Vehicle abnormal behavior detection method based on Beidou positioning
Through the Beidou positioning vehicle abnormal behavior detection system, combined with GPS electronic fence alarm and multi-source data integration, the abnormal behavior of the vehicle is identified, and the problem of insufficient positioning accuracy and stopping status monitoring in the existing technology is solved, and the risk of vehicle is achieved Effective control of behavior and improvement of supervision in the car rental industry.
Patent Information
- Application Number
- CN202411057620.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-12-03
AI Technical Summary
The existing GPS tracking system and traditional parking monitoring systems have shortcomings in positioning accuracy and vehicle parking status monitoring, and cannot effectively evaluate the risk behavior of the vehicle, especially in scenarios where signal coverage is insufficient or vehicle-intensive.
The vehicle abnormal behavior detection system based on Beidou positioning is adopted, combined with the GPS electronic fence alarm calculation module, the resident identification module and multi-source data integration, and cluster analysis is performed using K-Means, DBSCAN and KDE algorithms, combined with RNN/LSTM/Transformer The model conducts predictive modeling and risk scoring to achieve real-time monitoring and evaluation of vehicle abnormal behavior.
It has improved the efficiency of controlling the risk behavior of vehicles such as gathering multiple vehicles, long-term parking, and different locations, enhanced the supervision capabilities of the car rental industry, realized online and systematic risk control back-end management, and improved work efficiency.
Smart Images

Figure CN119068659A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of vehicle abnormal behavior detection, and more specifically, to a vehicle abnormal behavior detection method based on Beidou positioning. Background Art
[0002] Conventional GPS tracking systems in the car rental industry today: Although today’s GPS tracking systems can provide vehicle location information, their positioning accuracy may not be high due to issues such as signal blocking and multipath effects. This may lead to misjudgment of the vehicle’s actual location, especially in urban areas with tall buildings or insufficient signal coverage. In addition, the GPS tracking system cannot provide real-time monitoring of the vehicle’s parking status, nor can it effectively assess the risk behavior of the vehicle.
[0003] Traditional parking monitoring system: Traditional parking monitoring systems mainly rely on cameras and license plate recognition technology, which has a blind spot for monitoring the parked state of vehicles. Cameras may not be able to accurately identify the parked state of vehicles, especially in parking lots with dense vehicles or complex traffic scenes. In addition, traditional parking monitoring systems are usually unable to analyze and evaluate vehicle behavior in real time, nor can they effectively warn of risky vehicle behavior. Summary of the invention
[0004] The present invention provides a vehicle abnormal behavior detection method based on Beidou positioning, which solves the technical problems in related technologies and provides effective control methods for the lack of risk behaviors of vehicles such as multiple vehicles gathering, long-term parking, different locations, and leaving the place where vehicles are often parked / place of residence (prefectural-level city). It increases the supervision methods for rented vehicles, improves work efficiency, improves the online monitoring of post-loan risk control vehicles, and systematizes and standardizes the work of the risk control background group.
[0005] The present invention provides a vehicle abnormal behavior detection system based on Beidou positioning, comprising:
[0006] GPS electronic fence alarm calculation module, including:
[0007] A data receiving and collecting module, which is used to receive vehicle positioning data and obtain electronic fence data from a database;
[0008] An alarm module is used to classify the alarms into fence entry alarm, fence exit alarm and fence entry and exit alarm according to the change of the relative position of the vehicle to the fence;
[0009] Configure the dwell time to issue an alarm based on the change of the device signal relative to the fence and the dwell time;
[0010] A permanent location identification module, which is used to obtain the permanent location information of the vehicle;
[0011] Positioning difference alarm:
[0012] Calculate the positioning difference between the devices of each vehicle and issue an alarm when the difference is greater than the set threshold.
[0013] Furthermore, distributed deployment: the vehicle abnormal behavior detection system based on Beidou positioning is deployed on multiple servers and works collaboratively.
