Traffic anomaly processing method and system based on multi-source data space-time fusion

Abstract

The invention discloses a traffic abnormity processing method and system based on multi-source data space-time fusion, and relates to the technical field of traffic supervision, and the method comprises the steps: collecting traffic multi-modal data in parallel to construct a space-time correlation matrix, and analyzing the space-time correlation matrix according to a space-time correlation analysis model to determine an initial abnormal section; tracking abnormal points of the initial abnormal section according to a grid contraction strategy to determine a target abnormal section; determining an alarm screening strategy according to the abnormal sweep degree of the target abnormal section; matching an exception handling scheme according to the alarm screening strategy in combination with the alarm level of the target exception section; and monitoring the completion degree of the traffic dredging indexes in real time to determine the traffic dredging condition in combination with the abnormality processing scheme, and updating the traffic road condition broadcast in real time. According to the scheme, the problem of insufficient multi-source data fusion is solved, the whole-process intelligent management of traffic abnormity is realized, the abnormity positioning precision and the alarm processing efficiency are improved, and a real-time and accurate solution is provided for traffic management.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic supervision, and in particular to a method and system for handling traffic anomalies based on spatiotemporal fusion of multi-source data. Background Art

[0002] In intelligent transportation systems, the application of multi-source data is crucial for traffic management. Existing technologies have many shortcomings in the application of multi-source data in the transportation sector. On the one hand, the single method of multi-source data collection makes it difficult to fully acquire multi-dimensional information such as traffic flow, events, and the environment. On the other hand, the data fusion depth is insufficient, making it impossible to effectively explore the potential connections between different data sources, resulting in low accuracy in identifying traffic anomalies. Furthermore, existing technologies do not fully utilize the spatiotemporal characteristics of multi-source data and lack in-depth analysis of the spatiotemporal propagation patterns of traffic anomalies, resulting in rough positioning of anomaly sections and affecting the efficiency of anomaly handling. These issues have resulted in existing technical solutions being unable to meet the real-time and accuracy requirements of traffic management.

[0003] The above information disclosed in this Background section is only for enhancement of understanding of the background of the application and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0004] The purpose of the present invention is to address the problems in the prior art of insufficient depth of spatiotemporal fusion of multi-source data, rough anomaly positioning, and inefficient anomaly handling, which lead to insufficient accuracy and real-time performance of traffic anomaly management. A traffic anomaly handling method and system based on spatiotemporal fusion of multi-source data are proposed. By constructing a multimodal data distributed acquisition network, the parallel acquisition and spatiotemporal alignment of heterogeneous data are realized. The spatiotemporal correlation matrix and dynamic analysis model are used to accurately locate the abnormal section. The core abnormal points are tracked in combination with the grid contraction strategy. An intelligent matching mechanism of alarm screening rules and disposal solutions is constructed based on the degree of abnormal impact. Finally, the real-time update of road condition information is achieved through the joint monitoring of dredging indicators and disposal completion. The defects of insufficient multi-source data fusion and rough spatiotemporal analysis are overcome, thereby realizing the full-process intelligent management of traffic anomalies, improving the accuracy of abnormal positioning, the efficiency of alarm processing, and the scientific nature of disposal decisions, and providing a real-time and accurate solution for traffic management.

[0005] In a first aspect, a technical solution provided in an embodiment of the present invention is a method for handling traffic anomalies based on spatiotemporal fusion of multi-source data, comprising the following steps: Parallel collection of multimodal traffic data to construct a spatiotemporal correlation matrix, which is then analyzed using a spatiotemporal correlation analysis model to determine the initial abnormal section. According to the grid shrinkage strategy, the abnormal points in the initial abnormal section are tracked to determine the target abnormal section; Determine the alarm screening strategy based on the abnormal impact of the target abnormal section; match the abnormality handling plan based on the alarm screening strategy and the alarm level of the target abnormal section; Real-time monitoring of traffic flow indicators and the completion of abnormal handling plans to determine traffic flow conditions, and update traffic conditions reports in real time.

[0006] Preferably, the parallel collection of traffic flow data to construct a spatiotemporal correlation matrix comprises the following steps: Constructing a multimodal data distributed acquisition network, comprising a microwave radar array, video surveillance equipment, geomagnetic sensors, and mobile vehicle terminals deployed on the roadside; The collected multimodal traffic data is timestamped and stored in shards; the target road network is divided into three-dimensional grid cells according to geographic coordinates and a unique spatiotemporal index is configured for each grid cell; Feature extraction is performed on the multimodal traffic data within each grid cell to construct a multidimensional feature vector. Based on the multidimensional feature vector, the dynamic time warping algorithm is used to calculate the spatiotemporal similarity between different grid cells and construct a spatiotemporal correlation matrix. The multi-dimensional feature vector includes traffic flow features, abnormal event features, environmental features and historical data features.

[0007] Preferably, the step of analyzing the spatiotemporal correlation matrix according to the spatiotemporal correlation analysis model to determine the initial abnormal section comprises the following steps: The fluctuation factor of each element in the spatiotemporal correlation matrix is ​​determined based on the spatiotemporal correlation analysis model. The dynamic threshold of each element is calculated based on the fluctuation factor combined with the historical mean and standard deviation of the element. Elements with actual element values ​​greater than the dynamic threshold are marked as outliers and an outlier set is generated. The DBSCAN clustering algorithm is used to perform cluster analysis on the abnormal point set to determine the initial abnormal cluster set. The temporal continuity and spatial connectivity of each abnormal cluster in the initial abnormal cluster set are verified to determine the target abnormal cluster set. The abnormal clusters in the target abnormal cluster set are projected onto the actual road network. If the abnormal clusters span different road types, the initial abnormal segment is determined based on the road boundary. If the abnormal clusters contain discontinuous roads, the isolated segments are removed and the main connected roads are retained to determine the initial abnormal segment.

