Traffic event association method and device, storage medium and electronic equipment

Abstract

The invention provides a traffic event association method and device, a storage medium and electronic equipment, and the method comprises the steps: obtaining real-time view data which at least comprises road view data, vehicle view data and driver view data; inputting the road view data into the trained CV model, and determining a target traffic event corresponding to the road view data; updating the initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and updating the initial face file according to the driver view data to obtain an updated face file; and constructing an association relationship between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a total traffic file. According to the method, traffic incident detection generalization ability is high based on the CV large model, unseen image data can be processed, traffic incidents happening accidentally can be accurately detected, and support is provided for avoiding secondary accidents and improving the traffic safety level.

Description

Technical Field

[0001] The present application relates to the application field of artificial intelligence visual large models, and specifically, to a traffic event association method, a traffic event association device, a computer-readable storage medium, and an electronic device. Background Art

[0002] Traffic incident detection technology is an indispensable part of modern intelligent transportation systems (ITS). It integrates advanced technologies such as the Internet of Things, big data, machine learning and artificial intelligence, and significantly improves the efficiency and safety of traffic management. Traffic incident detection technology collects and analyzes traffic flow data, combined with machine learning and data analysis technology, to achieve real-time monitoring and response to traffic incidents.

[0003] Traffic flow data includes but is not limited to: vehicle speed (obtained through radar speed measurement, GPS tracking, etc.), vehicle density and flow (estimated using video surveillance, geomagnetic sensors, floating vehicle data, etc.), lane occupancy (using video recognition technology to determine whether the lane is occupied or blocked), traffic sign information (including traffic light status, road signs, construction information, etc.), weather conditions, traffic accident history records and other external environmental data.

[0004] Traditional event detection technology mainly relies on the changes in traffic flow parameters of upstream and downstream detectors, and manually sets thresholds to determine the occurrence of traffic events. This method has the following limitations: strong subjectivity, the setting of thresholds is often based on experience or expert judgment, lacking objectivity and accuracy; low recognition rate, unreasonable settings or complexity of traffic flow, which may lead to low event recognition rate, missed reports or false reports. In addition, the existing technology cannot accurately associate abnormal traffic events with vehicle and driver information. Summary of the invention

[0005] The main purpose of the present application is to provide a method for associating traffic events, a device for associating traffic events, a computer-readable storage medium and an electronic device, so as to at least solve the problem in the prior art that manually setting thresholds to judge the occurrence of traffic events is highly subjective, has a low recognition rate, and cannot accurately associate the information between traffic events, vehicles and drivers.

[0006] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for associating traffic events is provided, comprising: acquiring real-time view data, the real-time view data at least including road view data, vehicle view data and driver view data, the road view data at least including vehicle position data, pedestrian data, lane line data and indicator light data, the vehicle view data at least including license plate number data, and the driver view data at least including driver face data; inputting the road view data into a trained CV model to determine a target traffic event corresponding to the road view data; updating an initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and updating an initial face file according to the driver view data to obtain an updated face file; constructing an association relationship between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a full traffic file.

[0007] Optionally, the road view data is input into a trained CV model to determine a target traffic event corresponding to the road view data, including: determining a plurality of image data and a shooting time of each of the image data based on the road view data; performing feature extraction on the image data to obtain object features of each object on the road, wherein the objects include at least indicator lights, lane lines, pedestrians and vehicles on the road, and the object features include at least the type of the object and the position coordinates of the object; determining the position coordinates of each of the objects at different times based on the shooting time of the plurality of image data and the object features of each of the objects corresponding to the plurality of image data; and determining the target traffic event corresponding to the road view data based at least on the position coordinates of the plurality of objects at different times.

[0008] Optionally, there are multiple driver face data, and the initial face file is updated according to the driver view data to obtain an updated face file, including: a first comparison step, respectively comparing the multiple driver face data with the multiple face data in the initial face file to obtain a first comparison similarity of each driver face data, the first comparison similarity representing the similarity between the driver face data and one face data; a first determination step, determining a first cluster file and a first unarchived data according to all the driver face data and the corresponding first comparison similarity, and keeping the first cluster file The first face file is permanently stored in the initial face file to obtain a first updated face file; a second comparison step is performed at a preset time to compare the first unarchived data with multiple face data in the first updated face file to obtain a second comparison similarity; a second determination step is performed to determine a second cluster file and a second unarchived data according to the first unarchived data and the corresponding second comparison similarity, and the second cluster file is persistently stored in the initial face file to obtain a second updated face file; the first comparison step, the first determination step, the second comparison step and the second determination step are repeatedly performed with the target time length as a period.

[0009] Optionally, determining a first cluster archive and a first unarchived data according to all of the driver's facial data and the corresponding first comparison similarity includes: clustering the driver's facial data according to all of the driver's facial data and the corresponding first comparison similarity to obtain a plurality of first cluster data and a plurality of initial unarchived data, wherein each of the driver's facial data in the first cluster data corresponds to facial data of an initial facial archive, and a first comparison similarity between the driver's facial data in the first cluster data and the corresponding facial data is greater than a similarity threshold; clustering all of the initial unarchived data Clustering is performed to obtain a plurality of second cluster data, each of which corresponds to an unrecorded facial data, and the unrecorded facial data is facial data that does not exist in the initial facial file; when the amount of data in the second cluster data is greater than or equal to a preset amount, the plurality of first cluster data are respectively stored in the corresponding facial data files, and the second cluster data is separately stored in the newly added facial data file to obtain the first cluster file; when the amount of data in the second cluster data is less than a preset amount, the data in the second cluster data is determined as the first unarchived data.

[0010] Optionally, the multiple facial data of the driver are respectively compared with the multiple facial data in the initial facial file, including: when the driver facial data is greater than or equal to a preset data amount, the multiple facial data of the driver are respectively compared with the multiple facial data in the initial facial file; and / or, every preset time period, the multiple facial data of the driver are respectively compared with the multiple facial data in the initial facial file.

[0011] Optionally, before inputting the road view data into the trained CV model, the method further includes: constructing an initial CV model; training the initial CV model using multiple sets of training data to obtain the trained CV model, wherein each set of training data in the multiple sets of training data includes: historical road view data acquired within a historical time period and historical traffic events corresponding to the historical road view data.

