Shared parking space automatic matching and guiding navigation method based on passive internet of things label

Abstract

The application provides a shared parking space automatic matching and guiding navigation method based on a passive Internet of Things label, relates to the technical field of intelligent traffic, and comprises the following steps: receiving user terminal information, constructing a candidate parking space set based on a multi-layer search grid and density clustering, determining an optimal navigation path by using a road network connection state matrix and a congestion propagation prediction tree, and combining a parking space group state propagation model to calculate a maximum reservation waiting time, so that parking space reservation and verification are realized. The application improves the parking space resource utilization rate, reduces the time for users to find parking spaces, and effectively alleviates the urban traffic congestion problem.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent transportation technology, and in particular to a shared parking space automatic matching and guided navigation method based on passive Internet of Things tags. BACKGROUND

[0002] With the acceleration of urbanization and the rapid growth of car ownership, the problem of parking difficulty is increasingly prominent, especially in the core business district and densely populated areas of the city. To solve this problem, shared parking mode emerged as the times require, through the revitalization of private parking resources, to improve the utilization rate of parking spaces. The current shared parking system usually adopts the combination of mobile applications and various sensors to realize the functions of parking information collection, release and reservation. Passive Internet of Things tag technology, with its low cost, maintenance-free and easy deployment, is gradually applied to the field of parking management, which can effectively record and transmit parking information and provide data basis for parking management system.

[0003] However, the existing shared parking technology has some shortcomings. The traditional parking navigation method lacks a dynamic matching mechanism, usually only considers the distance factor and ignores real-time factors such as road conditions and traffic flow, resulting in that the recommended parking space often does not meet the actual needs of users, and the parking space may be occupied when the user arrives. The existing method generally lacks accurate parking state prediction ability, and cannot predict the availability of parking spaces according to historical data and the change rule of surrounding parking space state, making it difficult to provide reliable parking reservation service for users. Most of the matching methods use simple global search algorithm to find parking spaces, which has low computational efficiency in large-scale urban environment and cannot meet the real-time query demand in high-concurrency scenarios, affecting user experience. SUMMARY

[0004] The embodiment of the present application provides a shared parking space automatic matching and guided navigation method based on passive Internet of Things tags, which can solve the problems in the prior art.

[0005] In a first aspect, the embodiment of the present application provides a shared parking space automatic matching and guided navigation method based on passive Internet of Things tags, comprising:

[0006] Receiving destination location information and vehicle use time period information sent by a user terminal; based on the destination location information, obtaining parking space information within a preset range from a server, wherein the parking space information is stored through a passive Internet of Things tag installed in the parking space;

[0007] Based on the destination location information, a multi-layer search grid is constructed, the grid structure is dynamically adjusted based on density clustering by calculating parking score, and the candidate parking space set is generated by layer-by-layer expansion;

[0008] For the candidate parking space set, a road network connection state matrix is constructed, a candidate path set is generated, a congestion propagation prediction tree is constructed based on real-time traffic flow, a path score is calculated to determine an optimal navigation path;

[0009] According to the optimal navigation path, a target parking space is determined, a parking space group is divided based on a historical occupation state sequence of adjacent parking spaces, a state propagation path and a state propagation rate are calculated, and a maximum reservation waiting time is determined;

[0010] According to the optimal navigation path and the maximum reservation waiting time, parking reservation confirmation information is generated, and the parking reservation confirmation information is sent to a user terminal;

[0011] When the target parking space is reached, passive Internet of Things tag information is read and identity verification is performed, and the parking space usage state is updated after the verification is passed.

[0012] In an optional embodiment, a multi-layer search grid is constructed based on destination location information, a candidate parking space set is generated by calculating parking space scores and dynamically adjusting the grid structure based on density clustering, including:

[0013] Each layer of the grid adopts a regular hexagon structure and the side length increases by a preset multiple;

[0014] The state of the parking space in the grid is judged in real time, and the parking space score is calculated based on the parking space idle state, historical reservation data and turnover rate;

[0015] A density clustering method based on sector units is used to divide the grid, the density clustering area is determined by calculating the continuous distribution of the parking space score, and the grid structure is dynamically adjusted according to the shape characteristics of the density clustering area;

[0016] A layer-by-layer expansion method is used, when the number of parking spaces with a parking space score greater than a preset score threshold in a layer of the grid is less than a preset number of parking spaces, the next layer of the grid is expanded, until the preset number of parking spaces is met or the preset upper limit of the number of grid layers is reached, and a candidate parking space set is generated.

[0017] In an optional embodiment, a density clustering method based on sector units is used to divide the grid, the density clustering area is determined by calculating the continuous distribution of the parking space score, and the grid structure is dynamically adjusted according to the shape characteristics of the density clustering area, including:

[0018] The grid is divided into a plurality of sector units, and a density calculation grid is constructed in the sector unit, the grid size of the density calculation grid being proportional to the size of the parking space;

[0019] The parking space score in the density calculation grid is Gaussian smoothed, the smoothing range is controlled by setting the standard deviation of the Gaussian kernel function, and a continuous distribution of the parking space score is generated;

[0020] The local density value of the sector unit is calculated based on the continuous distribution of the parking space score, the density peak value point is extracted, the density clustering area is determined based on the area between adjacent density peak value points;

[0021] The shape features of the density clustering area are calculated, including the area, eccentricity and directionality, and the corresponding dominant expansion direction is determined according to the shape features;

[0022] The boundary fitting of the density clustering area of each sector unit is performed to generate the density contour line, and the density jump point is determined based on the curvature change of the density contour line;

[0023] The transition grid is inserted at the density jump point, and the grid side length is inversely proportional to the density gradient;

[0024] The change of the density distribution in the sector unit is monitored in real time, the merging operation is triggered when the minimum distance between adjacent density clustering areas is less than the preset distance threshold, the splitting operation is triggered when the maximum density difference in the density clustering area exceeds the preset density threshold, the grid division parameters of the corresponding sector unit are recalculated based on the results of merging or splitting, and the grid structure is dynamically updated.

[0025] In an optional embodiment, for the candidate parking space set, a road network connection state matrix is constructed, a candidate path set is generated, a congestion propagation prediction tree is constructed based on real-time traffic flow, and the path score is calculated to determine the optimal navigation path, which includes:

[0026] The road network topology structure is established according to the pre-acquired road level, the target road section passing capacity reference value is extracted, the turning constraint condition and the connection relationship between the target road sections are calculated, and the road network connection state matrix is generated;

[0027] In the road network connection state matrix, the candidate parking space set is taken as the terminal point, the reverse path search is performed according to the turning constraint condition, and the candidate path set is generated;

[0028] The congestion index is determined based on the ratio of the real-time traffic flow of the target road section to the passing capacity reference value, the road sections with the congestion index greater than the preset congestion threshold are marked, the traffic flow change rate between adjacent marked road sections is calculated, and the congestion spread vector is generated;

[0029] Based on the congestion spread vector, node expansion is performed in the road network connection state matrix to establish a congestion propagation prediction tree;

[0030] Based on the congestion propagation prediction tree, the passing probability and time cost of each path in the candidate path set are calculated, and the path score is determined by weighted summation, and the optimal navigation path is determined by selecting the path with the maximum path score.

[0031] In an alternative embodiment, based on the congestion spread vector, node expansion is performed in the road network connection state matrix to establish a congestion propagation prediction tree, comprising:

[0032] The congestion spread vector is decomposed into a direction component and a speed component, and the main propagation direction and the secondary propagation direction of the congestion propagation are determined according to the direction component and the speed component;

[0033] The target road segment is set as a root node, and node expansion is performed in the road network connection state matrix based on the main propagation direction and the secondary propagation direction, wherein each node contains a road segment identifier, a current traffic flow, an average vehicle speed, a number of lanes, and a turning probability matrix;

[0034] The propagation attenuation coefficient is calculated according to a distance attenuation function and a time attenuation function between adjacent nodes, and the congestion spread vector is modified based on the propagation attenuation coefficient to obtain a propagation weight between nodes;

[0035] Real-time traffic flow data of the nodes is collected, and when the real-time traffic flow data exceeds a preset traffic flow threshold, a real-time traffic flow change rate is calculated to determine whether a local update is triggered;

[0036] When the local update is triggered, an affected node set is determined, the affected node set is recalculated, and a local structure of the propagation prediction tree is updated;

[0037] According to the real-time traffic flow data of each node and the propagation weight, combined with a signal timing cycle, a turning delay between nodes is calculated, a weighted sum of the turning delay and a traffic flow propagation time is taken as a predicted congestion arrival time, and a congestion propagation prediction tree is generated.

