A driving rating method and system based on parking behavior node coupling

Abstract

The application discloses a driving evaluation method and system based on parking behavior node coupling, and belongs to the technical field of driving behavior evaluation. The method comprises the following steps: obtaining driver vehicle trajectory and parking data, performing time alignment and space correction, and obtaining a parking behavior data set. By using space clustering and time constraint, a parking behavior structure node set is generated, and a stability feature parameter is extracted. After receiving real-time trajectory data and task information, space matching and direction deviation calculation are performed, and a path deviation feature parameter is generated. Finally, the data is input into a risk evaluation model, and an evaluation grade of driving is output. The scheme fuses historical parking mode, stability and real-time trajectory deviation, outputs deviation risk value and driving grade, improves prediction accuracy and timeliness, facilitates task intervention, real-time warning and enhances interpretability.

Description

Technical Field

[0001] This application belongs to the field of driving behavior evaluation technology, specifically relating to a driving rating method and system based on parking behavior node coupling. Background Technology

[0002] With the rapid development of the logistics and transportation and ride-sharing industries, refined analysis of driver behavior has become a key requirement for improving operational efficiency and safety. By mining potential patterns in vehicle trajectory and parking behavior data, task scheduling can be optimized and safety risks reduced.

[0003] The current method first collects raw trajectory and parking data through in-vehicle equipment, samples at fixed time intervals and unifies the coordinate system; then, it independently clusters spatial location and timestamp to identify high-frequency parking areas and time periods; then, it matches real-time trajectory with preset path templates to calculate spatial deviation distance; finally, it compares statistical features with preset thresholds to generate risk levels.

[0004] However, the spatiotemporal correction of existing technologies relies on globally unified parameters, which makes it difficult to adapt to the differences in road network characteristics in different regions; and the separation of spatial and temporal clustering leads to the fragmentation of parking behavior structure nodes; at the same time, the evaluation method of static threshold comparison cannot capture the interaction between dynamic path deviation and driving habits, and the evaluation results are lagging and have low sensitivity to abnormal behavior. Summary of the Invention

[0005] To overcome the above-mentioned shortcomings, this invention is proposed to provide solutions or at least partially solve the technical problems of existing spatiotemporal correction relying on globally unified parameters, which makes it difficult to adapt to the differences in road network characteristics in different regions; the fragmentation of parking behavior structure nodes due to the separation of spatial and temporal clustering; and the inability of static threshold comparison evaluation method to capture the interaction between dynamic path deviation and driving habits, resulting in lagging evaluation results and low sensitivity to abnormal behavior.

[0006] In a first aspect, the present invention provides a driving rating method based on parking behavior node coupling, the method comprising: The system acquires vehicle trajectory data and raw parking data of the driver within a preset period. Based on the vehicle trajectory data, the raw parking data is time-aligned and spatially corrected to obtain a parking behavior dataset. Based on preset spatial clustering rules and preset temporal continuity constraints, a dual-constraint clustering analysis is performed on the parking behavior dataset to obtain a set of parking behavior structure nodes. The structural features of each parking behavior structure node in the parking behavior structure node set are independently extracted and stability analysis is performed to generate stable feature parameters for each parking behavior structure node. If the real-time trajectory data and task information within a preset time window are received from the mobile terminal of the vehicle where the driver is located, spatial matching and directional deviation calculation are performed on the real-time trajectory data based on the parking behavior structure node set and task information to generate path deviation feature parameters corresponding to the real-time trajectory data. The task information, stability feature parameters, and path deviation feature parameters are input into a preset risk assessment model for feature fusion calculation, and a deviation risk assessment value is output. Based on the deviation risk assessment value, the driver's driving evaluation level is generated.

[0007] In a second aspect, the present invention provides a driving rating system based on parking behavior node coupling, the system comprising: The parking behavior collection module is used to acquire vehicle trajectory data and raw parking data of drivers within a preset period. Based on the vehicle trajectory data, the raw parking data is time-aligned and spatially corrected to obtain a parking behavior dataset. The dual-constraint clustering analysis module is used to perform dual-constraint clustering analysis on the parking behavior dataset based on preset spatial clustering rules and preset temporal continuity constraints to obtain a set of parking behavior structure nodes. The stability analysis module is used to independently extract the structural features of each parking behavior structure node in the parking behavior structure node set and perform stability analysis to generate stable feature parameters for each parking behavior structure node. The path deviation feature analysis module is used to perform spatial matching and directional deviation calculation on the real-time trajectory data and task information within a preset time window transmitted by the mobile terminal of the driver's vehicle, based on the parking behavior structure node set and task information, to generate path deviation feature parameters corresponding to the real-time trajectory data. The driving evaluation module is used to input the task information, stability feature parameters, and path deviation feature parameters into a preset risk evaluation model for feature fusion calculation, output deviation risk evaluation value, and generate the driver's driving evaluation level based on the deviation risk evaluation value.

[0008] In a third aspect, an electronic device is provided, comprising a processor, a memory, and a program or instructions stored in the memory and executable on the processor, the program or instructions being loaded and run by the processor to perform the steps of the aforementioned driving rating method based on parking behavior node coupling.

[0009] In a fourth aspect, a computer-readable storage medium is provided, wherein a plurality of program codes are stored therein, the program codes being adapted to be loaded and run by a processor to perform the steps of the driving rating method based on parking behavior node coupling described above.

[0010] The above-described technical solutions of the present invention have at least one or more of the following beneficial effects: In implementing the technical solution of this invention, by fusing historical parking patterns, stability and real-time trajectory deviation, deviation risk values ​​and driving levels are output, improving prediction accuracy and timeliness, facilitating task-oriented intervention, real-time alarms and enhancing interpretability. Attached Figure Description

[0011] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Furthermore, similar numbers in the drawings are used to denote similar components, wherein: Figure 1 This is a schematic diagram of the first main steps of a driving rating method based on parking behavior node coupling according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the second main step of a driving rating method based on parking behavior node coupling according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the main structure of a driving rating system based on parking behavior node coupling according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0012] Some embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0013] In the description of this invention, "module" and "processor" can include hardware, software, or a combination of both. A module can include hardware circuitry, various suitable sensors, communication ports, memory, and may also include software components, such as program code, or a combination of software and hardware. A processor can be a central processing unit, microprocessor, image processor, digital signal processor, or any other suitable processor. The processor has data and / or signal processing capabilities. The processor can be implemented in software, in hardware, or a combination of both. Non-transitory computer-readable storage media include any suitable medium capable of storing program code, such as magnetic disks, hard disks, optical disks, flash memory, read-only memory, random access memory, etc. The term "A and / or B" means all possible combinations of A and B, such as only A, only B, or A and B. The terms "at least one A or B" or "at least one of A and B" have a similar meaning to "A and / or B" and can include only A, only B, or A and B. The singular terms "a" or "this" can also include plural forms.

[0014] See appendix Figure 1 , Figure 1 This is a schematic flowchart of the first main steps of a driving rating method based on parking behavior node coupling according to an embodiment of the present invention. Figure 1 As shown, a driving rating method based on parking behavior node coupling in an embodiment of the present invention mainly includes the following steps S101-S105.

[0015] Step S101: Obtain the driver's vehicle trajectory data and original parking data within a preset period. Based on the vehicle trajectory data, perform time alignment and spatial location correction on the original parking data to obtain a parking behavior dataset.

[0016] A preset period is a pre-defined time frame for data collection and analysis. The duration can be set to a fixed interval such as one week or one month, clearly defining the boundaries of data collection within the corresponding time period.

[0017] Vehicle trajectory data is information related to vehicle location and movement collected through GPS, sensors, or other positioning systems, including real-time vehicle location, timestamp, speed, acceleration, etc.

[0018] Raw parking data is the basic information generated during the parking process, including parking time, parking location, parking status, and parking duration.

[0019] The parking behavior dataset is a multi-dimensional collection of parking behavior information based on vehicle trajectory data and raw parking data, after time alignment and spatial location correction. It fully records detailed information for each parking session, including parking location, parking time, parking type, and parking duration.

[0020] Within a preset period, relying on the vehicle's onboard GPS positioning device, information such as the vehicle's latitude and longitude, speed, and direction of travel is collected in real time, continuously recorded at a frequency of once per second, gradually forming complete vehicle trajectory data. Raw parking data is collected through the vehicle's built-in parking sensors or parking lot equipment, focusing on recording the start and end times of parking events, the latitude and longitude coordinates of the parking location, and other information. This type of data collection is mostly achieved through interaction between the vehicle and the parking lot system, thereby obtaining the precise coordinates of the parking location and the actual parking duration. After data collection is completed, the timestamps of the vehicle trajectory data and the raw parking data need to be aligned to ensure that the start and end times of each parking action accurately match the trajectory data. For example, if a parking event starts at 08:30:00 and ends at 08:35:00, trajectory data within that time period is selected for matching, ensuring that the vehicle's status and trajectory information remain consistent during the parking period.

[0021] Spatial location correction methods are employed to process the data and address the bias issues caused by GPS positioning errors. For example, if there is a slight deviation between the actual parking location and the recorded trajectory location, the geographical distance between the two can be calculated, such as using Euclidean distance or spherical distance formulas, to correct the trajectory points and ensure that the vehicle trajectory matches the actual parking location. During the correction process, the position of the trajectory data can be adjusted using interpolation algorithms, or further corrections can be made by combining high-precision maps. After completing time alignment and spatial correction, the final dataset of parking behavior is generated.