[0008] Preferably, the method of tracking the abnormal points of the initial abnormal section according to the grid shrinkage strategy to determine the target abnormal section comprises the following steps: The abnormal points corresponding to the initial abnormal segments are extracted in sequence, and the true element values ​​corresponding to the abnormal points are normalized to obtain the abnormal intensity; Determine the importance of a segment based on anomaly intensity, spatial correlation, grid connectivity, and their corresponding weight factors; sort the anomaly segments from largest to smallest according to their importance, and select the top N anomaly segments as candidate segments; verify the spatial connectivity of the candidate segments; if there are isolated segments, shrink the candidate segments by a shrinkage factor a until the candidate segments form a single connected area, obtaining the target anomaly segment; The minimum convex hull of the filtered target abnormal segment is calculated to obtain the polygonal boundary; the spatial centroid of the polygonal boundary is used as the target abnormal point.

[0009] Preferably, an alarm screening strategy is determined according to the abnormal impact degree of the target abnormal section; and an abnormal handling plan is matched according to the alarm screening strategy and the alarm level of the target abnormal section, including the following steps: The spatial impact factor is determined based on the road length, number of lanes, and area covered by the target abnormal section; the temporal duration factor is determined based on the duration and frequency of the abnormal event; and the traffic impact factor is determined based on the traffic flow reduction rate, average speed reduction, and queue length increase. Determine the abnormal impact degree based on the spatial impact factor, time duration factor, traffic impact factor and their corresponding weight coefficients; use the quartile algorithm to classify the abnormal impact degree into alarm levels to determine the spatial information, time information and event information corresponding to the target abnormal point; Determine the key features of the alarm based on spatial information, time information and event information, extract the alarm features in the alarm information and compare them with the key features of the alarm, filter the alarm information based on the comparison results to determine the alarm point, determine the urgency of the abnormal point based on the alarm point and the alarm frequency of the alarm point; formulate corresponding abnormal handling plans based on the alarm level and its corresponding urgency.

[0010] Preferably, the real-time monitoring of traffic flow indicators and the completion of abnormality handling plans to determine traffic flow conditions and update traffic condition reports in real time includes the following steps: The traffic flow indicators of the target abnormal section are obtained through a multimodal data distributed acquisition network, and the traffic flow degree is calculated based on the traffic flow indicators. When the traffic flow degree exceeds the set threshold, the task list of the abnormal handling plan is retrieved, and the traffic congestion situation is determined based on the completion rate of the abnormal list. The traffic congestion situation is then simultaneously published to the traffic broadcast platform. The traffic flow indicators include traffic flow indicators, vehicle speed indicators and vehicle queue indicators.

[0011] As a preferred method, the formula for traffic flow is expressed as: Among them, α1, α2, and α3 are the weight coefficients of the corresponding indicators, Qt is the real-time traffic volume of each lane, Q0 is the historical baseline traffic volume, V t is the average speed, V0 is the free flow speed, L t is the queue length, L0 is the initial queue length.

[0012] In a second aspect, an embodiment of the present invention further provides a technical solution: a traffic anomaly processing system, applicable to a traffic anomaly processing method based on spatiotemporal fusion of multi-source data, comprising: Preprocessing module: collects multimodal traffic data in parallel to construct a spatiotemporal correlation matrix, and analyzes the spatiotemporal correlation matrix according to the spatiotemporal correlation analysis model to determine the initial abnormal section; Anomaly Location Module: Tracks anomaly points in the initial anomaly section based on the grid contraction strategy to determine the target anomaly section. Strategy Generation Module: Determines the alarm screening strategy based on the anomaly impact of the target anomaly section. Matches the anomaly handling plan based on the alarm screening strategy and the alarm level of the target anomaly section. Joint reporting module: Real-time monitoring of traffic flow indicators and the completion of abnormal handling plans to determine traffic flow conditions, and update traffic condition reports in real time.

[0013] In a third aspect, a technical solution provided in an embodiment of the present invention is: an electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps of a traffic anomaly handling method based on spatiotemporal fusion of multi-source data are implemented.

[0014] In a fourth aspect, a technical solution provided in an embodiment of the present invention is: a storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are loaded and executed by a processor, the steps of a traffic anomaly handling method based on spatiotemporal fusion of multi-source data are implemented.