[0038] In an alternative embodiment, according to the optimal navigation path, a target parking space is determined, adjacent parking spaces are divided into a parking space group based on a historical occupancy state sequence of the adjacent parking spaces, a state propagation path and a state propagation rate are calculated, and a maximum reservation waiting time is determined, comprising:

[0039] The position information of the target parking space and the spatial distribution information of the adjacent parking spaces are obtained, a parking space adjacency relationship matrix is established, a historical occupancy state sequence is collected, and a state transition probability between adjacent parking spaces is calculated;

[0040] Based on the state transition probability and the parking space adjacency relationship matrix, a spatial continuity threshold and a use mode similarity threshold are set, the parking spaces are grouped, the occupancy rate and the turnover frequency of the parking space group are calculated;

[0041] According to the historical occupancy state sequence of the parking space group, the state propagation path and the state propagation rate are calculated, and the key node parking space is determined; according to the turnover frequency and the state propagation rate of the parking space group, a reference waiting time is calculated;

[0042] Collecting real-time states of the key node parking spaces and real-time turnover data of the parking space group, dynamically correcting the reference waiting time according to a state propagation path and a state propagation rate, and obtaining a maximum reservation waiting time.

[0043] In an optional embodiment, according to a historical occupation state sequence of the parking space group, a state propagation path and a state propagation rate are calculated, and the key node parking space is determined to include:

[0044] Based on the historical occupation state sequence, a state change sequence of adjacent time windows in the parking space group is extracted, and a trigger relationship of state conversion of each parking space is obtained.

[0045] According to the trigger relationship, a state conversion weight between the parking spaces is calculated, and a parking space pair with a state conversion weight greater than a preset conversion threshold is determined as a chain occupation node pair.

[0046] Based on the chain occupation node pair, a state propagation directed graph is constructed, wherein a node corresponds to a parking space, an edge corresponds to a state propagation direction, and an edge weight corresponds to a state propagation probability.

[0047] In the state propagation directed graph, an edge with an edge weight greater than a preset probability threshold is extracted, and a state propagation path is determined.

[0048] The average state conversion time interval between adjacent node pairs on the state propagation path is calculated, and a state propagation rate is obtained.

[0049] According to the sum of the in-degree and out-degree of the parking space on the state propagation path, a parking space with the maximum degree is selected to determine the key node parking space.

[0050] In a second aspect of the embodiment of the application, a shared parking space automatic matching and guiding navigation system based on a passive Internet of Things label is provided, which includes:

[0051] A first unit is configured to receive destination location information and vehicle use time period information sent by a user terminal, obtain parking space information within a preset range from a server based on the destination location information, and store the parking space information through a passive Internet of Things label installed in the parking space.

[0052] A second unit is configured to construct a multi-layer search grid based on the destination location information, generate a candidate parking space set by expanding layer by layer through calculating parking space scores and dynamically adjusting a grid structure based on density clustering.

[0053] A third unit is configured to construct a road network connection state matrix for the candidate parking space set, generate a candidate path set, construct a congestion propagation prediction tree based on real-time traffic flow, calculate path scores to determine an optimal navigation path.

[0054] A fourth unit is configured to determine a target parking space based on the optimal navigation path, divide parking space groups based on a historical occupancy state sequence of adjacent parking spaces, calculate a state propagation path and a state propagation rate, and determine a maximum reservation waiting time;

[0055] A fifth unit is configured to generate parking space reservation confirmation information based on the optimal navigation path and the maximum reservation waiting time, and send the parking space reservation confirmation information to a user terminal.

[0056] A sixth unit is configured to read passive Internet of Things tag information and perform identity verification when the target parking space is reached, and update a parking space usage state after the identity verification is passed.

[0057] In a third aspect, an electronic device is provided, including:

[0058] a processor;

[0059] a memory for storing processor-executable instructions;

[0060] The processor is configured to invoke the instructions stored in the memory to execute the method described above.

[0061] In a fourth aspect, a computer-readable storage medium is provided, which stores computer program instructions, and the computer program instructions are executed by a processor to implement the method described above.

[0062] In the embodiments of the present application, the passive Internet of Things tag stores parking space information, and the multi-layer search grid technology is combined to achieve accurate identification and efficient matching of parking space resources, greatly improving the parking space utilization rate and user parking space searching efficiency, and effectively alleviating the urban parking difficulty problem. The navigation path generation method based on the road network connection state matrix and the congestion propagation prediction tree can dynamically adjust the optimal path according to real-time traffic flow, significantly reducing the time cost and fuel consumption of users searching for parking spaces, and reducing urban traffic congestion. The maximum reservation waiting time is determined by using the parking space group state propagation model, and the identity verification mechanism of the passive Internet of Things tag is combined to ensure the reliability of the parking space reservation and improve the turnover efficiency of the shared parking space, achieving intelligent management and accurate service of parking resources. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 FIG. 1 is a flowchart of a shared parking space automatic matching and guided navigation method based on a passive Internet of Things tag according to an embodiment of the present application;

[0064] Figure 2 FIG. 4 is a congestion propagation prediction flowchart. DETAILED DESCRIPTION

[0065] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0066] The technical solutions of the present application will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described in some embodiments.

[0067] Figure 1 The flowchart of the method for automatically matching and guiding navigation of shared parking spaces based on passive Internet of Things tags in the embodiments of the present application is shown in FIG. 1, which comprises the following steps. Figure 1

[0068] Receiving destination position information and use time period information sent by a user terminal; based on the destination position information, obtaining parking space information within a preset range from a server, wherein the parking space information is stored by a passive Internet of Things tag installed in the parking space;

[0069] Based on the destination position information, constructing a multi-layer search grid, calculating parking space scores, and dynamically adjusting the grid structure based on density clustering to generate a candidate parking space set by layer-by-layer expansion;

[0070] For the candidate parking space set, constructing a road network connection state matrix, generating a candidate path set, constructing a congestion propagation prediction tree based on real-time traffic flow, calculating path scores to determine an optimal navigation path;

[0071] Determining a target parking space according to the optimal navigation path, dividing parking space groups based on historical occupation state sequences of adjacent parking spaces, calculating state propagation paths and state propagation rates, and determining a maximum reservation waiting time;

[0072] Generating parking space reservation confirmation information according to the optimal navigation path and the maximum reservation waiting time, and sending the parking space reservation confirmation information to the user terminal;

[0073] When reaching the target parking space, reading passive Internet of Things tag information and performing identity verification, and updating the parking space use state after verification.

[0074] ​In a specific embodiment, after the user sends the destination location information and the use time period information through the terminal device, the system receives and processes these information. According to the received destination location information, the parking space information within a preset range is obtained from the server. These parking space information is stored through the passive Internet of Things label installed in each parking space, including the location coordinates, size specifications, usage rate, current occupancy state and other basic data of the parking space. The passive Internet of Things label uses radio frequency identification technology without battery power supply, which can be activated and transmit data by near field communication mode through the reading device.

[0075] After obtaining the parking space information, a multi-layer search grid is constructed based on the destination location information. The search grid adopts a center-outward expansion method, taking the destination as the center point, and constructing a multi-layer hexagonal grid structure. The side length of each layer of grid increases by a preset multiple, forming a search area from inside to outside. For each parking space in the grid, the parking space score is calculated, which considers factors such as parking space idle state, historical booking data and turnover rate. At the same time, the grid is dynamically adjusted by the density clustering method to identify parking space intensive areas, and the grid structure is optimized according to the shape characteristics of these areas. Through the layer-by-layer expansion method, when the number of parking spaces that meet the conditions in the inner layer grid is insufficient, the outer layer grid is expanded until the preset number of parking spaces is met or the upper limit of the preset grid layer is reached, and finally a candidate parking space set is generated.

[0076] For the generated candidate parking space set, a road network connection state matrix is constructed. Based on the pre-obtained road topology structure, the matrix records the turning constraint conditions and connectivity between target road segments. Taking the parking spaces in the candidate parking space set as the terminal point, the reverse path search is performed according to the turning constraint conditions to generate a candidate path set. The congestion index of each road segment is calculated using real-time traffic flow data, and the congestion road segments are marked. By calculating the traffic flow change rate between adjacent congestion road segments, a congestion spread vector is generated, and based on this, node expansion is performed in the road network connection state matrix to establish a congestion propagation prediction tree. The prediction tree can estimate the development trend of congestion conditions. Based on the congestion propagation prediction tree, the passing probability and time cost of each path in the candidate path set are calculated, the path score is determined by weighted summation, and the path with the highest score is selected as the optimal navigation path.

[0077] According to the determined optimal navigation path, the target parking space is determined. The position information of the target parking space and the spatial distribution information of the adjacent parking spaces are obtained, and a parking space adjacency relationship matrix is established. By analyzing the historical occupancy state sequence, the state transition probability between adjacent parking spaces is calculated, and based on the state transition probability and the parking space adjacency relationship, the parking spaces are divided into groups. For the divided parking space groups, the occupancy rate and the turnover frequency are calculated, and according to the historical occupancy state sequence, the state propagation path and the state propagation rate are calculated to determine the key node parking spaces. According to the turnover frequency of the parking space groups and the state propagation rate, the reference waiting time is calculated, and the reference waiting time is dynamically corrected in combination with the real-time state of the key node parking spaces and the real-time turnover data of the parking space groups, and finally the maximum reserved waiting time is determined.