[0022] Step S102: Based on preset spatial clustering rules and preset temporal continuity constraints, perform dual-constraint clustering analysis on the parking behavior dataset to obtain a parking behavior structure node set.

[0023] Preset spatial clustering rules are classification rules for similar parking locations based on their geographical location associations. Using geographical coordinates such as latitude and longitude as a basis, specific distance thresholds are set as judgment criteria. These thresholds are then used to determine whether different parking locations are sufficiently close, classifying them into the same group or cluster. For example, two parking locations with a spatial distance of less than 100 meters are classified into the same parking behavior cluster.

[0024] The preset time continuity constraint is a time-level continuity requirement set for parking behavior. It means that parking events must remain continuous within a specified time frame, and the time interval between two parking events cannot exceed a preset threshold. This condition is set to ensure the continuity of parking behavior and avoid incorrectly classifying unrelated or discontinuous parking events as the same behavior. For example, if the interval between two parking events exceeds 30 minutes, they will be considered two independent parking events.

[0025] The parking behavior structure node set is a group of parking behavior nodes obtained after performing spatial clustering and temporal continuity constraint analysis on parking data. Each node corresponds to a parking behavior aggregate, which contains all the feature information of that parking behavior, including parking location, parking time, parking duration, etc.

[0026] Based on preset spatial clustering rules, preliminary spatial partitioning of parking behavior data is performed. Specifically, a distance threshold, such as 100 meters, is first set, and then clustering is performed on the parking behavior data. For each parking record, its geographical distance to other parking points is calculated. If the distance between the parking point and an existing cluster center is less than the set threshold, the parking record is assigned to that node; if the distance is greater than the threshold, a new cluster node is created. After spatial clustering, parking behaviors are further filtered and organized according to preset time continuity constraints. The goal is to ensure the continuity of parking behaviors and ensure that the same parking event falls within a reasonable time range. For example, a 30-minute time threshold can be set. If the time interval between two parking behaviors exceeds this threshold, they are determined not to belong to the same parking behavior. With this constraint, parking records with excessively long time intervals can be filtered out, and behaviors belonging to the same parking event can be aggregated together. Specifically, the parking records need to be sorted according to the order of parking time, and then the time difference between each parking record and the previous parking record is calculated. If the time difference is greater than a preset threshold, they are determined to be different parking behavior nodes; otherwise, the two records are merged into one parking behavior node.

[0027] After spatial partitioning using preset spatial clustering rules and time filtering using preset temporal continuity constraints, parking behavior data is ultimately divided into multiple independent parking behavior structure nodes, which together constitute a parking behavior structure node set. Each node corresponds to a type of aggregated parking behavior pattern, containing detailed information about multiple parking events, such as parking location, parking time period, parking duration, and parking behavior patterns. For example, parking records generated by drivers at locations such as company entrances, residential areas, and shopping malls during the past month's work can be compiled into a parking behavior dataset. Combining spatial clustering rules, parking points within 100 meters are grouped into the same area. Based on temporal continuity constraints, parking consecutively or at intervals not exceeding 3 hours is considered the same behavior, thereby generating a parking behavior structure node set, such as the 7:00-9:00 AM node at the company entrance and the 12:00-2:00 PM node at the shopping mall.

[0028] Based on the above technical solution, optionally, a dual-constraint clustering analysis is performed on the parking behavior dataset based on preset spatial clustering rules and preset temporal continuity constraints to obtain a parking behavior structure node set, including: The spatial distance data between each parking point in the parking behavior dataset is calculated based on the preset spatial clustering rules, and a spatial distance matrix is ​​constructed based on the spatial distance data. Based on the preset spatial clustering rules, the spatial distance matrix is ​​filtered by threshold constraints to obtain a spatial adjacency relationship set. Based on the spatial adjacency relationship set, a spatial connectivity relationship graph that satisfies the preset spatial clustering rules is established. Spatial grouping is performed on the interconnected parking points in the spatial connectivity graph to form an initial set of spatial clustering units; Based on the initial spatial clustering unit set, extract the timestamp sequence of parking records within each clustering unit; Based on the preset time continuity constraint, the timestamp sequence of parking records within each initial spatial clustering unit is analyzed for continuity to obtain the continuity analysis results. Based on the continuity analysis results, the timestamp sequence is divided into breakpoints to form continuous time sub-units; Extract the spatial and temporal distribution features of each temporally continuous sub-unit, and construct a structural feature vector based on the spatial and temporal distribution features; Based on preset stability judgment rules and structural feature vectors, stability assessment and screening are performed on each time continuous sub-unit to form a set of parking behavior structural nodes.

[0029] In this scheme, a parking spot is the specific location where a vehicle is parked when parking occurs. It is represented by geographical coordinates such as latitude and longitude, and can also be supplemented with parking lot information, parking space information, or area markers.

[0030] Spatial distance data is a quantified value of the geometric distance between any two parking events within the same time window, which can reflect the difference in actual physical distance.

[0031] A spatial distance matrix is ​​a matrix that organizes the spatial distances between any two parking points into rows and columns. The matrix elements represent the spatial distances between parking points.

[0032] The spatial adjacency set is a record of all parking point pairs that form an adjacency relationship under a set threshold constraint, such as (i, j), which is used to represent the direct spatial association between parking points.

[0033] A spatial connectivity graph is constructed by using parking points as vertices and the adjacency relationships between parking points as edges. It can intuitively reflect the connectivity between parking points under given spatial constraints.

[0034] The initial spatial clustering unit set is formed by initially dividing the set of interconnected parking points based on the spatial connectivity graph. Each unit is a set of parking points that are spatially close to each other and interconnected in the graph.

[0035] A clustering unit is a specific collection of individual spatial clustering units, containing several parking points and the association information of each parking point within the unit, such as the unit center location, radius size, and parking point density.

[0036] Internal parking records are a collection of all parking events that occur within a specific cluster unit, including raw information such as parking time, parking location, and duration of stay.

[0037] A timestamp sequence is a sequence of time points formed by arranging all parking records within a cluster unit in chronological order. It records the start time, end time, or other key time points of parking.

[0038] Continuity analysis results are quantitative conclusions drawn from the analysis of the time-series coherence, interval regularity, and recurrence characteristics. Examples include whether time continuity meets a set threshold and the distribution of time intervals.

[0039] The time-continuous sub-unit is based on the analysis of time continuity, which divides the timestamp sequence into several sub-sequences according to the breakpoints. Each sub-sequence has a high degree of temporal continuity.

[0040] Spatial distribution characteristics refer to the geographical distribution features of parking spots within each continuous time unit, including the central location, distribution radius, point density, and cluster shape of the parking spots.

[0041] Temporal distribution characteristics refer to the temporal distribution of parking events within each continuous time unit. Examples include the mean and variance of parking duration, the daily distribution of parking events, and peak parking periods.

[0042] A structural feature vector integrates the spatial distribution characteristics, temporal distribution characteristics, and other multi-dimensional elements of a time-continuous subunit into a vector form, which is used to describe the comprehensive structural characteristics of that subunit.

[0043] Preset stability determination rules are a series of rules used to determine whether a time-continuous sub-unit has reached a stable state. They are usually formulated by combining thresholds, trends and statistical indicators such as spatial clustering, temporal regularity and residence consistency.

[0044] To process parking behavior datasets based on predefined spatial clustering rules, the first step is to calculate the spatial distances between parking points. The system first extracts the geographic coordinates of each parking point from the dataset. Then, using Geographic Information System (GIS) technology or simple mathematical methods like the Euclidean distance formula, it calculates the spatial distance between any two parking points. This distance is typically the straight-line distance between the two points. The calculation can be performed using different coordinate systems, such as geographic coordinates or projected coordinates, to ensure accuracy. After obtaining the distances between all parking points, this data is organized into a spatial distance matrix, where each element represents the spatial distance between two parking points. The spatial distance matrix presents the relative positional relationships between the parking points.

[0045] The generated spatial distance matrix is ​​then processed according to preset spatial clustering rules. Specifically, a threshold constraint filtering method is used to select point pairs that meet the rules from the matrix, that is, parking point pairs whose spatial distance is below the set threshold. The purpose is to filter out parking points that are too far apart and only retain point pairs that are spatially close, which have the potential to be grouped into the same spatial cluster in subsequent clustering. After the filtering is completed, a spatial adjacency relationship set is formed, where each element represents a pair of adjacent or connected parking points, clearly reflecting the spatial relationship between parking points. Based on this spatial adjacency relationship set, the system further constructs a spatial connectivity graph, with parking points as nodes in the graph and adjacency relationships as edges connecting nodes.

[0046] After the spatial connectivity graph is constructed, the system spatially groups the parking points in the graph. Grouping is based on the connectivity of the nodes in the graph, assigning all interconnected parking points to the same group, thus forming an initial set of spatial clustering units. Each clustering unit is a set of interconnected parking points, meaning these parking points are spatially close and may correspond to the same type of parking behavior pattern. This step is implemented using a graph connectivity algorithm, such as depth-first search or breadth-first search, traversing the nodes and edges in the graph to extract all connected subgraphs; each connected subgraph is an independent clustering unit.