[0014] Furthermore, by performing cluster calculation on the latitude and longitude data of the vehicle, the permanent location of the vehicle is obtained.
[0015] Furthermore, the K-Means clustering algorithm, DBSCAN algorithm, and KDE kernel density algorithm are used to obtain the vehicle's permanent location information.
[0016] Furthermore, a distance calculation function is used to calculate the distance between devices, and a determination is made as to whether to issue an alarm based on a set distance threshold.
[0017] Get detailed information about vehicles and equipment by querying related tables.
[0018] The present invention provides a method for detecting abnormal vehicle behavior based on Beidou positioning, comprising the following steps:
[0019] 101, obtaining electronic fence data from a database;
[0020] 102, used to classify the entry and exit types into fence alarm, fence exit alarm and entry and exit fence alarm according to the change of the relative position of the vehicle to the fence;
[0021] 103, an alarm is issued according to the change and the duration of the device signal relative to the fence.
[0022] Furthermore, it also includes:
[0023] 201, Data Fusion and Preprocessing
[0024] Multi-source data integration: Integrate vehicle sensor data, meteorological data, traffic flow information, and driver health monitoring data.
[0025] Feature extraction: Extract driving style indicators, driving route preferences, and time patterns from raw data.
[0026] 202, Anomaly Detection and Pattern Recognition
[0027] Identify unusual patterns in driving behavior;
[0028] Cluster analysis: cluster driving patterns through clustering algorithms to identify groups with high-risk driving behaviors;
[0029] 203, Predictive Modeling
[0030] Sequence prediction model: Use RNN or LSTM or Transformer architecture to predict future driving behavior based on time series data;
[0031] 204, Risk Scoring and Intervention Strategies
[0032] Dynamic risk scoring: Assign a dynamic risk score to each vehicle based on prediction results and real-time data;
[0033] Personalized intervention: Send customized safety recommendations or warnings to drivers based on risk scores.
[0034] Furthermore, an autoencoder algorithm is applied to identify abnormal patterns in driving behavior.
[0035] Furthermore, the calculation formula of the sequence prediction model is as follows:
[0036] Forget gate: f t =σ(W f ·[h t-1 ,x t ]+b f )
[0037] Among them, f t is the output of the forget gate, W f and b f is the weight matrix and bias term, σ is the sigmoid activation function, x t Represents the tth sequence unit of the input time series.
[0038] Input Gate:
[0039] i t =σ(W i ·[h t-1 ,x t ]+b i )
[0040]
[0041] i t is the output of the input gate, is the candidate cell state, and tanh is the hyperbolic tangent activation function.
[0042] Cell status update:
[0043]
[0044] in, Represents the Hadamard product (element-by-element product).
[0045] Output Gate:
[0046] ot =σ(W o [h t-1 ,x t ]+b o )
[0047] Hide status updates:
[0048] where h t represents the hidden state output at the tth time step;
[0049] L t =FC(h t )
[0050] Where FC means fully connected, L t It indicates the type of driving behavior predicted at the next moment at the tth moment.
[0051] In one embodiment of the present invention, a computer storage medium is provided, which stores computer-readable instructions. When a computer reads the computer-readable instructions, the aforementioned vehicle abnormal behavior detection method based on Beidou positioning can be executed.
