Tunnel fire risk assessment method and device, computer equipment and storage medium
Through video stream processing technology, vehicles in the tunnel are identified and tracked and their risks are evaluated, and the problems of untimely and inaccurate tunnel fire risk assessment in the prior art are solved, and real-time dynamic assessment and management of fire risk in the tunnel are realized.
Patent Information
- Application Number
- CN202411132613.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-05-31
- Filing Date
- 2024-08-16
- Publication Date
- 2024-11-19
AI Technical Summary
Existing tunnel safety technologies have failed to effectively evaluate and manage fire risks in tunnels, resulting in frequent fire accidents and lack of real-time dynamic assessment capabilities.
By obtaining the video stream of the tunnel entrance, computer vision and machine learning technology are used to identify and track vehicles entering the tunnel, predict their trajectory and speed, and combining vehicle types and risk indicators to generate a fire risk distribution map to provide real-time fire risk assessment.
Real-time and dynamic assessment of tunnel fire risks has been achieved, timeliness and accuracy of risk management has been improved, and decision-making support is provided for timely prevention and emergency response.
Smart Images

Figure CN118982779A_ABST
Abstract
Description
[0001] This application claims priority to U.S. Provisional Patent Application No. 63 / 654,251 filed on May 31, 2024, the entire contents of which are incorporated herein by reference. Technical Field
[0002] The present application relates to the field of tunnel safety technology, and in particular to a tunnel fire risk assessment method, device, computer equipment and storage medium. Background Art
[0003] As the number of tunnels increases, while bringing convenience, it also increases the risk of fire. Due to the closed and narrow space structure of the tunnel and the effects of "black hole" and "white hole", highway tunnels have become accident-prone areas. In highway tunnels with existing technologies, once a traffic accident occurs, it often leads to a fire accident, and even the fire spreads to multiple cars, and the smoke spreads to the entire tunnel. In existing tunnel safety technologies, there is no good continuous assessment of the fire risk in the tunnel. Summary of the invention
[0004] The purpose of this application is to solve at least one of the above-mentioned technical deficiencies, and in particular to propose a method for efficiently and accurately assessing the risk of tunnel fire.
[0005] In a first aspect, the present application provides a tunnel fire risk assessment method, comprising:
[0006] Acquire a first video stream; the first video stream is a video stream shot at a tunnel entrance;
[0007] Determine the type and trajectory of each vehicle in the tunnel based on the first video stream;
[0008] Based on each trajectory, the distribution of tunnel vehicles at different times is obtained;
[0009] Determine the vehicle set corresponding to each position in the tunnel according to the tunnel vehicle distribution, and determine the tunnel risk index corresponding to each position in the tunnel at the corresponding time according to the category of each vehicle in the vehicle set and the first corresponding relationship; the first corresponding relationship is the corresponding relationship between the category and the vehicle risk index;
[0010] According to the risk index of each tunnel, the fire risk distribution is obtained.
[0011] In one embodiment, determining the type and trajectory of each vehicle in the tunnel based on the first video stream includes:
[0012] Target recognition and tracking are performed based on the first video stream, thereby assigning a unique identifier to each vehicle entering the tunnel, and obtaining a category and a trajectory before entering the tunnel corresponding to the unique identifier;
[0013] Based on the trajectory before entry, the predicted average speed is obtained;
[0014] According to the predicted average speed, the predicted post-entry trajectory is obtained;
[0015] The predicted post-entry trajectory is associated with the unique identifier of the corresponding pre-entry trajectory to obtain the category and trajectory of the vehicle in the tunnel.
[0016] In one embodiment, target recognition and tracking are performed based on the first video stream, so as to assign a unique identifier to each vehicle entering the tunnel, and obtain the category and pre-entry trajectory corresponding to the unique identifier, including:
[0017] Inputting the current frame in the first video stream into the target detection model to obtain a recognition result corresponding to the current frame; the recognition result includes the categories and bounding box information of multiple vehicles to be matched;
[0018] For any unfinished pre-entry trajectory, generate the predicted position corresponding to the current frame, and match the predicted position with the bounding box information of each vehicle to be matched;
[0019] If the bounding box information of the to-be-matched vehicle successfully matches the predicted position of a pre-entry trajectory, the pre-entry trajectory is updated according to the bounding box information of the to-be-matched vehicle, and the unique identifier of the pre-entry trajectory is associated with the to-be-matched vehicle;
[0020] If the bounding box information of the to-be-matched vehicle does not have a successfully matched pre-entry trajectory, a new trajectory is created for the to-be-matched vehicle and a new unique identifier is assigned.
[0021] In one embodiment, before performing target recognition and tracking based on the first video stream, the method includes:
[0022] Performing a perspective transformation on each video frame in the first video stream to convert the perspective of each video frame into a top-down perspective;
[0023] According to the trajectory before entry, the predicted average speed is obtained, including:
[0024] Determine the pixel displacement between each two adjacent frames according to the trajectory before entering;
[0025] Convert each pixel displacement into physical displacement;
[0026] According to each physical displacement and frame interval, a plurality of first speeds are obtained;
[0027] Based on each first speed, a predicted average speed is obtained.