[0047] Based on the initial spatial clustering unit set, the timestamp sequence of parking records within each clustering unit is extracted. Internal parking records typically include parking start and end times, parking duration, etc., while the timestamp records the moment the parking event occurred. By sorting and organizing the parking point data within each clustering unit, the timestamp sequence corresponding to that clustering unit can be obtained. After obtaining the timestamp sequence of each clustering unit, the system performs analysis based on preset time continuity constraints. Specifically, a time interval threshold is set to determine whether the time interval between adjacent parking events meets the preset standard. For example, if the time interval between two adjacent parking events is less than the set threshold, they are considered to belong to a continuous phase of the same parking behavior; if the time interval is too long, it means that the parking behavior may be interrupted or discontinuous. Then, the continuity analysis results are obtained, clarifying whether each timestamp sequence meets the preset time continuity requirements.

[0048] The system then divides the timestamp sequence into time-continuous sub-units based on the continuity analysis results. This step identifies breakpoints or large time intervals in the timestamp sequence, splitting the time period corresponding to each cluster unit. Each time-continuous sub-unit represents a coherent stage of parking behavior in the time dimension. These time-continuous sub-units further help staff analyze the temporal distribution patterns and characteristics of parking behavior. After completing the division into time-continuous sub-units, the system extracts the spatial and temporal distribution features corresponding to each sub-unit. Integrating these features constructs a structural feature vector, which combines the characteristics of parking behavior in both spatial and temporal dimensions, enabling a quantitative description of the attributes of parking behavior.

[0049] Based on preset stability criteria, the system first evaluates the structural feature vector of each continuous time unit, focusing on consistency, stability, and repeatability. Consistency analysis primarily assesses the stability of parking behavior by calculating its spatial and temporal distribution, such as the magnitude of fluctuations in parking location and duration. Small fluctuations in parking location and minimal changes in parking duration indicate high consistency. Stability analysis utilizes time series autocorrelation to examine the temporal patterns of parking behavior. A high stability score is achieved when parking time intervals remain relatively constant and conform to expected behavior patterns. Repeatability analysis uses clustering algorithms to detect the repetition of parking behaviors. Repeated occurrences of the same type of parking behavior across different time periods or conditions indicate strong repeatability. The results of consistency, stability, and repeatability analyses are then weighted to calculate a comprehensive stability score; a higher score indicates better stability. Finally, the system sets a threshold to filter all parking behavior patterns. Patterns below the threshold are removed, while those meeting the threshold are included in the parking behavior structure node set.

[0050] In this scheme, representative and stable parking behavior patterns are effectively identified and extracted through spatial clustering, temporal continuity analysis, and stability assessment.

[0051] Step S103: Independently extract and analyze the structural features of each parking behavior structure node in the parking behavior structure node set to generate stable feature parameters for each parking behavior structure node.

[0052] A parking behavior structure node is a unit within a set of parking behavior structure nodes, corresponding to a driver's typical parking pattern within a specific geographical location and time period. For example, the node at the company entrance from 7:00 AM to 9:00 AM describes the driver's parking behavior characteristics within a certain location and time range.

[0053] Stable characteristic parameters are quantitative indicators obtained by analyzing the fluctuations of each parking behavior structure node in terms of spatial location, dwell time, parking frequency, etc. They describe the consistency and regularity of parking behavior at that node, such as the average dwell time, dwell time variance, and daily dwell frequency stability of a certain node.

[0054] For each node in the parking behavior structure node set, all parking event data corresponding to that node is extracted, specifically including the geographical location of each parking, start and end times of stay, stay duration, corresponding task type, road type, and traffic condition information. This data is organized chronologically, and the specific location of each parking event is marked on a map coordinate system. The center coordinates of all parking points are calculated, and the distance from each parking event to the center is measured. The minimum distance, maximum distance, average distance, and standard deviation are calculated to form a spatial concentration index for the node, intuitively reflecting the stability of the driver's parking position at that node. The stay duration of each parking is statistically analyzed, arranged chronologically, and the average, variance, and consecutive stay differences are calculated. The time intervals between adjacent parking events and their regularity are also recorded to form a time distribution index, reflecting the temporal stability of the driver's parking behavior at that node. The daily or weekly distribution of parking frequency, number of stay events, and stay duration is summarized to form a node behavior intensity index, reflecting the activity level of parking behavior at that node.

[0055] After completing the statistical analysis of spatial and temporal indicators, the spatial concentration index, temporal distribution index, and behavioral intensity index are integrated to form a multidimensional feature representation of the nodes. Then, the changing trends and fluctuation amplitudes of each indicator over a historical period are calculated. By comparing the historical average with the current data, any abnormal fluctuations or behavioral deviations are identified. The multidimensional feature representation is standardized and normalized to ensure that all indicators are on a uniform scale, allowing for direct comparative analysis and thus obtaining stable feature parameters that characterize the behavioral patterns of the nodes.

[0056] Based on the above technical solution, optionally, the structural features of each parking behavior structure node in the parking behavior structure node set are independently extracted and stability analysis is performed to generate stable feature parameters for each parking behavior structure node, including: Extract the spatial distribution parameters, temporal distribution parameters, and behavioral intensity parameters of each parking behavior structure node in the parking behavior structure node set, and construct a multidimensional structural parameter set for each parking behavior structure node based on the spatial distribution parameters, temporal distribution parameters, and behavioral intensity parameters; Based on the multidimensional structural parameter set, a parameter temporal evolution sequence is constructed. The parameter temporal evolution sequence is differentially calculated to obtain the numerical change between adjacent time points. Based on the numerical change, a parameter change gradient sequence for each parking behavior structural node is constructed. Change point detection is performed on the time-series evolution sequence of the parameters to determine the structural change nodes, and piecewise linear fitting is performed based on the structural change nodes to obtain the piecewise trend parameter set of each parking behavior structural node; Obtain the trend slope and segment interval length of each segment in the segmented trend parameter set, and calculate the comprehensive trend slope of each parking behavior structure node by performing a weighted average based on the trend slope and segment interval length. Based on the comprehensive trend slope and parameter change gradient sequence, the temporal fluctuation amplitude data and trend consistency index of each parking behavior structure node are calculated. Based on the time-series fluctuation amplitude data and trend consistency index, the time-series stability index of each parking behavior structure node is obtained by fusion calculation. Based on the time-series stability index and the preset stability judgment threshold, the stability state results of each parking behavior structure node are generated, and the time-series stability index and stability state results are structured, stored and uniformly represented to obtain the stability feature parameters of each parking behavior structure node.

[0057] In this scheme, spatial distribution parameters are quantitative indicators used to describe the spatial distribution characteristics of parking behavior structural nodes, including parking point location distribution, clustering degree, dispersion characteristics, etc., which can intuitively reflect the dense or sparse state of parking behavior in geographic space.

[0058] The time distribution parameter is a quantitative indicator used to describe the time distribution characteristics of parking behavior structural nodes. It includes elements such as parking time interval, occurrence frequency, and time pattern, and is mainly used to assess the periodicity and continuity of parking behavior.

[0059] Behavioral intensity parameters are parameters that reflect the intensity and frequency of parking behavior, including indicators such as parking duration and parking frequency, and measure the significance of parking behavior within a specific time and space range.

[0060] A multidimensional structural parameter set is a complete and systematic set of structural parameters that integrates multidimensional features such as spatial distribution parameters, temporal distribution parameters, and behavioral intensity parameters.

[0061] The parameter temporal evolution sequence is a sequence of data that records the dynamic changes of multidimensional structural parameters of parking behavior structure nodes over time. It can reflect the evolution process of nodes in the time dimension and includes temporal variation information of various parameters such as spatial distribution, temporal distribution, and behavioral intensity.

[0062] Numerical change is the difference between two adjacent time points in the parameter time series evolution sequence, which can be used to evaluate the specific magnitude of change of each parameter in the time dimension.

[0063] The parameter change gradient sequence is a sequence of data calculated based on numerical changes. It is used to characterize the rate of change of parking behavior structural nodes in different time periods and can reflect the speed and trend of parameter changes over time and the overall trend.

[0064] Structural change nodes are key nodes identified in the time-series evolution of parameters, marking significant changes or trend reversals in parking behavior structure, corresponding to time points with large parameter changes.

[0065] The segmented trend parameter set is a set of trend parameters for each segment obtained by performing segmented linear fitting on the parameter time series evolution sequence based on structural change nodes. It includes the slope and intercept of each segment of data and can reflect the changing trend of parking behavior in different time periods.

[0066] Trend slope is a core metric in the segmented trend parameter set. It is used to represent the rate of change of parking behavior over a certain period of time and can measure the upward or downward trend of the behavior.

[0067] The segmented interval length is the time length or data span corresponding to each data segment in the piecewise linear fitting process, which can reflect the persistence of a specific behavioral pattern in the time dimension.

[0068] The overall trend slope is a comprehensive value obtained by weighting the trend slope of each segment with the length of the segment interval, reflecting the changing trend of parking behavior over the overall time range.

