Traffic emergency mode rapid identification system and method based on visual question and answer technology

Abstract

The invention discloses a traffic emergency mode rapid identification system and method based on a visual question and answer technology, and relates to the technical field of intelligent traffic management, informatization technology and active safety prevention and control, and the traffic emergency mode rapid identification system based on the visual question and answer technology comprises an information reporting module and an information processing module, a real-time video calling module, a fuzzy matching module and an identification module; the information reporting module is used for sending traffic event information to the system; the information processing module is used for analyzing the traffic event information to obtain time and place information of emergencies and key feature information of the emergencies; the real-time video calling module is used for acquiring monitoring video data of corresponding time on a corresponding road section; the fuzzy matching module is used for matching a plurality of multi-dimensional feature basic normal forms according to the key feature information of the emergencies; and the identification module performs rapid and accurate identification based on the monitoring video data of the corresponding time on the corresponding road section and the multiple multi-dimensional feature basic normal forms to obtain a unique multi-dimensional feature basic normal form, so that corresponding event processing suggestions are obtained, emergencies can be processed in time, and the risk of derivative accidents is reduced.

Description

Technical Field

[0001] The present invention relates to the fields of intelligent traffic management, information technology and active safety prevention and control technology, and in particular to a traffic emergency pattern rapid recognition system based on visual question-answering technology. Background Art

[0002] With the rapid development of my country's expressway construction and the rapid growth of traffic volume, emergencies such as traffic accidents and vehicle failures occur frequently. Compared with non-emergency events such as traffic jams and road construction and maintenance, emergencies are characterized by suddenness, urgency and severity, which significantly increases the risk of derivative accidents. Taking a certain region as an example, from June 2016 to June 2021, there were 58,539 traffic accidents and 87,911 vehicle failures on expressways, of which derivative accidents caused by traffic accidents accounted for 2.86%. How to reduce the impact of emergencies and reduce the risk of derivative accidents in a timely manner has become a major challenge faced by the safe operation of expressways.

[0003] From a micro perspective, emergencies can lead to emergency situations such as damage to road safety facilities and closure of some lanes, disrupting the original stable traffic operation environment that meets road design standards. At the same time, when facing different emergency modes, drivers' driving behaviors will differ from normal driving states to varying degrees, leading to an imbalance in the inherent safety coupling mechanism of road design elements. From a macro perspective, the traffic flow state after an emergency will change dramatically, and the resulting derivative accident precursors, traffic flow state transitions, and traffic wave transmission laws will be different from those of conventional events. Depending on the emergency mode, the evolution law of traffic flow will also be different. At the same time, the reasonable intervention mechanism under normal circumstances will be disturbed by the influence of emergencies, making it difficult to control accident risks, leading to an imbalance in the inherent safety interaction mechanism of the human-vehicle-road-environment system.

[0004] Therefore, compared with routine events on highways, emergencies caused by road traffic accidents, vehicle breakdowns, spilled objects, etc. are more destructive to the inherent intrinsic safety mechanism of highways, and the resulting derivative accidents are often more harmful than ordinary accidents, forming a "short board effect" in the process of highway construction and safe operation, which seriously restricts the overall inherent safety level of my country's highways. Summary of the invention

[0005] In order to overcome the defects of the above-mentioned prior art in that emergencies are not handled in time and the risk of derivative accidents is increased, the present invention proposes a system and method for rapid recognition of traffic emergency patterns based on visual question answering technology.

[0006] To achieve the above object, the present invention adopts the following technical solutions, including:

[0007] A traffic emergency pattern rapid recognition system based on visual question answering technology, comprising: an information reporting module, an information processing module, a real-time video retrieval module, a fuzzy matching module and a recognition module;

[0008] The information reporting module is used to send traffic event information to the system; the information processing module is used to parse the traffic event information to obtain the time and location information of the emergency and the key feature information of the emergency; the real-time video retrieval module is used to obtain the monitoring video data of the corresponding time on the corresponding road section; the fuzzy matching module is used to match multiple multi-dimensional feature basic paradigms according to the key feature information of the emergency; the identification module obtains a unique multi-dimensional feature basic paradigm based on the monitoring video data of the corresponding time on the corresponding road section and identifies multiple multi-dimensional feature basic paradigms, thereby obtaining corresponding event handling opinions.

