A real-time parking space occupancy perception prediction system

Abstract

The application discloses a parking berth real-time occupation sensing and predicting system and belongs to the technical field of intelligent traffic. The system comprises the following modules: an event queue establishing module, a dynamic event influence sub-value obtaining module, a predictive increase coefficient generating module, a feedback correction module and a result generating module. The event queue establishing module is used for establishing an event queue; the dynamic event influence sub-value obtaining module is used for generating a dynamic event influence sub-value; the predictive increase coefficient generating module is used for generating a predictive increase coefficient coupled with real-time event information; the feedback correction module is used for feedback correction; and the result generating module is used for generating a real-time occupation predicting result. The application realizes single berth micro-prediction by establishing the association queue of events and city grids, quantifying the dynamic influence of events, generating the predictive coefficient coupled with real-time events, utilizing the residual feedback correction model and fusing data, achieves the precise sensing and predicting of parking berth occupation, improves the prediction adaptability and precision, and solves the problems of insufficient consideration of the influence of crowd gathering events, low prediction precision, lack of dynamic correction and difficulty in realizing single berth precise prediction in the prior art.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a real-time parking space occupancy perception and prediction system. Background Technology

[0002] With the acceleration of urbanization and the continuous growth of motor vehicle ownership, the contradiction between supply and demand of urban parking resources is becoming increasingly prominent. How to efficiently utilize limited parking resources and improve the utilization efficiency of parking spaces has become an important issue in alleviating urban traffic congestion and improving the travel experience. Real-time and accurate parking space occupancy perception and prediction can provide car owners with timely parking information guidance and optimize parking resource allocation, which is one of the core functions of intelligent parking management systems.

[0003] Existing parking space prediction systems mostly make predictions by analyzing historical parking data and combining time period characteristics, while some systems introduce external factors such as weather for auxiliary correction.

[0004] In existing technologies, most do not fully consider the sudden impact of various events with crowd gathering effects in cities (such as concerts, sports events, large-scale exhibitions, etc.) on parking demand, resulting in a large deviation between the prediction results and the actual situation when the event occurs. At the same time, existing predictions are mostly limited to the regional macro level, making it difficult to achieve accurate prediction of individual parking spaces. Furthermore, they lack dynamic feedback correction mechanisms and cannot adjust the prediction model in a timely manner according to the actual situation in real time, resulting in problems such as low prediction accuracy and weak adaptability. Summary of the Invention

[0005] This application provides a real-time parking space occupancy perception and prediction system, which solves the problems of insufficient consideration of the impact of crowd gathering events, prediction that remains at the macro level and lacks dynamic correction, resulting in low prediction accuracy and weak adaptability in the prior art. It realizes accurate prediction of parking spaces from macro grid to micro parking spaces, improving the accuracy and adaptability of prediction.

[0006] This application provides a real-time parking space occupancy perception and prediction system, including: an event queue establishment module: used to acquire historical event information and real-time event information with crowd gathering effect in the city, and to spatially map the event information with pre-divided urban grids to establish event queues associated with each grid; Dynamic Event Influence Sub-Value Acquisition Module: This module is used to obtain the baseline initial influence score of historical event information through multi-source heterogeneous data in historical event information, and generate dynamic event influence sub-values ​​for each city grid. Predictive amplification coefficient generation module: This module generates a predictive amplification coefficient coupled with real-time event information using a preset urban grid occupancy coefficient amplification model, and derives prediction results based on the predictive amplification coefficient. Feedback correction module: used to monitor the actual occupancy rate of each city grid in real time, and to perform feedback correction on the city grid occupancy coefficient increase model by combining the residual between the exported prediction results and the actual occupancy rate. The result generation module is used to fuse the corrected predictive amplification factor with the real-time status data of each parking space sensor in the parking lot, input it into the preset prediction model, and generate a real-time occupancy prediction result for a single parking space.