[0078] According to the optimal navigation path and the maximum reserved waiting time, parking reservation confirmation information is generated, and the information is sent to the user terminal. The reservation confirmation information includes the target parking space position, the navigation route, the estimated arrival time, the maximum reserved waiting time, the usage rate and the like. After the user arrives at the target parking space according to the navigation guide, the terminal equipment reads the passive Internet of Things tag information installed on the parking space. The identity of the read tag information is verified to verify whether the user identity matches the reservation information. After the verification is passed, the parking space usage state is updated and marked as occupied, and usage confirmation information is sent to the user terminal. If the user does not arrive within the maximum reserved waiting time, the reservation will be automatically cancelled, and the parking space state will be updated to available, and the parking space will be opened to other users.

[0079] In the embodiment, through the combination of passive Internet of Things tag technology and algorithms such as multi-layer grid search, density clustering, congestion prediction and state propagation analysis, intelligent matching and accurate navigation of shared parking spaces are realized, the utilization efficiency of urban parking resources is improved, the time cost of users searching for parking spaces is reduced, and through the parking reservation mechanism, the users can smoothly park after arriving at the destination.

[0080] In an optional implementation, a multi-layer search grid is constructed based on the destination position information, a candidate parking space set is generated by layer-by-layer expansion through calculating parking space scores and dynamically adjusting the grid structure based on density clustering, and the method comprises the following steps:

[0081] Each layer of the grid adopts a regular hexagon structure and the side length increases by a preset multiple;

[0082] The state of the parking spaces in the grid is judged in real time, and the parking space scores are calculated based on the parking space idle state, historical reservation data and turnover rate;

[0083] A density clustering method based on sector units is used to divide the grid, the density clustering areas are determined by calculating the continuous distribution of the parking space scores, and the grid structure is dynamically adjusted according to the shape characteristics of the density clustering areas;

[0084] The layer-by-layer expansion method is adopted. When the number of parking spaces with a score greater than a preset score threshold in a layer of grid is less than a preset number of parking spaces, the next layer of grid is expanded until the preset number of parking spaces is met or a preset upper limit of the number of layers of grid is reached, to generate a candidate parking space set.

[0085] In a specific embodiment, a destination position set by a user is obtained as a center point of a multi-layer search grid. A coordinate system with the center point as the origin is constructed, and a first layer of search grid is initialized. The grid adopts a regular hexagon structure, and the length of each side of the hexagon is a preset reference value, such as 100 meters. The regular hexagon structure has more uniform coverage characteristics than a traditional square grid, and can reduce search errors caused by directional bias.

[0086] For the first layer of grid, it is divided into 6 sector units, and each sector covers an angle range of 60 degrees. Real-time parking space information is obtained in each sector, including parking space position coordinates, current idle state, historical booking data, and turnover rate. For each parking space, a parking space score is calculated. For example, it is determined whether the parking space is currently idle, and an idle parking space is given a basic score of 10 points. According to the historical booking data, if the booking rate of the parking space in the past 24 hours is less than 30%, 5 points are added, if the booking rate is between 30% and 70%, 3 points are added, and if the booking rate is higher than 70%, 1 point is added. Then, the turnover rate of the parking space is considered. If the average parking time is less than 2 hours, 4 points are added, if the average parking time is between 2 and 4 hours, 2 points are added, and if the average parking time is more than 4 hours, 0 points are added. Finally, the three parts of the score are added together to obtain the comprehensive score of the parking space.

[0087] After completing the score calculation of all parking spaces in the first layer of grid, a sector unit-based density clustering method is applied to divide the grid. In the density clustering process, the density threshold is set to 50 parking spaces with a score greater than 12 points per square kilometer, and the clustering radius is 200 meters. When a region meets the above density conditions, it is marked as a high-density region. The shape characteristics of these high-density regions are analyzed. If an elliptical high-density region appears in a sector, and the included angle between the long axis direction of the elliptical high-density region and the radial direction of the sector is less than 15 degrees, the grid structure is dynamically adjusted along the long axis direction, and the extension boundary of the sector is additionally extended by 20% along the long axis direction.

[0088] After the first layer grid evaluation is completed, the number of parking spaces with a score greater than a preset score threshold (e.g., 15) is counted. If the number is less than a preset number of parking spaces (e.g., 20), the second layer grid is expanded. The second layer grid also adopts a regular hexagon structure, but the side length is twice that of the first layer, i.e., 200 meters. The same parking space scoring and density clustering process as the first layer is performed on the second layer. If the total number of parking spaces with a score greater than the threshold in the first and second layers is still less than the preset number of parking spaces, the third layer grid is expanded, and the side length is three times that of the first layer, i.e., 300 meters. At most, the expansion is up to a preset upper limit of the number of layers, e.g., 5 layers.

[0089] For example, a user sets a business district center as a destination. A first layer of hexagonal grid is constructed, with a side length of 100 meters. In the 6 sectors of the first layer grid, a total of 15 idle parking spaces are detected. The system calculates the score of each parking space: parking space A is currently idle (10 points), the historical booking rate is 20% (5 points are added), the average parking duration is 1.5 hours (4 points are added), and the total score is 19 points; parking space B is currently idle (10 points), the historical booking rate is 50% (3 points are added), the average parking duration is 3 hours (2 points are added), and the total score is 15 points; the scores of other parking spaces are calculated in the same way. According to the statistics, there are 8 parking spaces with a score greater than 15 in the first layer grid, which does not reach the preset 20 parking spaces, and the second layer grid is expanded.

[0090] The second layer grid has a side length of 200 meters, detects 35 idle parking spaces, and calculates the scores. An elliptical high-density area appears in the sector in the northeast direction, with a long axis direction at an angle of 10 degrees with the sector radial direction, meeting the dynamic adjustment condition. The extended boundary of the sector is additionally extended along the long axis direction by 20%. The total number of parking spaces with a score greater than 15 in the first and second layers reaches 25, which exceeds the preset 20, and the system stops expanding.

[0091] From the 25 high-score parking spaces, the 20 closest to the destination are selected to form a candidate parking space set, and presented to the user in order of distance from the destination. Through this method, the most suitable parking space for the user can be efficiently recommended, while ensuring that the number of recommended parking spaces meets the user's needs, improving the user's parking experience.

[0092] In an optional implementation, a sector unit-based density clustering method is used to divide the grid, a density clustering area is determined by calculating the continuous distribution of parking space scores, and the grid structure is dynamically adjusted according to the shape characteristics of the density clustering area, including:

[0093] The grid is divided into a plurality of sector units, and a density calculation grid is constructed in the sector unit, wherein the grid size of the density calculation grid is proportional to the size of the parking space;

[0094] Gaussian smoothing is performed on the parking space score in the density calculation grid, and the smoothing range is controlled by setting the standard deviation of the Gaussian kernel function to generate a continuous distribution of parking space scores.

[0095] The local density value of the sector unit is calculated based on the continuous distribution of the parking space score, the density peak points are extracted, and the density clustering area is determined based on the area between adjacent density peak points.

[0096] The shape features of the density clustering area, including the area, eccentricity and directionality, are calculated, and the corresponding dominant expansion direction is determined according to the shape features.

[0097] The boundary fitting of the density clustering area of each sector unit is performed to generate the density contour line, and the density jump point is determined based on the curvature change of the density contour line.

[0098] A transition grid is inserted at the density jump point, and the grid length is inversely proportional to the density gradient.

[0099] The change of the density distribution in the sector unit is monitored in real time, and the merging operation is triggered when the minimum distance between adjacent density clustering areas is less than the preset distance threshold, and the splitting operation is triggered when the maximum density difference within the density clustering area exceeds the preset density threshold. The grid division parameters of the corresponding sector unit are recalculated based on the results of merging or splitting, and the grid structure is dynamically updated.

[0100] In one specific embodiment, the area to be analyzed is divided into multiple sector units, for example, the entire parking lot area is divided into 8 sector units, each covering a 45-degree angle range. In each sector unit, a density calculation grid is constructed, and the grid size is proportional to the parking space size. Taking a standard parking space size of 5m x 2.5m as an example, the grid size can be set to 0.5m x 0.5m, so that each parking space area contains about 50 basic grid units, ensuring the balance between calculation accuracy and efficiency.

[0101] Gaussian smoothing is performed on the parking space score in the density calculation grid, and the smoothing range is controlled by setting the standard deviation of the Gaussian kernel function. For example, the standard deviation is set to 2.0, and the smoothing window size is 7x7, which ensures that the smoothed score distribution retains local features and eliminates noise interference. For example, the original parking space score of a grid point is 0.85, and after Gaussian smoothing, the scores of the point and its surrounding grid points are adjusted to a continuous distribution, forming a smooth transition area with a score range of 0.65 to 0.90.

[0102] Based on the smoothed parking space score continuous distribution, the local density value of each grid point in the sector unit is calculated. The density calculation adopts a neighborhood range with a radius of 3 grid units, and the proportion of the number of grid points with a score higher than 0.7 in the neighborhood is calculated as the density value. The local density maximum point is extracted as the density peak point, and when the density value of a certain grid point is greater than all its adjacent points and higher than the set threshold value 0.6, it is marked as a density peak point. Based on the area between adjacent density peak points, the region growing method is used to determine the density clustering area. Starting from the density peak point, when the density value of the neighborhood point is higher than the threshold value 0.4, it is included in the current clustering area until it cannot continue to expand.