[0009] Preferably, the key feature information includes at least one basic feature data of the emergency event; the fuzzy matching module receives the key feature information sent by the key feature information unit, and uses the association rule mining method to mine a plurality of fuzzy matching multi-dimensional feature basic paradigms containing the key feature information from the paradigm library;

[0010] The paradigm library is constructed according to historical emergencies, and the multi-dimensional feature basic paradigm is composed of basic feature data and treatment opinions of historical emergencies.

[0011] Preferably, the basic characteristic data includes event characteristic data, traffic flow data, driving behavior data, road design data, control measure data and weather environment data.

[0012] Preferably, the recognition module includes an image processing unit and a characteristic problem content unit; the image processing unit extracts features from the real-time monitoring video data of the corresponding time on the corresponding road section in combination with the time and location information to obtain the video features of the emergency;

[0013] The characteristic question content unit determines the characteristic question according to the multiple multi-dimensional characteristic basic paradigms matched by the fuzzy matching module; searches for basic features with different basic characteristic data values ​​from the multiple multi-dimensional characteristic basic paradigms, and uses the basic features as the characteristic question;

[0014] The recognition module uses question-answering technology to find answers to characteristic questions from the video features of emergencies, determine the values ​​of characteristic questions, and further determine the unique multidimensional characteristic basic paradigm. Based on the determined unique multidimensional characteristic basic paradigm, the corresponding event handling opinions are obtained.

[0015] Preferably, the information processing module uses information extraction technology in natural language processing to textually process the traffic event information, dividing it into time and place information and key feature information; the time and place information is stored in a time and place information unit, and the time and place information unit will send a video retrieval request to the real-time video retrieval module based on the time and place information; the key feature information is stored in a key feature information unit, and the key feature information unit will send the key feature information to the fuzzy matching module.

[0016] Preferably, the information reporting module sends the traffic event information to the information processing module.

[0017] Preferably, the real-time video retrieval module receives a video retrieval request, and calls a real-time video monitoring database according to time and location information to obtain real-time monitoring video data at a corresponding time on a corresponding road section.

[0018] A method for quickly identifying traffic emergency patterns based on visual question answering technology, applicable to a system for quickly identifying traffic emergency patterns based on visual question answering technology, comprising:

[0019] S1: Receive traffic incident information in real time, and parse the traffic incident information to obtain the time and location information of the emergency incident and key feature information of the emergency incident;

[0020] S2: Based on the time and location information of the emergency, obtain the surveillance video data of the corresponding time on the corresponding road section;

[0021] S3: Based on the key feature information of the emergency, fuzzy matching is used to obtain multiple multidimensional feature basic paradigms;

[0022] S4: Based on the surveillance video data of the corresponding road section at the corresponding time and multiple multi-dimensional feature basic paradigms, a unique multi-dimensional feature basic paradigm is identified to obtain corresponding event handling opinions.

[0023] A readable storage medium stores a computer program, which, when executed, implements a method for quickly identifying traffic emergency patterns based on visual question-answering technology.

[0024] A computer program product includes a computer program / instruction, which, when executed by a processor, implements a method for quickly identifying traffic emergency patterns based on visual question answering technology.

[0025] The advantages of the present invention are:

[0026] (1) The present invention receives traffic event information in real time, and parses the traffic event information to obtain the time and location information of the emergency and key feature information of the emergency; based on the time and location information of the emergency, obtains the monitoring video data of the corresponding time on the corresponding road section; based on the key feature information of the emergency, fuzzy matching is performed to obtain multiple multi-dimensional feature basic paradigms; based on the monitoring video data of the corresponding time on the corresponding road section and the multiple multi-dimensional feature basic paradigms, a unique multi-dimensional feature basic paradigm is quickly identified, so that corresponding event handling opinions can be obtained quickly and accurately, reducing the risk of derivative accidents.

[0027] (2) The present invention first extracts simple traffic event information to obtain time, location information and key feature information, and then fuzzy matches the basic paradigm of multidimensional features in the paradigm library based on the key feature information, so as to subsequently search for a unique basic paradigm of multidimensional features and obtain corresponding event processing opinions, thereby improving the search efficiency.

