distance between the start and end points of the candidate parking trajectory segment. If the distance is greater than a preset distance threshold, the candidate parking trajectory segment calculated in this instance is removed. Calculate the change in direction angle between consecutive trajectory points within a candidate parking trajectory segment, and calculate the sum and standard deviation. If the sum of the direction angle changes is greater than the first angle threshold or the standard deviation is greater than the second angle threshold, then the candidate parking trajectory segment calculated in this instance is removed.

5. The OSM road network parking space update method based on crowdsourced trajectory data according to claim 1, characterized in that, The parking space features include a first type of features designed based on the geometric shape of parking trajectory point clusters and a second type of features designed based on trajectory behavior information within the cluster; The first type of features includes at least: the area of ​​the minimum bounding rectangle of the cluster, the aspect ratio of the minimum bounding rectangle of the cluster, the ratio of the cluster outline area to its convex hull area, the angle between the main line direction of the cluster and the direction of the nearest road, the number of first-order neighboring clusters of the cluster in the main line direction and the short line direction, and the shortest distance from the center point of the minimum bounding rectangle of the cluster to the center line of its nearest road. The second type of features includes at least: the average number of trajectory points per trajectory within a cluster, the average speed of trajectory points within a cluster, the average duration of trajectories within a cluster, the average number of dwell points per trajectory within a cluster, the average dwell time of trajectories within a cluster, the proportion of U-turn or turning trajectories, the trajectory point direction diversity index calculated based on direction distribution, the overlap rate between trajectory time and commuting time within a cluster, and the trajectory point time diversity index calculated based on time distribution.

6. The OSM road network parking space update method based on crowdsourced trajectory data according to claim 5, characterized in that, The calculation of the trajectory point orientation diversity index is as follows: Divide the 0°~360° range into 30° intervals, calculate the proportion of trajectory points in each directional interval within the cluster to the total number of trajectory points, and use the Shannon diversity index to calculate the directional diversity of trajectory points in the cluster.

7. The OSM road network parking space update method based on crowdsourced trajectory data according to claim 6, characterized in that, The formula for calculating the aroma diversity index is as follows: ； in, N i Let N be the number of trajectory points within each directional interval of the cluster, and N be the total number of trajectory points. n This represents the number of directional intervals.

8. The OSM road network parking space update method based on crowdsourced trajectory data according to claim 5, characterized in that, The formula for calculating the overlap rate between the intra-cluster trajectory time and commuting time is as follows: ； Where k is the number of trajectories within the cluster, Let the duration of the i-th trajectory within the cluster coincide with the commuting time. This refers to commute time.

9. The OSM road network parking space update method based on crowdsourced trajectory data according to claim 5, characterized in that, The trajectory point temporal diversity index, calculated based on temporal distribution, is calculated using the following formula: ； in, P i N represents the number of trajectory points within each hourly interval from 0 to 24 hours within the cluster, and N is the total number of trajectory points.

10. The OSM road network parking space update method based on crowdsourced trajectory data according to claim 1, characterized in that, The process of classifying and identifying the parking space information under test using the trained parking space recognition model includes: The trained parking space recognition model is used to perform preliminary classification of daily parking trajectory point clusters. Weighted voting is then applied to the classification results of the same parking trajectory point cluster over multiple days, and the type with the most votes is taken as the preliminary classification type for that cluster. Let +1 represent "streetside parking space" and -1 represent "parking lot parking space". The formula for calculating the preliminary classification result Y is as follows: Y＝ ； Where sign(x) is the sign function, which returns +1 when x > 0, indicating a roadside parking space, -1 when x < 0, indicating a parking lot space, and x = 0, indicating a flat ticket, and randomly outputs the classification result; For the first i The classification result is given by day d, where d represents the number of days for classification. For the first i The weight of the classification results ; Then, if the total number of first-order neighbor clusters in the main and short directions of the cluster exceeds a preset number, a second vote is conducted based on the preliminary judgment results of the first-order neighbor clusters in the main and short directions of the cluster. The result with the most votes is taken as the final parking space type. The formula for calculating the final classification result is as follows: Y_final= ; Where n is the total number of first-order neighbor clusters in the main line and short line directions; Finally, based on the identification of roadside parking spaces and parking lot spaces, the average daily dwell time of parking spaces over multiple days is analyzed. If the median of the average dwell time over multiple days is less than 20 minutes, it is judged as a temporary parking space.

Images