[0069] Time-series fluctuation amplitude data describes the magnitude of fluctuations in parking behavior structural nodes over time. It is expressed using indicators such as the standard deviation and range of the data sequence, reflecting the magnitude of changes in parking behavior.

[0070] Trend consistency index is used to evaluate the stability and consistency of parking behavior over time. It is usually determined based on the degree of consistency of the trend slope or the change in the fluctuation range of each segment.

[0071] The temporal stability index is a comprehensive index obtained by integrating temporal fluctuation amplitude data and trend consistency index, which evaluates the stability of parking behavior structure nodes in the time dimension.

[0072] The preset stability threshold is a standard threshold set based on domain experience and professional knowledge during the stability assessment process. It is used to distinguish between stable and unstable states of parking behavior. If the temporal stability index exceeds this threshold, the parking behavior is determined to be in a stable state; otherwise, it is considered to be in an unstable state.

[0073] The stable state result is the result of the stability judgment of the parking behavior structure node after comparing the time series stability index with the preset stability judgment threshold. It is divided into two categories: "stable" and "unstable", which can reflect the persistence and consistency of the node in the time series.

[0074] Each parking behavior structure node in the parking behavior dataset is analyzed individually. For a single parking spot, the spatial distance between parking spots is calculated using Euclidean distance or other spatial distance metrics to clarify their distribution in geographic space and extract spatial distribution parameters. Based on the timestamp information of the parking spots, the duration of a single parking session and the frequency of parking stops are calculated to generate temporal distribution parameters. In the extraction of behavior intensity parameters, parking intensity is first defined as a comprehensive index of parking frequency and parking duration. When calculating parking frequency, the number of times each parking spot appears within a set time window is counted, while also considering the frequency differences in different time periods, such as weekdays versus weekends, and morning / evening peak hours versus off-peak hours. Based on the parking record timestamps, the duration of a single parking session is calculated to further derive the average parking duration for each parking spot. Using weighted calculation or simple multiplication, parking frequency and parking duration are merged to obtain preliminary parking intensity data. To eliminate the dimensional differences between parking frequency and parking duration and improve data comparability, the initial intensity data needs to be normalized or standardized. The resulting parking intensity index can clearly reflect the activity and importance of each parking location. By integrating spatial distribution parameters, temporal distribution parameters, and behavioral intensity parameters, a multi-dimensional structural parameter set for each parking behavior structural node is generated.

[0075] Based on a multidimensional structural parameter set, a parameter temporal evolution sequence is constructed. For each parking behavior structural node, the changes in its spatial distribution parameters, temporal distribution parameters, and behavior intensity parameters are tracked in chronological order. The parameter values ​​at each time point are recorded as multidimensional vectors, and all vectors are sorted by time to form the parameter temporal evolution sequence. This sequence can effectively capture the dynamic changes in parking behavior, providing data support for analyzing the temporal evolution patterns of parking behavior. After obtaining the parameter temporal evolution sequence, differential calculations are performed to solve for the numerical changes between adjacent time points, thereby clarifying the rate of change and fluctuation of each parking behavior parameter in the time dimension. Through differential processing, a parameter change gradient sequence is further generated, which intuitively reflects the fluctuation characteristics of parking behavior, such as the increase or decrease in parking duration and the fluctuation of parking frequency. Change point detection is performed on the parameter temporal evolution sequence, which is a key step in identifying changes in parking behavior structural nodes. By employing a statistically driven change point detection algorithm, the fluctuation characteristics of time-series data are analyzed to pinpoint time points where data fluctuations are significant or trends change markedly. These time points are termed structural change nodes, effectively identifying turning points and anomalies in parking behavior over time. Using these structural change nodes as segmentation criteria, piecewise linear fitting is performed on the parameter time-series evolution sequence. Each data segment corresponds to a continuous trend in parking behavior. Methods such as linear regression are used to fit the parameter variation patterns within each time period, generating a segmented trend parameter set. The trend slope in this parameter set characterizes the rate of change of parking behavior within the corresponding time period, while the intercept reflects the initial state of the behavior, presenting the changing trajectory of parking behavior across different time periods.

[0076] Two core indicators, trend slope and segment interval length, are extracted from the segmented trend parameter set. The trend slope measures the rate of change of parking behavior within a specific time period; a larger slope value indicates more drastic changes in parking behavior within that time period, while a smaller slope value indicates a relatively stable behavior. The segment interval length represents the duration of each fitted segment, reflecting the time stability of the corresponding behavior pattern. Based on the trend slope and segment interval length of each segment, a weighted average is calculated to obtain the comprehensive trend slope. In the weighted calculation process, different weights are assigned according to the segment interval length; the longer the time period, the higher its contribution to the overall trend.

[0077] Based on the gradient sequence of parameter changes, the temporal fluctuation amplitude data of parking behavior within each time period is calculated. Parking behavior data is segmented along the time dimension, and the amplitude of change in parking behavior within each segment is evaluated. If the parking intensity fluctuates significantly within a certain time period, the corresponding fluctuation amplitude value is high, indicating drastic changes in parking behavior during that period; conversely, it indicates a stable behavior state. The temporal fluctuation amplitude data allows for the quantification of the temporal instability and degree of change in parking behavior, laying the foundation for subsequent stability analysis. Based on the changing characteristics of parking behavior in each time period, a trend consistency index is calculated to measure whether the changing trend of parking behavior is consistent across different time periods. First, the overall direction of change in parking behavior in each time period is determined, i.e., the trend of increase or decrease in behavior; then, the direction of change in different time periods is compared. If the direction of change is consistent across multiple time periods, it indicates strong trend consistency in parking behavior; if the direction of change differs significantly, it indicates obvious behavioral fluctuation characteristics. The trend consistency index reflects the regularity and persistence of parking behavior; the higher the index value, the more stable the behavior state. The temporal fluctuation amplitude data and the trend consistency index are then fused to generate a temporal stability index. This indicator comprehensively considers the volatility and trend consistency of parking behavior, accurately assessing the stability of parking behavior structure nodes throughout the time series. The temporal stability index is compared with a preset stability threshold to determine the stable state of each parking behavior structure node. If the temporal stability index is higher than the threshold, the node is considered stable; otherwise, it is considered unstable. The stable state results and the temporal stability index are then stored in a structured manner to ultimately form the stable characteristic parameters of the parking behavior structure nodes.

[0078] This solution can provide effective evidence for parking management, behavior prediction, and anomaly detection, thereby improving decision-making accuracy and system stability.

[0079] Step S104: If the real-time trajectory data and task information within a preset time window are received from the mobile terminal of the vehicle where the driver is located, spatial matching and directional deviation calculation are performed on the real-time trajectory data based on the parking behavior structure node set and task information to generate path deviation feature parameters corresponding to the real-time trajectory data.

[0080] Mobile terminals are in-vehicle or personal electronic devices used by drivers. They are mainly used to collect vehicle status data, GPS location information, and driving trajectory, and can also transmit data with the system server.

[0081] The preset time window is a fixed time period set in advance by the system. Its function is to collect and process trajectory data in real time. For example, a window can be set every 5 minutes or 10 minutes, so that the path deviation and risk assessment results can be dynamically updated within a continuous time period.

[0082] Real-time trajectory data is the continuous position and motion information of a vehicle within a preset time window, including latitude and longitude coordinates, timestamps, vehicle speed, direction angle, etc., which can reflect the current driving trajectory of the vehicle.

[0083] Task information is data related to the orders or tasks currently being executed by the driver, including the origin and destination locations, the expected route, the task type, and the time requirements.

[0084] Path deviation feature parameters are quantitative indicators calculated from real-time trajectory data and parking behavior structure node sets. They are used to represent the deviation of the vehicle's actual trajectory from the target parking node and the preset route, including positional differences, differences between the driving direction and the target direction, etc.

[0085] Upon receiving real-time trajectory data and task information within a preset time window from the mobile terminal of the driver's vehicle, the first step is to perform temporal and spatial preprocessing on the received real-time trajectory data. Specifically, the timestamp of each point in the trajectory data is aligned one-to-one with the expected route time in the task information, correcting time deviations caused by GPS signal delays or noise. Simultaneously, the geographic coordinates of the trajectory data are smoothed and filtered to eliminate short-term jitter and abnormal jumps, ensuring more accurate and continuous positioning of each trajectory point. After preprocessing, combined with the task route and target parking point specified in the task information, one or more target parking behavior structure nodes most relevant to the current task are selected from the set of parking behavior structure nodes. Specifically, the spatial relationship between trajectory points and parking nodes, as well as the spatial distribution of the task route, needs to be compared, selecting the node closest to the task endpoint or key points along the route as the matching target. At the same time, the historical dwell patterns and stable characteristic parameters of the nodes are considered to ensure that the selected target nodes reflect the driver's frequently used or expected parking behavior areas.