[0028] (3) The present invention first extracts simple traffic event information to obtain time, location information and key feature information, obtains real-time monitoring video data of the corresponding road section at the corresponding time based on the time and location information, and then uses a number of feature questions to specifically view the monitoring video, so as to quickly obtain answers to the feature questions and match the unique multi-dimensional feature basic paradigm, so that the system can respond promptly and quickly.

[0029] (4) The recognition module of the present invention uses visual question-answering technology to integrate the extracted emergency video features and feature questions to form a unified input. Through the deep learning model, the relationship between the video features and the feature questions is analyzed to identify their similarities and correlations. Based on the matching results, the model obtains the processing plan related to the emergency, provides specific response suggestions, and finally outputs the corresponding event processing opinions to support the decision-making process and rapid response. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a system architecture diagram of the present invention;

[0031] Figure 2 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] like Figure 1As shown, the first aspect of the present invention proposes a traffic emergency pattern rapid recognition system based on visual question answering technology, comprising:

[0034] Information reporting module, information processing module, real-time video retrieval module, fuzzy matching module and recognition module;

[0035] The information reporting module is used to send traffic incident information to the system; the information processing module is used to parse the traffic incident information to obtain the time and location information of the emergency and the key feature information of the emergency; the real-time video retrieval module is used to obtain the monitoring video data of the corresponding time on the corresponding road section; the fuzzy matching module is used to match multiple multi-dimensional feature basic paradigms according to the key feature information of the emergency; the recognition module quickly and accurately identifies the monitoring video data of the corresponding time on the corresponding road section and multiple multi-dimensional feature basic paradigms, and obtains a unique multi-dimensional feature basic paradigm, thereby obtaining corresponding event handling opinions.

[0036] The real-time reporting information database for emergencies sends traffic event information to the information processing module. The traffic event information comes from the process of expressway operation, when drivers, traffic police, maintenance teams, road administration, and rescue team personnel use mobile phones and other electronic network devices to report to the real-time reporting information database for emergencies.

[0037] For example, at 17:38 on June 1, the traffic police called

[0038] G4001 Hefei Ring Expressway North Ring Section: Lu'an to Nanjing direction 65 km (Hefei West Hub, Changfeng County) a car crashed into the guardrail and flipped outside the guardrail. No road is occupied, the left lane, right lane and emergency lane are open. It is currently being handled. Rescue notice at 18:40: The accident was resolved.

[0039] The information processing module receives the traffic event information sent by the information reporting module, and uses the information extraction technology in natural language processing (Enhancedjieba-NER technology in this embodiment) to text the traffic event information, divide it into time and place information and key feature information, and store the time and place information in the time and place information unit. At the same time, the time and place information unit will send a video retrieval request to the real-time video retrieval module based on the time and place information, and the key feature information is stored in the key feature information unit. At the same time, the key feature information unit sends the key feature information to the fuzzy matching module; the key feature information includes no less than one basic feature data of the emergency.

[0040] For example, the traffic event information received by the information processing module:

[0041] At 17:38 on June 1, a car crashed into a guardrail and then flipped outside the guardrail on the G4001 Hefei Ring Expressway North Ring Section: 65 km from Lu'an to Nanjing (Hefei West Hub, in Changfeng County). The car did not occupy the road, and the left lane, right lane, and emergency lane were open to traffic. The accident is currently being handled. At 18:40, rescue notice: the accident was resolved.

[0042] The information processing module processes the traffic event information into text format and divides it into time and location information and key feature information:

[0043] 1) Time and location information: The time is 17:38-18:40 on June 1, G4001 Hefei Ring Expressway North Ring Section: 65 kilometers from Lu'an to Nanjing (Hefei West Hub, Changfeng County).

[0044] 2) Key feature information: traffic accident; no road occupation; left lane is passable; right lane is passable;

[0045] The emergency lane is open to traffic; the car hit the guardrail; the guardrail was destroyed; 1 car; a sedan.

[0046] The reported traffic event information includes time, location and key feature information. For example, if one of the reported information is that the collision type is hitting the guardrail, then this data belongs to the event feature data in the basic feature data. When matching, we need to see which paradigms in this category have the collision type of hitting the guardrail.