[0052] The beneficial effects of the present invention are: inventing an effective control method for the lack of risk behaviors of vehicles such as multiple vehicles gathering, long-term parking, different locations, leaving the regular parking place / place of residence (prefectural-level city), etc. It increases the supervision method of rented vehicles, improves work efficiency, improves the online monitoring of post-loan risk control vehicles, and systematizes and standardizes the work of the risk control background group. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] Figure 1 This is a schematic diagram of the operation results of the K-Means clustering algorithm of the present invention. Figure 1 ;
[0054] Figure 2 This is a schematic diagram of the operation results of the K-Means clustering algorithm of the present invention. Figure 2 ;
[0055] Figure 3 This is a schematic diagram of the operation results of the K-Means clustering algorithm of the present invention. Figure 3 ;
[0056] Figure 4 This is a schematic diagram of the operation results of the K-Means clustering algorithm of the present invention. Figure 4 ;
[0057] Figure 5 This is a schematic diagram of the operation results of the DBSCAN algorithm of the present invention. Figure 1 ;
[0058] Figure 6This is a schematic diagram of the operation results of the DBSCAN algorithm of the present invention. Figure 2 ;
[0059] Figure 7 This is a schematic diagram of the operation results of the DBSCAN algorithm of the present invention. Figure 3 ;
[0060] Figure 8 This is a schematic diagram of the operation results of the DBSCAN algorithm of the present invention. Figure 4 . DETAILED DESCRIPTION
[0061] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.
[0062] At least one embodiment of the present invention discloses a vehicle abnormal behavior detection system based on Beidou positioning, including:
[0063] GPS electronic fence alarm calculation module
[0064] ①Data reception and collection: Receive vehicle positioning data from the Kafka service and obtain electronic fence data from the MySQL database
[0065] ② Alarm: Alarms are divided into three types according to the entry and exit types: fence entry alarm, fence exit alarm, and fence entry and exit alarm. Alarms are issued when the relative position of the vehicle to the fence changes. There are two types according to the alarm scope: multi-vehicle alarm (for some selected devices according to the alarm configuration) and global alarm (applied to all devices). The stay time can be configured.
[0066] For example: If the alarm duration of device A entering fence A is configured as 0 minutes, an alarm will be issued immediately when the device signal changes from outside to inside relative to fence A. If it is configured as 10 minutes, an alarm will be issued if the GPS data continuously reported within 10 minutes after the device signal changes from outside to inside relative to fence A is all inside fence A.
[0067] ③Distributed deployment: The software can be deployed on multiple servers and work collaboratively.
[0068] GPS data example: {"id":null,"imei":"952092583999","sim":null,"gpsTime":"2023-07-31T05:56:01.036Z","receiveTime":"2023-07-31T05:56:01.036Z","lng":119348627,"lat":26091266,"course":88,"speed":37,"position":"Auxiliary road of Huagong Road, Gushan Town, Jin'an District, Fuzhou City, Fujian Province","gpsStatus":1,"latStatus":0,"lngStatus":0,"powerStatus":null,"sosStatus":null,"workStatus":null,"status":1,"serialNumber":null,"acc":1,"targetingStatus":1,"oilStatus":0,"speedAlarm":null,"vibrationAla rm":null,"powerAlarm":0,"fakeLbsAlarm":null,"cellId":null,"lac":null,"gsmSignal":23,"mcc":null,"mnc":nul l,"voltage":null,"gpsSignal":12,"ip":null,"port":null,"electricity":90,"workStyle":null,"source":0,"lowV oltageAlarm":null,"photosensitiveAlarm":null,"shieldAlarm":null,"disturbAlarm":null,"uploadCount":null," uploadTime":null,"dangerMark":null,"mileage":null,"deviceType":"FK008","gmtCreate":"2023-07-31T05:56:02.830Z","msgId":"200","alarmBits":null,"mac":null,"wifiSignal":null,"wifiSignalStrength":null,"changePositionStatus":null,"positionInfos":null,"key":"952092583999","dc_addr":null}.
[0069] Where imei is the unique identification number of the device, lat and lng are the longitude and latitude, and gpsTime is the reporting time.
[0070] 2. Permanent location algorithm: mainly uses K-Means clustering algorithm, DBSCAN algorithm, KDE kernel density algorithm and other algorithms to obtain the information of the car’s permanent location.
[0071] Ⅰ. Use the DBSCAN clustering algorithm to cluster the longitude and latitude data of all vehicle equipment and obtain the first four permanent locations.
[0072] Ⅱ. The longitude and latitude data of all vehicle equipment are clustered using the K-MEANS clustering algorithm to obtain four permanent locations.