[0086] The preprocessed real-time trajectory data is spatially matched with the selected target parking behavior structure nodes. Specifically, for each trajectory point in the real-time trajectory data, its Euclidean distance or perpendicular distance to the target node is calculated and compared with the reference position of the target node to obtain the spatial matching result for each trajectory point. The driving direction angle of each trajectory point is calculated based on the line vector connecting adjacent trajectory points, and the angle difference between these angles and the preset angle of the corresponding target node is calculated to obtain the directional deviation value for each trajectory point. This step captures both the degree to which the vehicle deviates from the target position and whether the vehicle's driving direction is consistent with the expected parking behavior. The calculation results of spatial deviation and directional deviation are integrated to generate path deviation feature parameters corresponding to the real-time trajectory data. These feature parameters typically include the spatial deviation distance, directional deviation angle, deviation trend, and overall consistency index of the trajectory matching with the target node for each trajectory point, forming a quantifiable parameter set. For example, within the time window of 08:00–08:05, the driver is performing the task of delivering goods from warehouse A to shopping mall B and parking and unloading at shopping mall B; the real-time trajectory data is also collected synchronously during this period. In the parking structure node set, shopping mall B has been marked as the target parking node C, with its coordinates approximately known, and this node has a high historical cluster density. After spatial matching with node C based on task information, it was found that the geometric distance from the current trajectory point to node C is 25 meters. Using the line vectors connecting adjacent trajectory points, the current vehicle's driving direction was calculated, which deviates by 12° from the preset parking direction of node C. Simultaneously, the spatial deviation of the current trajectory point relative to node C is 25 meters. These spatial and directional deviations are summarized and integrated to form the corresponding path deviation feature parameters.

[0087] Based on the above technical solution, optionally, spatial matching and directional deviation calculation are performed on the real-time trajectory data based on the parking behavior structure node set and task information to generate path deviation feature parameters corresponding to the real-time trajectory data, including: Based on task information and parking behavior structure node set, at least one target parking behavior structure node is determined, and spatial matching calculation is performed based on real-time trajectory data and target parking behavior structure node to obtain the spatial matching result between each trajectory point in real-time trajectory data and target parking behavior structure node. The driving direction angle of each trajectory point is generated based on the driving direction vector of each trajectory point by calculating the line vector connecting adjacent trajectory points based on the line vector connecting adjacent trajectory points. Calculate the angle difference between the driving direction angle of each trajectory point and the preset direction angle corresponding to the target parking behavior structure node to obtain the directional deviation value between each trajectory point and the target parking behavior structure node; Based on real-time trajectory data, determine the spatial location data of each trajectory point, calculate the geometric distance between the spatial location data of each trajectory point and the reference location data corresponding to the target parking behavior structure node, and obtain the spatial deviation value between each trajectory point and the target parking behavior structure node. Based on the directional deviation value and the spatial deviation value, path deviation feature parameters corresponding to the real-time trajectory data are generated.

[0088] In this scheme, the target parking behavior structure node is a parking node in the parking behavior dataset that meets the specific task requirements. It is usually a known parking target location, such as a parking spot or parking lot.

[0089] A trajectory point is a location point recorded based on real-time trajectory data during the vehicle's movement. It includes geographical coordinates and a timestamp, representing the vehicle's specific location at a certain moment.

[0090] The spatial matching result is the spatial matching calculation result between each trajectory point and the target parking behavior structure node. It is evaluated by measuring the spatial distance and relative position from the trajectory point to the target parking node, so as to determine the degree of matching between the vehicle's current position and the target parking position.

[0091] The vector connecting adjacent trajectory points is a vector between two adjacent trajectory points in the trajectory data. It is calculated from the spatial coordinate difference between these two trajectory points and is used to represent the vehicle's driving direction.

[0092] The driving direction angle is calculated based on the vector of the line connecting adjacent trajectory points. It is usually compared with the geographic coordinate system to clearly describe the specific direction of the vehicle's movement.

[0093] The preset direction angle is the preset direction angle of the target parking behavior structure node, which represents the expected direction of the parking target location and is usually associated with the road direction of the task target, the entrance and exit of the parking lot, etc.

[0094] The directional deviation value is the difference between the driving direction angle of each trajectory point and the preset direction angle of the target parking behavior structure node, which is used to indicate the degree of deviation between the vehicle's driving direction and the expected parking direction.

[0095] Spatial location data consists of the spatial coordinates of each trajectory point, including longitude, latitude, and timestamp, which clearly indicates the vehicle's geographical location at a specific moment.

[0096] The reference location data consists of the geographic coordinates of the target parking behavior structure nodes, which are the known coordinates of the target parking location. They serve as a benchmark for spatial matching and deviation calculation with the trajectory points.

[0097] The spatial deviation value is the geometric distance between the spatial location data of each trajectory point and the reference location data of the target parking behavior structure node, representing the degree of spatial deviation between the trajectory point and the parking target location.

[0098] First, based on the task information, one or more target parking behavior structure nodes are selected. The task information will explicitly specify the parking location or target area. The system matches this information with the node locations and historical parking behaviors in the parking behavior structure node set to select the target parking behavior structure node that best meets the current task requirements. This selection process calculates the geographical distance and matching degree between parking nodes, while taking into account historical stopping patterns and current task requirements. For example, if the task requires parking and unloading goods in a shopping mall, the system will select target parking behavior structure nodes that match the mall's location. Once the target parking behavior structure node is determined, spatial matching calculations are performed between the real-time trajectory data and the target node. The real-time trajectory data is transmitted in real time by the vehicle positioning system and includes the latitude and longitude coordinates of the vehicle's current location. The system processes each trajectory point in the trajectory data one by one, comparing the geographical coordinates of each trajectory point with the reference location data of the target parking behavior structure node, usually by calculating the Euclidean distance from the trajectory point to the target parking behavior structure node. The system considers the spatial relationship between the geographical location of the target parking behavior structure node and the current trajectory point, and records the spatial matching result between each trajectory point and the target parking behavior structure node to determine whether the trajectory point is close to the target parking behavior structure node.

[0099] The system calculates the driving direction for each trajectory point. Specifically, it calculates the vector connecting adjacent trajectory points, derived from the coordinate difference between two adjacent points, which clearly indicates the vehicle's driving direction from one trajectory point to the next. The calculation method involves taking the coordinate difference between two adjacent trajectory points to generate the corresponding direction vector. Then, the system compares this vector connecting adjacent trajectory points with a geographic coordinate system, for example, using true north as a reference, to obtain the driving direction angle.

[0100] The process of calculating the directional deviation value first involves obtaining the driving direction angle of each trajectory point, specifically by analyzing the vectors connecting adjacent trajectory points to deduce the vehicle's driving direction. Next, the system compares the calculated actual driving direction angle with the preset direction angle of the target parking behavior structure node. This preset direction angle is typically determined by considering the surrounding environmental characteristics and historical parking behavior of the target parking behavior structure node to arrive at the ideal parking direction. The system calculates the angle difference between the actual driving direction angle and the preset direction angle; this angle difference is the directional deviation value. A larger directional deviation value indicates a more significant deviation between the vehicle's driving direction and the ideal parking direction of the target parking behavior structure node; conversely, a smaller value indicates that the vehicle's driving direction is closer to the expected parking direction. The system then calculates the spatial deviation value for each trajectory point. Before calculation, the system extracts the spatial location data of each trajectory point, typically its latitude and longitude coordinates. Next, it calculates the geometric distance between this spatial location data and the reference location data of the target parking behavior structure node. This distance can be calculated using Euclidean distance or geographic coordinate distance; this distance value is the spatial deviation value. A larger spatial deviation value indicates a greater physical distance between the vehicle's current position and the target parking behavior structure node. Finally, the system integrates the calculated directional deviation and spatial deviation values ​​to generate path deviation feature parameters corresponding to the real-time trajectory data.

[0101] This solution can monitor the deviation of vehicles from their target parking positions in real time, helping to optimize parking strategies and improve the accuracy and efficiency of parking behavior.

[0102] Based on the above technical solution, optionally, after receiving real-time trajectory data and task information within a preset time window transmitted from the mobile terminal of the driver's vehicle, the method further includes: Obtain the driver's driving behavior data within the time window corresponding to the real-time trajectory data, and extract a set of key driving behavior features based on the driving behavior data; Calculate the risk index corresponding to each driving behavior feature in the set of key driving behavior features, and based on the risk index and preset driving behavior rules, conduct a risk assessment of the driver's driving behavior to generate a driving risk assessment result. If the driving risk assessment result is classified as high-risk, a driving warning message is generated based on the driving risk assessment result and sent to the mobile terminal of the driver's vehicle.

[0103] In this scheme, driving behavior data is various types of data directly related to the driver's driving operations that are collected and recorded within a set time window. These mainly include vehicle speed, acceleration, braking intensity, steering angle, driving trajectory, and the frequency of acceleration and deceleration.

[0104] The key driving behavior feature set is a set of features selected from driving behavior data that have a significant impact on driving safety. It can reveal the driver's driving habits and identify potential risky behaviors, including the frequency of sudden braking, the frequency of sudden acceleration, the amplitude of speed fluctuations, driving duration, and the degree of abruptness of turning.

[0105] Driving behavior characteristics quantify various specific driving operations of the driver into specific numerical indicators, such as peak acceleration, braking force, speed change rate, and frequency of direction change.

[0106] Risk indicators are quantitative values ​​calculated for each driving behavior characteristic, taking into account its potential impact on driving safety.