[0012] Optionally, obtaining road view data includes: obtaining initial view data, where the format of the initial view data is a first format, where the first format is an audio format or a video format; converting the format of the initial view data from the first format to a second format using an FFmpeg protocol to obtain the real-time view data, where the second format is a format recognizable by the trained CV model.

[0013] According to another aspect of the present application, a traffic event detection and archiving device is provided, including: an acquisition unit, used to acquire real-time view data, the real-time view data at least including road view data, vehicle view data and driver view data, the road view data at least including vehicle position data, pedestrian data, lane line data and indicator light data, the vehicle view data at least including license plate data, and the driver view data at least including driver face data; a determination unit, used to input the road view data into a trained CV model to determine a target traffic event corresponding to the road view data; an archiving unit, used to update an initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and to update an initial face file according to the driver view data to obtain an updated face file; a construction unit, used to construct an association relationship between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a full traffic file.

[0014] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the traffic event association methods.

[0015] According to another aspect of the present application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include an association method for executing any one of the traffic events described.

[0016] Applying the technical solution of the present application, the above-mentioned traffic event association method first obtains real-time view data, which at least includes road view data, vehicle view data and driver view data; then the road view data is input into the trained CV model to determine the target traffic event corresponding to the road view data; then the initial vehicle file is updated according to the vehicle view data to obtain an updated vehicle file, and the initial face file is updated according to the driver view data to obtain an updated face file; finally, the association relationship between any two of the target traffic event, the updated vehicle file and the updated face file is constructed to obtain the full traffic file. This method has a strong generalization ability for traffic event detection based on the CV large model, can process unseen image data, helps to accurately detect accidental traffic events, provides support for avoiding secondary accidents and improving traffic safety levels, and solves the problem of strong subjectivity and low recognition rate in the prior art of judging the occurrence of traffic events by manually setting thresholds, and cannot accurately associate the information between traffic events, vehicles and drivers.

[0017] According to another aspect of the present application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include an association method for executing any one of the traffic events described. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings constituting part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0019] Figure 1 A hardware structure block diagram of a mobile terminal for executing a method for associating traffic events provided in an embodiment of the present application is shown;

[0020] Figure 2 A schematic diagram of a flow chart of a method for associating traffic events provided according to an embodiment of the present application is shown;

[0021] Figure 3 A schematic diagram showing a flow chart of another method for associating traffic events provided according to an embodiment of the present application;

[0022] Figure 4 A schematic diagram of a traffic event correlation system provided according to an embodiment of the present application is shown;

[0023] Figure 5 A structural block diagram of a traffic event association device provided according to an embodiment of the present application is shown.

[0024] The above drawings include the following reference numerals:

[0025] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0026] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0027] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0030] CV large models, or computer vision large models, refer to neural network models with powerful visual processing capabilities trained through deep learning technology. These models usually have millions or even billions of parameters and can learn and optimize in large amounts of data to achieve highly accurate image recognition, object detection, semantic segmentation and other tasks.

[0031] Traffic incident detection is an important technology in the field of smart transportation. It analyzes the flow characteristics of traffic flow, determines potential, impending and existing traffic incidents, and identifies abnormal situations on the road, such as traffic accidents, traffic congestion, pedestrian intrusions, abnormal parking, etc., providing real-time information for traffic management, emergency response and drivers.

[0032] Pedestrian intrusion event detection: By identifying pedestrians on the road, timely detection and warning of pedestrians intruding into the motor vehicle lane.

[0033] Abnormal parking event detection: monitor the parking conditions of vehicles on the road and detect abnormal parking behaviors, such as illegal parking and long-term occupation of emergency lanes.

[0034] Wrong-travel event detection: Identify wrong-travel vehicles on the road and issue warnings in time to prevent traffic accidents.

[0035] Lane change event detection: monitor vehicle lane change behavior to ensure lane change safety and reduce traffic accidents caused by lane changes.

[0036] Congestion event detection: By analyzing traffic flow and speed data, the road congestion situation can be judged in real time, and congestion avoidance route suggestions can be provided to drivers.

[0037] Road spill detection: monitors whether there is spillage on the road, such as dropped goods, vehicle oil leakage, etc., and issues early warnings in time to prevent subsequent vehicle accidents.

[0038] Traffic accident detection: Through image analysis and video recognition technology, traffic accidents can be automatically detected and relevant departments can be notified in time for handling.

[0039] As introduced in the background technology, traditional event detection technology mainly relies on changes in traffic flow parameters of upstream and downstream detectors, and manually sets thresholds to determine the occurrence of traffic events. In order to solve the problems in the prior art of manually setting thresholds to determine the occurrence of traffic events, which are highly subjective, have a low recognition rate, and cannot accurately associate the information between traffic events, vehicles and drivers, the embodiments of the present application provide a traffic event association method, a traffic event association device, a computer-readable storage medium and an electronic device.

[0040] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0041] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a mobile terminal of a method for associating traffic events according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0042] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for associating traffic events in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the mobile terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The above-mentioned specific example of the network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0043] There are also many solutions in the existing technology that use machine learning and data analysis technology to deeply mine these data and establish traffic incident detection models. However, the following problems still exist:

[0044] 1. Sample imbalance problem: Event samples (such as traffic accidents, congestion, etc.) are far less than non-event samples, which makes the model insensitive to minority samples during training;

[0045] 2. Overfitting problem: Due to the limited total sample size, the model is prone to overfitting during the training process, resulting in poor performance in the test set or actual application;

[0046] 3. Poor model versatility: Existing traffic incident detection models are proprietary models for individual types of abnormal traffic. Using one proprietary model can only handle abnormal traffic accidents of the corresponding type. If you want to handle multiple abnormal types at the same time, you need to use multiple proprietary models. However, the types of abnormal traffic covered by proprietary models are also limited and cannot meet the needs. Moreover, when there are new traffic incidents that need to be analyzed, the existing proprietary models may not be directly applied, and it is necessary to re-collect relevant data and train new proprietary models, which results in a long time and high cost.

[0047] In this embodiment, a method for associating traffic events running on a mobile terminal, a computer terminal or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0048] Figure 2 FIG. 1 is a flow chart of a method for associating traffic events according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0049] Step S201, acquiring real-time view data, the real-time view data at least including road view data, vehicle view data and driver view data, the road view data at least including vehicle position data, pedestrian data, lane line data and indicator light data, the vehicle view data at least including license plate data, and the driver view data at least including driver face data;

[0050] Specifically, real-time view data is collected by cameras on the road.