[0073] III. Use the KDE grid algorithm to cluster the longitude and latitude data of all vehicle equipment to obtain a permanent location.
[0074] IV. Design vehicle permanent location algorithm based on prototype logic.
[0075] ①K-Means clustering algorithm
[0076] Data collection: Ensure that each device uploads GPS data every minute with timestamp and location information.
[0077] Data preprocessing: Clean the data, ensure the data is correct, and convert GPS coordinates into a unified format.
[0078] Data analysis: Use a clustering algorithm (K-Means) to identify areas where vehicles stay for a long time. Set a time threshold X to determine the permanent location. For example, if a vehicle stays in the same place for more than X minutes, it is considered a permanent location.
[0079] Result optimization: Considering that vehicles may be stuck in traffic, a speed threshold is set to filter out data points stuck in traffic congestion. For each device, analysis is performed separately and the results are merged to improve accuracy.
[0080] Results: Display the places that cars frequently visit in the form of a list or map (provide statistics on the frequency of visits and average stay time at each place). Figure 1-4 As shown:
[0081] ②DBSCAN algorithm
[0082] Data preparation: Collect vehicle GPS data, including timestamps and location coordinates.
[0083] DBSCAN algorithm: Use the DBSCAN algorithm to cluster the vehicle location data. Set appropriate parameters, such as the neighborhood radius (ε) and the minimum number of samples (MinPts). Based on the clustering results, find out the vehicle's permanent location.
[0084] Implementation steps:
[0085] 1) Initialize the DBSCAN model.
[0086] 2) Train the model and obtain clustering results.
[0087] 3) Identify core points and clusters.
[0088] 4) For each cluster, calculate the center of the cluster (average longitude and latitude), which is defined as the place where vehicles often go.
[0089] Result visualization: Plot the calculated permanent locations on a map.
[0090] like Figure 5-8 shown.
[0091] ③KDE kernel density algorithm
[0092] KDE kernel density estimation is a nonparametric method for estimating the probability density function of a random variable. In vehicle behavior monitoring, the KDE algorithm can be used to analyze vehicle location data to identify risk behaviors such as multiple vehicles gathering, long-term parking, different locations, and leaving frequently parked areas.
[0093] Positioning difference alarm: Calculate the positioning difference between the devices of each vehicle and issue an alarm when it is greater than the set threshold
[0094] Calculation logic:
[0095] Ⅰ Distance calculation function geodesic().meters
[0096] Ⅱ Calculate the distance between devices in each car_id. When any pair of devices exceeds the distance threshold, return the current car_id to the set, and finally return a set.
[0097] Ⅲ Query the joint table of location_device and car table according to the set collection. The query returns information such as car_no, device_id, position, etc.
[0098] The present invention provides a vehicle abnormal behavior detection method based on Beidou positioning, comprising the following steps:
[0099] 101, obtaining electronic fence data from a database;
[0100] 102, used to classify the entry and exit types into fence alarm, fence exit alarm and entry and exit fence alarm according to the change of the relative position of the vehicle to the fence;
[0101] 103, an alarm is issued according to the change and the duration of the device signal relative to the fence.
[0102] In one embodiment of the present invention, the vehicle abnormal behavior detection method based on Beidou positioning also includes:
[0103] 201, Data Fusion and Preprocessing
[0104] Multi-source data integration: Integrate vehicle sensor data (such as GPS trajectory, speed, acceleration), meteorological data, traffic flow information, driver health monitoring data, etc.
[0105] Feature extraction: Extract meaningful features from raw data, such as driving style indicators (frequency of sudden acceleration and braking), driving route preferences, time patterns, etc.
[0106] 202, Anomaly Detection and Pattern Recognition
[0107] Unsupervised learning: Apply algorithms such as autoencoders, Isolation Forest, and One-Class SVM to identify abnormal patterns in driving behavior.
[0108] Cluster analysis: Cluster driving patterns through K-means, DBSCAN and other algorithms to identify groups with high-risk driving behaviors.