[0047] The real-time video retrieval module receives the video retrieval request and calls the real-time video surveillance database according to the time and location information to obtain the real-time surveillance video data of the corresponding time on the corresponding road section. The real-time video surveillance database is composed of the video surveillance platform data of the highway management department, which can be used to view and retrieve real-time video surveillance and has a data docking port.

[0048] For example, the real-time video retrieval module receives the video retrieval request information:

[0049] An application was made to retrieve the video surveillance platform of the highway management department of the northern section of Hefei Ring Expressway, and the real-time surveillance video data of the section near 65 kilometers from Lu'an to Nanjing (Hefei West Hub, in Changfeng County) from 17:38 on June 1 to 18:40 on the same afternoon.

[0050] The real-time video retrieval module retrieves video information according to the time and location in the application information: the video surveillance platform of the highway management department of the northern section of Hefei Ring Expressway is retrieved, and the real-time surveillance video data of the section near 65 kilometers from Lu'an to Nanjing (Hefei West Hub, Changfeng County) from 17:38 on June 1 to 18:40 on the same afternoon.

[0051] The fuzzy matching module receives the key feature information sent by the key feature information unit, and adopts the association rule mining method based on the Apriori algorithm to mine the multidimensional feature basic paradigms containing multiple fuzzy matches of the key feature information from the paradigm library; the paradigm library is constructed according to historical emergencies, and the multidimensional feature basic paradigm is composed of the basic feature data and handling opinions of historical emergencies.

[0052] Among them, the basic characteristic data include event characteristic data, traffic flow data, driving behavior data, road design data, control measures data and weather environment data. The event characteristic data refers to some characteristic situation data of the emergency, such as the severity of the event, the number of vehicles involved, the type of vehicle collision involved, the casualties involved, the property loss involved, the spatiotemporal impact range of the event, the number of occupied lanes, the damage to facilities caused and other characteristic information of the emergency. The traffic flow data refers to the traffic situation data around the emergency, such as the flow rate, speed and occupancy rate of each lane every thirty seconds at the place where the emergency occurred and 500 meters, 1000 meters, 1500 meters and 2000 meters upstream at the time of the emergency and within 20 minutes before and after the emergency. The driving behavior data refers to some driving behavior data of the driver before and after the incident, such as whether the vehicle changed lanes or not. The data include driving behaviors such as crossing, following, and overtaking; road design data, i.e. road data of the relevant sections where the emergency occurs, such as the number of lanes, lane width, central dividing strip form, inner and outer shoulder styles, whether it is in the merging area or diverging area influence area, and other road geometric design parameters and safety facilities of the cross-section of the expressway at the place where the emergency occurs and 500 meters, 1000 meters, 1500 meters and 2000 meters upstream; control measures data, i.e. road setting restriction data within the relevant scope of the emergency occurrence, such as speed limit, flow control measures, induction measures and other control measures at the cross-section of the expressway at the place where the emergency occurs and 500 meters, 1000 meters, 1500 meters and 2000 meters upstream; weather environment data, i.e. weather conditions when the emergency occurs, such as whether it is sunny, rainy, foggy or snowy within the spatiotemporal impact range of the emergency.

[0053] Each paradigm in the paradigm library contains several or all of the six categories of basic characteristic data.

[0054] For example, the fuzzy matching module receives key feature information: traffic accident; no road occupation; left lane is passable; right lane is passable; emergency lane is passable; car hits guardrail; guardrail is destroyed; 1 car; sedan.

[0055] Based on these key feature information, matching is performed in the paradigm library, and finally three multidimensional feature basic paradigms are obtained. The matching results are shown in Table 1.

[0056] Table 1

[0057]

[0058]

[0059] Based on the event type, accident type, whether the road is occupied and other characteristic information, fuzzy matching is performed to obtain paradigms 1, 2, and 3.

[0060] The recognition module includes an image processing unit and a feature question content unit. First, the image processing unit uses deep learning visual detection technology (Hybrid-YOLOv8-Transformer algorithm is used in this embodiment) to combine the event time and location information to extract features of the real-time monitoring video data at the corresponding time on the corresponding road section to obtain the video features of the emergency event; the feature question content unit uses feature extraction and expression technology (IMG and BOW technology in this embodiment) to determine the feature problem according to multiple multidimensional feature basic paradigms; basic features with different basic feature data values ​​are found from multiple multidimensional feature basic paradigms, and the basic features are used as feature questions; then the recognition module uses visual question answering technology (AAVQA technology in this embodiment) to find the answer to the question feature from the emergency video features and determine the data value of the question feature, so as to further accurately find the unique multidimensional feature basic paradigm, and obtain the corresponding event handling opinion according to the determined unique multidimensional feature basic paradigm.