[0051] Wherein, obtaining real-time view data includes the following steps:

[0052] Step S2011, acquiring initial view data, where the format of the initial view data is a first format, and the first format is an audio format or a video format;

[0053] Step S2012: Using the FFmpeg protocol, the format of the initial view data is converted from the first format to a second format to obtain the real-time view data, wherein the second format is a format that can be recognized by the trained CV model.

[0054] Specifically, FFmpeg can convert one audio and video format into another, support data exchange between IoT devices, and improve compatibility. The scenario in the application process is to connect social cameras to the device center and connect them to the platform in the form of video streams or picture streams. The device center is the device management module in the system, and the view data refers to the video or picture data to be parsed. This describes the application scenario process of this technology. In terms of technical implementation, FFmpeg supports a variety of audio and video encoding formats, such as H.264, H.265, AAC, etc., which enables it to easily process audio and video data on IoT devices and perform encoding and decoding. FFmpeg supports the recording, playback and transmission of streaming media, which is very useful for application scenarios such as video surveillance and remote conferencing in IoT.

[0055] Before inputting the road view data into the trained CV model, the method further includes the following steps:

[0056] Step S301, constructing an initial CV model;

[0057] Step S302, using multiple sets of training data to train the above-mentioned initial CV model to obtain the above-mentioned trained CV model, wherein each set of training data in the multiple sets of training data includes: historical road view data obtained in a historical time period and historical traffic events corresponding to the above-mentioned historical road view data.

[0058] Specifically, the view parsing task and traffic incident detection task of the CV big model are trained, and the CV big model is put into use, so that the CV big model can obtain the corresponding traffic incident detection and warning results in real time according to the road traffic status in the view data based on the above two tasks that have been trained.

[0059] The process of training the CV large model with the above-mentioned view parsing task and traffic event detection task is as follows:

[0060] (1) Training view parsing tasks: using a variety of traffic events, manually labeling the multi-dimensional attributes of the traffic events to train the traffic event detection task; here, view parsing can be understood as large model parsing.

[0061] (2) Training traffic event detection tasks, using a variety of traffic events, through manual labeling of the multi-dimensional attributes of the traffic events, to train traffic event detection tasks; creating specific traffic event detection tasks, specific events such as pedestrian intrusion, motor vehicle reversing, abnormal parking, occupying the emergency lane, large bend and small turn, solid line lane change, jamming, crossing the line, road water accumulation, road damage, road congestion, road spillage, etc.; for example, "traffic event detection task" - pedestrians crossing the zebra crossing without courtesy to pedestrians: labeling its multi-dimensional attributes including: green light, zebra crossing, car, people, and the coordinates of people and cars and the time when people and cars are at each coordinate; based on the correlation between the above information, for example, when pedestrians are on the zebra crossing, cars also cross the zebra crossing, but the car passes or turns first, then pedestrians are not courtesy to pedestrians when crossing the zebra crossing. Therefore, training the CV model in this way enables the CV model to have the ability to identify the above "traffic event 1"; here, the above "traffic event detection task" can correspond to a large number of different traffic events.

[0062] Step S202, inputting the road view data into the trained CV model to determine a target traffic event corresponding to the road view data;

[0063] Specifically, after the above-mentioned CV big model is trained, the view data is recognized based on the CV big model, that is, the characteristics and attributes of all things in the view data are recognized based on the CV big model, and the road traffic status in the view data is converted and the corresponding traffic event detection and warning results are obtained in real time.

[0064] At the same time, in the process of training the CV model, when integrating the "target profile" in the above main concept, the "traffic event detection task" can be a normal event, such as pedestrian crossing events within 5 minutes, vehicle statistics events within 5 minutes, etc. (collectively referred to as profile events). Profile association means that the vehicle profile built based on license plate numbers and vehicle attributes can be associated with the events containing license plate numbers and vehicle attributes in the abnormal events we detect for subsequent application by customers.

[0065] The step of inputting the road view data into the trained CV model to determine the target traffic event corresponding to the road view data includes the following steps:

[0066] Step S2021, determining a plurality of image data and a shooting time of each of the image data according to the road view data;

[0067] Step S2022, extracting features from the image data to obtain object features of various objects on the road, wherein the objects at least include indicator lights, lane lines, pedestrians, and vehicles on the road, and the object features at least include the type of the object and the location coordinates of the object;

[0068] Step S2023, determining the position coordinates of each of the objects at different times according to the shooting times of the plurality of image data and the object features of each of the objects corresponding to the plurality of image data;

[0069] Step S2024: determining a target traffic event corresponding to the road view data at least based on the position coordinates of the plurality of objects at different times.

[0070] Specifically, traffic incident detection is performed based on the CV large model, which has strong generalization capabilities; the technology is combined with business scenarios to improve the efficiency and safety of traffic management.

[0071] Step S203, updating the initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and updating the initial face file according to the driver view data to obtain an updated face file;

[0072] Specifically, this can enrich the holographic content of the vehicle profile itself;

[0073] Wherein, there are multiple driver face data, and the initial face file is updated according to the driver face data to obtain the updated face file, including the following steps:

[0074] A first comparison step, comparing the plurality of driver face data with the plurality of face data in the initial face file, respectively, to obtain a first comparison similarity of each of the driver face data, wherein the first comparison similarity represents the similarity between the driver face data and one of the face data;

[0075] Among them, the facial data of the multiple drivers mentioned above are respectively compared with the facial data of the multiple faces in the initial facial file, including: when the facial data of the multiple drivers mentioned above is greater than or equal to the preset data amount, the facial data of the multiple drivers mentioned above are respectively compared with the facial data of the multiple faces in the initial facial file; and / or, every time a preset time period passes, the facial data of the multiple drivers mentioned above are respectively compared with the facial data of the multiple faces in the initial facial file.

[0076] Specifically, a vehicle / personnel file is constructed through archival events; based on the correlation between the target file and the above-mentioned abnormal traffic events, the target file is improved to provide data support for relevant departments through the target file.

