Trajectory analysis-based traffic abnormal event dynamic detection method and system

Abstract

The invention relates to the technical field of intelligent traffic, in particular to a traffic abnormal event dynamic detection method and system based on trajectory analysis, and the method comprises the steps: receiving original space-time trajectory data; preprocessing the original spatio-temporal trajectory data, and mapping the original spatio-temporal trajectory data to a road section sequence of a road network through a map matching algorithm; constructing a behavior model based on the track sequence after map matching, learning the track sequence of the normal behavior mode on the road section, and establishing an observation emission model and a state transition matrix; in the online stage, the log-likelihood value of an observation sequence and / or the similarity between the observation sequence and a normal behavior pattern cluster are / is calculated for a real-time track in a set sliding window, and when the log-likelihood value and / or the similarity exceed a set threshold value, the track is marked as abnormal; performing time-space aggregation on abnormal trajectories of the same road section or intersection in unit time according to single vehicle abnormality judgment, and triggering group abnormal event alarm when aggregation data exceed a preset value.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a method and system for dynamic detection of traffic anomalies based on trajectory analysis. Background Technology

[0002] In recent years, with the acceleration of urbanization and the rapid growth of motor vehicle ownership, traffic congestion, traffic accidents, and various traffic anomalies have occurred frequently, seriously impacting urban operation management, emergency response, and public travel safety. Existing traffic anomaly detection methods mainly include roadside fixed sensing (such as cameras and radar), macro-level traffic flow statistics (such as intersection signal sampling), and detection technologies based on floating cars / probe vehicles.

[0003] While existing traffic anomaly detection methods can detect congestion and accidents in specific scenarios, they still have some problems, such as limited sensing coverage and weak anti-interference ability: blind spots exist in the layout of roadside equipment, and the performance of cameras degrades under obstructed, low-light, or inclement weather conditions, making it difficult to fully perceive events; the deployment and maintenance costs of equipment such as geomagnetic coils are high, making it difficult to achieve large-scale coverage; it is difficult to balance spatiotemporal resolution and latency: macroscopic statistical methods are suitable for traffic flow analysis but are not sensitive enough to instantaneous violations (such as driving against traffic or temporary parking); while video-based human / visual algorithms usually require high computing resources and have recognition latency, making it difficult to meet the requirements of real-time early warning; the definition of anomalies is diverse and the detection scale is complex: traffic anomalies include single-vehicle violations (such as driving against traffic or illegal parking), short-term clustered events (small-scale collisions), and long-term congestion, etc., and the scale of events varies greatly in time and space, making it difficult for a single detection method to meet the needs of multi-scale and multi-type anomaly recognition.

[0004] The information disclosed in this background section is intended only to enhance the understanding of the general background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] This invention provides a method and system for dynamic detection of traffic anomalies based on trajectory analysis, thereby effectively solving the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: a dynamic detection method for traffic anomalies based on trajectory analysis, comprising the following steps:

[0007] Receive raw spatiotemporal trajectory data including vehicle-mounted positioning devices, mobile terminals, and roadside sensing devices; preprocess the raw spatiotemporal trajectory data and map it to road segment sequences of the road network using a map matching algorithm;

[0008] Based on the trajectory sequence after map matching, a behavior model is constructed, and the trajectory sequence of normal behavior patterns on the road segment is learned to establish an observation emission model and a state transition matrix.

[0009] During the online phase, the log-likelihood value and / or similarity to the normal behavior pattern cluster are calculated for the real-time trajectory within a set sliding window. When the log-likelihood value and / or similarity exceed a set threshold, the trajectory is marked as abnormal.

[0010] Based on the identification of single-vehicle anomalies, the system performs spatiotemporal aggregation on abnormal trajectories of the same road segment or intersection within a unit of time. When the aggregated data exceeds a preset value, a group anomaly event alarm is triggered.

[0011] Furthermore, the preprocessing of the original spatiotemporal trajectory data includes:

[0012] Kalman filtering was used to suppress noise in the original spatiotemporal trajectory data, and interpolation based on a uniform acceleration model was performed on the missing time points.

[0013] Furthermore, the road segment sequence mapped to the road network via the map matching algorithm includes:

[0014] A hidden Markov model map matching algorithm is adopted, taking into account projection error and turning cost, and outputs road segment sequence and corresponding timestamp mapping.

[0015] Furthermore, the observation launch model is modeled using a multivariate Gaussian distribution, which includes observation vectors comprising velocity, acceleration, rate of change of heading, and dwell time. The parameters of the observation launch model are trained using the Baum-Welch algorithm of a Hidden Markov Model.