[0015] Beneficial effects of the present invention: (1) In order to solve the technical problem of insufficient fusion depth caused by the spatiotemporal heterogeneity of multi-source traffic data, this application proposes a technical solution to achieve spatiotemporal fusion of multi-source data by constructing a spatiotemporal correlation matrix through a dynamic time warping algorithm and combining it with a spatiotemporal correlation analysis model. This solution realizes the parallel collection and spatiotemporal alignment of heterogeneous data through a multimodal data distributed acquisition network, and uses dynamic thresholds and clustering algorithms to identify outliers and divide sections. It breaks through the technical bottleneck of the existing technology that static weighted fusion cannot adapt to the non-stationary characteristics of traffic flow, and realizes the accurate modeling of the spatiotemporal dynamic correlation of traffic flow and the deep mining of abnormal characteristics. (2) In order to solve the technical problem that the traditional abnormal section positioning lacks road network topology constraints and leads to insufficient positioning accuracy, this application proposes a multi-factor evaluation and road network constraint correction technical solution based on the grid shrinkage strategy. The importance of the section is determined by weighted calculation of abnormal intensity, spatial correlation, and grid connectivity. Combined with adaptive threshold shrinkage and electronic map road network matching, dynamic tracking and boundary refinement of the target abnormal section are achieved, breaking through the technical limitation of the existing static threshold division method that cannot adapt to the road network topology characteristics, and significantly improving the spatiotemporal accuracy and physical rationality of abnormal section positioning; (3) In response to the technical problem of low decision-making efficiency caused by the semantic ambiguity of multi-source alarm information and the homogeneity of disposal plans, this application proposes a multi-dimensional evaluation and intelligent matching technical solution based on the degree of abnormal impact. The alarm level is determined by weighted calculation of spatial impact factor, time duration factor and traffic impact factor. The quartile algorithm and the plan template library are combined to realize dynamic matching of disposal plans, breaking through the technical bottleneck that the existing fixed plans cannot adapt to the dynamic changes of abnormalities, and realizing intelligent filtering of alarm information and precise scheduling of disposal resources.

[0016] The above content of the invention is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Other features, objects, and advantages of the present invention will become more apparent upon reading the detailed description of the non-limiting embodiments made with reference to the following drawings. The drawings are for the purpose of illustrating preferred embodiments only and are not to be construed as limiting the present invention. Like reference characters are used throughout the drawings to designate like parts.

[0018] Figure 1 This is a flow chart of the traffic anomaly processing method based on spatiotemporal fusion of multi-source data of the present invention.

[0019] Figure 2 This is a block diagram of the traffic anomaly handling system of the present invention. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0021] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be performed in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operations are completed, but can also have additional steps not included in the figures; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0022] Example 1: Figure 1 As shown, the traffic anomaly processing method based on spatiotemporal fusion of multi-source data includes the following steps: S1, parallel collection of traffic multimodal data to construct a spatiotemporal correlation matrix, and analysis of the spatiotemporal correlation matrix according to the spatiotemporal correlation analysis model to determine the initial abnormal section.

[0023] As an optional embodiment, the parallel collection of traffic flow data to construct a spatiotemporal correlation matrix includes the following steps: constructing a multimodal data distributed collection network, the multimodal data distributed collection network including a microwave radar array, video surveillance equipment, geomagnetic sensors, and mobile vehicle terminals deployed on the roadside; The collected multimodal traffic data is timestamped and stored in shards; the target road network is divided into three-dimensional grid cells according to geographic coordinates and a unique spatiotemporal index is configured for each grid cell; Feature extraction is performed on the multimodal traffic data within each grid cell to construct a multidimensional feature vector. Based on the multidimensional feature vector, the dynamic time warping algorithm is used to calculate the spatiotemporal similarity between different grid cells and construct a spatiotemporal correlation matrix. The multi-dimensional feature vector includes traffic flow features, abnormal event features, environmental features and historical data features.

[0024] It can be understood that this embodiment adopts the collaborative technical means of a multimodal data distributed acquisition network and a dynamic time warping algorithm. Traffic flow parameters, abnormal events and environmental data are captured in real time through multiple terminals such as roadside microwave radar arrays and video surveillance equipment. A unified data benchmark is constructed through timestamp labeling and three-dimensional grid spatiotemporal indexing. Feature vectors are then extracted from multiple dimensions such as traffic flow characteristics, abnormal event characteristics, environmental characteristics, and historical data characteristics. The degree of spatiotemporal correlation between different grid units is quantified with the help of a dynamic time warping algorithm, breaking through the technical bottleneck of the traditional single data source that is difficult to depict the spatiotemporal dynamic correlation of traffic flow, and providing a quantitative analysis basis with spatiotemporal consistency for the early identification of traffic anomalies and the deduction of the impact range.

[0025] For example, during a typical morning rush hour scenario in a city, when a vehicle breaks down on a section of a main road, a roadside microwave radar array captures the sudden drop in traffic speed in the area at millisecond frequency. Video surveillance equipment uses AI recognition to simultaneously generate an accident type tag, while geomagnetic sensors record abnormal fluctuations in lane occupancy. Simultaneously, a vehicle terminal equipped with a navigation app transmits real-time vehicle trajectory deviation information. This multi-source data is timestamped and stored in shards on edge servers. For example, the target road network is divided into three-dimensional units with spatiotemporal indexes (e.g., 116.4° East longitude, 39.9° North latitude, 7:30:15 AM).

[0026] Furthermore, during the feature extraction phase, each grid cell is aggregated to form a four-dimensional feature vector: traffic features include the difference between the average vehicle speed during that time period and the historical traffic flow during the same period; abnormal event features carry the accident code from video recognition; environmental features are linked to real-time meteorological data (such as whether it has rained); and historical data features introduce a baseline of congestion probability for the same time period. Subsequently, a dynamic time warping algorithm is used to align the feature sequences of different grid cells. For example, the similarity in speed variation between the grid 500 meters upstream of the anchor point and the accident grid in the temporal dimension is calculated. This constructs a matrix representing the strength of spatiotemporal correlations. Specifically, the higher the value in the matrix, the more it reflects the spatiotemporal propagation pattern, such as "ramp congestion → main road queue," thus providing a quantitative basis for the precise positioning of subsequent abnormal sections.

