City-level intelligent parking dynamic scheduling and driving risk early warning method and system

Abstract

The invention discloses a city-level intelligent parking dynamic scheduling and driving risk early warning method and system, and the method is executed by a city-level intelligent parking platform, and comprises the steps: obtaining vehicle behavior and parking state data collected by a sensing device, and weather, traffic flow and closed parking lot data; fusing multi-source data to construct a dynamic parking resource database; based on the database, utilizing an AI model to dynamically evaluate regional parking pressure and predict future parking demand distribution; meanwhile, weather, traffic flow, road structure and real-time parking pressure data are fused, and the driving risk level is quantified through a multi-factor risk assessment model; and according to the prediction and evaluation result, generating a parking guide strategy and driving risk early warning information, and issuing the parking guide strategy and the driving risk early warning information to the vehicle owner terminal. According to the invention, accurate prediction and active scheduling of the parking resources are realized, forward-looking early warning can be carried out on composite traffic risks including secondary risks caused by parking saturation, and the urban parking efficiency and the road safety level are effectively improved.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation technology, and in particular to a method for dynamic vehicle scheduling. Background Technology

[0002] With the acceleration of urbanization and the continuous increase in the number of motor vehicles, "parking difficulties" have become a common problem troubling city managers and citizens. Existing solutions mostly focus on the informatization and reservation functions of parking spaces, such as displaying available parking spaces through an app. However, these systems have significant shortcomings: First, they are mostly information silos, unable to achieve cross-site integration and coordinated scheduling of city-level parking resources; second, they lack the ability to accurately predict future parking demand in a region, resulting in lagging guidance strategies and easily causing "no parking spaces available" and traffic congestion; finally, existing systems have limited functionality, focusing only on parking itself and failing to link parking behavior with dynamic driving safety risks.

[0003] On the other hand, driving safety warning systems typically rely on vehicle sensors or simple traffic flow analysis, failing to effectively incorporate macro-environmental factors such as weather and real-time parking resource availability, resulting in incomplete and unforeseen risk identification. For example, during heavy rain, the risk of traffic accidents on surrounding roads in an area with extremely saturated parking spaces increases significantly, but existing systems cannot effectively warn of such complex risks. Summary of the Invention

[0004] This application provides a city-level intelligent parking dynamic scheduling and driving risk early warning method and system, which aims to solve the problems of lack of city-level resource integration, insufficient forward-looking prediction capability, and disconnection from macro-level driving safety risks in existing parking management systems.

[0005] Firstly, a method for dynamic scheduling and driving risk warning of urban-level smart parking is provided, executed by a city-level smart parking platform, the method comprising:

[0006] It acquires real-time vehicle behavior data and parking status information collected by sensing devices deployed in various urban areas, as well as weather data, real-time traffic flow data, and closed parking lot data from external data sources.

[0007] By integrating the vehicle behavior data, parking status information, weather data, real-time traffic flow data, and closed parking lot data, a dynamic parking resource database is constructed and continuously updated.

[0008] Based on the dynamic parking resource database, a pre-trained artificial intelligence model is used to dynamically assess regional parking pressure and predict the distribution of parking demand in each region during a specific future period.

[0009] Based on real-time weather data, real-time traffic flow data, road topology data, and real-time parking pressure data obtained from the parking analysis and prediction steps in the dynamic parking resource database, the comprehensive driving risk level of different road sections or intersections is calculated through a multi-factor risk assessment model.

[0010] Based on the predicted parking demand distribution, targeted parking guidance and scheduling strategies are generated; and based on the comprehensive driving risk level assessment results, high-risk road sections or intersections are identified.

[0011] The parking guidance and scheduling strategy, as well as the driving risk warning information for the high-risk road sections or intersections, are sent to the vehicle's onboard terminal or the owner's mobile terminal via the communication network.

[0012] Optionally, in the above scheme, the sensing device includes a high-position video camera and a geomagnetic sensor; the high-position video camera is used to identify license plate numbers, vehicle trajectories, and parking space occupancy status; the geomagnetic sensor is used to assist in detecting the occupancy status of parking spaces.

[0013] Optionally, in the above scheme, predicting the distribution of parking demand in various regions within a specific future time period specifically includes:

[0014] Historical parking data, real-time parking data, date type, surrounding points of interest data, and real-time traffic flow data are input into a deep learning model based on long short-term memory networks or temporal convolutional networks.

[0015] The deep learning model outputs a probability map of parking demand for each grid area within a preset future time period.