[0016] Furthermore, when calculating the similarity with the normal behavior pattern cluster, the trajectory similarity is calculated based on dynamic time warping (DTW), and the normal trajectory cluster is obtained by using density clustering algorithm in the similarity space.

[0017] Furthermore, the step of marking the trajectory as an anomaly when the log-likelihood value and / or similarity exceed a set threshold includes:

[0018] When the log-likelihood value is lower than the first set threshold and / or the similarity distance exceeds the second set threshold, the trajectory is marked as abnormal.

[0019] Furthermore, it also includes online incremental learning, including: incrementally updating the behavior model in batches for trajectory samples determined to be normal, and learning from scene changes.

[0020] The present invention also includes a dynamic detection system for traffic anomalies based on trajectory analysis, using the method described above, wherein the system comprises:

[0021] The data acquisition and matching unit is used to receive raw spatiotemporal trajectory data from vehicle-mounted positioning devices, mobile terminals, and roadside sensing devices; preprocess the raw spatiotemporal trajectory data; and map it to road segment sequences of the road network using a map matching algorithm.

[0022] The behavior modeling unit is used to construct a behavior model based on the trajectory sequence after map matching, learn the trajectory sequence of normal behavior patterns on the road segment, and establish an observation emission model and a state transition matrix.

[0023] The detection unit is used to calculate the log-likelihood value and / or similarity to the normal behavior pattern cluster of the observation sequence of the real-time trajectory within a set sliding window during the online phase. When the log-likelihood value and / or similarity exceeds a set threshold, the trajectory is marked as abnormal.

[0024] The aggregation unit is used to perform spatiotemporal aggregation of abnormal trajectories on the same road segment or intersection within a unit of time based on the single vehicle anomaly identification. When the aggregated data exceeds a preset value, a group anomaly event alarm is triggered.

[0025] The present invention also includes a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method as described above.

[0026] The present invention also includes a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described above.

[0027] The beneficial effects of this invention are as follows: By preprocessing the original spatiotemporal trajectory data and mapping it to the road segment sequence of the road network through a map matching algorithm, the trajectory data has a wide coverage range and is still effective in areas not visible to cameras; by constructing a behavior model, the trajectory sequence of normal behavior patterns on the road segment is learned, and through trajectory-level statistical learning and rule fusion, real-time detection and positioning of single-vehicle anomalies and group congestion / accidents can be achieved over a large area, covering camera blind spots, with strong adaptability and a certain degree of interpretability. It is suitable for deployment in urban intelligent transportation and emergency dispatch systems, can make full use of large-scale spatiotemporal trajectory data, and can take into account the identification of single-vehicle behavior anomalies and group events while ensuring timeliness, and has strong robustness and scalability. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a flowchart of the method of the present invention;

[0030] Figure 2 This is a schematic diagram of the system structure of the present invention;

[0031] Figure 3 This is a schematic diagram of the structure of the computer device of the present invention. Detailed Implementation

[0032] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0033] Example 1:

[0034] like Figure 1 As shown: A dynamic detection method for traffic anomalies based on trajectory analysis includes the following steps:

[0035] Receive raw spatiotemporal trajectory data from vehicle-mounted positioning devices, mobile terminals, and roadside sensing devices; preprocess the raw spatiotemporal trajectory data and map it to road segment sequences of the road network using a map matching algorithm;

[0036] Based on the trajectory sequence after map matching, a behavior model is constructed, and the trajectory sequence of normal behavior patterns on the road segment is learned to establish an observation emission model and a state transition matrix.

[0037] During the online phase, the log-likelihood value and / or similarity to the normal behavior pattern cluster are calculated for the real-time trajectory within a set sliding window. When the log-likelihood value and / or similarity exceed a set threshold, the trajectory is marked as abnormal.

[0038] Based on the identification of single-vehicle anomalies, the system performs spatiotemporal aggregation on abnormal trajectories of the same road segment or intersection within a unit of time. When the aggregated data exceeds a preset value, a group anomaly event alarm is triggered.

[0039] By preprocessing the raw spatiotemporal trajectory data and mapping it to road segment sequences in the road network using a map matching algorithm, the system leverages the wide coverage of the trajectory data, remaining effective even in areas invisible to cameras. By constructing a behavioral model and learning the trajectory sequences of normal behavioral patterns on the road segment, and through trajectory-level statistical learning and rule fusion, it can achieve real-time detection and location of abnormal single-vehicle behavior and group congestion / accidents over a large area. It covers camera blind spots, is highly adaptable, and has a certain degree of interpretability, making it suitable for deployment in urban intelligent transportation and emergency dispatch systems. It can fully utilize large-scale spatiotemporal trajectory data, ensuring timeliness while also recognizing abnormal single-vehicle behavior and group events, and possesses strong robustness and scalability.