[0103] The shape features of each determined density clustering area are calculated. The area of the region is determined by counting the number of grid points in the clustering area; the eccentricity is obtained by calculating the ratio of the long axis to the short axis of the minimum circumscribed ellipse of the region, for example, an eccentricity of 2.5 indicates that the region is in a clear strip-shaped distribution; the directionality is determined by principal component analysis to determine the principal axis direction of the region, expressed in degrees, such as a direction angle of 75 degrees indicating that the region mainly extends in that direction. According to these shape features, the corresponding dominant expansion direction is determined, and when the eccentricity is greater than 1.8 and the area of the region exceeds the preset threshold value (such as 100 grid units), the maximum principal component direction is determined as the dominant expansion direction.

[0104] The boundary fitting of the density clustering area of each sector unit is performed to generate the density contour line. The density threshold value is selected as 0.5, and the Marching Cubes algorithm is used to generate the contour line. Based on the curvature change of the density contour line, the density jump point is determined, and when the curvature change rate exceeds 0.3 and the local density gradient is greater than 0.2, it is marked as a density jump point. Five density jump points are detected in a certain sector unit, and these points are located in the region boundary where the density drops sharply from 0.6 to 0.3.

[0105] Transition grids are inserted at the density jump points, and the grid side length is inversely proportional to the density gradient. When the density gradient at a certain jump point is 0.25, the grid side length at that point is set to 0.8 times the size of the basic grid, i.e. 0.4 meters; when the density gradient is 0.5, the grid side length is set to 0.5 times the size of the basic grid, i.e. 0.25 meters. In this way, more detailed grid division is provided in the area with sharp density changes, improving the expression accuracy of the boundary area.

[0106] Real-time monitoring of the change of density distribution in the sector unit triggers the grid dynamic adjustment mechanism. When the minimum distance between adjacent density clustering regions is less than the preset distance threshold (such as 3 grid units), the merging operation is triggered, and the two clustering regions are merged into a larger region, and the shape feature is recalculated. When the maximum density difference in the density clustering region exceeds the preset density threshold 0.3, the splitting operation is triggered, and the region is split into two sub-regions along the direction of the maximum density gradient. For example, when the distance between two density regions is reduced to 2 grid units, the two regions are automatically merged and the grid parameters are recalculated, and the two grid regions of 8x10 and 9x7 are merged and adjusted to a single grid region of 16x12.

[0107] The grid division parameters of the corresponding sector unit are recalculated based on the results of merging or splitting, including grid density, direction and size. For the merged region, the grid density is the weighted average of the two original regions; for the split region, the grid density is reset according to the density peak value of each sub-region. Finally, the overall grid structure is dynamically updated to ensure that the grid division adapts to the current density distribution characteristics, improving resource allocation efficiency and data processing accuracy.

[0108] In an optional implementation, for the candidate parking space set, a road network connection state matrix is constructed, a candidate path set is generated, a congestion propagation prediction tree is constructed based on real-time traffic flow, and the optimal navigation path is determined by calculating path scores, which includes:

[0109] According to the pre-acquired road level, a road network topology structure is established, a target road section traffic capacity benchmark value is extracted, a turning constraint condition and a connection relationship between target road sections are calculated, and a road network connection state matrix is generated;

[0110] In the road network connection state matrix, the candidate parking space set is taken as the terminal point, and reverse path search is performed according to the turning constraint condition to generate a candidate path set;

[0111] A congestion index is determined based on the ratio of the real-time traffic flow of the target road section to the traffic capacity benchmark value, and the road sections with a congestion index greater than a preset congestion threshold are marked, the traffic flow change rate between adjacent marked road sections is calculated, and a congestion spread vector is generated;

[0112] Based on the congestion spread vector, node expansion is performed in the road network connection state matrix to establish a congestion propagation prediction tree;

[0113] Based on the congestion propagation prediction tree, the traffic probability and time cost of each path in the candidate path set are calculated, and the path scores are determined by weighted summation, and the optimal navigation path is determined by selecting the path with the maximum path score.

[0114] In one specific implementation, road class information is obtained from a geographic information database in advance, including categories such as trunk road, secondary trunk road, branch road, etc., and a road network topology is established based on this information. The road network topology is represented by a directed graph, in which nodes represent intersections or parking space locations, and edges represent road segments. For each road segment, its capacity reference value, i.e. the number of vehicles that can pass through per unit time under ideal traffic conditions, is extracted. For example, the capacity reference value of a two-way four-lane trunk road can reach 3600 vehicles per hour, while that of a two-way two-lane branch road is about 1200 vehicles per hour.

[0115] Turning constraints between road segments include traffic rule restrictions such as allowing left turns, allowing right turns, allowing straight-through, prohibiting left turns, etc. These restrictions can be identified by traffic signs or obtained from map data. Based on the road topology and turning constraints, the connectivity relationship between road segments is calculated, and a road network connection state matrix is generated. This matrix is a two-dimensional structure, with rows and columns representing nodes in the road network, and matrix element values representing the connectivity state and turning relationship between two nodes. For example, a matrix element value of 1 indicates that straight-through connectivity is allowed, a value of 2 indicates that left-turn connectivity is allowed, a value of 3 indicates that right-turn connectivity is allowed, and a value of 0 indicates that it is not connected. For complex intersections, the turning delay time also needs to be recorded, such as an average left-turn delay of 25 seconds and an average straight-through delay of 15 seconds.

[0116] After the road network connection state matrix is constructed, the parking space locations in the candidate parking space set are taken as the end nodes, and a reverse path search is performed in the road network connection state matrix. The reverse path search uses a breadth-first search algorithm, starting from the end node and expanding outward, following the turning constraints to build paths. During the search process, when the turning constraints are not met, the path branch will be pruned. For example, when the road prohibits left turns, the search will not continue to expand in that direction. Through the reverse path search, multiple feasible paths from the user's current location to each candidate parking space are obtained, forming a candidate path set. For each candidate parking space, usually 3-5 alternative paths are generated to cope with different traffic conditions.

[0117] To evaluate the path passing efficiency, the congestion index of each road section is calculated based on the real-time traffic volume data collected by traffic monitoring equipment. The congestion index is equal to the ratio of real-time traffic volume to the benchmark value of passing capacity. For example, the benchmark value of passing capacity of a certain two-way four-lane arterial road is 3600 vehicles per hour, and the current real-time traffic volume is 2880 vehicles per hour, so the congestion index is 0.8. When the congestion index is greater than the preset congestion threshold (usually set to 0.75), the road section is marked as a congestion road section. For the marked congestion road section, the traffic volume change rate between adjacent marked road sections is calculated, that is, the ratio of the change value of traffic volume per unit time to the benchmark value of passing capacity. For example, the traffic volume of a certain road section increases by 300 vehicles in 5 minutes, and the benchmark value of passing capacity is 3600 vehicles per hour, so the traffic volume change rate is 0.0833. Based on the traffic volume change rate and the spatial distribution of the congestion road section, a congestion spread vector is generated, which contains the direction and speed information of the congestion spread.

[0118] Based on the congestion spread vector, node expansion is performed in the road network connection state matrix to establish a congestion propagation prediction tree. The congestion propagation prediction tree takes the current congestion road section as the root node and expands the child nodes according to the direction of the congestion spread vector. During the expansion process, the probability of congestion spreading from one node to adjacent nodes is calculated considering road connectivity and traffic flow direction. For example, the probability of congestion spreading from the arterial road to the connected secondary arterial road is 0.65, while the probability of spreading to the branch road is 0.35. At the same time, according to the speed component of the congestion spread vector, the predicted time of congestion propagation to each child node is calculated. For important intersections, the influence of signal timing cycle on congestion propagation is considered, such as the signal cycle of 120 seconds and the red light time of 60 seconds, so the congestion propagation delay at the intersection is 60 seconds. Through iterative expansion, a complete congestion propagation prediction tree is constructed, and the structure of the congestion propagation prediction tree reflects the evolution trend of congestion conditions in a certain period of time (usually 30 minutes).

[0119] The passing efficiency of each path in the candidate path set is evaluated using the congestion propagation prediction tree. For each candidate path, its passing probability and time cost are calculated. The passing probability represents the probability that each road segment on the path is in a free state within the predicted arrival time, and is calculated according to the congestion probabilities of the nodes passed by the path in the congestion propagation prediction tree. For example, a path passes through three key nodes, and the congestion probabilities of these nodes at the predicted arrival time point are 0.2, 0.3 and 0.1 respectively. Therefore, the passing probability of the path is 0.8*0.7*0.9=0.504. The time cost includes normal driving time and congestion delay time. The normal driving time is calculated according to the path length and average driving speed. For example, if the path length is 5 kilometers and the average driving speed is 30 kilometers per hour, the normal driving time is 10 minutes. The congestion delay time is calculated according to the congestion state in the congestion propagation prediction tree. For example, if a 1-kilometer road segment on the path will be congested, causing the average speed to drop to 10 kilometers per hour, the congestion delay time of this road segment is 6 minutes.