[0016] Optionally, in the above scheme, the calculation of the comprehensive driving risk level through a multi-factor risk assessment model specifically includes:

[0017] Assign appropriate weighting factors to real-time weather data, real-time traffic flow data, historical accident data, road topology data, and real-time parking pressure data;

[0018] Calculate the comprehensive risk score based on the aforementioned weighting factors;

[0019] Based on preset risk score thresholds, risk levels are divided into low risk, medium risk, high risk, and extremely high risk.

[0020] Optionally, in the above scheme, the generation of targeted parking guidance and scheduling strategies specifically includes:

[0021] For areas predicted to experience high parking demand, a strategy is generated to divert traffic to surrounding parking lots or roadside parking areas with a higher proportion of vacant spaces, and an optimal navigation path is generated to the parking lots or roadside parking areas.

[0022] Optionally, in the above scheme, if a high-risk road segment or intersection is identified as being located on a vehicle's current navigation path during the strategy generation step, an alternative route is replanned for the vehicle to avoid the high-risk road segment or intersection, and the alternative route is sent out together with the driving risk warning information.

[0023] Secondly, a city-level intelligent parking dynamic scheduling and driving risk early warning system is provided to implement the above-mentioned method, the system comprising:

[0024] The perception layer consists of high-position video cameras and geomagnetic sensors deployed in various areas of the city, used to collect raw perception data on vehicle behavior and parking space status.

[0025] The data layer communicates with the perception layer and is configured with an interface to connect to external data sources. It is used to receive, clean, and fuse the original perception data, weather data, real-time traffic flow data, and closed parking lot data to build and maintain a unified dynamic parking resource database.

[0026] A platform layer, communicatively connected to the data layer, comprising:

[0027] The regional parking pressure dynamic assessment module is used to calculate and output regional parking pressure information based on the dynamic parking resource database;

[0028] The parking demand prediction module has a built-in artificial intelligence-based prediction model, which is used to process the data in the dynamic parking resource database and output parking demand distribution prediction information for future time periods.

[0029] The driving risk assessment module has a built-in multi-factor fusion risk assessment model, which is used to process the dynamic parking resource database and related data and output driving risk level information.

[0030] The application layer, which communicates with the platform layer, is used to receive the parking demand distribution prediction information and the driving risk level information, generate publishable parking guidance and scheduling instructions and driving risk warning instructions, and output them to the outside world through the information publishing module.

[0031] Optionally, in the above scheme, the parking demand prediction module may specifically adopt a deep learning model architecture based on Long Short-Term Memory Network (LSTM) or Temporal Convolutional Network (TCN).

[0032] Optionally, in the above scheme, the multi-factor fusion risk assessment model in the driving risk assessment module is configured to perform weighted calculations on the input real-time weather data, real-time traffic flow data, historical accident data, road topology data, and parking resource tension data from the regional parking pressure dynamic assessment module, so as to output a quantified risk level.

[0033] Optionally, in the above scheme, the information publishing module in the application layer can establish a communication connection with at least one of the following terminal forms: Variable Message Service (VMS), vehicle terminal, smartphone application, and mini-program, in order to push information.

[0034] Compared with the prior art, this application has at least the following beneficial effects:

[0035] Based on further analysis and research of existing technical problems, this application recognizes that solving the problems of "parking difficulty" and "driving risk" cannot rely on isolated information systems or a single early warning dimension. This application constructs a unified data platform that deeply integrates city-level multi-source traffic and environmental data, and uses an artificial intelligence model to simultaneously achieve forward-looking parking demand prediction and multi-factor driving risk assessment that integrates real-time parking status. This achieves the dual effect of coordinating and optimizing parking resource allocation from the source and proactively warning of driving risks, thereby improving urban parking efficiency while significantly enhancing road traffic safety.

[0036] This application also has at least the following beneficial effects:

[0037] 1. Global optimization and forward-looking scheduling: Through city-level data integration and prediction, a shift from "passive display" to "active guidance" has been achieved, effectively balancing regional parking load, reducing parking-seeking traffic flow, and alleviating congestion at the source.

[0038] 2. Proactive safety and risk prevention: By incorporating parking status into the risk assessment system, accurate early warnings of secondary traffic safety issues caused by parking difficulties are achieved, thereby improving the overall prevention and control capabilities of road safety.