[0061] Among them, the Hybrid-YOLOv8-Transformer algorithm is an innovative target detection method for real-time video surveillance and intelligent traffic management. It combines the efficient detection architecture of YOLOv8 (You Only Look Once v8) and the powerful feature extraction capability of the Transformer network. Its main function is to enhance the feature extraction capability of video data so as to quickly identify traffic emergencies.

[0062] AAVQA visual question answering technology, its main function is to process emergency video features and feature questions, and obtain corresponding event handling opinions. The extracted emergency video features and feature questions are integrated to form a unified input. Through the deep learning model, the relationship between video features and feature questions is analyzed, their similarities and relevance are identified, and based on the matching results, the model generates emergency-related handling solutions and provides specific response suggestions. Finally, the corresponding event handling opinions are output to support the decision-making process and rapid response.

[0063] The recognition module uses the AAVQA algorithm to extract emergency video features from the video to answer uncertain questions in fuzzy matching and accurately match the unique multi-dimensional feature basic paradigm.

[0064] The video features of emergencies include the number of vehicles involved, lane occupancy, damage to safety facilities, and visibility;

[0065] For example, the image processing unit in the recognition module receives surveillance video from 17:38 to 18:40 on June 1st, 65 kilometers from Lu'an to Nanjing (Hefei West Hub, Changfeng County), and obtains the video features of the emergency; the feature question content unit receives three multi-dimensional feature basic paradigms, and extracts and sorts out several feature questions based on these three multi-dimensional feature basic paradigms: number of deaths, number of injuries, visibility, average vehicle speed, etc.

[0066] Then the recognition module matches the emergency video features with the characteristic questions to obtain a unique multi-dimensional feature basic paradigm, thereby obtaining the corresponding event handling opinions.

[0067] The answer to the emergency question obtained through the emergency video features is: 0 deaths, 1 injured, high visibility, average speed 92km / h, etc. The unique multidimensional feature basic paradigm is obtained through matching the determined emergency answer, that is, paradigm 2. The current event features show that this event is a minor accident, which will be handled in time and the scope of the accident will not expand. The specific situation is subject to actual inspection.

[0068] Get event handling advice through Paradigm 2:

[0069] a) Treatment of personnel: If there are injured personnel, arrange for medical personnel to conduct further examination;

[0070] b) Road handling: The accident vehicle does not occupy the lane. No lane management is required;

[0071] c) Vehicle handling: If the vehicle is damaged and cannot be driven, please arrange towing and repair;

[0072] d) Speed ​​limit: When the average number of vehicles is high, the speed of the lane is limited.

[0073] like Figure 2 As shown, the second aspect of the present invention proposes a method for rapid recognition of traffic emergency patterns based on visual question answering technology, which is applicable to the above-mentioned rapid recognition system for traffic emergency patterns based on visual question answering technology, comprising:

[0074] S1: Receive traffic incident information in real time, and parse the traffic incident information to obtain the time and location information of the emergency incident and key feature information of the emergency incident;

[0075] S2: Based on the time and location information of the emergency, obtain the surveillance video data of the corresponding time on the corresponding road section;

[0076] S3: Based on the key feature information of the emergency, fuzzy matching is used to obtain multiple multidimensional feature basic paradigms;

[0077] S4: Based on the surveillance video data of the corresponding time on the corresponding road section and multiple multi-dimensional feature basic paradigms, a unique multi-dimensional feature basic paradigm is quickly and accurately identified to obtain the corresponding event handling opinions.

[0078] Of course, it is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, but also includes the same or similar structures that can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-limiting, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any reference numerals in the claims should not be regarded as limiting the claims involved.

[0079] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

[0080] The techniques, shapes, and structural parts not described in detail in the present invention are all well-known techniques.