[0107] Preset driving behavior rules are based on industry safety standards, referencing historical safety data and driving safety requirements. They set clear thresholds for different driving behaviors and are mainly used to quickly identify dangerous driving behaviors. For example, if the vehicle speed exceeds a certain value, it will be directly judged as a high-risk behavior.

[0108] The driving risk assessment results are conclusions drawn from a comprehensive judgment that combines key driving behavior feature sets, risk indicators, and preset driving behavior rules. They are divided into three risk levels: low, medium, and high. The core function is to classify the degree of danger of a driver's driving behavior.

[0109] Driving warning messages are automatically generated reminders when the system determines that driving behavior is high-risk, with the aim of reminding drivers to pay attention to safety and operate in a standardized manner.

[0110] To acquire driver behavior data within a corresponding time window based on real-time trajectory data, the system first performs time slicing on the trajectory data transmitted from the in-vehicle device. The system extracts driving behavior data for the corresponding time period from the in-vehicle computer or other in-vehicle devices according to the specified time window. Then, the timestamps of the real-time trajectory data are aligned with the behavior records of the in-vehicle device. The system also needs to clean the data, removing noisy data, such as using interpolation to fill in missing speed data or removing outlier data, to ensure that the extracted driving behavior data accurately reflects the driver's actual driving situation. Based on the extracted driving behavior data, the system further refines a key driving behavior feature set. This feature set includes core driving behavior features such as rapid acceleration, sudden braking, and sharp turns. Specifically, rapid acceleration is determined by calculating the rate of change of acceleration, sudden braking is assessed by changes in braking force and braking rate, and sharp turns are confirmed by sudden changes in steering angle and vehicle speed during the turn. Each driving behavior feature is quantified using an appropriate algorithm to generate specific feature values, and these feature values ​​together constitute the key driving behavior feature set.

[0111] After the key driving behavior feature set is generated, the system begins to calculate the risk index corresponding to each driving behavior feature. For example, for rapid acceleration, the system calculates the frequency and magnitude of acceleration changes; for sudden braking, it calculates the risk index based on braking force and braking duration; for sharp turns, it conducts a risk assessment based on extreme steering angles and vehicle speed. The risk index for each driving behavior feature is based on historical driving data and industry standards. Thresholds are set to define the risk level of the behavior. If the risk index of a feature exceeds the high-risk threshold, it is judged as a high-risk behavior, and vice versa.

[0112] The system combines these risk indicators with preset driving behavior rules to conduct a comprehensive risk assessment. Specifically, the system weights and sums the risk indicators of various driving behavior characteristics to obtain a comprehensive risk value. If the comprehensive risk value exceeds a preset high-risk threshold, the driving behavior is judged as high-risk, resulting in a high-risk driving risk assessment result. Once the driving risk assessment result is high-risk, the system generates driving warning information, which clearly indicates the specific risky behavior, such as frequent sudden braking, speeding, or excessively sharp turns. It also selects an appropriate push notification method based on the driver's preferences. Finally, the driving warning information is sent to the driver's device via the vehicle's mobile terminal.

[0113] This solution monitors driving behavior in real time and generates risk assessments, enabling timely identification of dangerous driving behaviors and the issuance of warnings. This improves driving safety and enhances road safety management.

[0114] Step S105: Input the task information, stability feature parameters and path deviation feature parameters into the preset risk assessment model for feature fusion calculation, output deviation risk assessment value, and generate the driver's driving assessment level based on the deviation risk assessment value.

[0115] The preset risk assessment model is a built-in calculation module of the system, which is used to integrate input data such as task information, stable feature parameters and path deviation feature parameters, and output deviation risk assessment value after completing the assessment.

[0116] The deviation risk assessment score is a quantitative result output by the risk assessment model. It is generally a continuous score from 0 to 100, or a standardized score, used to measure the degree of deviation of the vehicle from the target task path or parking point during the current driving process and the potential risk. The higher the score, the greater the deviation risk.

[0117] Driving performance ratings are a comprehensive label that combines deviation risk assessment values ​​with preset grading rules to derive a rating of a driver's driving behavior. Common rating formats include letter ratings, such as A, B, C, etc., as well as descriptive ratings, such as Excellent, Good, Average, Poor, or Low, Medium, High Risk.

[0118] After receiving task information, stable feature parameters, and path deviation feature parameters, the system first performs consistency verification and alignment on these three types of inputs. It checks the correspondence between timestamps and time windows one by one to ensure that the time periods of the task path, parking points, and path deviation features are mapped simultaneously. The stable feature parameters and path deviation feature parameters undergo scaling processing, using normalization or standardization methods to map each dimension to the 0–1 interval or convert it into a form with zero mean and unit variance, ensuring the comparability of features with different dimensions in subsequent fusion stages.

[0119] The system uses task information as context, mapping the target parking point, task path, historical cluster density, stability parameters, and other information to a multi-dimensional feature vector space to construct an initial feature vector combination. Then, the stable feature vector, path deviation vector, and information from the task information such as geometric constraints, time constraints, distance, and route similarity are concatenated and integrated to generate a unified high-dimensional feature matrix. Feature fusion is then performed row-by-row for each time point or time period. In the fusion phase, the system first assigns initial weights to key dimensions such as spatial deviation distance, directional deviation angle, alignment confidence, task path similarity, and remaining distance to the destination based on domain knowledge. These weights are then fine-tuned using a data-driven approach to ensure that the contribution of each dimension to the final result matches the actual application scenario. The weight-tuned feature matrix is ​​then input into a pre-defined risk assessment model, ultimately outputting a deviation risk assessment value. After the model outputs the result, this deviation risk assessment value is mapped to the corresponding driving rating level. For example, a deviation risk assessment value of 68 corresponds to a driving rating level B.

[0120] The training process for the pre-defined risk assessment model is as follows: First, training samples are collected from historical data sources, including historical task information, stable feature parameters, path deviation feature parameters, and corresponding historical deviation risk assessment value labels. During the collection process, it is necessary to ensure alignment of temporal and spatial contexts, unify the granularity of the time window and the coordinate system, and appropriately fill in or directly remove missing values.

[0121] Next, the raw data is cleaned and standardized, outlier trajectory points are removed, and data units and dimensions are standardized. The input feature vectors are normalized or standardized, including historical task information, stable feature vectors, and path deviation feature vectors. Stratified sampling is performed according to dimensions such as task type, time period, and road scenario to construct training, validation, and test sets. In the feature processing stage, necessary derived features are constructed, such as route similarity, remaining distance to the destination, and cluster density weights. Correlation analysis and dimensionality reduction are performed on the features to improve the model's robustness.

[0122] Then, the output format and architecture of the model are determined. The output format can be one or more options, such as a regression outputting continuous deviation risk values ​​or a discrete driving level obtained through threshold mapping. The model architecture can also be one or more options, such as linear regression, GBDT, random forest, XGBoost, neural networks, etc. During the training phase, samples from a historical window are used as input, and real historical deviation risk values ​​are used as supervision labels to perform end-to-end fitting.

[0123] The training process involves several key operations, including hyperparameter search (using grid, random, or Bayesian optimization methods); implementing regularization strategies; adjusting sample weights to mitigate potential imbalance; and conducting hierarchical cross-validation to evaluate the model's generalization ability across different task types and contexts. During model evaluation, appropriate metrics are selected based on the output format. For regression outputs, metrics such as MAE, RMSE, and R² are used; for discrete-level outputs, metrics such as accuracy, F1, and AUC are used. Simultaneously, diagnostic analysis of the model's residual distribution, sources of bias, and feature importance is conducted. If necessary, model output calibration and post-processing are performed, such as distribution alignment for continuous outputs, setting reasonable threshold intervals to map to driving levels, providing feature contribution and confidence intervals, and enhancing model interpretability. Finally, the trained model is deployed to the online environment, establishing an iterative mechanism for offline retraining and online continuous learning to ensure the model can adapt to changes in task information and environmental conditions.

[0124] Based on steps S101-S105 above, by fusing historical parking patterns, stability and real-time trajectory deviation, deviation risk value and driving level are output, improving prediction accuracy and timeliness, facilitating task-oriented intervention, real-time alarms and enhancing interpretability.

[0125] Based on the above technical solution, optionally, after generating the driver's driving evaluation level based on the deviation risk assessment value, the method further includes: If the driving evaluation level is lower than the preset safety level, an alarm message is generated based on the deviation risk evaluation value and the driving evaluation level, and the alarm message is sent to the mobile terminal of the vehicle where the driver is located. Accordingly, after sending the alarm information to the mobile terminal of the driver's vehicle, the method further includes: The driver's real-time trajectory data within the next preset time window is obtained, and the real-time trajectory data is re-spatial matched and directional deviation calculated based on the parking behavior structure node set and task information to generate path deviation feature parameters corresponding to the real-time trajectory data. The task information, stability feature parameters, and path deviation feature parameters are re-inputted into the preset risk assessment model for feature fusion calculation, and the deviation risk assessment value is output. Based on the deviation risk assessment value, the driver's driving evaluation level is regenerated. Repeat the above steps if the driving evaluation level is higher than the preset safety level, until the task completion information is received from the mobile terminal of the driver's vehicle.

[0126] In this solution, the alarm information is generated based on the degree of deviation from the driving behavior to remind the driver to correct the deviation and ensure safety. The reminders can be delivered via voice or text.