[0039] 3. System integration and efficiency enhancement: The deep integration of the originally independent parking management and traffic safety early warning systems has generated a synergistic effect of "1+1>2", providing a unified and efficient technical platform for smart urban traffic management. Attached Figure Description

[0040] Figure 1 A flowchart illustrating a city-level smart parking dynamic scheduling and driving risk early warning method provided in one embodiment of this application;

[0041] Figure 2A flowchart illustrating a city-level smart parking dynamic scheduling and driving risk early warning method provided in another embodiment of this application;

[0042] Figure 3 A schematic diagram of the overall architecture of a city-level intelligent parking dynamic scheduling and driving risk early warning system provided in one embodiment of this application;

[0043] Figure 4 A schematic diagram illustrating the working principle of a parking demand prediction module provided in one embodiment of this application;

[0044] Figure 5 This is a schematic diagram of multi-factor fusion of a driving risk assessment model provided in one embodiment of this application. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0046] In the description of this application, unless otherwise stated, the terms "including", "comprising", "having", etc., also mean "not limited to" (certain units, components, materials, steps, etc.).

[0047] In the technical solution of this application, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with the relevant laws, regulations, and standards of the relevant countries and regions, and necessary confidentiality measures have been taken. They do not violate public order and good morals, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0048] The purpose of this application is to overcome the shortcomings of the existing technology and provide a city-level intelligent parking dynamic scheduling and driving risk early warning method and system to achieve efficient utilization of parking resources and proactive prevention and control of driving risks.

[0049] In some embodiments, a city-level smart parking dynamic scheduling and driving risk warning method is provided, executed by a city-level smart parking platform, the method comprising:

[0050] It acquires real-time vehicle behavior data and parking status information collected by sensing devices deployed in various urban areas, as well as weather data, real-time traffic flow data, and closed parking lot data from external data sources.

[0051] By integrating the vehicle behavior data, parking status information, weather data, real-time traffic flow data, and closed parking lot data, a dynamic parking resource database is constructed and continuously updated.

[0052] Based on the dynamic parking resource database, a pre-trained artificial intelligence model is used to dynamically assess regional parking pressure and predict the distribution of parking demand in each region during a specific future period.

[0053] Based on real-time weather data, real-time traffic flow data, road topology data, and real-time parking pressure data obtained from the parking analysis and prediction steps in the dynamic parking resource database, the comprehensive driving risk level of different road sections or intersections is calculated through a multi-factor risk assessment model.

[0054] Based on the predicted parking demand distribution, targeted parking guidance and scheduling strategies are generated; and based on the comprehensive driving risk level assessment results, high-risk road sections or intersections are identified.

[0055] The parking guidance and scheduling strategy, as well as the driving risk warning information for the high-risk road sections or intersections, are sent to the vehicle's onboard terminal or the owner's mobile terminal via the communication network.

[0056] In some embodiments, the present application's solution upgrades the traditional, isolated parking information query service into a proactive, forward-looking, and safety-oriented city-level smart transportation service through systematic and intelligent data processing and decision-making processes. The following provides further explanation of each step:

[0057] Step 1: Comprehensive Data Acquisition. This method does not rely on a single data source. Instead, it uses sensing devices deployed throughout the city (such as high-position video cameras and geomagnetic sensors) and external data interfaces to simultaneously acquire data on vehicle behavior, parking space status, weather, real-time traffic flow, and parking spaces in closed parking lots. This forms the data foundation for the system's operation, enabling multi-dimensional real-time perception of the urban parking and traffic environment.

[0058] Step Two: Dynamic Database Construction. By integrating heterogeneous data from multiple sources, the system constructs and continuously updates a dynamic parking resource database, essentially creating a comprehensive, real-time, interconnected digital twin view of "parking-traffic-environment." This database is not merely a simple aggregation of information, but rather the core data support for achieving cross-site collaborative scheduling and complex risk assessment.

[0059] Step 3: Intelligent Prediction and Assessment. The system uses pre-trained AI models (such as LSTM, TCN, etc.) to assess regional parking pressure and predict demand. This step combines historical patterns, real-time status, and environmental characteristics to output a forward-looking parking demand distribution map, enabling the scheduling strategy to shift from "passive response" to "proactive guidance."

[0060] Step 4: Integrated Driving Risk Assessment. The innovation of this step lies in the fact that the driving risk assessment not only considers weather, traffic flow, and road structure, but also introduces real-time parking pressure as a key risk factor for the first time. Through multi-factor model quantitative calculation, the system can identify complex high-risk scenarios such as "rainy / snowy weather + saturated parking areas," achieving more comprehensive and realistic risk warnings.