[0040] In this embodiment, preprocessing of the raw spatiotemporal trajectory data includes:

[0041] Kalman filtering was used to suppress noise in the original spatiotemporal trajectory data, and interpolation based on a uniform acceleration model was performed on the missing time points.

[0042] The sequence of road segments mapped to the road network using map matching algorithms includes:

[0043] A hidden Markov model map matching algorithm is adopted, taking into account projection error and turning cost, and outputs road segment sequence and corresponding timestamp mapping.

[0044] The observation launch model is modeled using a multivariate Gaussian distribution, which includes observation vectors. These vectors include velocity, acceleration, rate of change of heading, and dwell time. The parameters of the observation launch model are trained using the Baum-Welch algorithm of a hidden Markov model.

[0045] In this embodiment, when calculating the similarity with the normal behavior pattern cluster, the trajectory similarity is calculated based on dynamic time warping (DTW), and the normal trajectory cluster is obtained using a density clustering algorithm within the similarity space.

[0046] When the log-likelihood value and / or similarity exceed a set threshold, the trajectory is marked as an anomaly, including:

[0047] When the log-likelihood value is lower than the first set threshold and / or the similarity distance exceeds the second set threshold, the trajectory is marked as abnormal.

[0048] This also includes online incremental learning, which involves incrementally updating the behavior model in batches for trajectory samples that are judged to be normal, and learning from scene changes.

[0049] like Figure 2 As shown, this embodiment also includes a dynamic detection system for traffic anomalies based on trajectory analysis. Using the method described above, the system includes:

[0050] The data acquisition and matching unit is used to receive raw spatiotemporal trajectory data from vehicle-mounted positioning devices, mobile terminals, and roadside sensing devices; preprocess the raw spatiotemporal trajectory data; and map it to the road segment sequence of the road network through a map matching algorithm.

[0051] The behavior modeling unit is used to construct a behavior model based on the trajectory sequence after map matching, learn the trajectory sequence of normal behavior patterns on the road segment, and establish an observation emission model and a state transition matrix.

[0052] The detection unit is used to calculate the log-likelihood value and / or similarity to the normal behavior pattern cluster of the observation sequence of the real-time trajectory within a set sliding window during the online phase. When the log-likelihood value and / or similarity exceeds a set threshold, the trajectory is marked as abnormal.

[0053] The aggregation unit is used to perform spatiotemporal aggregation of abnormal trajectories on the same road segment or intersection within a unit of time based on the single vehicle anomaly identification. When the aggregated data exceeds a preset value, a group anomaly event alarm is triggered.

[0054] Example 2:

[0055] This embodiment includes a dynamic detection method that can make full use of large-scale spatiotemporal trajectory data, take into account both abnormal bicycle behavior and group event identification while ensuring timeliness, and has strong robustness and scalability. It has important practical significance and application value.

[0056] Its core idea is:

[0057] 1. Preprocess the original trajectory and match it with the map to obtain a discrete state sequence based on the road network;

[0058] 2. Construct a multi-layered behavior model (macro-flow model + fine-grained trajectory pattern model), and learn "normal" behavior patterns through probabilistic models (HMM / GMM) or similarity-based trajectory clustering (DTW / road segment sequence frequency);

[0059] 3. During the online phase, calculate the likelihood of the behavior or the distance from the group center for the trajectory within the sliding window; low likelihood or high distance is considered anomaly.

[0060] 4. Introduce multi-scale detection and rule triggering (speed threshold, dwell time, reverse driving recognition), and combine it with spatiotemporal aggregation to realize automatic escalation of alarms for single vehicle anomalies and group events (such as accidents and congestion).

[0061] This method includes a training (offline or continuous learning) module, a real-time detection module, and an alarm / visualization interface. Trajectory: Represented as T = {(x i ,y i ,t i)}, i=1,2,…,N. Map Matching: M(T)=S={s j}, j=1,2,…,L, where s j This refers to the segment ID or a discrete unit on the segment. Behavior vector: the observation o at time i. i =[v i ,a i ,Δθ i [,τ_stop,…] represents speed, acceleration, heading change, dwell time, etc.