[0127] If the driving evaluation level is lower than the preset safety level, an alarm message is generated based on the deviation risk evaluation value and the driving evaluation level, and then sent to the mobile terminal. This message is delivered via voice or text; for example, the mobile terminal might display an alarm message: "Deviation risk evaluation value is 98, driving evaluation level is high risk." After the broadcast, data is collected for the next preset window. If the driving evaluation level is higher than the preset safety level, data collection continues for the next preset window, and a new evaluation level is generated. If it is lower than the preset safety level, an alarm message is generated; if it is higher than the preset safety level, data collection continues, and the evaluation is re-evaluated, until the task completion message is received from the mobile terminal of the driver's vehicle.

[0128] In this solution, through real-time monitoring and timely push of alarm information, the system can effectively remind drivers to correct deviation behavior, reduce safety hazards, improve driving safety, avoid potential risks, and ensure the smooth completion of tasks.

[0129] See appendix Figure 2 , Figure 2 This is a schematic flowchart of the second main step of a driving rating method based on parking behavior node coupling according to an embodiment of the present invention. Figure 2 As shown, a driving rating method based on parking behavior node coupling in an embodiment of the present invention mainly includes the following steps S201-S208.

[0130] Step S201: Obtain the driver's vehicle trajectory data and original parking data within a preset period. Based on the vehicle trajectory data, perform time alignment and spatial position correction on the original parking data to obtain a parking behavior dataset.

[0131] Step S202: Based on preset spatial clustering rules and preset temporal continuity constraints, perform dual-constraint clustering analysis on the parking behavior dataset to obtain a parking behavior structure node set.

[0132] Step S203: Independently extract and perform stability analysis on the structural features of each parking behavior structure node in the parking behavior structure node set to generate stable feature parameters for each parking behavior structure node.

[0133] Step S204: If the real-time trajectory data and task information within a preset time window are received from the mobile terminal of the vehicle where the driver is located, spatial matching and directional deviation calculation are performed on the real-time trajectory data based on the parking behavior structure node set and task information to generate path deviation feature parameters corresponding to the real-time trajectory data.

[0134] Step S205: Input the task information, stability feature parameters and path deviation feature parameters into the preset risk assessment model for feature fusion calculation, output deviation risk assessment value, and generate the driver's driving assessment level based on the deviation risk assessment value.

[0135] Step S206: If the driving evaluation level is higher than the preset safety level, obtain the driver's real-time trajectory data in the next preset time window, and re-perform spatial matching and directional deviation calculation on the real-time trajectory data based on the parking behavior structure node set and task information to generate path deviation feature parameters corresponding to the real-time trajectory data.

[0136] Step S207: Re-input the task information, stability feature parameters, and path deviation feature parameters into the preset risk assessment model for feature fusion calculation, output deviation risk assessment value, and regenerate the driver's driving assessment level based on the deviation risk assessment value.

[0137] Step S208: If the driving evaluation level is still higher than the preset safety level, repeat the above steps until the task completion information is received from the mobile terminal of the driver's vehicle.

[0138] In this embodiment, the preset safety level is a judgment standard based on the deviation risk assessment value, used to determine the degree of deviation between the vehicle and the target task path or parking point. When the driver's driving behavior significantly deviates from the expected path or parking point, and the corresponding deviation risk assessment value exceeds the preset safety level threshold, the driving behavior will be identified as a high-risk behavior.

[0139] Task completion information is data uploaded to the system by the mobile terminal in the driver's vehicle, used to inform the driver that the currently designated task or objective has been completed.

[0140] When the driving evaluation level is higher than the preset safety level, the system automatically acquires the driver's real-time trajectory data for the next time window. The system combines the parking behavior structure node set with task information to perform spatial matching and directional deviation calculations, generating corresponding path deviation feature parameters. Then, the system inputs the task information, stability feature parameters, and path deviation feature parameters into a preset risk assessment model. After the model calculates, it outputs a deviation risk assessment value and regenerates the driving evaluation level based on this value. If the newly generated driving evaluation level is still higher than the preset safety level, the system will repeat the entire process. This process continues until the system receives task completion information, at which point the entire driving behavior assessment process officially ends.

[0141] Based on steps S201-S208 above, by continuously monitoring and evaluating the driver's driving behavior, the system can identify potential risks in real time and provide timely feedback, helping the driver correct dangerous driving behaviors, improving driving safety, reducing the probability of traffic accidents, and ensuring a safer and more reliable driving process.

[0142] It should be noted that although the steps in the above embodiments are described in a specific order, those skilled in the art will understand that in order to achieve the effects of the present invention, different steps do not necessarily have to be executed in such an order. They can be executed simultaneously (in parallel) or in other orders, and these variations are all within the scope of protection of the present invention.

[0143] Furthermore, the present invention also provides a driving rating system based on parking behavior node coupling.

[0144] See appendix Figure 3 , Figure 3 This is a main structural block diagram of a driving rating system based on parking behavior node coupling according to an embodiment of the present invention. Figure 3 As shown, it specifically includes: The parking behavior collection module 301 is used to acquire vehicle trajectory data and original parking data of the driver within a preset period, and to perform time alignment and spatial position correction on the original parking data based on the vehicle trajectory data to obtain a parking behavior dataset. The dual-constraint clustering analysis module 302 is used to perform dual-constraint clustering analysis on the parking behavior dataset based on preset spatial clustering rules and preset temporal continuity constraints to obtain a parking behavior structure node set. The stability analysis module 303 is used to independently extract the structural features of each parking behavior structure node in the parking behavior structure node set and perform stability analysis to generate stable feature parameters for each parking behavior structure node. The path deviation feature analysis module 304 is used to perform spatial matching and directional deviation calculation on the real-time trajectory data based on the parking behavior structure node set and task information, and generate path deviation feature parameters corresponding to the real-time trajectory data, if it receives real-time trajectory data and task information transmitted by the mobile terminal of the vehicle where the driver is located within a preset time window. The driving evaluation module 305 is used to input the task information, stability feature parameters and path deviation feature parameters into a preset risk evaluation model for feature fusion calculation, output deviation risk evaluation value, and generate the driver's driving evaluation level based on the deviation risk evaluation value.

[0145] The driving rating system based on parking behavior node coupling provided in this application embodiment can achieve... Figure 1 The various processes implemented in the method implementation examples will not be described again here to avoid repetition.

[0146] Those skilled in the art will understand that all or part of the processes in the method of the above embodiment of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium can include any entity or device capable of carrying the computer program code, a medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the content included in the computer-readable storage medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0147] Furthermore, the present invention also provides an electronic device 400, including a processor 401, a memory 402, and a program or instructions stored in the memory 402 and executable on the processor 401. When the program or instructions are executed by the processor 401, they implement the various processes of the above-described embodiment of a driving rating method based on parking behavior node coupling and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0148] It should be noted that the electronic devices in the embodiments of this application include the mobile electronic devices and non-mobile electronic devices described above.

[0149] Furthermore, the present invention also provides a computer-readable storage medium. In one embodiment of the computer-readable storage medium according to the present invention, the computer-readable storage medium can be configured to store a program that executes a driving rating method based on parking behavior node coupling of the above-described method embodiments. This program can be loaded and run by a processor to implement the above-described driving rating method based on parking behavior node coupling. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. The computer-readable storage medium can be a storage device comprising various electronic devices. Optionally, in the embodiments of the present invention, the computer-readable storage medium is a non-transitory computer-readable storage medium.

[0150] Furthermore, it should be understood that since the various modules are only provided to illustrate the functional units of the device of the present invention, the physical devices corresponding to these modules may be the processor itself, or a part of the processor's software, hardware, or a combination of software and hardware. Therefore, the number of modules shown in the figures is merely illustrative.

[0151] Those skilled in the art will understand that the various modules in the device can be adaptively split or combined. Such splitting or combining of specific modules will not cause the technical solution to deviate from the principles of the present invention; therefore, the technical solutions after splitting or combining will fall within the protection scope of the present invention.

[0152] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A driving rating method based on parking behavior node coupling, characterized in that, The method includes: The system acquires vehicle trajectory data and raw parking data of the driver within a preset period. Based on the vehicle trajectory data, the raw parking data is time-aligned and spatially corrected to obtain a parking behavior dataset. Based on preset spatial clustering rules and preset temporal continuity constraints, a dual-constraint clustering analysis is performed on the parking behavior dataset to obtain a set of parking behavior structure nodes; this includes calculating the spatial distance data between each parking point in the parking behavior dataset based on preset spatial clustering rules, and constructing a spatial distance matrix based on the spatial distance data; Based on the preset spatial clustering rules, the spatial distance matrix is ​​filtered by threshold constraints to obtain a spatial adjacency relationship set. Based on the spatial adjacency relationship set, a spatial connectivity relationship graph that satisfies the preset spatial clustering rules is established. Spatial grouping is performed on the interconnected parking points in the spatial connectivity graph to form an initial set of spatial clustering units; Based on the initial spatial clustering unit set, extract the timestamp sequence of parking records within each clustering unit; Based on the preset time continuity constraint, the timestamp sequence of parking records within each initial spatial clustering unit is analyzed for continuity to obtain the continuity analysis results. Based on the continuity analysis results, the timestamp sequence is divided into breakpoints to form continuous time sub-units; Extract the spatial and temporal distribution features of each temporally continuous sub-unit, and construct a structural feature vector based on the spatial and temporal distribution features; Based on the preset stability judgment rules and structural feature vectors, the stability of each time continuous sub-unit is evaluated and screened to form a set of parking behavior structural nodes. The structural features of each parking behavior structure node in the parking behavior structure node set are independently extracted and stability analysis is performed to generate stable feature parameters for each parking behavior structure node. If the real-time trajectory data and task information within a preset time window are received from the mobile terminal of the vehicle where the driver is located, spatial matching and directional deviation calculation are performed on the real-time trajectory data based on the parking behavior structure node set and task information to generate path deviation feature parameters corresponding to the real-time trajectory data. The task information, stability feature parameters, and path deviation feature parameters are input into a preset risk assessment model for feature fusion calculation, and a deviation risk assessment value is output. Based on the deviation risk assessment value, the driver's driving evaluation level is generated.