[0061] Step 5: Integrated Strategy Generation. Based on the prediction and evaluation results, the system simultaneously generates parking guidance strategies and driving risk warnings. This means that the system not only solves the problem of "where to park," but also informs users "whether it is safe along the way," demonstrating the deep integration of parking scheduling and driving safety at the decision-making level.

[0062] Step 6: Precise Information Service Delivery. Through various channels such as VMS, vehicle terminals, or mobile apps, the system will accurately and promptly push guidance and warning information to affected vehicles, forming a closed loop from cloud-based decision-making to user reach, ensuring that the technical solution is implemented as a practical service.

[0063] This embodiment fully demonstrates how this application constructs a smart transportation system that can both optimize the utilization of parking resources and proactively prevent driving risks through data fusion, AI prediction, risk modeling, and information service linkage, thereby solving the complex urban traffic problems of "difficulty in finding parking" and "driving risks" in practical applications.

[0064] In some embodiments, the sensing device includes a high-position video camera and a geomagnetic sensor; the high-position video camera is used to identify license plate numbers, vehicle trajectories, and parking space occupancy status; the geomagnetic sensor is used to assist in detecting the occupancy status of parking spaces.

[0065] To achieve accurate perception at the data acquisition source, the system employs a multi-sensor fusion and collaborative technology strategy, and clearly defines the core devices and their respective roles in the perception layer:

[0066] 1. High-position video cameras, as a primary non-contact sensing device, have the core advantage of acquiring rich visual information. This provides crucial data for vehicle identification, parking time and fee calculation, and evidence collection of illegal parking. They can analyze vehicle paths, speeds, and flows within the urban road network, forming the basis for understanding macro-level traffic flow and micro-level driving behavior. Through video analysis algorithms, they can directly visually determine whether parking spaces are occupied, providing the most intuitive information on parking space occupancy.

[0067] 2. As an embedded, contact-based (or short-range) sensing device, the core advantage of geomagnetic sensors lies in their stable monitoring and minimal susceptibility to environmental interference (such as light and weather). By detecting changes in the geomagnetic field caused by the metal body of a vehicle, they can reliably and accurately determine the occupancy or vacancy status of the parking space above it. Their data can be cross-validated and calibrated with visual judgments from cameras, ensuring the continuity and accuracy of parking space status data, especially at night, in rainy or foggy weather, or under conditions of visual obstruction.

[0068] By combining high-position video cameras with geomagnetic sensors, a redundant and complementary system of "visual perception + status sensing" is formed. This not only significantly improves the accuracy and robustness of parking space status detection, laying a solid foundation for building a reliable dynamic parking resource database, but also provides license plate and trajectory information from the video data, offering indispensable multi-dimensional data input for subsequent vehicle behavior analysis, parking demand prediction, and related driving risk assessment.

[0069] In some embodiments, predicting the distribution of parking demand in different areas over a specific future time period specifically includes:

[0070] Historical parking data, real-time parking data, date type, surrounding points of interest data, and real-time traffic flow data are input into a deep learning model based on long short-term memory networks or temporal convolutional networks.

[0071] The deep learning model outputs a probability map of parking demand for each grid area within a preset future time period.

[0072] The key shift in this application from traditional "information retrieval" based on the current state to "demand prediction" based on intelligent computing lies in:

[0073] 1. Multi-dimensional fusion of model input features.

[0074] The predictive model does not rely solely on historical parking records, but integrates data from five key dimensions, which form the basis for its accurate predictions:

[0075] Historical parking data reveals long-term parking patterns in different areas over periods (such as weekdays and weekends) and trends.

[0076] Real-time parking data reflects the current parking occupancy status of various parts of the city and serves as the starting point for prediction.

[0077] Date type: As an important time context, it helps the model distinguish parking behavior differences under different date patterns such as weekdays, weekends, and holidays.

[0078] Surrounding points of interest data: This provides regional functional attributes (such as commercial areas, hospitals, schools, and residential areas), which is key to understanding the generation and attraction sources of parking demand.

[0079] Real-time traffic flow data: Traffic congestion directly affects the speed at which vehicles reach their destinations, which in turn relates to the accumulation of parking demand over time.

[0080] 2. Advanced deep learning model architecture.

[0081] Choosing Long Short-Term Memory (LSTM) networks or Temporal Convolutional Networks (TCN) as the core model has clear technical advantages:

[0082] LSTM excels at capturing and memorizing long-term dependencies in time-series data, making it ideal for learning complex patterns in parking demand that evolve over time (such as morning and evening rush hours and holiday cycles).