[0007] Furthermore, the steps of spatially mapping event information to pre-divided city grids and establishing event queues associated with each grid include: Obtain geographic coordinate data within the city limits, divide the city area into several regular grid units according to a preset grid size, and assign a unique grid identifier code to each grid unit; Based on the received historical and real-time event information, the geographical coordinates of the events are extracted, and the grid cell to which each event belongs is determined by a coordinate matching algorithm; Create an independent event queue data structure for each grid cell, and store the events that match the grid in chronological order into the corresponding queue.

[0008] Furthermore, by using multi-source heterogeneous data from historical event information to obtain the baseline initial influence score of historical event information, the steps for generating dynamic event influence sub-values ​​for each city grid include: Collect and integrate multi-source heterogeneous data of historical event information, and preprocess the data to convert it into feature vectors; Based on the transformed feature vector, the actual parking demand impact of each historical event is obtained, and the baseline initial impact score of historical events of different types and scales is obtained. When a new event appears in the real-time event queue, the characteristics of the new event are matched with the characteristics of historical events to obtain the characteristics of the historical events that have been successfully matched. The baseline initial influence score corresponding to the successfully matched historical event features is called, and then combined with the real-time dynamic features of the new event to generate a new event influence sub-value, which is then assigned to each city grid affected by the event.

[0009] Furthermore, the pre-defined expression for the urban grid occupancy coefficient increase model is as follows: ; In the formula, For urban grid units In the future The predictive occupancy factor at any given time, i.e., the predictive increase factor. For urban grid units exist The baseline historical occupancy factor at any given time. In order to be in At any given moment, its influence can radiate to the grid cells. The total number of events. For the first in the queue The dynamic event influence sub-value of an event. For the spatiotemporal weight decay function, For the event Center point and grid cell The spatial distance between the center points, for Time and Event The time difference between the start and end times.

[0010] Furthermore, the steps for deriving the prediction results based on the predictive amplification factor include: For each city grid that needs to be predicted, extract its predictive amplification factor value; From the pre-stored urban infrastructure database, query and retrieve the list of parking spaces associated with the unique identifier of the urban grid, obtain the total number of parking spaces in the urban grid area, and obtain a preliminary prediction of the number of parking spaces occupied.

[0011] Based on the preliminary predicted number of berths occupied, the results are converted into a predicted occupancy rate. The predicted occupancy rate of the city grid, along with the corresponding city grid identifier and the predicted time point information, is stored as a data record.

[0012] Furthermore, the steps of real-time monitoring of the actual occupancy rate of each city grid and combining the residuals between the derived prediction results and the actual occupancy rate include: The system continuously receives real-time status data from all berth sensors within each urban grid through a berth sensor network. For each city grid, the status of all berths within it is aggregated in real time to obtain the number of berths currently occupied, and the actual occupancy rate of the city grid at the current moment is obtained. Based on the obtained predicted occupancy percentage, combined with the actual occupancy rate, the residual between the predicted result and the actual occupancy rate is obtained.

[0013] Furthermore, the steps for generating real-time occupancy predictions down to a single berth include: Obtain the predictive amplification factor of the city grid after feedback correction; The occupancy status of each parking space at the current moment is obtained through parking space sensors in the parking lots of this urban grid. A micro-feature vector is constructed for each berth. The corrected predictive amplification factor is then fused with the micro-feature vector of each berth to obtain multi-dimensional input data that includes demand pressure and the berth's own attributes. The multidimensional input data is transmitted to a preset prediction model, which outputs the probability value that the single berth will be occupied at a specified time in the future.

[0014] Furthermore, the formula for obtaining the baseline initial influence score is as follows: ; In the formula, For the event The baseline initial influence score, For the event of 3D standardized feature vector For the first Standardized values ​​of each feature for 3D feature weight vector For the first The weight coefficients of each feature, The total number of features.

[0015] Furthermore, the formula for obtaining the influence sub-value of a dynamic event is: ; In the formula, For the event At any moment The dynamic event influence sub-value, For the event The baseline initial influence score, This is a social media popularity amplification factor. For the event exist Real-time social media buzz As the baseline value for normalized heat, The coupling coefficient is the aggregation strength. For the event exist The cumulative aggregation intensity over time, Threshold for determining the aggregation effect, This is the time decay factor.