[0027] It should be noted that the core principle of the Dynamic Time Warping (DTW) algorithm is to solve the nonlinear alignment problem of time series through a dynamic programming strategy. Its core goal is to construct optimal matching paths between time series, allowing for local stretching or compression on the time axis to overcome the limitations of traditional Euclidean distance in dealing with sequences of unequal length or time misalignment. In the transportation sector, this algorithm is primarily used to model the spatiotemporal correlations of heterogeneous data from multiple sources. For example, when traffic flow data collected by roadside microwave radar (high-frequency sampling), video surveillance (fixed frame rate), and on-board terminals (asynchronous upload) have different sampling frequencies, the DTW algorithm can dynamically align time series such as speed and flow across different grid cells to quantify their spatiotemporal similarity. Typical application scenarios include capturing the spatiotemporal propagation of congestion from interchange ramps to main roads during the morning rush hour, or analyzing the impact of sudden accidents on upstream and downstream road networks. DTW calculates the matching degree of speed series between the accident grid and the grid 500 meters upstream, constructing a matrix representing the strength of the "spatial proximity-temporal delay" correlation, providing a quantitative basis for the spatiotemporal dynamic analysis of traffic anomalies.

[0028] As an optional embodiment, the method of analyzing the spatiotemporal correlation matrix according to the spatiotemporal correlation analysis model to determine the initial abnormal section includes the following steps: The fluctuation factor of each element in the spatiotemporal correlation matrix is ​​determined based on the spatiotemporal correlation analysis model. The dynamic threshold of each element is calculated based on the fluctuation factor combined with the historical mean and standard deviation of the element. Elements with actual element values ​​greater than the dynamic threshold are marked as outliers and an outlier set is generated. The DBSCAN clustering algorithm is used to perform cluster analysis on the abnormal point set to determine the initial abnormal cluster set. The temporal continuity and spatial connectivity of each abnormal cluster in the initial abnormal cluster set are verified to determine the target abnormal cluster set. The abnormal clusters in the target abnormal cluster set are projected onto the actual road network. If the abnormal clusters span different road types, the initial abnormal segment is determined based on the road boundary. If the abnormal clusters contain discontinuous roads, the isolated segments are removed and the main connected roads are retained to determine the initial abnormal segment.

[0029] It is understandable that during the morning rush hour on a city's main roads, when a multi-vehicle rear-end collision occurs on an overpass ramp, this embodiment achieves intelligent identification and precise segmentation of abnormal sections through the collaborative technical means of the spatiotemporal correlation analysis model and the DBSCAN clustering algorithm. For example, based on real-time traffic data collected by microwave radar, the system first calculates the fluctuation factor of each element in the spatiotemporal correlation matrix (i.e., the degree of deviation between the real-time value and the historical mean value for the same period), and dynamically generates an anomaly determination threshold using the mean plus 2 times the standard deviation (the anomaly determination threshold is expressed as: T = μ + λ × g × σ, where σ is the standard deviation of historical data, λ is the empirical coefficient, g is the fluctuation factor, and μ is the historical mean value for the same period). For example, when the traffic mutation rate fluctuation factor of a grid cell exceeds the threshold, it is marked as an anomaly. Subsequently, the outliers were clustered using the DBSCAN algorithm, setting a spatial neighborhood radius of 100 meters and a minimum number of points of 3. Consecutive outliers at ramp entrances and upstream of main roads were clustered into initial outlier clusters. The results verified that their timestamp differences were less than 15 minutes (temporal continuity) and that there were no gaps between adjacent grids (spatial connectivity), forming the target outlier cluster. When this cluster was projected onto the actual road network, if the outlier cluster spanned ramps and main roads (different road types), the segment was cut based on the administrative boundary of the road. If it included non-continuous roads separated by intersections, the isolated grids were removed, and only the main connected roads (e.g., the east-west section of the main road from K12+000 to K12+800) were retained as the initial outlier segments. This process dynamically characterizes the degree of data deviation through a fluctuation factor, adaptively adjusts the threshold based on the standard deviation, and combines the road network topology to constrain the clustering boundaries. This overcomes the limitations of traditional static threshold methods, which are unable to adapt to the spatiotemporal dynamics of traffic flow and the road network structure, and achieves a precise mapping of outlier segments to actual road units.

[0030] S2. Track the abnormal points in the initial abnormal section according to the grid shrinkage strategy to determine the target abnormal section.

[0031] As an optional embodiment, the method of tracking the abnormal points of the initial abnormal section according to the grid shrinkage strategy to determine the target abnormal section includes the following steps: The abnormal points corresponding to the initial abnormal segments are extracted in sequence, and the true element values ​​corresponding to the abnormal points are normalized to obtain the abnormal intensity; Determine the importance of a segment based on anomaly intensity, spatial correlation, grid connectivity, and their corresponding weight factors; sort the anomaly segments from largest to smallest according to their importance, and select the top N anomaly segments as candidate segments; verify the spatial connectivity of the candidate segments; if there are isolated segments, shrink the candidate segments by a shrinkage factor a until the candidate segments form a single connected area, obtaining the target anomaly segment; The minimum convex hull of the filtered target abnormal segment is calculated to obtain the polygonal boundary; the spatial centroid of the polygonal boundary is used as the target abnormal point.