[0083] TCN: Employs dilated causal convolution, which can efficiently process long sequences and has stable parallel computing capabilities, performing well in scenarios requiring fast response and real-time updates.

[0084] Both models can effectively process the aforementioned multidimensional time-series input data and extract nonlinear, high-dimensional patterns of parking demand evolution.

[0085] 3. Output results in the form of a "probability graph".

[0086] The model output is not a simple binary judgment of "available / unavailable," but rather a probability map of parking demand in each grid area within a preset future time period. Using grids as units, it provides detailed spatial distribution of parking demand intensity across the city; it clearly points to a specific future time (e.g., the next 30 minutes), providing a time window for early intervention and scheduling; and by expressing demand intensity in probabilistic form, it can more delicately reflect uncertainty and the gradient of demand changes, providing richer decision-making basis for generating dynamic scheduling strategies than simple threshold judgments.

[0087] This application utilizes deep learning technology to model multi-source data that integrates spatiotemporal and environmental features, thereby transforming parking demand prediction from an empirical estimation problem into a computable, quantifiable, and spatiotemporally resolved intelligent prediction task. This is the core technological support for realizing the fundamental shift from "passively displaying available parking spaces" to "actively guiding vehicle flow".

[0088] In some embodiments, calculating the comprehensive driving risk level using a multi-factor risk assessment model specifically includes:

[0089] Assign appropriate weighting factors to real-time weather data, real-time traffic flow data, historical accident data, road topology data, and real-time parking pressure data;

[0090] Calculate the comprehensive risk score based on the aforementioned weighting factors;

[0091] Based on preset risk score thresholds, risk levels are divided into low risk, medium risk, high risk, and extremely high risk.

[0092] By systematically integrating the core dimensions affecting driving safety, a multi-dimensional risk assessment framework has been constructed:

[0093] Real-time weather data reflects physical conditions that directly affect driving safety, such as environmental visibility and road surface adhesion coefficient.

[0094] Real-time traffic flow data: characterizes the real-time load and congestion level of roads, and traffic flow anomalies are often related to the probability of accidents.

[0095] Historical accident data: revealing the inherent "black spots" and accident tendencies of specific road sections or intersections from a statistical perspective.

[0096] Road topology data describes the impact of inherent road geometric features such as curves, gradients, number of lanes, and intersection complexity on safety.

[0097] Real-time parking pressure data: quantitatively reflects the secondary risk behaviors such as additional traffic flow, frequent starts and stops, and illegal temporary parking around the target area caused by searching for parking spaces.

[0098] By assigning appropriate weights to different factors (a process that can be determined based on data mining of historical accident data or expert knowledge in the field of traffic safety), the model can reflect the contribution of each factor to the overall risk in a differentiated manner.

[0099] By combining the real-time quantified values ​​of each factor with their weights through a clear mathematical model (such as linear weighting or nonlinear fusion), a unified comprehensive risk score is calculated, realizing a single numerical measure of complex risk conditions, making risk levels under different spatiotemporal conditions comparable and rankable.

[0100] Based on preset risk score thresholds (usually based on statistical analysis of historical data or safety standard settings), continuous risk scores are mapped to discrete "low, medium, high, and extremely high" risk levels. This makes the risk assessment results intuitive and easy to understand, facilitating the generation of subsequent early warning information and the formulation of tiered response strategies. It not only provides a direct basis for issuing differentiated early warning information (such as color and intensity) to drivers, but also provides clear decision support for traffic managers to conduct regional risk situation assessments and resource allocation.

[0101] In some embodiments, generating targeted parking guidance and scheduling strategies specifically includes:

[0102] For areas predicted to experience high parking demand, a strategy is generated to divert traffic to surrounding parking lots or roadside parking areas with a higher proportion of vacant spaces, and an optimal navigation path is generated to the parking lots or roadside parking areas.

[0103] The strategy generation is not based on the current idle status, but rather on areas of high demand "predicted to occur." This reflects the foresight and proactiveness of this application's solution. Its core logic is "diversion-redirection": before predicting that a certain "hotspot" area is about to become saturated, the system intervenes in advance, diverting vehicles that may flood into that area to suitable "reservoirs" (other parking lots or roadside parking areas with sufficient vacant spaces) in the surrounding area. This directly regulates the destination distribution of vehicles at the source, aiming to prevent "no parking spaces available" and the resulting concentration of parking-seeking traffic in local areas. It is a key technical action for achieving city-level parking load balancing.