[0016] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. By establishing an event-driven, gridded dynamic prediction model, city-level macro events, such as concerts and sporting events, are coupled with micro-level parking space occupancy status, thereby achieving a precise response to sudden crowd gathering effects and effectively solving the problem of delayed parking demand prediction caused by ignoring the impact of real-time events in existing technologies.

[0017] 2. By spatially mapping event information with crowd gathering effects to urban grids and establishing a queue of related events, the spatial positioning of the impact of events on parking demand can be achieved, thereby improving the spatial correlation of parking space occupancy prediction and effectively solving the problem of insufficient prediction accuracy caused by the ambiguity of the event impact range in existing technologies.

[0018] 3. By extracting the initial impact score of historical events from multi-source heterogeneous data and combining it with real-time event characteristics to generate dynamic event impact sub-values, dynamic quantitative assessment of the impact of events on parking demand is achieved, thereby enabling the prediction model to adapt to different types and scales of events. Attached Figure Description

[0019] Figure 1 This is a structural diagram of a real-time parking space occupancy perception and prediction system provided in an embodiment of this application. Detailed Implementation

[0020] This application provides a real-time parking space occupancy perception and prediction system, which solves the problems of insufficient consideration of the impact of crowd gathering events, low prediction accuracy, lack of dynamic correction, and difficulty in achieving accurate prediction of a single parking space in the prior art. By establishing a correlation queue between events and urban grids, quantifying the dynamic influence of events, generating prediction coefficients for coupled real-time events, using residual feedback to correct the model, and fusing data to achieve micro-prediction, the system achieves accurate perception and prediction of parking space occupancy, improving prediction adaptability and accuracy.

[0021] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0022] like Figure 1 The diagram shown is a structural diagram of a real-time parking space occupancy perception and prediction system provided in an embodiment of this application. It includes: an event queue establishment module: used to acquire historical and real-time event information with crowd gathering effects occurring within the city, and spatially map the event information to pre-divided urban grids to establish event queues associated with each grid; the event queues are updated in real-time using real-time event information, and the event queues are used to provide a time-sequential event context for each urban grid to support the calculation of the baseline initial influence score and the real-time updating and correction of dynamic event influence sub-values.

[0023] The steps to determine whether an event has a crowd gathering effect are as follows: A comparison model is established based on multi-source data such as historical passenger flow, vehicle entry and exit volume, and TPI (Traffic Flow Index) acquired in real time through physical devices such as deployed traffic flow detectors, roadside radar, and video traffic monitoring systems. The real-time monitoring data during the current event is compared with the historical benchmark data of the same location and time period. When the difference value exceeds the preset gathering threshold and the duration reaches the set period, it can be determined that the event has a crowd gathering effect.

[0024] Dynamic Event Influence Sub-value Acquisition Module: This module obtains the baseline initial influence score of historical event information through multi-source heterogeneous data in historical event information, and generates dynamic event influence sub-values ​​for each city grid by combining historical parking behavior data of each city grid. Predictive amplification coefficient generation module: This module generates a predictive amplification coefficient coupled with real-time event information using a preset urban grid occupancy coefficient amplification model, and derives prediction results based on the predictive amplification coefficient. Feedback correction module: used to monitor the actual occupancy rate of each city grid in real time, and to perform feedback correction on the city grid occupancy coefficient increase model by combining the residual between the exported prediction results and the actual occupancy rate. The actual occupancy rate of each city grid can be obtained through the following steps: acquiring vehicle entry, exit, and parking status data through sensing devices (such as geomagnetic sensors, video recognition cameras, ultrasonic detectors, etc.) deployed around parking lots or roads, and collecting and statistically analyzing the occupancy signals according to the spatially divided grid units. In each monitoring cycle, the proportion of occupied parking spaces in the grid to the total number of parking spaces is calculated, which is the actual occupancy rate of the grid. This occupancy rate will be updated and stored in real time through edge computing or cloud platforms for subsequent analysis of crowd gathering effects and occupancy prediction models.