[0077] That is, after the CV model is trained, the characteristics of the vehicles and the structured attributes of the vehicles such as vehicle brand, license plate number, and license plate color can be obtained according to the appearance of vehicles in the view data; clustering and archiving of vehicles can be realized based on the vehicle clustering algorithm to form vehicle files; based on progressive relationship deduction, the correlation between vehicle files and traffic events can be obtained to establish a more complete vehicle file; the driver's identity information can be obtained through face recognition technology based on the driver and passengers bound to the vehicle; rich data support can be provided to the traffic management department to help make intelligent decisions and optimization.

[0078] A first determination step, determining a first cluster file and a first unarchived data according to all the driver face data and the corresponding first comparison similarity, and persistently storing the first cluster file in the initial face file to obtain a first updated face file;

[0079] According to all the driver face data and the corresponding first comparison similarity, determining the first cluster archive and the first unarchived data includes the following steps:

[0080] Step S401, clustering the driver face data according to all the driver face data and the corresponding first comparison similarities to obtain a plurality of first cluster data and a plurality of initial unarchived data, wherein each driver face data in the first cluster data corresponds to face data of the initial face archive, and a first comparison similarity between the driver face data in the first cluster data and the corresponding face data is greater than a similarity threshold;

[0081] Step S402, clustering all the above-mentioned initial unarchived data to obtain a plurality of second clustered data, each of the above-mentioned second clustered data corresponds to an unrecorded face data, and the above-mentioned unrecorded face data is the face data that does not exist in the above-mentioned initial face archive;

[0082] Step S403, when the number of data in the second cluster data is greater than or equal to the preset number, storing the plurality of first cluster data in the corresponding face data files respectively, and storing the second cluster data separately in the newly added face data file, to obtain the first cluster file;

[0083] Step S404: when the amount of data in the second cluster data is less than a preset amount, determine the data in the second cluster data as the first unarchived data.

[0084] Specifically, in this way, data and files can be continuously clustered and archived to enrich vehicle files and face files. That is, when the initial face file contains face data of a person who is the same as the driver's face data, the driver's face data is stored in the face file of the corresponding person.

[0085] A second comparison step, comparing the first unarchived data with the plurality of face data in the first updated face archive at a preset time to obtain a second comparison similarity;

[0086] A second determination step, determining a second cluster archive and a second unarchived data according to the first unarchived data and the corresponding second comparison similarity, and persistently storing the second cluster archive in the initial face archive to obtain a second updated face archive;

[0087] Repeat the execution steps, taking the target duration as a period, and repeatedly execute the first comparison step, the first determination step, the second comparison step and the second determination step.

[0088] Specifically, repeating the clustering and archiving steps multiple times can improve the matching accuracy. Real-time clustering and archiving may be affected by computing power, resulting in inaccurate matching. Therefore, set a preset time when the system computing power is larger, repeat the clustering and archiving of data, and improve the matching accuracy. The preset time is generally 2 a.m. every day.

[0089] In some examples, the aggregation process includes:

[0090] 1. The system runs and imports the real-name face database, monitors the video / image stream, and generates snapshot images to record the snapshot database.

[0091] 2. When the number of snapshots exceeds the set threshold, the system automatically starts initializing the archiving task.

[0092] 3. Initialize the clustering task, pull the full amount of captured data, compare the quality data, shard according to the rules, call the clustering worker capability, generate clustering archives and unarchived data, and persist the generated clustering archives into storage.

[0093] 4. The snapshot is written to the snapshot storage database in real time, and the real-time archiving worker is automatically started.

[0094] 5. Real-time clustering obtains real-time snapshot data 1vN and compares the clustering archives and face database (real-name archives) to archive the snapshots that meet the comparison threshold, and those that do not meet the threshold are classified as unarchived snapshots.

[0095] 6. The daily (or custom time) task scheduling module starts the offline clustering worker to work, obtains the unarchived snapshots within the time period, calls the clustering worker for clustering, and generates today's archives and today's scatter points.

[0096] 7. The daily (or custom time) task scheduling module starts the archive merging worker, obtains the newly added archive data and the full archive data in the newly added time period, compares the full cluster archives 1vN, archives the unarchived archives that meet the comparison threshold, and the archives that do not meet the threshold fall into the unarchived archives.

[0097] 8. Obtain the generated unarchived archives and the full amount of unarchived data, call the archive merging worker, 1vN compare the full amount of clustered archives, archive the unarchived archives that meet the comparison threshold, and the archives that do not meet the threshold fall into the unarchived archives.

[0098] 9. Compare the cluster files with the face database to match the cluster files with the real-name database. Add real-name information.

[0099] 10. Repeat steps 5-9 periodically.

[0100] Step S204, constructing an association relationship between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a full traffic file.

[0101] Specifically, we use progressive deduction to unearth clues between traffic incidents and vehicle files, which can enrich the holographic content of the vehicle files themselves on the one hand; on the other hand, we use facial recognition technology to obtain the identity information of the driver and passengers bound to the vehicle, improve the identity information confirmation process of the driver of the illegal vehicle, improve the efficiency of traffic police law enforcement, and even achieve remote law enforcement.

[0102] With the rise and rapid development of general large model technology, CV (computer vision) large models are gradually entering the application stage. However, there are indeed many manufacturers in the industry that have CV large models, but they have not been able to effectively implement them in actual application scenarios, resulting in most of these models still remaining in the retrieval demonstration stage.

[0103] The above example solves the problem of judging the occurrence of traffic incidents by manually setting thresholds, but it has the limitations of strong subjectivity and low recognition rate; using machine learning and data analysis technology to conduct in-depth mining of these data and establish a traffic incident detection model faces problems such as sample imbalance, data overfitting, and poor model generality; the large CV model fails to be effectively combined with actual application scenarios for implementation.

[0104] The specific effects to be achieved in the above examples include: (1) Traffic incident detection based on the CV large model has strong generalization ability and can process unseen image data; (2) Once a traffic incident is detected, the system can immediately trigger an alarm to notify the traffic management department to respond in real time and dispatch the police quickly; (3) It helps to accurately detect accidental traffic incidents, provide support for avoiding secondary accidents and improving traffic safety levels; (4) Traffic incidents integrate vehicle file data to provide rich data support for traffic management departments to help make intelligent decisions and optimization.