[0120] The path score is calculated by weighted combination of the passing probability and the time cost. The weight configuration can be set according to user preferences. For example, time-sensitive users can set the time cost weight to 0.7 and the passing probability weight to 0.3; while stability preference users can set the passing probability weight to 0.6 and the time cost weight to 0.4. The calculated path score is used for sorting, and the path with the highest score is selected as the optimal navigation path. For example, in a navigation, there are three candidate paths, and their passing probabilities are 0.75, 0.6 and 0.8 respectively, and the predicted total travel times are 15 minutes, 12 minutes and 18 minutes respectively. After calculation according to the weight configuration, the path scores are 0.645, 0.612 and 0.636 respectively, so the first path is selected as the optimal navigation path.

[0121] After the optimal navigation path is determined, the path information is associated with the corresponding target parking space information to generate a complete navigation guide, which is displayed to the user through the user terminal. The navigation guide includes path direction indication, distance information, predicted arrival time, real-time traffic condition prompt and other contents, guiding the user to efficiently arrive at the target parking space position. At the same time, according to the changes of real-time traffic conditions, the congestion propagation prediction tree is dynamically updated, and the path score is recalculated and the navigation route is updated if necessary, to ensure the real-time optimality of the navigation path.

[0122] In an optional embodiment, based on the congestion propagation vector, node expansion is performed in the road network connection state matrix to establish the congestion propagation prediction tree, which includes:

[0123] The congestion propagation vector is decomposed into a direction component and a speed component, and the primary propagation direction and the secondary propagation direction of the congestion propagation are determined according to the direction component and the speed component;

[0124] The target link is set as a root node, and node expansion is performed in a link network connection state matrix based on a primary propagation direction and a secondary propagation direction, wherein each node contains a link identification, a current traffic flow, an average vehicle speed, a number of lanes, and a turning probability matrix;

[0125] A propagation attenuation coefficient is calculated according to a distance attenuation function and a time attenuation function between adjacent nodes, and a congestion propagation vector is corrected based on the propagation attenuation coefficient to obtain a propagation weight between nodes;

[0126] Real-time traffic flow data of the nodes is collected, and when the real-time traffic flow data exceeds a preset traffic flow threshold, a real-time traffic flow change rate is calculated to determine whether local updating is triggered;

[0127] When local updating is triggered, an affected node set is determined, the affected node set is recalculated, and a local structure of the propagation prediction tree is updated.

[0128] According to real-time traffic flow data and a propagation weight of each node, a turning delay between nodes is calculated in combination with a signal timing cycle, a weighted sum of the turning delay and a traffic flow propagation time is taken as an estimated congestion arrival time, and a congestion propagation prediction tree is generated.

[0129] In a specific embodiment, the congestion propagation vector contains direction and speed information of congestion diffusion, which is decomposed into a direction component and a speed component. The direction component is expressed in degrees, for example, the northeast direction can be expressed as 45 degrees, and the southwest direction can be expressed as 225 degrees. The speed component is expressed as the distance of congestion diffusion per unit time, for example, 100 meters per minute. Based on the decomposed direction component, a primary propagation direction and a secondary propagation direction of congestion propagation are determined. The primary propagation direction is the main direction of the direction component, and the secondary propagation direction is the direction adjacent to the primary propagation direction. For example, when the direction component is the northeast direction (45 degrees), the primary propagation direction is northeast, and the secondary propagation direction is east and north. Exemplarily, the weight of the primary propagation direction is usually set to 0.6, and the weights of the secondary propagation directions are set to 0.2 respectively, to ensure the direction accuracy of congestion prediction.

[0130] The target road section is set as the root node of the congestion propagation prediction tree, and the tree structure is started to be built. In the road network connection state matrix, the node is expanded according to the primary propagation direction and the secondary propagation direction. Each node contains rich attribute information: the road section identifier such as "RS-10086" is used to uniquely identify a road section; the current traffic flow is, for example, 2500 vehicles per hour; the average speed is, for example, 35 kilometers per hour; the number of lanes is, for example, 4 lanes in both directions; and the turning probability matrix records the probability of vehicles selecting each direction at the node, such as straight 0.7, left turn 0.2, and right turn 0.1. During the node expansion process, the primary propagation direction is preferred, followed by the secondary propagation direction, and then the other directions. For example, for a certain intersection node, if the primary propagation direction is east and the secondary propagation direction is northeast and southeast, then the expansion order is the eastbound road section, the northeastbound road section, the southeastbound road section, and finally the northbound, southbound, and westbound road sections.

[0131] During the node expansion process, the attenuation effect of congestion propagation needs to be considered. The distance attenuation function and the time attenuation function between adjacent nodes are used to calculate the propagation attenuation coefficient. The distance attenuation function represents the law that the congestion influence decreases with the increase of distance, such as two nodes 500 meters apart, the distance attenuation value is 0.85; when the distance is 1000 meters, the distance attenuation value decreases to 0.6. The time attenuation function represents the law that the congestion influence decreases with the passage of time, such as a time interval of 5 minutes, the time attenuation value is 0.8; a time interval of 15 minutes, the time attenuation value decreases to 0.4. The propagation attenuation coefficient is calculated by combining the distance attenuation value and the time attenuation value, for example, the weight of each is 0.5, then the propagation attenuation coefficient is the weighted average of the two. Based on the propagation attenuation coefficient, the congestion spread vector is modified to obtain the propagation weight between nodes. For example, the original congestion spread vector speed component is 100 meters per minute, after being modified by the propagation attenuation coefficient 0.7, the speed component in the node-to-node propagation weight is reduced to 70 meters per minute.

[0132] The real-time traffic data of the node is continuously collected by the traffic monitoring device. When the real-time traffic data exceeds the preset traffic threshold, the real-time traffic change rate is calculated. For example, the traffic threshold of a node is set to 2000 vehicles per hour. When the real-time traffic reaches 2100 vehicles, the calculation is triggered. If the traffic of the node increases by 300 vehicles in the past 10 minutes, the real-time traffic change rate is 1800 vehicles per hour. When the traffic change rate exceeds the preset change rate threshold (such as 1500 vehicles per hour), the local update mechanism is triggered. After triggering the local update, the affected node set needs to be determined. The affected node set includes the node that triggers the update and its downstream nodes, usually extending 3 to 5 nodes along the main propagation direction and the secondary propagation direction. For example, a node on a main road triggers an update, and its affected node set includes the node and its 12 related nodes in the east, northeast, and southeast directions. For each node in the affected node set, the congestion state and propagation parameters are recalculated, and the local structure of the propagation prediction tree is updated.

[0133] To accurately predict the congestion propagation time, multiple factors need to be considered. According to the real-time traffic data of each node and the propagation weight, the traffic propagation time is calculated. For example, two nodes 800 meters apart, if the speed component in the propagation weight is 60 meters per minute, the traffic propagation time is about 13.3 minutes. At the same time, the signal timing cycle has a significant impact on congestion propagation. For example, a crossroads has a signal cycle of 120 seconds, with a green light time of 70 seconds and a red light time of 50 seconds in the east-west direction, so the average turning delay in this direction is 25 seconds. The turning delay and traffic propagation time are combined by weighting to obtain the predicted congestion arrival time. In practical applications, the turning delay weight can be set to 0.3 and the traffic propagation time weight to 0.7. For example, if the turning delay of a path is 25 seconds and the traffic propagation time is 13.3 minutes, the predicted congestion arrival time is 9.55 minutes. Based on the predicted congestion arrival time, the congestion propagation prediction tree is constructed. The congestion propagation prediction tree takes time as the vertical axis and space as the horizontal axis, directly showing the spatio-temporal evolution process of congestion state in the road network.

[0134] After the prediction tree is constructed, it can be used to evaluate the traffic efficiency of each navigation path. For each path in the candidate path set, the congestion condition when the user arrives at each key point on the path is determined based on the prediction tree, and the possible delay time is calculated. If a user plans to pass through a road segment in 10 minutes, and the prediction tree shows that congestion will occur on the road segment in 8 minutes and last for 20 minutes, the user is likely to encounter congestion and needs to consider alternative routes. Through comprehensive evaluation of each path, the optimal navigation path is selected and provided to the user.

[0135] In actual operation, the prediction tree needs to be continuously updated. Real-time traffic data is re-collected every 2 minutes to check whether local update needs to be triggered. For important road sections or traffic bottleneck locations, a higher sampling frequency is set, such as every 30 seconds, to ensure the timeliness of the prediction. When the local structure of the prediction tree is updated, the traffic efficiency of the affected paths is re-evaluated, and the navigation recommendations are dynamically adjusted if necessary to ensure that users obtain the optimal route.