[0032] Furthermore, the anomaly intensity reflects the relative significance of the abnormal characteristics of the grid unit. It is obtained by normalizing the real-time monitoring values ​​(such as vehicle speed and traffic mutation rate). The anomaly intensity formula of the i-th grid unit is: Among them, X i is the true element value of the i-th grid (such as the speed reduction), X max and X min They represent the maximum and minimum values ​​of the indicator in the initial abnormal segment respectively.

[0033] Furthermore, the spatial correlation degree represents the spatial proximity between the grid unit and the geometric center of the segment. The inverse of the Euclidean distance is taken. The spatial correlation degree formula of the i-th grid unit is: Among them, (x i ,y i ) is the Gaussian projection coordinate of the i grid cell; (x c ,y c ) are the geometric center coordinates of the initial abnormal section.

[0034] Furthermore, the grid connectivity c represents the number of abnormal grids in the eight-neighborhood, reflecting the closeness of spatial association. With grid i as the center, the number of grids marked as abnormal in the upper, lower, left, right and four diagonal neighborhoods is counted.

[0035] Therefore, the calculation formula for segment importance is: I i =β1×f i +β2×d i +β3×c i , β1, β2, and β3 are preset weight coefficients (the default values ​​are β1 = 0.5, β2 = 0.3, and β3 = 0.2), which are determined through expert experience or historical data training.

[0036] It is understandable that in the scenario of a sudden multi-vehicle rear-end collision on a city's main road during the morning rush hour, this embodiment uses a grid contraction strategy and a synergistic technical approach to achieve dynamic optimization of abnormal sections from rough positioning to precise locking. For example, the system first extracts abnormal points from each grid cell within the initial abnormal section, normalizes the real element values ​​such as vehicle speed drops and traffic flow mutations to an abnormal intensity between 0 and 1 (for example, a grid cell with a speed below 20% of the threshold is normalized to 0.7). The spatial correlation is defined by combining the inverse of the distance from the grid to the section center and the grid connectivity by the number of abnormal grids in the eight-neighborhood area. The importance of the section is calculated using the weighted formula I (corresponding to the weights of abnormal intensity / spatial correlation / connectivity, respectively). Taking an initial section as an example, the top 30% of the grids are selected as candidate sections after being sorted by importance. If there are isolated sub-regions separated by intersections, the boundary is iteratively contracted by a contraction factor of 0.9 until a single connected region is formed along the main road. Finally, the minimum convex hull algorithm was used to fit the segment boundaries, and the polygon centroid (e.g., 39.9°N, 116.4°E) was used as the target anomaly point. This process dynamically screens the core anomaly area using multi-dimensional weights. Combining spatial connectivity constraints with network topology corrections, this overcomes the limitations of traditional clustering algorithms, which cannot remove edge noise. This improves anomaly location accuracy from hundreds of meters in the initial segment to 50 meters in the core impact area, providing a spatiotemporal benchmark for the precise scheduling of emergency resources.

[0037] It should be noted that the core of the minimum convex hull algorithm is to solve the minimum convex polygon containing a given point set. Its mathematical essence is to determine the outer boundary of the point set through geometric construction. In the traffic anomaly processing scenario, the minimum convex hull algorithm is used for boundary refinement and spatial positioning of the target abnormal section. Technical personnel in this field can, based on the algorithm principles, correct the spatial boundaries of the abnormal grid after screening out the core abnormal grid through the grid shrinkage strategy, effectively solving the problems of blurred boundaries of abnormal sections and interference from noise points, and providing an accurate spatial benchmark for subsequent matching of disposal plans.

[0038] S3. Determine an alarm screening strategy based on the abnormal impact degree of the target abnormal section; match an abnormality handling plan based on the alarm screening strategy and the alarm level of the target abnormal section.

[0039] As an optional embodiment, an alarm screening strategy is determined according to the abnormal impact degree of the target abnormal section; and an abnormality handling plan is matched according to the alarm screening strategy and the alarm level of the target abnormal section, including the following steps: The spatial impact factor is determined based on the road length, number of lanes, and area covered by the target abnormal section; the temporal duration factor is determined based on the duration and frequency of the abnormal event; and the traffic impact factor is determined based on the traffic flow reduction rate, average speed reduction, and queue length increase. Determine the abnormal impact degree based on the spatial impact factor, time duration factor, traffic impact factor and their corresponding weight coefficients; use the quartile algorithm to classify the abnormal impact degree into alarm levels to determine the spatial information, time information and event information corresponding to the target abnormal point; Determine the key features of the alarm based on spatial information, time information and event information, extract the alarm features in the alarm information and compare them with the key features of the alarm, filter the alarm information based on the comparison results to determine the alarm point, determine the urgency of the abnormal point based on the alarm point and the alarm frequency of the alarm point; formulate corresponding abnormal handling plans based on the alarm level and its corresponding urgency.

[0040] It should be noted that the spatial spread factor represents the coverage (spread range) of the abnormal event in the geographic space. The road length, number of lanes and regional area are normalized separately, and the normalized parameters are weighted and summed to obtain the spatial spread factor; the temporal duration factor represents the temporal severity of the anomaly, and the duration directly reflects the degree of continuous interference of the anomaly on the traffic system; the duration and frequency are normalized separately and then weighted calculated to obtain the temporal duration factor; the traffic impact factor directly reflects the degree of damage to the traffic flow. The traffic flow decline rate, average speed reduction and queue length increase are normalized and then weighted calculated to obtain the traffic impact factor.