[0104] The criterion for selecting guidance targets is "a high proportion of vacant parking spaces in the surrounding area," rather than simply "the closest location." Encouraging vehicles to disperse to areas with relatively abundant resources promotes the overall optimization of parking resources across the entire urban road network, rather than a localized relocation that may cause new congestion.

[0105] The system not only tells users "where to stop," but also simultaneously generates a navigation route showing "the optimal way to get there." It provides users with a one-stop solution from decision-making to execution, greatly improving the acceptability and efficiency of guidance strategies and ensuring that scheduling intentions are effectively implemented.

[0106] By generating a comprehensive guidance strategy that integrates target selection logic and route navigation, the system is no longer a passive information board, but an intelligent agent that proactively allocates resources and organizes traffic flow. This effectively solves the problem of "lagging guidance strategies" in the background technology. Through proactive diversion and precise route guidance, it alleviates parking pressure in the target area while also optimizing the overall traffic flow in the region.

[0107] In some embodiments, during the strategy generation step, if a high-risk road segment or intersection is identified as being located on a vehicle's current navigation path, an alternative route is replanned for the vehicle to avoid the high-risk road segment or intersection, and the alternative route is sent out together with the driving risk warning information.

[0108] When overlapping risks are detected, the system does not merely issue a warning, but automatically triggers a dynamic route replanning algorithm. This algorithm uses "avoiding identified high-risk road sections / intersections" as one of its core constraints (while also considering traditional optimization objectives such as distance and time), and calculates a safe alternative route in real time. This action elevates the system's role from "risk announcer" to "safety navigator," directly translating risk mitigation decisions into actionable navigation instructions, providing users with the most direct and effective risk avoidance solutions.

[0109] In some embodiments, reference Figure 3 A city-level intelligent parking dynamic scheduling and driving risk early warning system is provided to implement the method provided in the above embodiments. The system includes:

[0110] The perception layer consists of high-position video cameras and geomagnetic sensors deployed in various areas of the city, used to collect raw perception data on vehicle behavior and parking space status.

[0111] The data layer communicates with the perception layer and is configured with an interface to connect to external data sources. It is used to receive, clean, and fuse the original perception data, weather data, real-time traffic flow data, and closed parking lot data to build and maintain a unified dynamic parking resource database.

[0112] A platform layer, communicatively connected to the data layer, comprising:

[0113] The regional parking pressure dynamic assessment module is used to calculate and output regional parking pressure information based on the dynamic parking resource database;

[0114] The parking demand prediction module has a built-in artificial intelligence-based prediction model, which is used to process the data in the dynamic parking resource database and output parking demand distribution prediction information for future time periods.

[0115] The driving risk assessment module has a built-in multi-factor fusion risk assessment model, which is used to process the dynamic parking resource database and related data and output driving risk level information.

[0116] The application layer, which communicates with the platform layer, is used to receive the parking demand distribution prediction information and the driving risk level information, generate publishable parking guidance and scheduling instructions and driving risk warning instructions, and output them to the outside world through the information publishing module.

[0117] In some embodiments, a city-level intelligent parking dynamic scheduling and driving risk early warning system includes:

[0118] The perception layer consists of IoT devices such as high-position cameras and geomagnetic sensors deployed along urban roads, at parking lot entrances and exits, and inside the parking lot. It is responsible for collecting raw data such as vehicle trajectories, license plate information, and parking space occupancy status around the clock.

[0119] Data Layer: As the system's data hub, this layer receives and cleans raw data from the perception layer. It also accesses weather data from meteorological departments, real-time traffic flow data from traffic management departments, and parking space data from various enclosed parking lots via API interfaces. This layer builds and maintains a unified, real-time updated dynamic parking resource database.

[0120] Platform layer: This is the system's "intelligent brain," containing three core AI models:

[0121] Regional parking pressure dynamic assessment module: calculates the parking resource occupancy rate and turnover rate of each region in real time and generates a heat map.

[0122] Parking demand prediction module: It uses deep learning algorithms to comprehensively analyze historical patterns and real-time dynamics to predict the spatiotemporal distribution of parking demand in the near future.

[0123] Driving risk assessment module: Construct a multi-factor fusion model to quantitatively analyze the comprehensive risk level of different road sections under specific weather, traffic and parking pressures.

[0124] Application layer: As the interaction interface between the system and users, it accurately pushes parking guidance suggestions, dispatch instructions and risk warning information generated by the platform layer to drivers and managers through channels such as VMS, mobile APP, and vehicle system.