[0136] Traditional congestion prediction techniques are mainly based on historical data statistics and simple linear propagation models, which cannot accurately capture the nonlinear propagation characteristics of congestion in complex road networks. In the prior art, congestion prediction often treats road sections as independent units, ignoring the influence of road network connection on congestion propagation, resulting in insufficient prediction accuracy. At the same time, traditional methods usually use a global update strategy, which consumes a lot of computing resources and has slow response speed, making it difficult to meet real-time navigation needs. The method of the present embodiment introduces direction and speed components, considers primary and secondary propagation directions, and accurately describes the spatial diffusion characteristics of congestion. By introducing distance and time decay functions, the decay law of congestion influence with space and time is quantified. A local update mechanism is used to recompute only the affected area, reducing computational complexity and improving response speed. The signal timing cycle and turning delay are considered to improve prediction accuracy.

[0137] In an optional implementation, the target parking space is determined according to the optimal navigation path, the parking space groups are divided based on the historical occupation state sequence of the adjacent parking spaces, the state propagation path and the state propagation rate are calculated, and the maximum reserved waiting time is determined, including:

[0138] The position information of the target parking space and the spatial distribution information of the adjacent parking spaces are obtained, a parking space adjacency relationship matrix is established, a historical occupation state sequence is collected, and the state transition probability between adjacent parking spaces is calculated;

[0139] Based on the state transition probability and the parking space adjacency relationship matrix, a spatial continuity threshold and a usage mode similarity threshold are set, the parking spaces are grouped, the occupancy rate and the turnover frequency of the parking space groups are calculated;

[0140] According to the historical occupation state sequence of the parking space groups, the state propagation path and the state propagation rate are calculated, and the key node parking spaces are determined; according to the turnover frequency and the state propagation rate of the parking space groups, the reference waiting time is calculated;

[0141] Real-time states of the key node parking spaces and real-time turnover data of the parking space groups are collected, the reference waiting time is dynamically corrected according to the state propagation path and the state propagation rate, and the maximum reserved waiting time is obtained.

[0142] In one specific embodiment, after determining the target parking space according to the optimal navigation path, the parking space usage characteristics are further analyzed to determine a reasonable reservation waiting time. The precise location information of the target parking space is read through the passive Internet of Things tag, including latitude and longitude coordinates, parking area identification, etc. At the same time, the spatial distribution information of adjacent parking spaces is obtained, and the relative position relationship and distance between each parking space are recorded. Based on this information, a parking space adjacency relationship matrix is established, and the elements in the parking space adjacency relationship matrix represent the adjacency degree between parking spaces. For example, for parking spaces within 5 meters apart, the adjacency degree is set to 1; for parking spaces 5 to 10 meters apart, the adjacency degree is set to 0.8; for parking spaces 10 to 20 meters apart, the adjacency degree is set to 0.5; for parking spaces more than 20 meters apart, the adjacency degree is set to 0. In a certain parking area, assuming that there are 10 parking spaces arranged in two rows, 5 in each row, and the distance between adjacent parking spaces in the same row is 2.5 meters, and the distance between the two rows is 6 meters, the adjacency degree of adjacent parking spaces in the same row is 1, the adjacency degree of corresponding parking spaces in different rows is 0.8, and the adjacency degree of non-corresponding parking spaces in different rows is 0.5 or 0.

[0143] Through the historical data recorded by the passive Internet of Things tag, the occupancy state sequence of each parking space in the past period of time (such as the past 30 days) is collected. The state sequence is recorded in time windows, and the time window is usually set to 15 minutes or 30 minutes. In each time window, the parking space state is marked as occupied or idle. For example, a parking space is in the occupancy state for 25 days and in the idle state for 5 days in the 9:00-9:15 time window on weekdays in the past 30 days, and the occupancy probability of this time window is 0.83. Based on the historical occupancy state sequence, the state transition probability between adjacent parking spaces is calculated, i.e. the probability that the state of an adjacent parking space changes after the state of a parking space changes.

[0144] Based on the state transition probability and the parking space adjacency relationship matrix, the parking spaces are divided into groups. The spatial continuity threshold and the usage mode similarity threshold are set to determine whether the parking spaces belong to the same group. The spatial continuity threshold represents the minimum adjacency degree requirement between parking spaces, and is usually set to 0.5; the usage mode similarity threshold represents the minimum similarity requirement of the usage mode of the parking spaces, and is usually set to 0.7. The usage mode similarity is calculated by comparing the occupancy probability difference of two parking spaces in the corresponding time window. For example, in the 9:00-9:15 time window on weekdays, the occupancy probability of parking space 1 is 0.83, and the occupancy probability of parking space 2 is 0.78, so the similarity of this time window is 0.95; the overall usage mode similarity is calculated by calculating the average similarity of all time windows. When the adjacency degree of two parking spaces is greater than the spatial continuity threshold and the usage mode similarity is greater than the usage mode similarity threshold, they are divided into the same group. For example, in a parking lot in an office area, "morning to night shift worker parking space group", "afternoon to night shift worker parking space group", "temporary visitor parking space group" and other parking space groups with different characteristics are formed.

[0145] For each divided parking group, its occupancy rate and turnover frequency are calculated. The occupancy rate represents the average proportion of occupied parking spaces in the group, for example, the average occupancy rate of a group is 0.75 on weekdays. The turnover frequency represents the average number of times the state of a parking space changes from occupied to idle per unit time, for example, the turnover frequency of a group is 0.5 times per hour on weekdays, that is, on average, a parking space is released every 2 hours. Based on the historical occupancy state sequence, the propagation characteristics of state changes in the parking group are analyzed, and the state propagation path and state propagation rate are calculated. The state propagation path represents the propagation order of state changes in the group, for example, in a straight parking group, state changes usually propagate from one end to the other end. The state propagation rate represents the speed of state changes propagating between parking spaces, for example, in a group, state changes propagate to the next adjacent parking space on average every 15 minutes. By analyzing the state propagation path and key positions, key node parking spaces are determined. Key node parking spaces are parking spaces that play a key role in the propagation process of state changes, and their state changes have a strong indicative effect on other parking spaces in the group. In a circularly arranged parking group, the parking space at the entrance is usually a key node parking space, and its state change leads other parking spaces.

[0146] Based on the turnover frequency and state propagation rate of the parking group, the benchmark waiting time is calculated. The benchmark waiting time is the longest time a user needs to wait in a typical situation, and the calculation method is to multiply the reciprocal of the turnover frequency by the maximum number of steps on the state propagation path, and then multiply by an adjustment coefficient. For example, the turnover frequency of a group is 0.5 times per hour, the maximum number of steps on the state propagation path is 3, and the adjustment coefficient is 0.8, then the benchmark waiting time is 0.8×3×(1 / 0.5)=4.8 hours. The adjustment coefficient is set according to different time periods and scenarios, such as 0.6 for peak hours, 0.8 for flat peak hours, and 1.0 for low peak hours.

[0147] The state of the key node parking space and the turnover data of the parking space group are collected in real time, which is used to dynamically correct the benchmark waiting time. The real-time state of the key node parking space is collected by passive Internet of Things tags, including whether it is currently occupied, the duration of occupation, etc. The real-time turnover data of the parking space group includes the turnover frequency in the last hour, the average occupation duration, etc. According to the real-time state of the key node parking space, the possibility of parking space release on the state propagation path is predicted. For example, if the key node parking space is just occupied and the average occupation duration is 2 hours, it is predicted that the parking space will be released after 2 hours, and the subsequent parking spaces on the state propagation path will be released in succession. According to the ratio of real-time turnover frequency to historical average turnover frequency, the benchmark waiting time is corrected. For example, if the real-time turnover frequency is 1.2 times the historical average, the benchmark waiting time should be multiplied by 0.83 (i.e. 1 / 1.2) for correction. By comprehensively considering the state propagation path, the state propagation rate and the real-time adjustment factor, the maximum reserved waiting time is finally obtained. For example, the original benchmark waiting time is 4.8 hours, which is corrected to 4 hours after the real-time turnover frequency correction, and is further corrected to 3.5 hours according to the state prediction of the key node parking space, and the maximum reserved waiting time is finally determined as 3.5 hours.

[0148] The maximum reserved waiting time is used for generating parking reservation confirmation information. The reservation confirmation information contains target parking space location, navigation route, estimated arrival time, maximum reserved waiting time, etc. After being sent to the user terminal, the user can know that if he does not arrive within 3.5 hours, the reservation will be automatically cancelled. In actual application, different maximum reserved waiting time adjustment coefficients can be set according to the user credit level, such as an additional 20% reserved time for users with high credit level. After the user arrives at the target parking space location, the passive Internet of Things tag information is read by the terminal device for identity verification, and the parking space usage state is updated after verification. If the user does not arrive within the maximum reserved waiting time, the reservation will be automatically cancelled, and the parking space state will be updated to available, so that it can be reserved by other users. In addition, the system also updates the historical occupation state sequence according to the latest usage record, continuously optimizes the parking space group division and waiting time calculation model, and improves the service accuracy.

[0149] As shown in Figure 2 , a congestion propagation prediction flowchart is shown.