[0105] The above scheme uses a pre-trained CV big model to enable it to identify various types of abnormal traffic events, so as to give prompts to relevant departments (such as traffic management departments) based on abnormal traffic events; at the same time, it can also determine the attribute information of targets related to traffic events (the targets are mainly people and vehicles) based on the above CV big model, and integrate the target files of the above targets based on the above attribute information (the target files here can be files with people as ids, or files with vehicles as ids), and based on the association between the target files and the above abnormal traffic events, improve the target files to provide data support for relevant departments through the target files.

[0106] The above-mentioned traffic event association method of the present application first obtains real-time view data, which at least includes road view data, vehicle view data and driver view data; then the road view data is input into the trained CV model to determine the target traffic event corresponding to the road view data; then the initial vehicle file is updated according to the vehicle view data to obtain an updated vehicle file, and the initial face file is updated according to the driver view data to obtain an updated face file; finally, the association relationship between any two of the target traffic event, the updated vehicle file and the updated face file is constructed to obtain the full traffic file. This method has strong generalization ability for traffic event detection based on the CV large model, can process unseen image data, helps to accurately detect accidental traffic events, provides support for avoiding secondary accidents and improving traffic safety levels, and solves the problem of strong subjectivity and low recognition rate in the prior art of judging the occurrence of traffic events by manually setting thresholds, and cannot accurately associate the information between traffic events, vehicles and drivers.

[0107] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the method for associating traffic events of the present application will be described in detail below in conjunction with specific embodiments.

[0108] This embodiment relates to a specific method for associating traffic events, such as Figure 3As shown, firstly, the view data is connected to the device center, and then the view data is transmitted to the analysis center for analysis, and then the analyzed data is input into the CV model, and the attributes of each object in the view data are obtained as the attributes of all things. Traffic events are identified according to the attributes of all things and the traffic event recognition algorithm, and an early warning is issued when an abnormal event is identified. At the same time, the vehicle captured image is obtained according to the analyzed view data recognition, and the vehicle file is obtained by using the vehicle clustering algorithm. The identified traffic events and vehicle files are associated with the relationship deduction algorithm to obtain the vehicle holographic file, and the vehicle holographic file is applied to vehicle trajectory query and identity confirmation, or other applications.

[0109] It should be noted that the main idea of ​​the above technical solution is to use the CV big model to warn of abnormal traffic incidents and to archive vehicles; and, in this technical solution, abnormal traffic incidents are identified because the model architecture of our CV big model can support big data processing, and this technical solution uses the "traffic incident detection task" and annotates its attribute features as training samples and conducts model training; in addition, because our CV big model uses the above-mentioned "attribute feature annotation as training sample" method, it is also possible to build vehicle / personnel files through the "traffic incident detection task" and improve the files in combination with "abnormal traffic incidents" to provide reference for relevant departments. For example, when a "black car" has a traffic accident, the traffic control department not only has to deal with the traffic accident but also needs to check the relevant information of the "black car". Our technology can provide strong data support for the traffic control department.

[0110] The above scheme uses a CV large model to detect traffic incidents. It is a process of determining potential, upcoming and existing traffic incidents by analyzing the flow characteristics of traffic flow. At the same time, it explores the correlation between traffic incidents and vehicle files, provides rich data support for traffic management and other departments, and helps to make intelligent decisions and optimization. Real-time monitoring of common events in daily life, such as pedestrian intrusion, motor vehicle reversal, abnormal parking, occupation of emergency lanes, large bends and small turns, solid line lane changes, jamming, line pressure, road waterlogging, road damage, road congestion, road spillage, etc., information acquisition, confirmation, remote law enforcement of traffic violators, early warning of frequent traffic incidents, and traffic road condition detection are all examples of scenarios applicable to the present invention.

[0111] The above scheme is divided into five stages in terms of process: equipment access, real-time analysis, control and early warning, relationship deduction, and business application. The equipment access stage mainly selects the video camera or offline video and picture data to be analyzed, and directly connects to the national standard platform through the GB 28181 protocol or completes the view data docking through the GA / T 1400 interface protocol; the real-time analysis stage mainly performs target recognition and analysis based on the CV large model. This system mainly includes vehicle feature extraction in view data, vehicle structured attribute recognition, and scene-based traffic event detection; the control and early warning stage mainly monitors the traffic events that are about to occur or have occurred in real time, and once discovered, quickly warns relevant personnel, and quickly dispatches police or makes corresponding decisions; the relationship deduction stage mainly performs progressive relationship analysis based on vehicle characteristics, vehicle attributes, and people, vehicles, and objects in traffic events, and obtains the deep relationship between traffic events and vehicle files; the voluntary application stage mainly performs further research and analysis based on various types of data generated by the present invention, including but not limited to identity verification of violators, remote law enforcement, etc.

[0112] In some instances, a certain road is a main traffic artery, and the large number of vehicles every day leads to frequent traffic incidents. Once they occur, they will cause road congestion. The speed of traffic police response and violation handling affects the duration of traffic abnormalities.

[0113] To this end, all cameras on the road can be connected to the present invention, and real-time detection and analysis of the roads, vehicles, traffic events, and road traffic status can be performed. Once a traffic event occurs, a relevant event warning is immediately generated to the designated police officer to quickly dispatch the police. At the same time, during the dispatch process, the present invention will automatically obtain the identity information of the illegal or traffic-violating person based on the vehicle and driver information of the traffic event. Depending on the impact of the traffic event, it can be selected whether to remotely handle it. The handling result is notified to the driver through a text message or WeChat, and the road traffic is restored as soon as possible.

[0114] In other examples, traffic incidents frequently occur in a city, and it is desired to analyze the causes and solutions. To this end, all cameras in the city can be connected to the present invention to perform real-time traffic incident detection, analyze the high-frequency roads and environmental conditions when they occur, avoid secondary accidents, and provide support for improving traffic safety.