[0028] In one embodiment, the tunnel fire risk assessment method further includes:
[0029] Based on the second video stream, the identification track of each vehicle in the shooting range is determined by using a cross-camera identification method, and the identification track is associated with an existing unique identifier, wherein the second video stream is a video stream shot at more than one different position in the tunnel;
[0030] Using the corresponding recognition trajectory, the corresponding predicted entry trajectory is corrected.
[0031] In one embodiment, the vehicle risk index includes the vehicle fire load and / or the vehicle heat release rate, and the tunnel risk index includes the single point fire load and / or the single point heat release rate; and according to each tunnel risk index, a fire risk distribution is obtained, including:
[0032] A first coordinate system is formed with the distance from the tunnel entrance as the horizontal axis and the size of the tunnel risk index as the vertical axis;
[0033] The tunnel risk index at the same time is plotted in the first coordinate system to obtain the first fire risk distribution at the corresponding time.
[0034] In one embodiment, the vehicle risk index includes vehicle fire load, and the tunnel risk index includes single-point fire load; and according to each tunnel risk index, a fire risk distribution is obtained, including:
[0035] Add up all single-point fire loads at the same time to get the total fire load at the corresponding time;
[0036] The second coordinate system is formed with time as the horizontal axis and the size of the fire load as the vertical axis;
[0037] The total fire load at multiple consecutive moments is plotted in the second coordinate system to obtain a second fire risk distribution.
[0038] In one embodiment, the tunnel fire risk assessment method further includes:
[0039] According to the actual fire occurrence point, determine the available evacuation points of the tunnel and predict the available safe evacuation time;
[0040] For any position corresponding to a non-empty vehicle set, the nearest available evacuation point is used as the target evacuation point, and the estimated evacuation time corresponding to the position is determined according to the distance between the target evacuation point and the position;
[0041] The maximum value among the estimated evacuation times is determined as the necessary safe evacuation time;
[0042] Determine the evacuation risk based on the required safe evacuation time and the preset available safe evacuation time.
[0043] In one embodiment, the tunnel fire risk assessment method further includes:
[0044] For any position corresponding to a non-empty vehicle set, the number of people to be evacuated corresponding to the position is determined according to the category of each vehicle in the corresponding vehicle set and the second corresponding relationship; the second corresponding relationship is the corresponding relationship between the category and the passenger capacity;
[0045] According to the number of people to be evacuated and the estimated evacuation time corresponding to each position in the tunnel, the remaining number of people to be evacuated at each moment is determined and plotted in a third coordinate system with time as the horizontal axis and the remaining number of people to be evacuated as the vertical axis to obtain the evacuation situation distribution.
[0046] In a second aspect, the present application provides a tunnel fire risk assessment device, comprising:
[0047] A data acquisition module, used to acquire a first video stream; the first video stream is a video stream shot at a tunnel entrance;
[0048] an identification module, for determining the type and trajectory of each vehicle in the tunnel based on the first video stream;
[0049] A vehicle distribution determination module is used to obtain the tunnel vehicle distribution at different times based on each trajectory;
[0050] A tunnel risk index determination module is used to determine the vehicle set corresponding to each position in the tunnel according to the tunnel vehicle distribution, and determine the tunnel risk index corresponding to each position in the tunnel at the corresponding time according to the category of each vehicle in the vehicle set and the first corresponding relationship; the first corresponding relationship is the corresponding relationship between the category and the vehicle risk index;
[0051] The fire risk distribution determination module is used to obtain the fire risk distribution according to the risk indicators of each tunnel.
[0052] In a third aspect, the present application provides a computer device comprising one or more processors and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the steps of the tunnel fire risk assessment method in any of the above-mentioned embodiments are executed.
[0053] In a fourth aspect, the present application provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the tunnel fire risk assessment method in any of the above embodiments.
[0054] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:
[0055] The fire risk assessment method based on the present application begins with acquiring the first video stream at the entrance of the tunnel, and then uses computer vision and machine learning techniques to extract the type of vehicles entering the tunnel and their trajectories in the tunnel from these data. Next, the predicted trajectory information is converted into a vehicle distribution map at different times in the tunnel. On this basis, the system will conduct a detailed analysis of each location in the tunnel, determine the vehicle composition around the location, and calculate the specific tunnel risk index for each location based on the pre-defined correspondence between the vehicle type and the vehicle risk index. Finally, by integrating these local risk indicators, the system generates a fire risk distribution for the entire tunnel and provides a comprehensive fire risk assessment. This method achieves real-time and dynamic assessment of tunnel fire risks at a low deployment and implementation cost, greatly improves the timeliness and accuracy of risk management, and provides valuable decision-making support for timely preventive measures and emergency responses. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0057] Figure 1 A schematic diagram of a flow chart of a tunnel fire risk assessment method provided in one embodiment of the present application;
[0058] Figure 2 A schematic diagram of a process for determining the type and trajectory of a vehicle in a tunnel in one embodiment of the present application;
[0059] Figure 3 This is a schematic diagram of a process of performing target recognition and tracking based on a first video stream in one embodiment of the present application;
[0060] Figure 4 A schematic diagram of a process for determining a predicted average speed in one embodiment of the present application;
[0061] Figure 5 This is a schematic diagram of a first fire risk distribution in one embodiment of the present application;
[0062] Figure 6 A schematic diagram of a second fire risk distribution in one embodiment of the present application;
[0063] Figure 7 A schematic diagram of a process for determining evacuation risk in one embodiment of the present application;
[0064] Figure 8 This is a schematic diagram of the layout of evacuation points in a tunnel in one embodiment of the present application;
[0065] Fig. 9 A schematic diagram of the distribution of evacuation conditions in one embodiment of the present application;
[0066] Fig.10 An internal structure diagram of a computer device provided for one embodiment of the present application. DETAILED DESCRIPTION
[0067] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions 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 are within the scope of protection of this application.