[0125] Through four tightly coupled and specialized levels, the system constructs an integrated technology platform that enables seamless data flow, precise intelligent analysis, and proactive closed-loop services. It not only physically integrates various sensors and data sources but also logically merges the previously separate functions of parking resource scheduling and driving safety early warning through advanced AI models, achieving a synergistic effect of "1+1>2." This provides complete hardware and software architecture support for systematically solving the complex problems of urban "parking difficulties" and "driving risks."

[0126] In some embodiments, the parking demand prediction module specifically adopts a deep learning model architecture based on a Long Short-Term Memory (LSTM) network or a Temporal Convolutional Network (TCN).

[0127] In some embodiments, the parking demand prediction module specifically employs a deep learning model architecture based on a Long Short-Term Memory (LSTM) network or a Temporal Convolutional Network (TCN). This choice is based on the strong spatiotemporal sequence dependency characteristic of parking demand data. LSTM models excel at capturing long-term temporal dependencies and can effectively learn parking patterns on a daily or weekly basis; TCN models, with their advantages of causal convolution and parallel computing, perform excellently in processing long-sequence data and achieving fast real-time prediction.

[0128] In some embodiments, the multi-factor fusion risk assessment model in the driving risk assessment module is configured to perform weighted calculations on the input real-time weather data, real-time traffic flow data, historical accident data, road topology data, and parking resource tension data from the regional parking pressure dynamic assessment module, so as to output a quantified risk level.

[0129] In some embodiments, the multi-factor fusion risk assessment model in the driving risk assessment module is specifically configured to perform weighted fusion calculations on the input real-time weather data, real-time traffic flow data, historical accident data, road topology data, and parking resource tension data from the regional parking pressure dynamic assessment module. The model assigns appropriate weights to each factor and comprehensively calculates a quantified comprehensive risk score, which is then mapped to risk levels such as "low, medium, high, and extremely high," thereby achieving a quantitative assessment of complex traffic risks, including secondary risks caused by parking difficulties.

[0130] In some embodiments, the information publishing module in the application layer establishes a communication connection with at least one terminal form among Variable Message Service (VMS), vehicle terminal, smartphone application, and mini-program to push information.

[0131] In some embodiments, to ensure that obstacle warning information can effectively reach users, the information publishing module in the application layer is designed to have multi-channel publishing capabilities. This module establishes communication connections and protocol interfaces with at least one terminal form among variable information signs (VMS), vehicle terminals, smartphone applications, and mini-programs, thereby enabling it to select the most suitable channel for accurate push of parking guidance and risk warning information based on the scenario and user reach efficiency.

[0132] Example 1

[0133] refer to Figure 3The system's perception layer deployed 50 high-position cameras and 300 geomagnetic sensors in a commercial area. The data layer, through the city's data platform, accessed real-time rainfall data from the municipal meteorological bureau and real-time traffic speed data from the traffic control center. The platform layer's parking demand prediction module (using an LSTM model) predicted, based on historical data, that the area would reach 95% saturation during the weekend evening rush hour. Simultaneously, the driving risk assessment module detected moderate rain expected at that time.

[0134] The application layer then pushes a message to vehicles heading into the area via the region's VMS and partner navigation apps: "[Warning and Guidance] Parking spaces in the commercial area ahead are about to be full, and the roads are slippery due to rain, posing a high risk. We suggest you go to the nearby Parking Lot A (150 spaces remaining) and click on one-click navigation."

[0135] Example 2

[0136] A driver reserved a roadside parking space via a mobile app. En route, the driving risk assessment module, based on real-time data, determined that an intersection on their planned route was congested due to a sudden traffic accident, raising the risk level. The system immediately sent a warning to the driver via the app: "[Route Risk Warning] An accident has occurred at the XX intersection on your planned route, causing severe congestion. Risk level: High. A new route has been planned for you, expected to save 8 minutes. Would you like to switch?" Simultaneously, the system ensured the reserved parking space was held for the driver to avoid exceeding the time limit due to detours.

[0137] Examples 1 and 2 demonstrate that the solution proposed in this application effectively addresses the problems of fragmented urban parking management and passive traffic safety early warning. Through data-driven approaches and large-scale model empowerment, it achieves intelligent management of urban traffic resources and a significant improvement in safety levels.