[0115] At this point, a complete specific implementation example is as follows. ① The present invention supports the access of real-time video cameras through the GB 28181 standard to the video networking platform or the REST direct connection method, and also supports the access of the bayonet-type image stream camera through the GA / T 1400 protocol to achieve real-time traffic road surface analysis, and supports uploading offline videos and picture packages to complete post-analysis; ② Create a parsing task based on the CV general large model to obtain the parsing results and structured attributes of various targets. Obtain vehicle characteristics and structured attributes by identifying vehicle targets in the view data, and obtain the current traffic conditions by analyzing the traffic road scene; ③ Create a specific traffic event detection task to obtain the corresponding traffic event detection warning in real time. The current traffic violations and illegal events supported by this system include: pedestrian intrusion, motor vehicle reverse driving, abnormal parking, occupying the emergency lane, large bends and small turns, solid line lane changes, jamming, line pressing, road waterlogging, road damage, road congestion, road spillage, etc. By performing event detection on the results of traffic road scene analysis, it is known whether the configured traffic abnormality event has occurred at present. Once the event is detected, a relevant warning will be generated; ④ Based on the vehicle capture results and structured attribute clustering and archiving, the vehicle file is obtained. The automatic vehicle clustering algorithm of this system can complete the clustering and archiving of vehicle capture based on the comprehensive analysis of the vehicle capture characteristics, license plate number, and license plate color, form a unique file for the vehicle, and record the full trajectory of the vehicle ⑤ Based on the license plate number and license plate color, the vehicle file and traffic events are associated to obtain the recent full trajectory of the vehicle involved in the incident. By analyzing the warning vehicle and progressively associating the vehicle file, a more comprehensive vehicle trajectory and event information is formed ⑥ According to the vehicle-bound driver and passenger, the driver's identity information is obtained for subsequent traffic event handling. When performing real-time vehicle analysis, the relationship between the vehicle and the driver and passenger will also be obtained by different frame binding according to time and space conditions. The driver's identity information is obtained through the face recognition technology of the main driver, which provides convenience for subsequent search and disposal ⑦ At the same time, the vehicle file also records the traffic event information of the vehicle to enrich the content of the vehicle file.

[0116] Figure 4 For the system architecture diagram, Figure 4 As shown in the figure, the associated system of traffic events includes application layer, service layer, parsing layer and access layer. The application layer includes all things analysis, human and vehicle control, traffic events, historical warning, captured data, vehicle files, personnel identity, data fusion, equipment access, system management and resource management, etc. The service layer includes control and warning, target retrieval, target files and multi-algorithm fusion. The parsing layer adopts all things recognition algorithm, including scene detection, long tail algorithm and face human machine non-identification. The access layer includes Rtsp protocol, GB28181 and GA / T 1400 protocol.

[0117] The traffic incident detection model used in the above embodiment is a CV large model, which can capture and represent complex image features, has strong generalization ability, and can process unseen image data. Compared with the prior art, the subsequent implementation of new traffic incident detection consumes less time, has low cost, and has strong versatility.

[0118] The above-mentioned embodiments use progressive deduction to dig out clues between traffic incidents and vehicle files, which enriches the holographic content of the vehicle files themselves on the one hand; on the other hand, the identity information of the drivers and passengers bound to the vehicle is obtained through face recognition technology, which also improves the process of confirming the identity information of the drivers of illegal vehicles, improves the efficiency of traffic police law enforcement, and even achieves remote law enforcement. It helps to accurately detect accidental traffic incidents, provide support for avoiding secondary accidents and improving traffic safety levels. Provide rich data support for traffic management departments to help make intelligent decisions and optimization. It not only improves the efficiency and safety of traffic management, but also provides important support for the development of intelligent transportation systems.

[0119] The present application embodiment also provides a kind of association device of traffic incident, it should be noted that, the association device of traffic incident of the present application embodiment can be used for executing the association method for traffic incident provided by the present application embodiment.This device is used for realizing the above-mentioned embodiment and preferred implementation mode, and the description has been made no more.As used below, the term "module" can realize the combination of software and / or hardware of predetermined function.Although the device described in the following embodiment is preferably realized with software, the realization of hardware, or the combination of software and hardware is also possible and conceived.

[0120] The following is an introduction to the traffic event association device provided in the embodiment of the present application.

[0121] Figure 5 Schematic diagram of a traffic event association device according to an embodiment of the present application. Figure 5As shown, the device includes an acquisition unit 10, a determination unit 20, an archiving unit 30 and a construction unit 40, the acquisition unit 10 is used to acquire real-time view data, the real-time view data at least includes road view data, vehicle view data and driver view data, the road view data at least includes vehicle position data, pedestrian data, lane line data and indicator light data, the vehicle view data at least includes license plate data, and the driver view data at least includes driver face data; the determination unit 20 is used to input the road view data into the trained CV model to determine the target traffic event corresponding to the road view data; the archiving unit 30 is used to update the initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and update the initial face file according to the driver view data to obtain an updated face file; the construction unit 40 is used to construct an association relationship between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a full traffic file.

[0122] The above-mentioned traffic event association device of the present application includes an acquisition unit, a determination unit, an archiving unit and a construction unit. The acquisition unit is used to acquire real-time view data, and the real-time view data at least includes road view data, vehicle view data and driver view data; the determination unit is used to input the road view data into the trained CV model to determine the target traffic event corresponding to the road view data; the archiving unit is used to update the initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and to update the initial face file according to the driver view data to obtain an updated face file; the construction unit is used to construct an association relationship between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a full traffic file. The device has a strong generalization ability for traffic event detection based on the CV large model, can process unseen image data, helps to accurately detect accidental traffic events, provides support for avoiding secondary accidents and improving traffic safety levels, and solves the problem of strong subjectivity and low recognition rate in the prior art of judging the occurrence of traffic events by manually setting thresholds, and cannot accurately associate the information between traffic events, vehicles and drivers.

[0123] In some examples, the determination unit includes a first determination module, an extraction module, a second determination module and a third determination module. The first determination module is used to determine multiple image data and the shooting time of each of the above image data according to the above road view data; the extraction module is used to extract features from the above image data to obtain object features of each object on the road, the above objects at least include indicator lights, lane lines, pedestrians and vehicles on the above road, and the above object features at least include the type of the above object and the position coordinates of the above object; the second determination module is used to determine the position coordinates of each of the above objects at different times according to the shooting time of the multiple image data and the object features of each of the above objects corresponding to the multiple image data; the third determination module is used to determine the target traffic event corresponding to the above road view data at least according to the position coordinates of the multiple objects at different times. Traffic event detection based on the CV large model has strong generalization ability; the technology is combined with business scenarios to improve the efficiency and safety of traffic management.