[0062] Data Acquisition and Preprocessing:

[0063] Data sources include vehicle GPS / OBU, mobile phone positioning, taxi dispatch data, and roadside perception, which are fused to form a trajectory stream. Preprocessing includes: temporal interpolation (linear or based on a uniform acceleration model), noise filtering (Kalman or particle filtering), and sampling rate normalization.

[0064] Map matching: Map GPS points to road network edges (using Hidden Markov Model matching, considering projection error and turning costs). The output is a road segment sequence S and a timestamp mapping.

[0065] Behavioral modeling - Hidden Markov Model (HMM) probability model;

[0066] 1. State definition: The state set mathematical{Q} represents the fine-grained behavioral units on the road segment (e.g., normal driving - speed limit section, slow driving - intersection, parking - parking lane, etc.). Each road segment can have multiple behavioral states.

[0067] 2. Observation model: Observation o i Modeled using Gaussian distribution:

[0068] P(o i |q i =s)=N(o i μ s ,Σ s );

[0069] 3. Transition probability: $A = [a_{uv}]$ is estimated from road network topology and historical behavior statistics.

[0070] a uv =P(q) t+1 =v|q t =u)∝α·I(u→v exists)+(1-α)·f_hist(u,v);

[0071] Where f_hist is the historical observation frequency and α is the topology priority weight.

[0072] 4. Initial distribution: estimated by the frequency of entry into the road segment.

[0073] 5. Training: The Baum-Welch (EM) algorithm is used to estimate {A, μ}. s ,Σ s Alternatively, segmented learning (training separately by road segment) can be used to reduce dimensionality.

[0074] 6. Anomaly Detection: Given a trajectory (or sliding window W), calculate the trajectory likelihood:

[0075] L(T)=P(O|λ)=Σ_q P(O,q|λ);

[0076] The log-likelihood of the trajectory, log L(T), is calculated using a forward algorithm. If log L(T) < τ_lik (threshold), the trajectory is considered abnormal. Simultaneously, the most probable state path is obtained using Viterbi; any high-risk transitions (e.g., sudden stop -> wrong-way driving) within this path also trigger an anomaly.

[0077] Behavioral Modeling: Trajectory Similarity and Clustering

[0078] 1. After segmenting the trajectory into a sliding window, use Dynamic Time Warping (DTW) or Frechet distance to calculate the similarity between trajectories.

[0079] 2. Use density clustering (DBSCAN / HDBSCAN) to cluster normal trajectory patterns within the similarity space. Isolated points (low density) are considered anomalous.

[0080] 3. Advantages: More intuitive, fewer assumptions about distribution; Disadvantages: High computational complexity, requires approximation or index structure (e.g., LB_Keogh lower bound, or trajectory fingerprinting).

[0081] Fusion of multi-scale and rule-based detection:

[0082] Instantaneous violations (rule detection): instantaneous speed exceeding the limit, reverse driving judgment (detection of contradiction between driving direction and road segment direction after map matching), and abnormal dwell time t_stop exceeding the threshold.

[0083] Group event detection: If the aggregated count of abnormal trajectories at a certain road segment or intersection exceeds a threshold k within a unit of time, a group event (possible accident or congestion) is reported. Spatiotemporal aggregation can be achieved using heatmaps and connectivity analysis.

[0084] Threshold and parameter selection:

[0085] Sliding window length W: It is recommended to select based on the event type; W = 5-15 seconds for instantaneous violations, and W = 60-300 seconds for congestion detection. Likelihood threshold t_Jik: Calculate the ROC curve using the validation set and select the point that maximizes £1. Aggregate threshold k: Can be dynamically set according to the importance of the road segment (smaller k for main roads, larger k for side roads).

[0086] Deployment architecture:

[0087] 1. Data layer: Kafka / Kinesis is used to access the trajectory stream, and a time-series database (TimescaleDB) is used to store the trajectory history;

[0088] 2. Preprocessing layer: Streaming map matching service (supports batch / stream), road network index (R-tree) based on local caching;

[0089] 3. Model layer: HMM / GMM service + clustering index service (approximate nearest neighbor), supporting online incremental learning;

[0090] 4. Alarm layer: rule engine + aggregator, which escalates anomalies to events and sends them to the operations / scheduling end;

[0091] 5. Visualization: Map-based heatmaps, trajectory playback, and alarm timelines.

[0092] This embodiment utilizes the wide coverage of trajectory data, which remains effective even in areas invisible to cameras; it combines probabilistic models with similarity clustering, balancing interpretability and flexibility; it supports online incremental learning and group event escalation, making it suitable for real-time scheduling.