[0068] This application provides a tunnel fire risk assessment method, see Figure 1 , including steps S102 to S110.
[0069] S102, obtaining a first video stream. The first video stream is a video stream shot at a tunnel entrance.
[0070] It can be understood that the first video stream refers to the video data formed by the image acquisition device installed at the tunnel entrance continuously capturing and recording the situation of vehicles entering the tunnel. The video data includes a continuous image sequence, usually containing multiple video frames per second, and can fully record the dynamic process of vehicles entering the tunnel. Generally speaking, vehicles in different directions can enter both sides of the tunnel, and the videos recorded by the image acquisition devices installed on both sides of the tunnel can be the first video stream.
[0071] S104: Determine the type and trajectory of each vehicle in the tunnel based on the first video stream.
[0072] It can be understood that vehicle classification refers to identifying the specific type of each vehicle entering the tunnel, such as cars, trucks, buses, etc. This is usually achieved through deep learning models, especially models with architectures such as YOLO, RCNN, and SSD. Deep learning models can automatically learn the features of images to accurately classify vehicles. During the training process, the model will learn the typical features of various vehicle models, such as body shape, size ratio, and headlight position. In order to improve the classification accuracy, it is necessary to use a large amount of labeled vehicle image data for model training, and consider various factors that may affect recognition, such as different angles and lighting conditions.
[0073] Trajectory tracking is to track the position changes of each vehicle in continuous video frames. This usually involves target detection and multi-target tracking algorithms. Common methods include Kalman filter, particle filter, SORT or DeepSORT algorithms. These algorithms can establish a corresponding relationship between vehicles between continuous frames, and can keep tracking even when the vehicle is temporarily blocked or slightly deformed. The trajectory generated in this step does not need to be the change of the specific position of the vehicle in the tunnel, but only needs to reflect the distance between the vehicle and the tunnel entrance at each time. It is equivalent to the change of the vehicle position over time, and the connection formed by the change on the number axis with the tunnel entrance as the origin and the tunnel length direction as the extension direction. Therefore, this embodiment is based on the movement behavior characteristics of the vehicle before entering the tunnel, such as speed, to predict the movement behavior of the vehicle after entering the tunnel, so as to obtain the distance between each vehicle in the tunnel and the tunnel entrance at each time. Among them, if an image acquisition device is also installed in the tunnel, the predicted trajectory can also be corrected based on the video stream captured by the image acquisition device in the tunnel, so as to obtain more accurate trajectory information.
[0074] In practical applications, vehicle classification and trajectory tracking need to be used in combination. For example, you can first use the target detection algorithm to locate the vehicle in each frame, then classify the detected vehicles, and finally use the tracking algorithm to assign a unique identifier to the same vehicle in different frames, and connect the positions with the same unique identifier in different frames to form complete trajectory information.
[0075] S106, obtaining the tunnel vehicle distribution at different times according to each trajectory.
[0076] It can be understood that this step is to statistically analyze the vehicle trajectory information obtained in step S104 to obtain the vehicle distribution status in the tunnel, that is, the presence of vehicles at different positions in the tunnel at each moment. Specifically, each trajectory reflects the position of the corresponding vehicle at different moments. For the moment when the tunnel vehicle distribution needs to be determined, the position of the corresponding vehicle at that moment can be extracted from each trajectory. The vehicles contained in each position within the tunnel range can be counted to obtain the tunnel vehicle distribution at that moment. In addition, each position mentioned here can be a point marked at a fixed interval on the number axis with the tunnel entrance as the origin and the tunnel length direction as the extension direction. The distance from each point is within half of the fixed interval, which belongs to the vehicle contained in the position. In addition to recording which vehicles are contained in each position, the tunnel vehicle distribution will also include relevant information of these vehicles, such as category, speed, etc. The number distribution of tunnel vehicles can also be visualized by the horizontal axis being the distance from the tunnel entrance and the vertical axis being the number of vehicles.
[0077] S108, determining a vehicle set corresponding to each position in the tunnel according to the tunnel vehicle distribution, and determining a tunnel risk index corresponding to each position in the tunnel at a corresponding time according to the category of each vehicle in the vehicle set and the first corresponding relationship.
[0078] It can be understood that the vehicle set refers to the set of all vehicles existing in the corresponding area of each location in the tunnel. The first correspondence is a predefined mapping that associates different categories of vehicles with corresponding vehicle risk indicators. Since vehicles in the tunnel are potential fire sources, the risks brought by the combustion of each type of vehicle are different, and the vehicle risk indicator quantifies the degree of contribution to the fire risk from a specific perspective, which reflects the risk that this type of vehicle may cause in the event of a fire. This correspondence is usually based on a large amount of historical data analysis and expert knowledge. For example, a truck carrying flammable items may be assigned a higher vehicle risk indicator, while an ordinary car has a relatively low indicator. Commonly used vehicle risk indicators include vehicle fire load, vehicle heat release rate, etc. The tunnel risk indicator combines all vehicle risk indicators of the corresponding vehicle set at the location, reflecting the overall fire hazard level at the location.