[0124] In some instances, there are multiple facial data of the above-mentioned driver, and the archiving unit includes a first comparison module, a fourth determination module, a second comparison module, a fifth determination module and an execution module. The first comparison module is used to compare the multiple facial data of the above-mentioned driver with the multiple facial data in the above-mentioned initial facial archive respectively to obtain a first comparison similarity of each of the above-mentioned driver facial data, and the above-mentioned first comparison similarity represents the similarity between the above-mentioned driver facial data and one of the above-mentioned facial data; the fourth determination module is used to determine a first cluster archive and a first unarchived data according to all the above-mentioned driver facial data and the corresponding first comparison similarity, and keep the above-mentioned first cluster archive. The first face file is permanently stored in the above-mentioned initial face file to obtain a first updated face file; the second comparison module is used to compare the above-mentioned first unfiled data with the multiple face data in the above-mentioned first updated face file at a preset time to obtain a second comparison similarity; the fifth determination module is used to determine the second cluster file and the second unfiled data according to the above-mentioned first unfiled data and the corresponding second comparison similarity, and store the above-mentioned second cluster file persistently in the above-mentioned initial face file to obtain a second updated face file; the execution module is used to repeatedly execute the above-mentioned first comparison step, the above-mentioned first determination step, the above-mentioned second comparison step and the above-mentioned second determination step with the target time length as a period. Repeating the cluster archiving step multiple times can improve the matching accuracy.

[0125] In some instances, the fourth determination module includes a first clustering module, a second clustering module, a storage module and a determination submodule, the first clustering module is used to cluster the driver facial data according to all the above-mentioned driver facial data and the corresponding first comparison similarity, to obtain multiple first clustering data and multiple initial unarchived data, each of the above-mentioned driver facial data in the above-mentioned first clustering data corresponds to facial data of the above-mentioned initial facial file, and the first comparison similarity between the above-mentioned driver facial data in the above-mentioned first clustering data and the corresponding facial data is greater than the similarity threshold; the second clustering module is used to cluster all the above-mentioned initial unarchived data, A plurality of second cluster data are obtained, each of which corresponds to an unrecorded face data, and the unrecorded face data is face data that does not exist in the initial face file; the storage module is used to store the plurality of first cluster data in the corresponding face data files respectively when the number of data in the second cluster data is greater than or equal to the preset number, and store the second cluster data separately in the newly added face data file to obtain the first cluster file; the determination submodule is used to determine the data in the second cluster data as the first unrecorded data when the number of data in the second cluster data is less than the preset number. In this way, data and files can be continuously clustered and archived to enrich vehicle files and face files.

[0126] In some examples, the first comparison module includes a first comparison submodule and a second comparison submodule, wherein the first comparison submodule is used to compare the plurality of driver face data with the plurality of face data in the initial face file respectively when the driver face data is greater than or equal to a preset data amount; and the second comparison submodule is used to compare the plurality of driver face data with the plurality of face data in the initial face file respectively every time a preset time period passes. Based on the correlation between the target file and the abnormal traffic event, the target file is improved to provide data support for relevant departments through the target file.

[0127] In some examples, the apparatus further includes a construction module and a training module. The construction module is used to construct an initial CV model before inputting the road view data into the trained CV model. The training module is used to train the initial CV model using multiple sets of training data to obtain the trained CV model. Each set of training data in the multiple sets of training data includes: historical road view data and historical traffic events corresponding to the historical road view data acquired in a historical time period. The CV large model can obtain corresponding traffic event detection and warning results in real time according to the road traffic status in the view data based on the trained two tasks.

[0128] In some examples, the acquisition unit includes an acquisition module and a format conversion module, the acquisition module is used to acquire the initial view data, the format of the initial view data is a first format, the first format is an audio format or a video format; the format conversion module is used to convert the format of the initial view data from the first format to a second format using the FFmpeg protocol to obtain the real-time view data, the second format is a format that can be recognized by the trained CV model. FFmpeg can convert one audio and video format to another format, can support data exchange between IoT devices, and improve compatibility.

[0129] The above-mentioned traffic event association device includes a processor and a memory, and the above-mentioned acquisition unit and the like are stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize the corresponding functions. The above-mentioned modules are all located in the same processor; or, the above-mentioned modules are located in different processors in any combination.

[0130] The processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set, and the kernel parameters can be adjusted to solve the problem that the existing technology of manually setting thresholds to judge the occurrence of traffic events is highly subjective, has a low recognition rate, and cannot accurately associate the information between the traffic event, the vehicle, and the driver.

[0131] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0132] An embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the traffic event association method.

[0133] An embodiment of the present invention provides a processor, and the processor is used to run a program, wherein the program executes the traffic event association method when running.

[0134] An embodiment of the present invention provides a device, the device including a processor, a memory, and a program stored in the memory and executable on the processor, and when the processor executes the program, at least the following steps are implemented:

[0135] Step S201, acquiring real-time view data, the real-time view data at least including road view data, vehicle view data and driver view data, the road view data at least including vehicle position data, pedestrian data, lane line data and indicator light data, the vehicle view data at least including license plate data, and the driver view data at least including driver face data;

[0136] Step S202, inputting the road view data into the trained CV model to determine a target traffic event corresponding to the road view data;

[0137] Step S203, updating the initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and updating the initial face file according to the driver view data to obtain an updated face file;

[0138] Step S204, constructing an association relationship between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a full traffic file.

[0139] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0140] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing at least the following method steps:

[0141] Step S201, acquiring real-time view data, the real-time view data at least including road view data, vehicle view data and driver view data, the road view data at least including vehicle position data, pedestrian data, lane line data and indicator light data, the vehicle view data at least including license plate data, and the driver view data at least including driver face data;

[0142] Step S202, inputting the road view data into the trained CV model to determine a target traffic event corresponding to the road view data;

[0143] Step S203, updating the initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and updating the initial face file according to the driver view data to obtain an updated face file;

[0144] Step S204, constructing an association relationship between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a full traffic file.

[0145] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0146] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0147] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0148] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0149] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0150] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0151] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0152] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0153] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0154] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0155] 1) The above-mentioned traffic event association method of the present application first obtains real-time view data, which at least includes road view data, vehicle view data and driver view data; then the road view data is input into the trained CV model to determine the target traffic event corresponding to the road view data; then the initial vehicle file is updated according to the vehicle view data to obtain an updated vehicle file, and the initial face file is updated according to the driver view data to obtain an updated face file; finally, the association relationship between any two of the target traffic event, the updated vehicle file and the updated face file is constructed to obtain the full traffic file. This method has a strong generalization ability for traffic event detection based on the CV large model, can process unseen image data, helps to accurately detect accidental traffic events, provides support for avoiding secondary accidents and improving traffic safety levels, and solves the problem of strong subjectivity and low recognition rate in the prior art of judging the occurrence of traffic events by manually setting thresholds, and cannot accurately associate the information between traffic events, vehicles and drivers.