[0025] The result generation module is used to fuse the corrected predictive amplification factor with the real-time status data of each parking space sensor in the parking lot, input it into the preset prediction model, and generate a real-time occupancy prediction result for a single parking space.

[0026] Furthermore, the steps of spatially mapping event information to pre-divided city grids and establishing event queues associated with each grid include: Obtain geographic coordinate data within the city limits, divide the city area into several regular grid units according to a preset grid size, and assign a unique grid identifier code to each grid unit; Based on the received historical and real-time event information, the geographical coordinates of the events are extracted, and the grid cell to which each event belongs is determined by a coordinate matching algorithm; Create an independent event queue data structure for each grid cell, and store the events that match the grid in chronological order into the corresponding queue; By maintaining dynamic updates to the event queue, new events are automatically added to the tail of the queue, and expired events are removed from the head of the queue, ensuring that the event information stored in each grid queue remains timely and complete.

[0027] Furthermore, by using multi-source heterogeneous data from historical event information to obtain the baseline initial influence score of historical event information, the steps for generating dynamic event influence sub-values ​​for each city grid include: Collect and integrate multi-source heterogeneous data of historical event information, and preprocess the data to convert it into feature vectors. The multi-source heterogeneous data includes the official classification of the event, media coverage, social network discussion volume, traffic control notices, and historical berth occupancy data of the corresponding grid during the event period.

[0028] By cleaning and standardizing multi-source heterogeneous data, data with different structures such as text, numerical data, and time series data can be transformed into unified, quantifiable feature vectors.

[0029] Based on the transformed feature vector, the actual parking demand impact of each historical event is obtained, and the baseline initial impact score of historical events of different types and scales is obtained. When a new event appears in the real-time event queue, the characteristics of the new event are matched with the characteristics of historical events to obtain the characteristics of the historical events that have been successfully matched. The baseline initial influence score corresponding to the successfully matched historical event features is called, and then combined with the real-time dynamic features of the new event to generate a new event influence sub-value, which is then assigned to each city grid affected by the event.

[0030] Furthermore, the pre-defined expression for the urban grid occupancy coefficient increase model is as follows: ; In the formula, For urban grid units In the future The final predictive occupancy factor at any given time, i.e., the predictive increase factor. For urban grid units exist The baseline historical occupancy coefficient for each time period is derived from regular berth occupancy data under normal berth occupancy conditions without special events. In order to be in At any given moment, its influence can radiate to the grid cells. The total number of events. For the first in the queue The dynamic event influence sub-value of an event is used to quantify the driving force of that independent event on parking demand. This is a spatiotemporal weight decay function used to calculate events. The diminishing effect of influence over space and time. For the event Center point and grid cell The spatial distance between the center points, for Time and Event The time difference between the start and end times.

[0031] Furthermore, the steps for deriving the prediction results based on the predictive amplification factor include: For each city grid that needs to be predicted, extract its predictive amplification factor value; From the pre-stored urban infrastructure database, query and retrieve the list of parking spaces associated with the unique identifier of the urban grid to obtain the total number of parking spaces in the urban grid area. By performing a multiplication operation, multiply the predictive amplification factor of the urban grid with the actual total number of parking spaces to obtain the preliminary predicted number of parking spaces occupied.

[0032] Based on the preliminary predicted number of berths occupied, a division operation is further performed, dividing the predicted number of occupied berths by the total number of berths, and the result is converted into the predicted occupancy rate. The predicted occupancy rate of the city grid, along with the corresponding city grid identifier and the predicted time point information, is stored as a data record.