[0079] S110, obtaining fire risk distribution according to the risk indicators of each tunnel.
[0080] It can be understood that fire risk distribution is a comprehensive concept. The tunnel risk indicators mentioned here include tunnel risk indicators at various positions in the tunnel at the same time, and also include tunnel risk indicators at various positions in the tunnel at different times. Therefore, by integrating the tunnel risk indicators at various positions in the tunnel at the same time, we can obtain the characteristics of the fire risk level in the tunnel changing with space. By integrating the tunnel risk indicators at various positions in the tunnel at different times, we can obtain the characteristics of the overall fire risk level in the tunnel changing with time. It should be noted that the risks mentioned in this application focus more on the consequences of fire rather than the possibility of fire.
[0081] The fire risk assessment method in this embodiment begins by acquiring the first video stream at the tunnel entrance, and then uses computer vision and machine learning techniques to extract the types of vehicles entering the tunnel and their trajectories in the tunnel from these data. Next, the predicted trajectory information is converted into a vehicle distribution map at different times in the tunnel. On this basis, the system will conduct a detailed analysis of each location in the tunnel, determine the vehicle composition around the location, and calculate the specific tunnel risk index for each location based on the pre-defined correspondence between vehicle types and vehicle risk indicators. Finally, by integrating these local risk indicators, the system generates a fire risk distribution for the entire tunnel and provides a comprehensive fire risk assessment. This method achieves real-time and dynamic assessment of tunnel fire risks at a low deployment and implementation cost, greatly improves the timeliness and accuracy of risk management, and provides valuable decision support for timely preventive measures and emergency responses.
[0082] In one embodiment, see Figure 2 , based on the first video stream, determining the category and trajectory of each vehicle in the tunnel, including steps S202 to S208.
[0083] S202, target recognition and tracking are performed based on the first video stream, so as to assign a unique identifier to each vehicle entering the tunnel, and obtain the category and pre-entry trajectory corresponding to the unique identifier.
[0084] It can be understood that this step is the basis of the entire process and involves two core tasks in computer vision: target recognition and target tracking. Target recognition refers to locating and classifying specific objects in an image or video frame, in this case, it is to identify vehicles. Target tracking is to track the movement of identified objects in continuous video frames. This embodiment requires the use of target recognition and target tracking in combination. The target recognition method can be used to locate the vehicle in each frame first, and then the detected vehicles are classified. Finally, the target tracking method is used to assign a unique identifier to the same vehicle in different frames, and the positions with the same unique identifier in different frames are connected to form complete trajectory information. The assignment of unique identifiers is a key point in this step. Whenever the system recognizes that a new vehicle enters the field of view, it will assign a unique identifier to it. This identifier can be a simple increasing integer or a more complex encoding containing a timestamp or other metadata. This unique identifier will be used throughout the processing process to distinguish different vehicles. In addition, since the first video stream captures the image of the path before the tunnel entrance, the pre-entry trajectory before entering the tunnel will be directly generated according to the first video stream, reflecting the position change of the vehicle from entering the field of view of the image acquisition device to entering the tunnel entrance. This trajectory is usually represented as a series of timestamps and corresponding coordinates.
[0085] S204, obtaining a predicted average speed based on the trajectory before entering.
[0086] It can be understood that in this embodiment, the feature that needs to be extracted from the movement behavior of the vehicle before entering the tunnel is the speed. This embodiment will predict the average speed of the vehicle after entering the tunnel based on the speed of the vehicle before entering the tunnel, that is, obtain the predicted average speed, which provides an important basis for predicting the position change of the vehicle in the tunnel.
[0087] S206, obtaining a predicted post-entry trajectory according to the predicted average speed.
[0088] It can be understood that since the trajectory required in this embodiment is mainly the change in distance between the vehicle and the tunnel entrance, after obtaining the predicted average speed, it is only necessary to record the moment when the unique identifier corresponding to the vehicle disappears among all the unique identifiers identified by the first video stream, and then subtract multiple moments from this moment respectively, and multiply the obtained time difference by the predicted average speed to obtain the change in distance between the vehicle and the tunnel entrance after entering the tunnel, that is, the predicted trajectory after entry.
[0089] S208 , associating the predicted post-entry trajectory with the unique identifier of the corresponding pre-entry trajectory to obtain the type and trajectory of the vehicle in the tunnel.
[0090] It can be understood that this step is to integrate the results of the previous steps to form a complete information description data for each vehicle entering the tunnel. The vehicle category, trajectory before entry, predicted trajectory after entry, predicted average speed, etc. can be associated by establishing and maintaining a coherent data structure. Specifically, a data structure needs to be designed to store and retrieve the information description data, which needs to include fields such as unique identification, vehicle category, trajectory before entry, predicted average speed, and predicted trajectory after entry. An index mechanism can also be established based on the unique identification so that the information of each vehicle can be quickly retrieved and updated. Specifically, an index mechanism can be established based on a hash table or a balanced tree structure.