[0109] 203, Predictive Modeling
[0110] Sequence prediction model: Use RNN (recurrent neural network), LSTM (long short-term memory network) or Transformer architecture to predict future driving behavior based on time series data. Driving behavior includes: various types of car rental defaults.
[0111] Situational awareness: Combine real-time road conditions, weather changes and other factors to adjust the weight of the prediction model to make it closer to the actual driving environment.
[0112] 204, Risk Scoring and Intervention Strategies
[0113] Dynamic Risk Scoring: Assign a dynamic risk score to each vehicle based on prediction results and real-time data.
[0114] Personalized intervention: Send customized safety suggestions or warnings to drivers based on risk scores, such as speed reduction reminders, rest suggestions, etc.
[0115] 205, Continuous Learning and Model Optimization
[0116] Online learning: Using incremental learning techniques, the model can continuously update itself as new data flows in.
[0117] Feedback loop: Collect the driver’s response to the intervention strategy and subsequent actual driving performance for iterative optimization of the model.
[0118] Feature engineering is a key part of machine learning projects, which involves extracting and constructing features from raw data that help the model learn. In vehicle violation prediction, feature engineering involves sampling the collected information through a sliding time window to obtain sequence information. The sequence unit includes:
[0119] Geospatial features:
[0120] Road section types: expressways, urban roads, rural roads, etc.
[0121] The rate of change of the longitude and latitude of the vehicle's position.
[0122] Driving behavior characteristics:
[0123] The number of sudden acceleration and braking.
[0124] Average and maximum speed.
[0125] Driver reaction time.
[0126] Environmental characteristics:
[0127] Real-time weather conditions: rain, fog, snow, etc.
[0128] Traffic flow: congestion levels, accident reports.
[0129] Driver status characteristics:
[0130] Fatigue index: Based on eyelid closure time.
[0131] Distraction: Frequency of cell phone use.
[0132] Sequence prediction models, especially those based on recurrent neural networks (RNN), long short-term memory networks (LSTM), and Transformer architectures, are ideal for processing time series data. These models are able to capture long-term dependencies in sequences and are well suited for predicting possible future violations of vehicles.
[0133] LSTM (Long Short-Term Memory) model formula:
[0134] The core of the LSTM unit lies in its "gate" mechanism that can control the flow of information, including input gate, output gate and forget gate. The following is the mathematical expression of the LSTM unit:
[0135] Forget gate: f t =σ(W f ·[h t-1 ,x t ]+b f )
[0136] Among them, f t is the output of the forget gate, W f and b f is the weight matrix and bias term, σ is the sigmoid activation function, x t Represents the tth sequence unit of the input time series.
[0137] Input Gate:
[0138] i t =σ(W i ·[h t-1 , x t ]+b i )
[0139]
[0140] i t is the output of the input gate, is the candidate cell state, and tanh is the hyperbolic tangent activation function.
[0141] Cell status update:
[0142]
[0143] in, Represents the Hadamard product (element-by-element product).
[0144] Output Gate:
[0145] o t =σ(W o ·[h t-1 ,x t]+b o )
[0146] Hide status updates:
[0147]
[0148] where h t represents the hidden state output at the tth time step;
[0149] L t =FC(h t )
[0150] Where FC means fully connected, L t It indicates the type of driving behavior predicted at the next moment at the tth moment.
[0151] In one embodiment of the present invention, a computer storage medium is provided, which stores computer-readable instructions. When a computer reads the computer-readable instructions, the aforementioned vehicle abnormal behavior detection method based on Beidou positioning can be executed.
[0152] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.
Claims
1. A vehicle abnormal behavior detection system based on Beidou positioning, characterized in that: include: GPS electronic fence alarm calculation module, including: A data receiving and collecting module, which is used to receive vehicle positioning data and obtain electronic fence data from a database; An alarm module is used to classify the alarms into fence entry alarm, fence exit alarm and fence entry and exit alarm according to the change of the relative position of the vehicle to the fence; Configure the dwell time to issue an alarm based on the change of the device signal relative to the fence and the dwell time; A permanent location identification module, which is used to obtain the permanent location information of the vehicle; Positioning difference alarm: Calculate the positioning difference between the devices of each vehicle and issue an alarm when the difference is greater than the set threshold.