[0033] Furthermore, the steps involved in combining the residuals between the derived prediction results and the actual occupancy rate include: The system continuously receives real-time status data from all berth sensors within each urban grid through a berth sensor network. For each city grid, the status of all berths within it is aggregated in real time to obtain the number of berths currently occupied. By dividing the number of occupied berths by the total number of berths in that grid, the actual occupancy rate of the city grid at the current moment is obtained. Based on the obtained predicted occupancy percentage, combined with the actual occupancy rate, the predicted occupancy rate is subtracted from the calculated actual occupancy rate to obtain the residual between the predicted result and the actual occupancy rate. By binding the residual between the predicted results and the actual occupancy rate with the corresponding city grid identifier and timestamp, and storing it in a time series database specifically for recording errors, data input is provided for the subsequent feedback correction module.

[0034] Furthermore, the steps for generating real-time occupancy predictions down to a single berth include: Obtain the predictive amplification factor of the city grid after feedback correction; The occupancy status of each parking space at the current moment is obtained through parking space sensors in the parking lots of this urban grid. A micro-feature vector is constructed for each parking space. This vector contains not only its current occupancy status but also its static attributes, such as the parking space's specific location within the parking lot, its proximity to entrances and exits, and its historical usage frequency. The corrected predictive amplification factor is then fused with the micro-feature vector of each parking space to obtain multi-dimensional input data that includes demand pressure and the parking space's own attributes. The multidimensional input data is transmitted to a preset prediction model, which independently calculates the input data for each berth and outputs the probability value of the single berth being occupied at a specified time in the future, thereby achieving a refined and differentiated prediction of the future occupancy status of each specific berth.

[0035] Furthermore, the formula for obtaining the baseline initial influence score is as follows: ; In the formula, For the event The baseline initial influence score, For the event of 3D standardized feature vector For the first Standardized values ​​of each feature for 3D feature weight vector For the first The weight coefficients of each feature, The total number of features.

[0036] Furthermore, the formula for obtaining the influence sub-value of a dynamic event is: ; In the formula, For the event At any moment The dynamic event influence sub-value, For the event The baseline initial influence score, This is a social media popularity amplification factor. For the event exist Real-time social media buzz As the baseline value for normalized heat, The coupling coefficient is the aggregation strength. For the event exist The cumulative aggregation intensity over time, Threshold for determining the aggregation effect, This is the time decay factor.

[0037] Those skilled in the art should understand that the above formulas are all dimensionless numerical calculations. The formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0038] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0039] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0040] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0041] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0042] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A real-time parking space occupancy sensing and prediction system, characterized in that, include: Event queue establishment module: used to acquire historical and real-time event information with crowd gathering effects in the city, and to spatially map the event information to pre-divided urban grids to establish event queues associated with each grid; Dynamic Event Influence Sub-Value Acquisition Module: This module is used to obtain the baseline initial influence score of historical event information through multi-source heterogeneous data in historical event information, and generate dynamic event influence sub-values ​​for each city grid. The steps for generating dynamic event influence sub-values ​​for each city grid include: Collect and integrate multi-source heterogeneous data of historical event information, and preprocess the data to convert it into feature vectors; Based on the transformed feature vector, the actual parking demand impact of each historical event is obtained, and the baseline initial impact score of historical events of different types and scales is obtained. When a new event appears in the real-time event queue, the characteristics of the new event are matched with the characteristics of historical events to obtain the characteristics of the historical events that have been successfully matched. The baseline initial influence score corresponding to the successfully matched historical event features is called, and then combined with the real-time dynamic features of the new event to generate a new event influence sub-value, and the influence sub-value is assigned to each city grid affected by the event. Predictive amplification coefficient generation module: This module generates a predictive amplification coefficient coupled with real-time event information based on the dynamic event influence sub-value and a preset urban grid occupancy coefficient amplification model, and then derives the prediction results based on the predictive amplification coefficient. The preset expression for the urban grid occupancy coefficient increase model is: ； In the formula, For urban grid units In the future The predictive occupancy factor at any given time, i.e., the predictive increase factor. For urban grid units exist The baseline historical occupancy factor at any given time. In order to be in At any given moment, its influence can radiate to the grid cells. The total number of events. For the first in the queue The dynamic event influence sub-value of an event. For the spatiotemporal weight decay function, For the event Center point and grid cell The spatial distance between the center points, for Time and Event The time difference between the start times; Feedback correction module: used to monitor the actual occupancy rate of each city grid in real time, and to perform feedback correction on the city grid occupancy coefficient increase model by combining the residual between the exported prediction results and the actual occupancy rate. The results generation module is used to fuse the corrected predictive amplification factor with the micro-feature vector of each parking space in the parking lot, input it into the preset prediction model, and generate real-time occupancy prediction results for each individual parking space.