[0091] In one embodiment, see Figure 3 , target recognition and tracking are performed based on the first video stream, thereby assigning a unique identifier to each vehicle entering the tunnel, and obtaining the category and pre-entry trajectory corresponding to the unique identifier, including steps S302 to S308.
[0092] S302: Input the current frame in the first video stream into the target detection model to obtain a recognition result corresponding to the current frame. The recognition result includes the categories and bounding box information of multiple vehicles to be matched.
[0093] It can be understood that target detection refers to locating and classifying specific objects in an image. In this embodiment, the specific object is a vehicle. The target detection model usually uses a deep learning algorithm, such as YOLO, Faster R-CNN or SSD. These models can simultaneously complete object location and classification in a single forward propagation. The results identified by the target detection model include the category of each vehicle and the bounding box information reflecting the location of the vehicle. The bounding box information includes the bounding box itself that encircles the vehicle, as well as the coordinate information of the bounding box. When collecting the first video stream, the continuously updated current frame will be continuously input into the target detection model for target detection.
[0094] The training of the target detection model requires the use of a large number of images of different categories of vehicles, and it is necessary to pay attention to balancing the number of vehicles of each category in the data set. In a specific embodiment, the categories of vehicles can be divided into ten categories, namely motorcycles, small cars, medium-sized cars, large cars, small buses, medium-sized buses, large buses, small trucks, medium-sized trucks and large trucks. The selected model structure can be YOLOv7, and the training can be carried out in a batch training manner. The batch size can be set to 16, the training cycle is 300, the initial learning rate is 0.01, and a gradient descent optimizer is used.
[0095] S304: For any unfinished pre-entry trajectory, a predicted position corresponding to the current frame is generated, and matching is performed with the bounding box information of each to-be-matched vehicle according to the predicted position.
[0096] It can be understood that the pre-entry trajectory will end after the corresponding vehicle enters the tunnel, while the pre-entry trajectory of the corresponding vehicle that has not entered the tunnel needs to continue to extend. Considering that each pre-entry trajectory is a known historical trajectory, the current position point of the trajectory can be predicted based on these historical situations to generate a predicted position corresponding to the current frame. The method of generating the predicted position can be to use the Kalman filter algorithm, and the generated predicted position can also be in the form of a bounding box. The bounding box information of each vehicle to be matched reflects the position of all vehicles appearing in the current frame. By matching the predicted position with the bounding box information of each vehicle to be matched, a bounding box closest to the predicted position can be found. The specific matching algorithm can be performed using the Hungarian algorithm.
[0097] S306: If the bounding box information of the to-be-matched vehicle successfully matches the predicted position of a pre-entry track, the pre-entry track is updated according to the bounding box information of the to-be-matched vehicle, and the unique identifier of the pre-entry track is associated with the to-be-matched vehicle.
[0098] It can be understood that if the bounding box information of the vehicle to be matched successfully matches a pre-entry track, it means that the vehicle to be matched is the same vehicle as the vehicle that forms the pre-entry track, and the bounding box information of the vehicle to be matched can be regarded as reflecting the current position point of the pre-entry track. The bounding box information is used to generate the latest position point for the pre-entry track, and connected with the previous pre-entry track to form an updated pre-entry track. In addition, it is also necessary to associate the unique identifier of the pre-entry track with the vehicle to be matched, so as to facilitate the distinction between each vehicle in each frame.
[0099] S308: If the bounding box information of the to-be-matched vehicle does not have a successfully matched pre-entry track, a new track is created for the to-be-matched vehicle, and a new unique identifier is assigned.
[0100] It can be understood that if the bounding box information of the vehicle to be matched is not successfully matched with a pre-entry track, it means that the vehicle to be matched is a new vehicle entering the field of view, and the bounding box information of the vehicle to be matched can be regarded as reflecting the starting point of a new pre-entry track. A new track is created for the vehicle to be matched using the bounding box information. In addition, a new unique identifier needs to be assigned to the pre-entry track and the vehicle to be matched to facilitate the distinction between the vehicles in each frame.
[0101] In one embodiment, in order to better determine the speed of each vehicle, considering that the viewing angle of the device shooting the first video stream is generally obliquely downward, it is easy to have distortion and other problems. Therefore, it is necessary to transform the viewing angle of each video frame in the first video stream to convert the viewing angle of each video frame into a top-down viewing angle.
[0102] Based on the pre-entry trajectory, the predicted average speed is obtained, see Figure 4 , including steps S402 to S408.
[0103] S402, determining the pixel displacement between each two adjacent frames according to the trajectory before entering.
[0104] It can be understood that the pre-entry trajectory is usually a data sequence consisting of a series of timestamps and corresponding image coordinates. In this step, we need to calculate the position change of the vehicle between consecutive frames. That is, the pixel displacement between each two adjacent frames can be obtained based on the difference in position coordinates between two adjacent frames.
[0105] S404, converting each pixel displacement into a physical displacement.
[0106] It can be understood that this step involves the conversion from image space to actual physical space, which is the key to calculating the true speed. The pixel displacement reflects the movement of the vehicle on the image plane and needs to be converted according to the corresponding relationship between the pixel displacement and the actual physical displacement.