[0041] It is understandable that in the scenario of a sudden multi-vehicle rear-end collision on a city's main road, this embodiment uses a collaborative technical approach of multi-dimensional factor weighted evaluation and quartile grading to achieve intelligent screening of abnormal alarms and precise matching of treatment plans. For example, the system first determines the spatial impact factor based on the spatial characteristics of the accident section, which covers 2 kilometers of road and 4 lanes. It then generates a time duration factor based on the frequency of the abnormality lasting 30 minutes and occurring twice a month during the same period in history. It then constructs a traffic impact factor based on traffic impact data such as a 35% drop in traffic volume, a 50% drop in vehicle speed, and a 1-kilometer increase in queue length. The degree of abnormal impact is calculated using a weighted formula (e.g., a spatial / temporal / traffic impact weighting of 4:3:3). Using a quartile algorithm, the severity of the impact is divided into three alarm levels: red (top 25%), yellow (25%-75%), and blue (bottom 25%). Key features of the alarm are extracted, including the spatial coordinates of the accident zone (e.g., 39.9°N, 116.4°E), the time window (7:30-8:00 AM), and the event type (rear-end collision). When receiving multiple alarms, the system automatically compares the alarm features with key features for spatial proximity (e.g., within 500 meters), temporal consistency (±15 minutes), and event matching, filtering out valid alarm points and determining the level of urgency based on alarm frequency (e.g., three or more). For example, a red alarm level combined with a high-frequency alarm automatically matches a Level I response plan, triggering measures such as police arrival within five minutes and detour guidance on main road guidance screens. This process quantifies the impact of anomalies through multi-dimensional factors, scientifically classifies using the quartile algorithm, and intelligently screens through feature comparison. It breaks through the limitations of traditional fixed plans that are unable to adapt to the dynamic characteristics of anomalies, and realizes intelligent decision-making throughout the entire process from abnormal feature analysis to the generation of disposal plans, providing precise resource scheduling strategies for traffic emergency management.

[0042] S4. Real-time monitoring of traffic flow indicators and the completion of abnormal handling plans to determine traffic flow conditions, and update traffic condition reports in real time.

[0043] As an optional embodiment, the real-time monitoring of traffic flow indicators and the completion of abnormality handling plans to determine traffic flow conditions and update traffic condition reports in real time includes the following steps: The traffic flow indicators of the target abnormal section are obtained through a multimodal data distributed acquisition network, and the traffic flow degree is calculated based on the traffic flow indicators. When the traffic flow degree exceeds the set threshold, the task list of the abnormal handling plan is retrieved, and the traffic congestion situation is determined based on the completion rate of the abnormal list. The traffic congestion situation is then simultaneously published to the traffic broadcast platform. The traffic flow indicators include traffic flow indicators, vehicle speed indicators and vehicle queue indicators.

[0044] As an optional embodiment, the formula for traffic flow is expressed as: Among them, α1, α2, and α3 are the weight coefficients of the corresponding indicators, Q t is the real-time traffic volume of each lane, Q0 is the historical baseline traffic volume, V t is the average speed, V0 is the free flow speed, L t is the queue length, L0 is the initial queue length.

[0045] It is understood that in the handling of sudden traffic accidents on urban main roads, this embodiment uses a multimodal data distributed collection network and a joint assessment of traffic flow and completion to achieve real-time quantification of the effectiveness of traffic anomaly handling and dynamic updating of road condition information. For example, when the accident handling enters the obstacle clearance phase, the roadside microwave radar and the on-board terminal continuously collect traffic flow indicators such as the flow recovery rate, vehicle speed recovery rate, and queue length reduction in the target section. The system uses a preset model to map these into a traffic flow degree in the range of 0-1 (for example, when traffic returns to 60% of normal levels and vehicle speeds increase to 40 km / h, the calculated flow degree is 0.6). When the degree of unblocking exceeds a threshold (e.g., 0.5), the system automatically retrieves the task list for the disposal plan, compares the availability of on-site police personnel, the progress of the tow truck, and other task completion rates (e.g., 2 out of 3 disposal tasks have been completed), and comprehensively determines the congestion status. For example, if the degree of unblocking reaches 0.7 and the task completion rate is 80%, a message is generated stating that "the accident section has been partially unblocked; it is recommended to maintain safe distance between vehicles." This message is also pushed to the navigation app and traffic broadcast platform. By dynamically coupling real-time monitoring of unblocking indicators with the progress of disposal, this solution overcomes the limitations of traditional manual inspections, which often suffer from delayed feedback and ambiguous evaluation of disposal results. It achieves fully automated, closed-loop management of the entire process, from exception handling to road condition updates, providing the public with traffic information services accurate to the minute.