[0150] In an optional implementation, according to the historical occupation state sequence of the parking space group, the state propagation path and the state propagation rate are calculated, and the key node parking space includes:

[0151] Based on the historical occupation state sequence, the state change sequence of adjacent time windows in the parking space group is extracted, and the trigger relationship of each parking space state transition is obtained;

[0152] According to the trigger relationship, the state transition weight between the parking spaces is calculated, and the parking space pair with a state transition weight greater than a preset transition threshold is determined as a chain occupation node pair;

[0153] Based on the chain occupation node pair, a state propagation directed graph is constructed, wherein the nodes correspond to the parking spaces, the edges correspond to the state propagation directions, and the edge weights correspond to the state propagation probabilities;

[0154] In the state propagation directed graph, edges with an edge weight greater than a preset probability threshold are extracted to determine a state propagation path;

[0155] The average state transition time interval between adjacent node pairs on the state propagation path is calculated to obtain a state propagation rate;

[0156] According to the sum of the in-degree and out-degree of the parking space on the state propagation path, the parking space with the largest degree is selected as a key node parking space.

[0157] In a specific embodiment, after determining the parking space group, the propagation rule of the state change of the parking spaces in the group is analyzed in depth to accurately calculate the state propagation path and rate. Based on the historical occupation state sequence collected by the passive Internet of Things label, the state change sequence of adjacent time windows in the parking space group is extracted. The time window is usually set to 15 minutes, for example, each day is divided into 96 time windows. For each parking space, the state (occupied or idle) and the state change (from occupied to idle or from idle to occupied) of the parking space in each time window are recorded. By comparing the state changes of different parking spaces in adjacent time windows, the trigger relationship of the state transition of each parking space is obtained. The trigger relationship describes the influence of the state change of one parking space on the state change of another parking space. For example, in a parking space group in an office parking lot, it is observed that after parking space P1 changes from occupied to idle, parking space P2 also changes from occupied to idle in the subsequent 1-3 time windows for 24 times within 30 days, and it is considered that the state change of parking space P1 may trigger the state change of parking space P2.

[0158] According to the above trigger relationship, the state transition weight between the parking spaces is calculated. The state transition weight represents the degree of influence of the state change of one parking space on the state change of another parking space, which can be calculated by counting the frequency of the trigger event. For example, after parking space P1 changes from occupied to idle, the probability that parking space P2 also changes from occupied to idle within a certain time range is 0.8, and the state transition weight from P1 to P2 is 0.8. In actual application, a preset transition threshold is set, which is usually between 0.6 and 0.7, and the parking space pair with a state transition weight greater than the threshold is determined as a chain occupation node pair. For example, in a group containing 20 parking spaces, 15 chain occupation node pairs such as (P1, P2), (P2, P4), (P4, P7), etc. may be identified, which represent parking spaces with obvious state propagation relationship.

[0159] Based on the identified pairs of chain occupation nodes, a state propagation directed graph is constructed. In the directed graph, nodes correspond to parking spaces, edges represent state propagation directions, and edge weights correspond to state propagation probabilities. For example, in a parking space group in a certain commercial district, the constructed state propagation directed graph contains 30 nodes (parking spaces) and 25 edges, with edge weights ranging from 0.65 to 0.9. The directed graph visually demonstrates the propagation network of state changes within the parking space group. In the state propagation directed graph, edges with weights greater than a preset probability threshold are extracted to determine state propagation paths. The preset probability threshold is usually set to 0.75 to ensure that the selected propagation paths have high reliability. For example, from the original 25 edges, 15 edges with weights greater than 0.75 are selected to form the main state propagation paths. In a certain linearly arranged parking lot, a state propagation path similar to "P1→P3→P5→P8→P12" may be formed; while in a circularly arranged parking lot, a circular or radial propagation path may be formed.

[0160] The average state transition time interval between adjacent node pairs on the state propagation path is calculated to obtain the state propagation rate. The state transition time interval refers to the time required for a parking space to change its state after the adjacent parking space changes its state. For example, in a certain office district parking lot, it is observed that after parking space P1 changes from occupied to idle, parking space P3 also changes from occupied to idle on average 25 minutes later; 20 minutes later, parking space P5 changes from occupied to idle, and so on. The average state transition time interval of all adjacent node pairs on the propagation path is calculated to obtain the state propagation rate, such as 25 minutes to propagate one node. The state propagation rate reflects the dynamic characteristics of state changes within the parking space group and is an important parameter for calculating the reservation waiting time.

[0161] According to the importance of the parking space on the state propagation path, the key node parking space is determined. The importance of a parking space can be measured by the sum of its in-degree and out-degree in the state propagation directed graph. The in-degree represents the number of edges pointing to the parking space, reflecting its degree of influence from other parking spaces; the out-degree represents the number of edges starting from the parking space, reflecting its influence on other parking spaces. The sum of the in-degree and the out-degree is called the degree of the parking space. The greater the degree, the more important the role of the parking space in the state propagation network. The parking space with the largest degree is selected as the key node parking space. For example, in the state propagation directed graph of a certain circular parking lot, parking space P5 has an in-degree of 3, an out-degree of 4, and a degree of 7, which is the largest among all parking spaces, so it is determined as the key node parking space. The state change of the key node parking space has an important indicative effect on the state evolution of the entire parking space group, and its intensive monitoring helps to improve the prediction accuracy.

[0162] In practical applications, key node parking spaces are usually set with more frequent state monitoring frequency. For example, ordinary parking spaces collect state data every 10 minutes, while key node parking spaces collect state data every 3 minutes. Based on the real-time state of the key node parking space P5, the state change trend of the entire parking space group is predicted. If the key node parking space P5 changes from occupied to idle, according to the state propagation path and propagation rate, it is predicted that the parking space P8 may be released after 25 minutes, and the parking space P12 may be released after another 25 minutes. This prediction information is used to optimize the parking space recommendation and navigation strategy, and to improve the probability of users finding idle parking spaces.

[0163] As the parking space usage pattern changes, the state propagation directed graph needs to be updated regularly. In daily use, the state propagation directed graph is updated once a week to adapt to the seasonal changes in parking space usage patterns; in special periods (such as before and after holidays), the update cycle is shortened to once a day to adapt to sudden changes in usage patterns. During the update process, the weight of historical data is 0.7 and the weight of new data is 0.3, ensuring the stability and adaptability of the state propagation model. Through the continuously optimized state propagation directed graph, more accurate decision-making basis is provided for parking reservation and navigation.

[0164] Traditional parking space occupancy prediction techniques mainly perform time series analysis based on historical data of individual parking spaces, ignoring the spatial correlation and state propagation characteristics between parking spaces. Existing technologies usually treat parking spaces as independent individuals, making it difficult to capture group behavior patterns, resulting in insufficient prediction accuracy, especially during peak periods and special time periods. At the same time, traditional methods lack a mechanism for identifying key nodes, making it impossible to prioritize monitoring and predicting parking spaces that have the greatest impact on the overall state, resulting in low resource utilization efficiency. The method of the present embodiment introduces state transition trigger relationship analysis to quantify the influence degree of state changes between parking spaces; by constructing a state propagation directed graph, the propagation network and path of state changes are visually displayed; based on the degree centrality index, key node parking spaces are identified to achieve focused monitoring and prediction. By analyzing the propagation rules of state changes, the accuracy of the prediction and the resource utilization efficiency are improved.

[0165] The shared parking space automatic matching and guided navigation system based on passive Internet of Things tags according to the embodiment of the present application comprises:

[0166] The first unit is configured to receive destination location information and vehicle usage time period information sent by a user terminal, obtain parking space information within a preset range from a server based on the destination location information, and store the parking space information through passive Internet of Things tags installed in the parking spaces.

[0167] The second unit is configured to construct a multi-layer search grid based on the destination location information, dynamically adjust the grid structure by calculating parking space scores and based on density clustering, and expand the candidate parking space set layer by layer.

[0168] The third unit is configured to, for the candidate parking space set, construct a road network connection state matrix, generate a candidate path set, construct a congestion propagation prediction tree based on real-time traffic flow, calculate path scores to determine an optimal navigation path;

[0169] The fourth unit is configured to determine a target parking space according to the optimal navigation path, divide parking space groups based on a historical occupation state sequence of adjacent parking spaces, calculate a state propagation path and a state propagation rate, and determine a maximum reservation waiting time.

[0170] The fifth unit is configured to generate parking reservation confirmation information according to the optimal navigation path and the maximum reservation waiting time, and send the parking reservation confirmation information to a user terminal.

[0171] The sixth unit is configured to, when the target parking space is reached, read passive Internet of Things tag information and perform identity verification, and update a parking space use state after verification.

[0172] The third aspect of the embodiment of the present application provides an electronic device, comprising:

[0173] a processor;

[0174] a memory for storing processor-executable instructions;

[0175] The processor is configured to invoke the instructions stored in the memory to execute the method described above.

[0176] The fourth aspect of the embodiment of the present application provides a computer-readable storage medium having computer program instructions stored thereon, and the computer program instructions are executed by a processor to implement the method described above.