[0107] S406, obtaining a plurality of first speeds according to the physical displacements and the frame intervals.
[0108] It can be understood that this step is to calculate the instantaneous speed using the known physical displacement and the corresponding time. The frame interval refers to the time difference between two adjacent frames, which can be obtained according to the inverse of the frame rate of the first video stream. The time required for each physical displacement is the interval between two image frames, that is, the frame interval. By dividing the physical displacement by the frame interval, the corresponding first speed is obtained.
[0109] S408: Obtain a predicted average speed based on each first speed.
[0110] It can be understood that this step can be to directly calculate the average value of each first speed. This method is simple and direct, but may be sensitive to noise. A weighted average method can also be used. The later the first speed is, the higher the weight is given, so that the latest state of the vehicle can be better captured. In addition, many tunnels are currently equipped with image acquisition devices. These image acquisition devices can capture the actual movement of the vehicle in the tunnel. The second video stream captured by these image acquisition devices can be used to correct the predicted average speed, thereby updating the predicted trajectory after entry.
[0111] Specifically, target recognition and tracking can be performed based on the second video stream to obtain the average speed of each vehicle in the shooting range, and each vehicle in the shooting range can be associated with an existing unique identifier. The method of obtaining the average speed of each vehicle based on the second video stream is similar to the method of obtaining the predicted average speed in the above text. Associating vehicles in the shooting range of the second video stream with existing unique identifiers is a cross-camera data association problem. This may involve matching multiple features, such as the appearance features, size, color, etc. of the vehicle. A deep learning model can be used to extract these features of each vehicle identified in the first video stream and associate them with the unique identifier. The same method is used to extract features for each vehicle identified in the second video stream. The features corresponding to each vehicle in the second video stream are matched with the features corresponding to each unique identifier, so that each vehicle identified in the second video stream is associated with an existing unique identifier, and then the predicted average speed can be corrected by using the recognized average speed, which can be directly overwritten. The predicted post-entry trajectory may be updated by retaining the predicted post-entry trajectory corresponding to the time period before the predicted average speed is updated, and re-predicting the predicted post-entry trajectory for the time period after the predicted average speed is updated using the new predicted average speed.
[0112] In one embodiment, the vehicle risk index includes vehicle fire load and / or vehicle heat release rate, and the tunnel risk index includes single point fire load and / or single point heat release rate. Vehicle fire load refers to the total heat that may be released by a complete combustion of a vehicle. This is related to factors such as the size, material composition, fuel type and load of the vehicle. After the vehicles are divided into ten categories, the vehicle fire load of each category of vehicles can be set with reference to relevant research, such as 1GJ for motorcycles, 2GJ for small cars, 5GJ for medium-sized cars, 9GJ for large cars, 14GJ for small buses, 20GJ for medium-sized buses, 35GJ for large buses, 20GJ for small trucks, 55GJ for medium-sized trucks and 140GJ for large trucks. The vehicle heat release rate reflects the rate at which the vehicle releases heat per unit time during the fire process. This rate can be the peak heat release rate (peakHeat Release Rate, pHRR) or the heat release rate per unit area (Heat Release Rate PerUnit Area, HRRPUA). Taking the peak heat release rate as an example, after dividing the vehicles into ten categories, the vehicle heat release rate of each category of vehicles can be set with reference to relevant research, such as 1MW for motorcycles, 2MW for small cars, 4MW for medium-sized cars, 8MW for large cars, 12MW for small buses, 18MW for medium-sized buses, 30MW for large buses, 16MW for small trucks, 50MW for medium-sized trucks and 120MW for large trucks. Among them, the tunnel risk index corresponding to the vehicle fire load is the single-point fire load, which reflects the total energy that may be released by the complete combustion of all vehicles within a range centered on a location point. The tunnel risk index corresponding to the vehicle heat release efficiency is the single-point heat release rate, which reflects the heat release rate or fire source power of all vehicles within a range centered on a location point during the fire process.
[0113] Converting the tunnel risk index into a fire risk distribution is a key step. This embodiment will consider the design options for the fire risk distribution from the perspective of spatial distribution. Specifically, a first coordinate system is formed with the distance from the tunnel entrance as the horizontal axis and the size of the tunnel risk index as the vertical axis. The tunnel risk index at the same time is plotted in the first coordinate system to obtain the first fire risk distribution at the corresponding time. That is, for tunnels at different times, the corresponding first fire risk distribution is generated respectively. The first fire risk distribution can intuitively display the high-risk areas and relatively safe areas in the tunnel. In a dynamic environment, this distribution will change over time, such as Figure 5 As shown in the figure, the first fire risk distribution of a tunnel at four moments, namely 60 seconds, 120 seconds, 240 seconds and 500 seconds, is shown respectively. By generating a series of distribution graphs at time points, the fire risk in the tunnel can be dynamically evaluated.
[0114] In one embodiment, the vehicle risk index includes the vehicle fire load, and the tunnel risk index includes the single-point fire load. This embodiment will consider the design selection of the fire risk distribution from the perspective of time distribution. Specifically, it is necessary to add all the single-point fire loads at the same time to obtain the total fire load at the corresponding time. A second coordinate system is formed with time as the horizontal axis and the size of the fire load as the vertical axis. The total fire load at multiple consecutive moments is plotted in the second coordinate system to obtain a second fire risk distribution.