Claims

1. A traffic emergency pattern rapid recognition system based on visual question answering technology, characterized in that: include: Information reporting module, information processing module, real-time video retrieval module, fuzzy matching module and recognition module; The information reporting module is used to send traffic incident information to the system; the information processing module is used to parse the traffic incident information to obtain the time and location information of the emergency incident and the key feature information of the emergency incident; The real-time video retrieval module is used to obtain the surveillance video data of the corresponding time on the corresponding road section; the fuzzy matching module is used to match multiple multi-dimensional feature basic paradigms according to the key feature information of the emergency; The recognition module obtains a unique multi-dimensional feature basic paradigm based on the surveillance video data of the corresponding time on the corresponding road section and identifies multiple multi-dimensional feature basic paradigms, thereby obtaining corresponding event processing opinions.

2. A traffic emergency pattern rapid recognition system based on visual question answering technology as claimed in claim 1, characterized in that: The key feature information includes at least one basic feature data of the emergency event; the fuzzy matching module receives the key feature information sent by the key feature information unit, and uses the association rule mining method to mine a plurality of fuzzy matching multi-dimensional feature basic paradigms containing the key feature information from the paradigm library; The paradigm library is constructed according to historical emergencies, and the multi-dimensional feature basic paradigm is composed of basic feature data and treatment opinions of historical emergencies.

3. A traffic emergency pattern rapid recognition system based on visual question answering technology as claimed in claim 2, characterized in that: The basic characteristic data include event characteristic data, traffic flow data, driving behavior data, road design data, control measures data and weather environment data.

4. The traffic emergency pattern rapid recognition system based on visual question answering technology as claimed in claim 1, characterized in that: The recognition module includes an image processing unit and a characteristic problem content unit; the image processing unit extracts features from the real-time monitoring video data of the corresponding time on the corresponding road section in combination with the time and location information to obtain the video features of the emergency event; The characteristic question content unit determines the characteristic question according to the multiple multi-dimensional characteristic basic paradigms matched by the fuzzy matching module; searches for basic features with different basic characteristic data values ​​from the multiple multi-dimensional characteristic basic paradigms, and uses the basic features as the characteristic question; The recognition module uses question-answering technology to find answers to characteristic questions from the video features of emergencies, determine the values ​​of characteristic questions, and further determine the unique multidimensional characteristic basic paradigm. Based on the determined unique multidimensional characteristic basic paradigm, the corresponding event handling opinions are obtained.

5. The traffic emergency pattern rapid recognition system based on visual question answering technology as claimed in claim 1, characterized in that: The information processing module uses the information extraction technology in natural language processing to textually process the traffic event information, dividing it into time and location information and key feature information; the time and location information is stored in the time and location information unit, and the time and location information unit will send a video retrieval request to the real-time video retrieval module based on the time and location information; the key feature information is stored in the key feature information unit, and the key feature information unit will send the key feature information to the fuzzy matching module.

6. A traffic emergency pattern rapid recognition system based on visual question answering technology as claimed in claim 1, characterized in that: The information reporting module sends the traffic event information to the information processing module.

7. The traffic emergency pattern rapid recognition system based on visual question answering technology as claimed in claim 1 is characterized in that: The real-time video retrieval module receives the video retrieval request and calls the real-time video monitoring database according to the time and location information to obtain the real-time monitoring video data of the corresponding time on the corresponding road section.

8. A method for rapid identification of traffic emergency patterns based on visual question answering technology, characterized in that: A traffic emergency pattern rapid identification system based on visual question answering technology applicable to any one of claims 1 to 7, comprising: S1: Receive traffic incident information in real time, and parse the traffic incident information to obtain the time and location information of the emergency incident and key feature information of the emergency incident; S2: Based on the time and location information of the emergency, obtain the surveillance video data of the corresponding time on the corresponding road section; S3: Based on the key feature information of the emergency, fuzzy matching is used to obtain multiple multidimensional feature basic paradigms; S4: Based on the surveillance video data of the corresponding road section at the corresponding time and multiple multi-dimensional feature basic paradigms, a unique multi-dimensional feature basic paradigm is identified to obtain corresponding event handling opinions.

9. A readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, the method for quickly identifying traffic emergency patterns based on visual question-answering technology as described in claim 8 is implemented.

10. A computer program product, characterized in that It includes a computer program / instruction, which, when executed by a processor, implements the method for quickly identifying traffic emergency patterns based on visual question-answering technology as described in claim 8.

Images