[0177] The present application can be a method, device, system and / or computer program product. The computer program product can include a computer readable storage medium having computer readable program instructions loaded thereon for executing various aspects of the present application.

[0178] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for automatic matching and navigation of shared parking spaces based on passive IoT tags, characterized in that, include: Receive destination location information and vehicle usage time information sent by the user terminal; Based on the destination location information, parking space information within a preset range is obtained from the server, and the parking space information is stored through passive IoT tags installed in the parking spaces; A multi-layered search grid is constructed based on destination location information. Parking space scores are calculated, and the grid structure is dynamically adjusted based on density clustering. The grid expands layer by layer to generate a candidate parking space set, including: Each layer of the grid uses a regular hexagonal structure with the side length increasing by a preset multiple; Real-time assessment of parking space status within the grid, and calculation of parking space scores based on parking space vacancy status, historical reservation data, and turnover rate; A sector-based density clustering method is used to divide the grid. The density clustering region is determined by calculating the continuous distribution of parking space scores. The grid structure is dynamically adjusted according to the shape characteristics of the density clustering region. The method of expanding layer by layer is adopted. When the number of parking spaces with a score greater than the preset score threshold in a layer of grid is less than the preset number of parking spaces, the method expands to the next layer of grid until the preset number of parking spaces is met or the preset upper limit of the number of grid layers is reached, and a candidate parking space set is generated. For the candidate parking space set, a road network connectivity state matrix is ​​constructed, a candidate path set is generated, a congestion propagation prediction tree is constructed based on real-time traffic flow, and a path score is calculated to determine the optimal navigation path, including: Establish a road network topology based on the pre-acquired road grades, extract the benchmark values ​​of the traffic capacity of the target road segments, calculate the turning constraints and connectivity relationships between the target road segments, and generate a road network connection state matrix. In the road network connection state matrix, with the candidate parking space set as the endpoint, a reverse path search is performed according to the turning constraint to generate a candidate path set. The congestion index is determined based on the ratio of real-time traffic flow to the baseline value of traffic capacity of the target road segment. Road segments with congestion indices greater than the preset congestion threshold are marked, and the rate of change of traffic flow between adjacent marked road segments is calculated to generate a congestion spread vector. Based on the congestion propagation vector, node expansion is performed in the road network connection state matrix to establish a congestion propagation prediction tree; Based on the congestion propagation prediction tree, the probability of passage and time cost of each path in the candidate path set are calculated, and the results are weighted and summed to determine the path score. The path with the highest path score is selected as the optimal navigation path. The target parking space is determined based on the optimal navigation path. Parking space groups are divided based on the historical occupancy status sequence of adjacent parking spaces. The state propagation path and state propagation rate are calculated, and the maximum reserved waiting time is determined. Based on the optimal navigation route and the maximum reserved waiting time, a parking space reservation confirmation message is generated and sent to the user terminal; Upon arrival at the target parking space, the passive IoT tag information is read and identity verification is performed. Once verification is successful, the parking space usage status is updated.

2. The method according to claim 1, characterized in that, A sector-based density clustering method is used to divide the grid. Density clustering regions are determined by calculating the continuous distribution of parking space scores. The grid structure is dynamically adjusted based on the shape characteristics of these density clustering regions. The grid is divided into multiple sector units, and a density calculation grid is constructed within each sector unit. The size of the density calculation grid is proportional to the size of the parking space. Gaussian smoothing is applied to the parking space scores within the density calculation grid. The smoothing range is controlled by setting the standard deviation of the Gaussian kernel function, thereby generating a continuous distribution of parking space scores. The local density value of sector units is calculated based on the continuous distribution of parking space scores, density peak points are extracted, and density clustering regions are determined based on the regions between adjacent density peak points. Calculate the shape characteristics of density clustering regions, including region area, eccentricity, and orientation, and determine the corresponding dominant expansion direction based on the shape characteristics; Boundary fitting is performed on the density clustering regions of each sector unit to generate density contour lines, and density jump points are determined based on the curvature changes of the density contour lines. A transition grid is inserted at the density jump point, and the grid side length is inversely proportional to the density gradient. The system monitors changes in density distribution within sector units in real time. When the minimum distance between adjacent density clustering regions is less than a preset distance threshold, a merging operation is triggered. When the maximum density difference within a density clustering region exceeds a preset density threshold, a splitting operation is triggered. Based on the results of merging or splitting, the meshing parameters of the corresponding sector units are recalculated, and the mesh structure is dynamically updated.

3. The method according to claim 1, characterized in that, Based on the congestion propagation vector, node expansion is performed in the road network connection state matrix to establish a congestion propagation prediction tree, including: The congestion propagation vector is decomposed into directional and velocity components, and the main and secondary propagation directions of congestion propagation are determined based on the directional and velocity components. The target road segment is set as the root node, and the nodes are expanded in the road network connection state matrix based on the main propagation direction and the secondary propagation direction. Each node contains the road segment identifier, current traffic flow, average vehicle speed, number of lanes, and turning probability matrix. The propagation attenuation coefficient is calculated based on the distance attenuation function and the time attenuation function between adjacent nodes. The congestion spread vector is then corrected based on the propagation attenuation coefficient to obtain the propagation weight between nodes. Collect real-time traffic flow data from the nodes. When the real-time traffic flow data exceeds the preset traffic flow threshold, calculate the real-time traffic flow change rate and determine whether to trigger a local update. When a local update is triggered, the set of affected nodes is determined, the set of affected nodes is recalculated, and the local structure of the propagation prediction tree is updated. Based on the real-time traffic flow data and propagation weights of each node, the turning delay between nodes is calculated in conjunction with the signal timing cycle. The weighted sum of the turning delay and the traffic flow propagation time is used as the expected congestion arrival time to generate a congestion propagation prediction tree.

4. The method according to claim 1, characterized in that, The target parking space is determined based on the optimal navigation path. Parking spaces are grouped based on the historical occupancy status sequence of adjacent spaces. The state propagation path and state propagation rate are calculated. The maximum reserved waiting time is determined, including: Obtain the location information of the target parking space and the spatial distribution information of adjacent parking spaces, establish a parking space adjacency matrix, collect historical occupancy state sequences, and calculate the state transition probability between adjacent parking spaces; Based on the state transition probability and the parking space adjacency matrix, a spatial continuity threshold and a usage pattern similarity threshold are set to divide the parking spaces into groups and calculate the occupancy rate and turnover frequency of the parking space groups. Based on the historical occupancy status sequence of parking space groups, calculate the state propagation path and state propagation rate to determine key node parking spaces; and calculate the baseline waiting time based on the turnover frequency and state propagation rate of parking space groups. The real-time status of the key node parking spaces and the real-time turnover data of the parking space group are collected. The baseline waiting time is dynamically corrected according to the status propagation path and status propagation rate to obtain the maximum reserved waiting time.

5. The method according to claim 4, characterized in that, Based on the historical occupancy status sequence of parking space groups, the state propagation path and state propagation rate are calculated to determine the key node parking spaces, including: Based on the historical occupancy status sequence, extract the status change sequence of adjacent time windows within the parking space group to obtain the triggering relationship of each parking space status transition; Based on the triggering relationship, calculate the state transition weight between parking spaces, and determine the parking space pairs with state transition weights greater than the preset transition threshold as chain occupancy node pairs; Based on the chain of occupancy node pairs, a directed graph of state propagation is constructed, where nodes correspond to parking spaces, edges correspond to the direction of state propagation, and edge weights correspond to the probability of state propagation. In the directed graph of state propagation, edges with weights greater than a preset probability threshold are extracted to determine the state propagation path; The average state transition time interval between adjacent node pairs on the state propagation path is calculated to obtain the state propagation rate. Based on the sum of the in-degree and out-degree of the parking space on the state propagation path, the parking space with the largest degree is selected to determine the key node parking space.

6. A shared parking space automatic matching and guidance navigation system based on passive IoT tags, used to implement the method of any one of claims 1-5, characterized in that, include: The first unit is used to receive destination location information and vehicle usage time information sent by the user terminal; Based on the destination location information, parking space information within a preset range is obtained from the server, and the parking space information is stored through passive IoT tags installed in the parking spaces; The second unit is used to construct a multi-layer search grid based on destination location information. It calculates parking space scores and dynamically adjusts the grid structure based on density clustering to generate a candidate parking space set by expanding layer by layer. The third unit is used to construct a road network connection state matrix for the candidate parking space set, generate a candidate path set, construct a congestion propagation prediction tree based on real-time traffic flow, and calculate path scores to determine the optimal navigation path. The fourth unit is used to determine the target parking space based on the optimal navigation path, divide the parking space into groups based on the historical occupancy status sequence of adjacent parking spaces, calculate the state propagation path and state propagation rate, and determine the maximum reserved waiting time. The fifth unit is used to generate parking space reservation confirmation information based on the optimal navigation path and the maximum reserved waiting time, and send the parking space reservation confirmation information to the user terminal; The sixth unit is used to read the passive IoT tag information and perform identity verification when the target parking space is reached. After successful verification, the parking space usage status is updated.

7. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 5.

Images