2. The driving rating method based on parking behavior node coupling according to claim 1, characterized in that, in, The structural features of each parking behavior structure node in the parking behavior structure node set are independently extracted and stability analysis is performed to generate stable feature parameters for each parking behavior structure node, including: Extract the spatial distribution parameters, temporal distribution parameters, and behavioral intensity parameters of each parking behavior structure node in the parking behavior structure node set, and construct a multidimensional structural parameter set for each parking behavior structure node based on the spatial distribution parameters, temporal distribution parameters, and behavioral intensity parameters; Based on the multidimensional structural parameter set, a parameter temporal evolution sequence is constructed. The parameter temporal evolution sequence is differentially calculated to obtain the numerical change between adjacent time points. Based on the numerical change, a parameter change gradient sequence for each parking behavior structural node is constructed. Change point detection is performed on the time-series evolution sequence of the parameters to determine the structural change nodes, and piecewise linear fitting is performed based on the structural change nodes to obtain the piecewise trend parameter set of each parking behavior structural node; Obtain the trend slope and segment interval length of each segment in the segmented trend parameter set, and calculate the comprehensive trend slope of each parking behavior structure node by performing a weighted average based on the trend slope and segment interval length. Based on the comprehensive trend slope and parameter change gradient sequence, the temporal fluctuation amplitude data and trend consistency index of each parking behavior structure node are calculated. Based on the time-series fluctuation amplitude data and trend consistency index, the time-series stability index of each parking behavior structure node is obtained by fusion calculation. Based on the time-series stability index and the preset stability judgment threshold, the stability state results of each parking behavior structure node are generated, and the time-series stability index and stability state results are structured, stored and uniformly represented to obtain the stability feature parameters of each parking behavior structure node.

3. The driving rating method based on parking behavior node coupling according to claim 1, characterized in that, in, Based on the parking behavior structure node set and task information, spatial matching and directional deviation calculation are performed on the real-time trajectory data to generate path deviation feature parameters corresponding to the real-time trajectory data, including: Based on task information and parking behavior structure node set, at least one target parking behavior structure node is determined, and spatial matching calculation is performed based on real-time trajectory data and target parking behavior structure node to obtain the spatial matching result between each trajectory point in real-time trajectory data and target parking behavior structure node. The driving direction angle of each trajectory point is generated based on the driving direction vector of each trajectory point by calculating the line vector connecting adjacent trajectory points based on the line vector connecting adjacent trajectory points. Calculate the angle difference between the driving direction angle of each trajectory point and the preset direction angle corresponding to the target parking behavior structure node to obtain the directional deviation value between each trajectory point and the target parking behavior structure node; Based on real-time trajectory data, determine the spatial location data of each trajectory point, calculate the geometric distance between the spatial location data of each trajectory point and the reference location data corresponding to the target parking behavior structure node, and obtain the spatial deviation value between each trajectory point and the target parking behavior structure node. Based on the directional deviation value and the spatial deviation value, path deviation feature parameters corresponding to the real-time trajectory data are generated.

4. The driving rating method based on parking behavior node coupling according to claim 1, characterized in that, in, After receiving real-time trajectory data and task information within a preset time window transmitted from the mobile terminal of the driver's vehicle, the method further includes: Obtain the driver's driving behavior data within the time window corresponding to the real-time trajectory data, and extract a set of key driving behavior features based on the driving behavior data; Calculate the risk index corresponding to each driving behavior feature in the set of key driving behavior features, and based on the risk index and preset driving behavior rules, conduct a risk assessment of the driver's driving behavior to generate a driving risk assessment result. If the driving risk assessment result is classified as high-risk, a driving warning message is generated based on the driving risk assessment result and sent to the mobile terminal of the driver's vehicle.

5. The driving rating method based on parking behavior node coupling according to claim 1, characterized in that, in, After generating the driver's driving rating based on the deviation risk assessment value, the method further includes: If the driving evaluation level is higher than the preset safety level, the driver's real-time trajectory data in the next preset time window is obtained, and the real-time trajectory data is re-spatial matched and directional deviation calculated based on the parking behavior structure node set and task information to generate path deviation feature parameters corresponding to the real-time trajectory data. The task information, stability feature parameters, and path deviation feature parameters are re-inputted into the preset risk assessment model for feature fusion calculation, and the deviation risk assessment value is output. Based on the deviation risk assessment value, the driver's driving evaluation level is regenerated. If the driving evaluation level is still higher than the preset safety level, repeat the above steps until the task completion information is received from the mobile terminal of the driver's vehicle.

6. The driving rating method based on parking behavior node coupling according to claim 1, characterized in that, in, After generating the driver's driving rating based on the deviation risk assessment value, the method further includes: If the driving evaluation level is lower than the preset safety level, an alarm message is generated based on the deviation risk evaluation value and the driving evaluation level, and the alarm message is sent to the mobile terminal of the vehicle where the driver is located. Accordingly, after sending the alarm information to the mobile terminal of the driver's vehicle, the method further includes: The driver's real-time trajectory data within the next preset time window is obtained, and the real-time trajectory data is re-spatial matched and directional deviation calculated based on the parking behavior structure node set and task information to generate path deviation feature parameters corresponding to the real-time trajectory data. The task information, stability feature parameters, and path deviation feature parameters are re-inputted into the preset risk assessment model for feature fusion calculation, and the deviation risk assessment value is output. Based on the deviation risk assessment value, the driver's driving evaluation level is regenerated. Repeat the above steps if the driving evaluation level is higher than the preset safety level, until the task completion information is received from the mobile terminal of the driver's vehicle.

7. A driving rating system based on parking behavior node coupling, characterized in that, The system includes: The parking behavior collection module is used to acquire vehicle trajectory data and raw parking data of drivers within a preset period. Based on the vehicle trajectory data, the raw parking data is time-aligned and spatially corrected to obtain a parking behavior dataset. The dual-constraint clustering analysis module is used to perform dual-constraint clustering analysis on the parking behavior dataset based on preset spatial clustering rules and preset time continuity constraints to obtain a set of parking behavior structure nodes; wherein, it includes calculating the spatial distance data between each parking point in the parking behavior dataset based on preset spatial clustering rules, and constructing a spatial distance matrix based on the spatial distance data; Based on the preset spatial clustering rules, the spatial distance matrix is ​​filtered by threshold constraints to obtain a spatial adjacency relationship set. Based on the spatial adjacency relationship set, a spatial connectivity relationship graph that satisfies the preset spatial clustering rules is established. Spatial grouping is performed on the interconnected parking points in the spatial connectivity graph to form an initial set of spatial clustering units; Based on the initial spatial clustering unit set, extract the timestamp sequence of parking records within each clustering unit; Based on the preset time continuity constraint, the timestamp sequence of parking records within each initial spatial clustering unit is analyzed for continuity to obtain the continuity analysis results. Based on the continuity analysis results, the timestamp sequence is divided into breakpoints to form continuous time sub-units; Extract the spatial and temporal distribution features of each temporally continuous sub-unit, and construct a structural feature vector based on the spatial and temporal distribution features; Based on the preset stability judgment rules and structural feature vectors, the stability of each time continuous sub-unit is evaluated and screened to form a set of parking behavior structural nodes. The stability analysis module is used to independently extract the structural features of each parking behavior structure node in the parking behavior structure node set and perform stability analysis to generate stable feature parameters for each parking behavior structure node. The path deviation feature analysis module is used to perform spatial matching and directional deviation calculation on the real-time trajectory data and task information within a preset time window transmitted by the mobile terminal of the driver's vehicle, based on the parking behavior structure node set and task information, to generate path deviation feature parameters corresponding to the real-time trajectory data. The driving evaluation module is used to input the task information, stability feature parameters, and path deviation feature parameters into a preset risk evaluation model for feature fusion calculation, output deviation risk evaluation value, and generate the driver's driving evaluation level based on the deviation risk evaluation value.

8. An electronic device comprising a processor, a memory, and a program or instructions stored in the memory and executable on the processor, characterized in that, The program or instructions are adapted to be loaded and run by the processor to perform a driving rating method based on parking behavior node coupling as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by a processor to perform a driving rating method based on parking behavior node coupling as described in any one of claims 1 to 6.

Images