[0115] It can be understood that the core of this method is to transform discrete tunnel risk indicators into continuous fire risk distribution. By adding up all single-point fire loads at the same moment, we get the total fire load of the entire tunnel at that moment. This process is actually a spatial integration, which aggregates the risks distributed at different locations in the tunnel into an overall indicator. Using time as the horizontal axis and the size of the fire load as the vertical axis to form a second coordinate system, this representation method allows us to observe the changes in the total fire risk over time, so that it can reflect the impact of factors such as changes in traffic flow and changes in vehicle composition on fire risk, such as Figure 6 As shown in the figure, the highest peak at around 540 seconds is because several large vehicles entered the tunnel, causing its total fire load to increase to the peak value.
[0116] In one embodiment, in the fire risk assessment, the risk can also be assessed from the perspective of personnel evacuation. Figure 7 , also including steps S702 to S708.
[0117] S702: According to the actual fire occurrence point, determine the available evacuation points of the tunnel and predict the available safe evacuation time.
[0118] It can be understood that the actual fire occurrence point refers to the location of the fire point obtained by the fire detection sensor in the tunnel. The available evacuation point refers to the exit or safe area that is still safely accessible during the development of the fire. This may include the main exit of the tunnel, the emergency exit, or the safe shelter. The evaluation of the available evacuation points can be simplified, for example, by screening based on the distance between the actual fire occurrence point and the evacuation point. If the distance is less than the distance threshold, it will be excluded from the range of available evacuation points, such as Figure 8 As shown in the figure, in a 1000-meter-long tunnel, an evacuation point is set every 250 meters, and both sides of the tunnel can also be used as evacuation points. The actual fire occurred at 500 meters, and the evacuation point at 500 meters cannot be used because it is too close to the actual fire. The available evacuation points are 250 meters, 500 meters, and both sides of the tunnel. The selection of available evacuation points can also be combined with the results of fire simulation to evaluate the spread of the fire in real time to evaluate the availability of the evacuation points.
[0119] Available Safe Egress Time (ASET) refers to the time from the occurrence of a fire to the time when the environment in the tunnel becomes too harsh to evacuate. In the field of fire simulation, there are many mature model programs that can simulate the available safe evacuation time, and the relevant parameters of the tunnel and the actual fire occurrence point can be directly input for simulation.
[0120] S704: For any position corresponding to a non-empty vehicle set, the nearest available evacuation point is used as a target evacuation point, and an estimated evacuation time corresponding to the position is determined according to the distance between the target evacuation point and the position.
[0121] It can be understood that after a fire occurs, there are people who need to be evacuated at locations where there are non-empty vehicle sets. Each location will select a suitable evacuation point according to the situation. In this embodiment, the target evacuation point is selected for each location that needs to be evacuated based on the principle of the shortest distance. The evacuation time required for the evacuation of people from these locations, that is, the expected evacuation time, consists of multiple parts, including fire detection time, pre-action time and personnel evacuation movement time. Among them, in order to simplify the analysis, the fire detection time and pre-action time can be set to a fixed value, and the variable in the expected evacuation time is the personnel evacuation movement time. The personnel evacuation movement time includes the personnel movement time. The personnel movement time will be determined by the ratio between the distance between the starting point and the end point (that is, the distance between the location to be evacuated and the corresponding target evacuation point) and the pedestrian speed. The pedestrian speed can be set to a uniform and fixed speed, such as 0.5 meters per second or 1 meter per second. It can be seen that according to the distance between the target evacuation point and the location and the preset pedestrian speed, the personnel movement time can be calculated, and then the personnel evacuation movement time can be obtained. By combining the personnel evacuation movement with the preset fire detection time and pre-action time, the estimated evacuation time corresponding to each location can be obtained.
[0122] In addition, in order to further refine the analysis of personnel evacuation movement, the time to pass the door and the time to enter the evacuation channel can be added to the personnel evacuation movement. The time to pass the door is the time required for the personnel inside the vehicle to leave the vehicle, and the time to enter the evacuation channel is the time required for the personnel to enter the target evacuation point. The unit time to pass the door can be set to 0.3 seconds per person, and the unit time to pass the evacuation channel can be set to 2.25 seconds per person. Combined with the total number of people corresponding to the vehicle set, each corresponding time to pass the door and the time to pass the evacuation channel can be calculated. Specifically, the number of people to be evacuated corresponding to the location can be determined based on the category of each vehicle in the corresponding vehicle set and the second corresponding relationship. The second corresponding relationship is the correspondence between the category and the passenger capacity. By multiplying the number of people to be evacuated with the preset unit time to pass the door and the unit time to pass the evacuation channel, the time to pass the door and the time to pass the evacuation channel of the location can be obtained.
[0123] S706: Determine the maximum value of each estimated evacuation time as the required safe evacuation time.
[0124] It is understandable that the overall security depends on the weakest link. Selecting the maximum value as the required safe evacuation time ensures that even the most disadvantaged people have enough time to evacuate.
[0125] S708: Determine the evacuation risk according to the required safe evacuation time and the available safe evacuation time.