2. The parking space real-time occupancy perception and prediction system as described in claim 1, characterized in that, The steps of spatially mapping event information to pre-divided city grids and establishing event queues associated with each grid include: Obtain geographic coordinate data within the city limits, divide the city area into several regular grid units according to a preset grid size, and assign a unique grid identifier code to each grid unit; Based on the received historical and real-time event information, the geographical coordinates of the events are extracted, and the grid cell to which each event belongs is determined by a coordinate matching algorithm; Create an independent event queue data structure for each grid cell, and store the events that match the grid in chronological order into the corresponding queue.

3. The parking space real-time occupancy perception and prediction system as described in claim 1, characterized in that, The steps for deriving prediction results based on the predictive amplification factor include: For each city grid that needs to be predicted, extract its predictive amplification factor value; From the pre-stored urban infrastructure database, query and retrieve the list of parking spaces associated with the unique identifier of the urban grid, obtain the total number of parking spaces in the urban grid area, and obtain the preliminary predicted number of parking spaces occupied. Based on the preliminary predicted number of berths occupied, the results are converted into a predicted occupancy rate. The predicted occupancy rate of the city grid, along with the corresponding city grid identifier and the predicted time point information, is stored as a data record.

4. The parking space real-time occupancy perception and prediction system as described in claim 1, characterized in that, The steps for real-time monitoring of the actual occupancy rate of each city grid and combining the residuals between the derived prediction results and the actual occupancy rate include: The system continuously receives real-time status data from all berth sensors within each urban grid through a berth sensor network. For each city grid, the status of all berths within it is aggregated in real time to obtain the number of berths currently occupied, and the actual occupancy rate of the city grid at the current moment is obtained. Based on the obtained predicted occupancy percentage, combined with the actual occupancy rate, the residual between the predicted result and the actual occupancy rate is obtained.

5. The parking space real-time occupancy perception and prediction system as described in claim 1, characterized in that, The steps for generating real-time occupancy predictions down to a single berth include: Obtain the predictive amplification factor of the city grid after feedback correction; The occupancy status of each parking space at the current moment is obtained through parking space sensors in the parking lots of this urban grid. A micro-feature vector is constructed for each berth. The corrected predictive amplification factor is then fused with the micro-feature vector of each berth to obtain multi-dimensional input data that includes demand pressure and the berth's own attributes. The multidimensional input data is transmitted to a preset prediction model, which outputs the probability value that the single berth will be occupied at a specified time in the future.

6. The parking space real-time occupancy perception and prediction system as described in claim 1, characterized in that, The formula for obtaining the initial baseline influence score is: ； In the formula, For the event The baseline initial influence score, For the event of A dimensional standardized feature vector is a set of quantized features derived from multi-source heterogeneous data. For the first Standardized values ​​of each feature for dimensional feature weight vector, For the first The weight coefficients of each feature, The total number of features.

7. The parking space real-time occupancy perception and prediction system as described in claim 1, characterized in that, The formula for obtaining the influence sub-value of a dynamic event is: ； In the formula, For the event At any moment The dynamic event influence sub-value, For the event The baseline initial influence score, This is a social media popularity amplification factor. For the event exist Real-time social media buzz As the baseline value for normalized heat, The coupling coefficient is the aggregation strength. For the event exist The cumulative aggregation intensity over time, Threshold for determining the aggregation effect, This is the time decay factor.

Images