[0156] 2) The above-mentioned traffic event association device of the present application includes an acquisition unit, a determination unit, an archiving unit and a construction unit. The acquisition unit is used to acquire real-time view data, and the real-time view data at least includes road view data, vehicle view data and driver view data; the determination unit is used to input the road view data into the trained CV model to determine the target traffic event corresponding to the road view data; the archiving unit is used to update the initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and to update the initial face file according to the driver view data to obtain an updated face file; the construction unit is used to construct an association relationship between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a full traffic file. The device has a strong generalization ability for traffic event detection based on the CV large model, can process unseen image data, helps to accurately detect accidental traffic events, provides support for avoiding secondary accidents and improving traffic safety levels, and solves the problem of strong subjectivity and low recognition rate in the prior art of judging the occurrence of traffic events by manually setting thresholds, and cannot accurately associate the information between traffic events, vehicles and drivers.

[0157] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for associating traffic events, characterized in that: include: Acquire real-time view data, the real-time view data at least including road view data, vehicle view data and driver view data, the road view data at least including vehicle position data, pedestrian data, lane line data and indicator light data, the vehicle view data at least including license plate number data, and the driver view data at least including driver face data; Inputting the road view data into the trained CV model to determine a target traffic event corresponding to the road view data; updating the initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and updating the initial face file according to the driver view data to obtain an updated face file; An association relationship is established between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a full traffic file.

2. The method according to claim 1, characterized in that: Inputting the road view data into the trained CV model, and determining a target traffic event corresponding to the road view data, including: Determining a plurality of image data and a shooting time of each of the image data according to the road view data; Extracting features from the image data to obtain object features of various objects on the road, wherein the objects include at least indicator lights, lane lines, pedestrians, and vehicles on the road, and the object features include at least the type of the object and the position coordinates of the object; Determining the position coordinates of each of the objects at different times according to the shooting times of the plurality of image data and the object features of each of the objects corresponding to the plurality of image data; A target traffic event corresponding to the road view data is determined at least according to the position coordinates of the plurality of objects at different times.

3. The method according to claim 1, characterized in that The driver face data has multiple data, and the initial face file is updated according to the driver face data to obtain an updated face file, including: A first comparison step, comparing the plurality of driver face data with the plurality of face data in the initial face file, respectively, to obtain a first comparison similarity of each driver face data, wherein the first comparison similarity represents the similarity between the driver face data and one of the face data; A first determination step, determining a first cluster file and a first unarchived data according to all the driver face data and the corresponding first comparison similarity, and persistently storing the first cluster file in the initial face file to obtain a first updated face file; a second comparison step, comparing the first unarchived data with the plurality of face data in the first updated face archive at a preset time to obtain a second comparison similarity; A second determination step, determining a second cluster archive and a second unarchived data according to the first unarchived data and the corresponding second comparison similarity, and persistently storing the second cluster archive in the initial face archive to obtain a second updated face archive; The first comparing step, the first determining step, the second comparing step and the second determining step are repeatedly performed with the target duration as a period.

4. The method according to claim 3, characterized in that: Determining a first cluster archive and a first unarchived data according to all the driver face data and the corresponding first comparison similarity, including: Clustering the driver face data according to all the driver face data and the corresponding first comparison similarities to obtain a plurality of first cluster data and a plurality of initial unarchived data, wherein each driver face data in the first cluster data corresponds to face data of an initial face archive, and a first comparison similarity between the driver face data in the first cluster data and the corresponding face data is greater than a similarity threshold; Clustering all the initial unarchived data to obtain a plurality of second clustered data, each of which corresponds to an unrecorded face data, where the unrecorded face data is face data that does not exist in the initial face archive; When the amount of data in the second cluster data is greater than or equal to a preset amount, respectively storing a plurality of the first cluster data in the corresponding files of the face data, and separately storing the second cluster data in the newly added face data file, to obtain the first cluster file; When the amount of data in the second cluster data is less than a preset amount, the data in the second cluster data is determined as the first unarchived data.

5. The method according to claim 3, characterized in that: Comparing the plurality of driver face data with the plurality of face data in the initial face file respectively includes: When the driver's facial data is greater than or equal to a preset data amount, comparing the plurality of driver's facial data with the plurality of facial data in the initial facial file respectively; and / or, Each time a preset period of time passes, the plurality of driver face data are compared with the plurality of face data in the initial face file.

6. The method according to claim 1, characterized in that Before inputting the road view data into the trained CV model, the method further comprises: Build an initial CV model; The initial CV model is trained using multiple sets of training data to obtain the trained CV model, wherein each set of training data in the multiple sets of training data includes: historical road view data and historical traffic events corresponding to the historical road view data acquired within a historical time period.

7. The method according to claim 1, characterized in that Get real-time view data, including: Acquire initial view data, where a format of the initial view data is a first format, and the first format is an audio format or a video format; The format of the initial view data is converted from the first format to a second format using the FFmpeg protocol to obtain the real-time view data, where the second format is a format recognizable by the trained CV model.

8. A traffic incident detection and archiving device, characterized in that: include: an acquisition unit, configured to acquire real-time view data, wherein the real-time view data includes at least road view data, vehicle view data and driver view data, wherein the road view data includes at least vehicle position data, pedestrian data, lane line data and indicator light data, wherein the vehicle view data includes at least license plate number data, and wherein the driver view data includes at least driver face data; a determining unit, configured to input the road view data into a trained CV model to determine a target traffic event corresponding to the road view data; an archiving unit, configured to update an initial vehicle file according to the vehicle view data to obtain an updated vehicle file, and to update an initial face file according to the driver view data to obtain an updated face file; The construction unit is used to construct an association relationship between any two of the target traffic event, the updated vehicle file and the updated face file to obtain a full traffic file.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for associating traffic events as described in any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for associating traffic events as described in any one of claims 1 to 7.

Images