[0046] Embodiment 2: A technical solution also provided in the embodiment of the present invention is a traffic anomaly processing system, such as Figure 2 Shown include: Preprocessing module 101: collects multimodal traffic data in parallel to construct a spatiotemporal correlation matrix, and analyzes the spatiotemporal correlation matrix according to a spatiotemporal correlation analysis model to determine the initial abnormal section; Anomaly Location Module 102: Tracks anomaly points in the initial anomaly section according to the grid contraction strategy to determine the target anomaly section. Strategy Generation Module 103: Determines an alarm screening strategy based on the anomaly impact of the target anomaly section. Matches an anomaly handling plan based on the alarm screening strategy and the alarm level of the target anomaly section. Joint reporting module 104: real-time monitoring of traffic flow indicators and the degree of completion of abnormality handling plans to determine traffic flow conditions, and update traffic condition reports in real time.

[0047] This embodiment has at least the following substantial technical effects: the pre-processing module acquires microwave radar traffic data, video surveillance images and vehicle trajectory information in parallel through a multimodal data distributed acquisition network, constructs a spatiotemporal correlation matrix with the help of a dynamic time warping algorithm, and accurately identifies the initial abnormal section in combination with a spatiotemporal correlation analysis model, thus breaking through the spatiotemporal heterogeneity limitation of a single data source; the abnormal positioning module uses a grid contraction strategy, and through a multi-factor evaluation of abnormal intensity, spatial correlation, and grid connectivity and DBSCAN clustering, combined with a minimum convex hull algorithm, improves the positioning accuracy from hundreds of meters in the initial section to 50 meters (or even smaller) in the core impact area, effectively eliminating edge noise; the strategy generation module uses a spatial sweep factor (road length) to accurately identify the initial abnormal section and effectively eliminate ... It uses a three-dimensional assessment of factors (abnormality / number of lanes / area), time duration factor (abnormality duration / frequency of occurrence), and traffic impact factor (change in flow rate / vehicle speed / queue length), divides alarm levels using the quartile algorithm, and intelligently matches differentiated handling plans from traffic diversion to emergency obstacle removal; the joint reporting module collects traffic flow recovery rate, vehicle speed recovery rate and other traffic flow indicators in real time, and dynamically updates road condition reports based on the completion of handling tasks. Through the closed-loop collaboration of the four major modules, the system realizes full-chain automation from anomaly detection and precise positioning to handling decision-making and information release, breaking through the technical bottlenecks of data fragmentation and response lag in traditional traffic management, and providing urban traffic emergency management with an intelligent solution with temporal and spatial consistency.

[0048] Embodiment 3: A technical solution provided in the embodiment of the present invention is: an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, it implements the steps of a traffic anomaly handling method based on spatiotemporal fusion of multi-source data.

[0049] Embodiment 4: A technical solution provided in an embodiment of the present invention is: a storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are loaded and executed by a processor, the steps of a traffic anomaly handling method based on spatiotemporal fusion of multi-source data are implemented.

[0050] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional modules as needed, that is, the internal structure of the specific device can be divided into different functional modules to complete all or part of the functions described above.

[0051] In the embodiments provided in this application, it should be understood that the disclosed structures and methods can be implemented in other ways. For example, the embodiments of the structure described above are merely illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another structure, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, structure or unit, which can be electrical, mechanical or other forms.

[0052] Units described as separate components may or may not be physically separate, and components shown as units may be one physical unit or multiple physical units, that is, they may be located in one place or distributed in multiple places. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0053] In addition, the functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or software functional units.

[0054] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a device (which can be a single-chip microcomputer, chip, etc.) or a processor (processor) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0055] The specific implementation described above is a preferred implementation of the traffic anomaly handling method and system based on spatiotemporal fusion of multi-source data of the present invention, and is not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific implementation. Any equivalent changes made in accordance with the shape and structure of the present invention are within the scope of protection of the present invention.

Claims

1. Traffic anomaly processing method based on spatiotemporal fusion of multi-source data, characterized by: The steps include: Parallel collection of multimodal traffic data to construct a spatiotemporal correlation matrix, which is then analyzed using a spatiotemporal correlation analysis model to determine the initial abnormal section. According to the grid shrinkage strategy, the abnormal points in the initial abnormal section are tracked to determine the target abnormal section; Determine the alarm screening strategy based on the abnormal impact of the target abnormal section; match the abnormality handling plan based on the alarm screening strategy and the alarm level of the target abnormal section; Real-time monitoring of traffic flow indicators and the completion of abnormal handling plans to determine traffic flow conditions, and update traffic conditions reports in real time.

2. The traffic anomaly processing method based on spatiotemporal fusion of multi-source data according to claim 1 is characterized in that: The parallel collection of traffic flow data to construct a spatiotemporal correlation matrix includes the following steps: Constructing a multimodal data distributed acquisition network, comprising a microwave radar array, video surveillance equipment, geomagnetic sensors, and mobile vehicle terminals deployed on the roadside; The collected multimodal traffic data is timestamped and stored in shards; the target road network is divided into three-dimensional grid cells according to geographic coordinates and a unique spatiotemporal index is configured for each grid cell; Extract features from multimodal traffic data within each grid cell to construct a multi-dimensional feature vector; Based on the multi-dimensional feature vector, the spatiotemporal similarity between different grid cells is calculated by the dynamic time warping algorithm to construct the spatiotemporal correlation matrix; The multi-dimensional feature vector includes traffic flow features, abnormal event features, environmental features and historical data features.