[0093] Please see Figure 3 The diagram shows a structural schematic of a computer device provided in an embodiment of this application. An embodiment of this application provides a computer device 400, including a processor 410 and a memory 420. The memory 420 stores a computer program executable by the processor 410. When the computer program is executed by the processor 410, it performs the method described above.

[0094] This application embodiment also provides a storage medium 430, on which a computer program is stored, and the computer program is executed by a processor 410 to perform the above method.

[0095] The storage medium 430 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0096] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. "A plurality of" means two or more, unless otherwise explicitly specified.

[0097] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0098] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0099] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0100] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0101] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0102] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it includes one or a combination of the steps of the method embodiments.

[0103] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A dynamic detection method for traffic anomalies based on trajectory analysis, characterized in that, Includes the following steps: Receive raw spatiotemporal trajectory data including vehicle-mounted positioning devices, mobile terminals, and roadside sensing devices; preprocess the raw spatiotemporal trajectory data and map it to road segment sequences of the road network using a map matching algorithm; Based on the trajectory sequence after map matching, a behavior model is constructed, and the trajectory sequence of normal behavior patterns on the road segment is learned to establish an observation emission model and a state transition matrix. During the online phase, the log-likelihood value and / or similarity to the normal behavior pattern cluster are calculated for the real-time trajectory within a set sliding window. When the log-likelihood value and / or similarity exceed a set threshold, the trajectory is marked as abnormal. Based on the identification of single-vehicle anomalies, the system performs spatiotemporal aggregation on abnormal trajectories of the same road segment or intersection within a unit of time. When the aggregated data exceeds a preset value, a group anomaly event alarm is triggered.

2. The method for dynamic detection of traffic anomalies based on trajectory analysis according to claim 1, characterized in that, The preprocessing of the original spatiotemporal trajectory data includes: Kalman filtering was used to suppress noise in the original spatiotemporal trajectory data, and interpolation based on a uniform acceleration model was performed on the missing time points.

3. The method for dynamic detection of traffic anomalies based on trajectory analysis according to claim 1, characterized in that, The road segment sequence mapped to the road network by the map matching algorithm includes: A hidden Markov model map matching algorithm is adopted, taking into account projection error and turning cost, and outputs road segment sequence and corresponding timestamp mapping.

4. The method for dynamic detection of traffic anomalies based on trajectory analysis according to claim 1, characterized in that, The observation launch model is modeled using a multivariate Gaussian distribution, which includes observation vectors. These observation vectors include velocity, acceleration, rate of change of heading, and dwell time. The parameters of the observation launch model are trained using the Baum-Welch algorithm of a Hidden Markov Model.

5. The method for dynamic detection of traffic anomalies based on trajectory analysis according to claim 1, characterized in that, When calculating the similarity with normal behavior pattern clusters, trajectory similarity is calculated based on dynamic time warping (DTW), and normal trajectory clusters are obtained using density clustering algorithm within the similarity space.

6. The method for dynamic detection of traffic anomalies based on trajectory analysis according to claim 5, characterized in that, The step of marking the trajectory as an anomaly when the log-likelihood value and / or similarity exceed a set threshold includes: When the log-likelihood value is lower than a first set threshold and / or the similarity distance exceeds a second set threshold, the trajectory is marked as abnormal.

7. The method for dynamic detection of traffic anomalies based on trajectory analysis according to claim 1, characterized in that, It also includes online incremental learning, including: incrementally updating the behavior model in batches for trajectory samples that are judged to be normal, and learning from scene changes.

8. A dynamic detection system for traffic anomalies based on trajectory analysis, characterized in that, Using the method of any one of claims 1 to 7, the system comprises: The data acquisition and matching unit is used to receive raw spatiotemporal trajectory data from vehicle-mounted positioning devices, mobile terminals, and roadside sensing devices; preprocess the raw spatiotemporal trajectory data; and map it to road segment sequences of the road network using a map matching algorithm. The behavior modeling unit is used to construct a behavior model based on the trajectory sequence after map matching, learn the trajectory sequence of normal behavior patterns on the road segment, and establish an observation emission model and a state transition matrix. The detection unit is used to calculate the log-likelihood value and / or similarity to the normal behavior pattern cluster of the observation sequence of the real-time trajectory within a set sliding window during the online phase. When the log-likelihood value and / or similarity exceeds a set threshold, the trajectory is marked as abnormal. The aggregation unit is used to perform spatiotemporal aggregation of abnormal trajectories on the same road segment or intersection within a unit of time based on the single vehicle anomaly identification. When the aggregated data exceeds a preset value, a group anomaly event alarm is triggered.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1-7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images