2. The vehicle abnormal behavior detection system based on Beidou positioning according to claim 1 is characterized in that: Distributed deployment: The vehicle abnormal behavior detection system based on Beidou positioning is deployed on multiple servers and works collaboratively.
3. The vehicle abnormal behavior detection system based on Beidou positioning according to claim 1 is characterized in that: The permanent location of the vehicle is obtained by clustering the latitude and longitude data of the vehicle.
4. The vehicle abnormal behavior detection system based on Beidou positioning according to claim 3 is characterized in that: The K-Means clustering algorithm, DBSCAN algorithm, and KDE kernel density algorithm are used to obtain the vehicle's permanent location information.
5. The method for detecting abnormal vehicle behavior based on Beidou positioning according to claim 1, characterized in that: Use the distance calculation function to calculate the distance between devices and determine whether to issue an alarm based on the set distance threshold; Get detailed information about vehicles and equipment by querying related tables.
6. A vehicle abnormal behavior detection method based on Beidou positioning, characterized in that: The following steps are involved: 101, obtaining electronic fence data from a database; 102, used to classify the entry and exit types into fence alarm, fence exit alarm and entry and exit fence alarm according to the change of the relative position of the vehicle to the fence; 103, an alarm is issued according to the change and the duration of the device signal relative to the fence.
7. The method for detecting abnormal vehicle behavior based on Beidou positioning according to claim 1, characterized in that: Also includes: 201, Data Fusion and Preprocessing Multi-source data integration: integrating vehicle sensor data, meteorological data, traffic flow information, and driver health monitoring data; Feature extraction: extract driving style indicators, route preferences, and time patterns from raw data; 202, Anomaly Detection and Pattern Recognition Identify unusual patterns in driving behavior; Cluster analysis: cluster driving patterns through clustering algorithms to identify groups with high-risk driving behaviors; 203, Predictive Modeling Sequence prediction model: Use RNN or LSTM or Transformer architecture to predict future driving behavior based on time series data; 204, Risk Scoring and Intervention Strategies Dynamic risk scoring: Assign a dynamic risk score to each vehicle based on prediction results and real-time data; Personalized intervention: Send customized safety recommendations or warnings to drivers based on risk scores.
8. The method for detecting abnormal vehicle behavior based on Beidou positioning according to claim 7 is characterized in that: Applying autoencoder algorithms to identify abnormal patterns in driving behavior.
9. The method for detecting abnormal vehicle behavior based on Beidou positioning according to claim 7, characterized in that: The calculation formula of the sequence prediction model is as follows: Forget gate: f t =σ(W f ·[h t-1 ,x t ]+b f ) Among them, f t is the output of the forget gate, W f and b f is the weight matrix and bias term, σ is the sigmoid activation function, x t Represents the tth sequence unit of the input time series; Input Gate: i t =σ(W i ·[h t-1 ,x t ]+b i ) i t is the output of the input gate, is the candidate cell state, tanh is the hyperbolic tangent activation function; Cell status update: in, Represents the Hadamard product (element-by-element product); Output Gate: the t =σ(W o ·[h t-1 ,x t ]+b o ) Hide status updates: where h t represents the hidden state output at the tth time step; L t =FC(h t ) Where FC means fully connected, L t It indicates the type of driving behavior predicted at the next moment at the tth moment.
10. A computer storage medium, characterized in that: It stores computer-readable instructions, and when a computer reads the computer-readable instructions, it can execute the vehicle abnormal behavior detection method based on Beidou positioning as described in any one of claims 6-9.
Citation Information
Cited By
Airport crossing person and vehicle management method and system
CN120220426A