3. The traffic anomaly processing method based on spatiotemporal fusion of multi-source data according to claim 1 or 2 is characterized in that: The method of analyzing the spatiotemporal correlation matrix according to the spatiotemporal correlation analysis model to determine the initial abnormal section comprises the following steps: The fluctuation factor of each element in the spatiotemporal correlation matrix is ​​determined based on the spatiotemporal correlation analysis model. The dynamic threshold of each element is calculated based on the fluctuation factor combined with the historical mean and standard deviation of the element. Elements with actual element values ​​greater than the dynamic threshold are marked as outliers and an outlier set is generated. The DBSCAN clustering algorithm is used to perform cluster analysis on the abnormal point set to determine the initial abnormal cluster set. The temporal continuity and spatial connectivity of each abnormal cluster in the initial abnormal cluster set are verified to determine the target abnormal cluster set. The abnormal clusters in the target abnormal cluster set are projected onto the actual road network. If the abnormal clusters span different road types, the initial abnormal segment is determined based on the road boundary. If the abnormal clusters contain discontinuous roads, the isolated segments are removed and the main connected roads are retained to determine the initial abnormal segment.

4. The traffic anomaly processing method based on spatiotemporal fusion of multi-source data according to claim 1 is characterized in that: The method of tracking the abnormal points of the initial abnormal section according to the grid shrinkage strategy to determine the target abnormal section includes the following steps: sequentially extracting the abnormal points corresponding to the initial abnormal section, and normalizing the real element values ​​corresponding to the abnormal points to obtain the abnormal intensity; Determine the importance of a segment based on anomaly intensity, spatial correlation, grid connectivity, and their corresponding weight factors; sort the anomaly segments from largest to smallest according to their importance, and select the top N anomaly segments as candidate segments; verify the spatial connectivity of the candidate segments; if there are isolated segments, shrink the candidate segments by a shrinkage factor a until the candidate segments form a single connected area, obtaining the target anomaly segment; The minimum convex hull of the filtered target abnormal segment is calculated to obtain the polygonal boundary; the spatial centroid of the polygonal boundary is used as the target abnormal point.

5. The traffic anomaly processing method based on spatiotemporal fusion of multi-source data according to claim 1 is characterized in that: Determine the alarm screening strategy based on the abnormal impact degree of the target abnormal section; match the abnormality handling plan based on the alarm screening strategy and the alarm level of the target abnormal section, including the following steps: The spatial spread factor is determined based on the road length, number of lanes, and area covered by the target abnormal section, and the temporal duration factor is determined based on the duration of the abnormal event and the frequency of abnormal occurrence; Determine the traffic impact factor based on the traffic flow reduction rate, average vehicle speed reduction, and queue length increase; Determine the abnormal impact degree based on the spatial impact factor, time duration factor, traffic impact factor and their corresponding weight coefficients; The quartile algorithm is used to classify the alarm level of the abnormal impact to determine the spatial information, time information and event information corresponding to the target abnormal point; Determine the key features of the alarm based on spatial information, time information and event information, extract the alarm features in the alarm information and compare them with the key features of the alarm, filter the alarm information based on the comparison results to determine the alarm point, determine the urgency of the abnormal point based on the alarm point and the alarm frequency of the alarm point; formulate corresponding abnormal handling plans based on the alarm level and its corresponding urgency.

6. The traffic anomaly processing method based on spatiotemporal fusion of multi-source data according to claim 1 or 5, characterized in that: The real-time monitoring of traffic flow indicators and the completion of abnormality handling plans to determine traffic flow conditions and update traffic condition reports in real time includes the following steps: The traffic flow indicators of the target abnormal section are obtained through a multimodal data distributed acquisition network, and the traffic flow degree is calculated based on the traffic flow indicators. When the traffic flow degree exceeds the set threshold, the task list of the abnormal handling plan is retrieved, and the traffic congestion situation is determined based on the completion rate of the abnormal list. The traffic congestion situation is then simultaneously published to the traffic broadcast platform. The traffic flow indicators include traffic flow indicators, vehicle speed indicators and vehicle queue indicators.

7. The traffic anomaly processing method based on spatiotemporal fusion of multi-source data according to claim 6 is characterized in that: The formula for traffic flow is: Among them, α1, α2, and α3 are the weight coefficients of the corresponding indicators, Q t is the real-time traffic volume of each lane, Q0 is the historical baseline traffic volume, V t is the average speed, V0 is the free flow speed, L t is the queue length, and L0 is the initial queue length.

8. A traffic anomaly processing system, applicable to the traffic anomaly processing method based on spatiotemporal fusion of multi-source data as claimed in any one of claims 1 to 7, characterized in that: include: Preprocessing module: collects multimodal traffic data in parallel to construct a spatiotemporal correlation matrix, and analyzes the spatiotemporal correlation matrix according to the spatiotemporal correlation analysis model to determine the initial abnormal section; Anomaly location module: Tracks the anomaly points in the initial anomaly section according to the grid contraction strategy to determine the target anomaly section; Strategy generation module: determines the alarm screening strategy based on the abnormal impact of the target abnormal section; matches the abnormal handling plan based on the alarm screening strategy and the alarm level of the target abnormal section; Joint reporting module: Real-time monitoring of traffic flow indicators and the completion of abnormal handling plans to determine traffic flow conditions, and update traffic condition reports in real time.

9. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, the method implements the steps of the traffic anomaly processing method based on spatiotemporal fusion of multi-source data according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores computer-executable instructions. When the computer-executable instructions are loaded and executed by the processor, the steps of the traffic anomaly processing method based on spatiotemporal fusion of multi-source data according to any one of claims 1 to 7 are implemented.

Images