[0138] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A city-level intelligent parking dynamic scheduling and driving risk early warning method, characterized in that, The method, executed by a city-level smart parking platform, includes: It acquires real-time vehicle behavior data and parking status information collected by sensing devices deployed in various urban areas, as well as weather data, real-time traffic flow data, and closed parking lot data from external data sources. By integrating the vehicle behavior data, parking status information, weather data, real-time traffic flow data, and closed parking lot data, a dynamic parking resource database is constructed and continuously updated. Based on the dynamic parking resource database, a pre-trained artificial intelligence model is used to dynamically assess regional parking pressure and predict the distribution of parking demand in each region during a specific future period. Based on real-time weather data, real-time traffic flow data, road topology data, and real-time parking pressure data obtained from the parking analysis and prediction steps in the dynamic parking resource database, the comprehensive driving risk level of different road sections or intersections is calculated through a multi-factor risk assessment model. Based on the predicted parking demand distribution, targeted parking guidance and scheduling strategies are generated; and based on the comprehensive driving risk level assessment results, high-risk road sections or intersections are identified. The parking guidance and scheduling strategy, as well as the driving risk warning information for the high-risk road sections or intersections, are sent to the vehicle's onboard terminal or the owner's mobile terminal via the communication network.

2. The method according to claim 1, characterized in that, The sensing device includes a high-position video camera and a geomagnetic sensor; the high-position video camera is used to identify license plate numbers, vehicle trajectories, and parking space occupancy status; the geomagnetic sensor is used to assist in detecting the occupancy status of parking spaces.

3. The method according to claim 1, characterized in that, The prediction of parking demand distribution in various regions within a specific future time period specifically includes: Historical parking data, real-time parking data, date type, surrounding points of interest data, and real-time traffic flow data are input into a deep learning model based on long short-term memory networks or temporal convolutional networks. The deep learning model outputs a probability map of parking demand for each grid area within a preset future time period.

4. The method according to claim 1, characterized in that, The calculation of the comprehensive driving risk level using a multi-factor risk assessment model specifically includes: Assign appropriate weighting factors to real-time weather data, real-time traffic flow data, historical accident data, road topology data, and real-time parking pressure data; Calculate the comprehensive risk score based on the aforementioned weighting factors; Based on preset risk score thresholds, risk levels are divided into low risk, medium risk, high risk, and extremely high risk.

5. The method according to claim 1, characterized in that, The generation of targeted parking guidance and scheduling strategies specifically includes: For areas predicted to experience high parking demand, a strategy is generated to divert traffic to surrounding parking lots or roadside parking areas with a higher proportion of vacant spaces, and an optimal navigation path is generated to the parking lots or roadside parking areas.

6. The method according to claim 1, characterized in that, In the strategy generation step, if a high-risk road segment or intersection is identified as being located on a vehicle's current navigation path, an alternative route is replanned for the vehicle to avoid the high-risk road segment or intersection, and the alternative route is sent out together with the driving risk warning information.

7. A city-level intelligent parking dynamic scheduling and driving risk early warning system, characterized in that, The system for implementing the method of any one of claims 1 to 6 comprises: The perception layer consists of high-position video cameras and geomagnetic sensors deployed in various areas of the city, used to collect raw perception data on vehicle behavior and parking space status. The data layer communicates with the perception layer and is configured with an interface to connect to external data sources. It is used to receive, clean, and fuse the original perception data, weather data, real-time traffic flow data, and closed parking lot data to build and maintain a unified dynamic parking resource database. A platform layer, communicatively connected to the data layer, comprising: The regional parking pressure dynamic assessment module is used to calculate and output regional parking pressure information based on the dynamic parking resource database; The parking demand prediction module has a built-in artificial intelligence-based prediction model, which is used to process the data in the dynamic parking resource database and output parking demand distribution prediction information for future time periods. The driving risk assessment module has a built-in multi-factor fusion risk assessment model, which is used to process the dynamic parking resource database and related data and output driving risk level information. The application layer, which communicates with the platform layer, is used to receive the parking demand distribution prediction information and the driving risk level information, generate publishable parking guidance and scheduling instructions and driving risk warning instructions, and output them to the outside world through the information publishing module.

8. The system according to claim 7, characterized in that, The parking demand prediction module specifically adopts a deep learning model architecture based on Long Short-Term Memory Network (LSTM) or Temporal Convolutional Network (TCN).

9. The system according to claim 7, characterized in that, The multi-factor fusion risk assessment model in the driving risk assessment module is configured to perform weighted calculations on the input real-time weather data, real-time traffic flow data, historical accident data, road topology data, and parking resource tension data from the regional parking pressure dynamic assessment module, so as to output a quantified risk level.

10. The system according to claim 7, characterized in that, The information publishing module in the application layer communicates with at least one of the following terminal forms: Variable Message Service (VMS), vehicle terminal, smartphone application, and mini-program, to push information.

Images