[0126] It can be understood that when the required safe evacuation time is greater than the available safe evacuation time, it means that all personnel in the tunnel can be evacuated safely and there is no evacuation risk. Otherwise, it means that there may be evacuation risks for personnel in the tunnel.
[0127] In one embodiment, a tunnel fire risk assessment method is provided. Figure 8 , also includes steps S802 to S804.
[0128] S802: For any position corresponding to a non-empty vehicle set, determine the number of people to be evacuated corresponding to the position according to the category of each vehicle in the corresponding vehicle set and the second corresponding relationship. The second corresponding relationship is the corresponding relationship between the category and the passenger capacity.
[0129] It can be understood that this step is to convert vehicle information into number of people information, which is the basis for evaluating the evacuation situation. The vehicle set refers to the set of all vehicles existing in a specific location of the tunnel and its surrounding area. The second correspondence can be used to convert vehicle categories into estimated passenger capacity. For each vehicle included in a vehicle set, the corresponding passenger capacity is found respectively, and these passenger capacities are added together to obtain the number of people to be evacuated at the corresponding location.
[0130] S804, according to the number of people to be evacuated and the estimated evacuation time corresponding to each position in the tunnel, determine the remaining number of people to be evacuated at each time, and plot it in a third coordinate system with time as the horizontal axis and the remaining number of people to be evacuated as the vertical axis to obtain the evacuation situation distribution.
[0131] It can be understood that this step is a model of the evacuation situation that changes over time, and it is displayed in a graphical form. After the fire occurs, as the evacuation progresses, the number of people who need to be evacuated in the tunnel should gradually decrease. Therefore, taking the occurrence of the fire as the time origin, combined with the number of people who need to be evacuated and the estimated evacuation time corresponding to each position in the tunnel, if the current moment is greater than the estimated evacuation time corresponding to a position, the remaining number of people who need to be evacuated in the tunnel should be reduced by the number of people who need to be evacuated corresponding to that position. As the current moment gradually increases, the remaining number of people who need to be evacuated at each moment can be obtained. In order to visualize the situation, the evacuation situation distribution can be obtained in the third coordinate system with the evacuation time as the horizontal axis and the remaining number of people who need to be evacuated as the vertical axis, such as Fig. 9As shown, the corresponding evacuation situation distribution after a simulated fire occurs in a tunnel at different times is shown. Fig. 9 At the time of (a), there are 84 people in the tunnel, and the required safe evacuation time is 338.2 seconds. At the time of (b), there are 33 people in the tunnel, and the required safe evacuation time is 282.7 seconds. At the time of (c), there are 51 people in the tunnel, and the required safe evacuation time is 333.3 seconds. At the time of (d), there are 97 people in the tunnel, and the required safe evacuation time is 292 seconds.
[0132] The present application provides a tunnel fire risk assessment device, including a data acquisition module, an identification module, a vehicle distribution determination module, a tunnel risk index determination module and a fire risk distribution determination module. The data acquisition module is used to obtain a first video stream. The first video stream is a video stream shot at the tunnel entrance. The identification module is used to determine the category and trajectory of each vehicle in the tunnel based on the first video stream. The vehicle distribution determination module is used to obtain the tunnel vehicle distribution at different times according to each trajectory. The tunnel risk index determination module is used to determine the vehicle set corresponding to each position in the tunnel according to the tunnel vehicle distribution, and determine the tunnel risk index corresponding to each position in the tunnel at the corresponding time according to the category of each vehicle in the vehicle set and the first corresponding relationship. The first corresponding relationship is the corresponding relationship between the category and the vehicle risk index. The fire risk distribution determination module is used to obtain the fire risk distribution according to each tunnel risk index.
[0133] For the specific limitations of the tunnel fire risk assessment device, please refer to the limitations of the tunnel fire risk assessment method above, which will not be repeated here. The various modules in the above-mentioned tunnel fire risk assessment device can be implemented in whole or in part through software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation.
[0134] The present application provides a computer device, including one or more processors and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the following are performed: obtaining a first video stream; the first video stream is a video stream shot at a tunnel entrance; based on the first video stream, determining the category and trajectory of each vehicle in the tunnel; obtaining the tunnel vehicle distribution at different times according to each trajectory; determining a vehicle set corresponding to each position in the tunnel according to the tunnel vehicle distribution, and determining a tunnel risk index corresponding to each position in the tunnel at a corresponding time according to the category of each vehicle in the vehicle set and a first corresponding relationship; the first corresponding relationship is a corresponding relationship between a category and a vehicle risk index; and obtaining a fire risk distribution according to each tunnel risk index.
[0135] In one embodiment, when the computer-readable instructions are executed by one or more processors, the steps of the tunnel fire risk assessment method in any of the above embodiments are performed.
[0136] Indicatively, if Fig.10 As shown, Fig.10 A schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. Fig.10 The computer device 1000 includes a processing component 1002, which further includes one or more processors, and a memory resource represented by a memory 1001 for storing instructions executable by the processing component 1002, such as an application. The application stored in the memory 1001 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1002 is configured to execute instructions to perform the steps of the tunnel fire risk assessment method of any of the above embodiments.
[0137] The computer device 1000 may further include a power supply component 1003 configured to perform power management of the computer device 1000 , a wired or wireless network interface 1004 configured to connect the computer device 1000 to a network, and an input / output (I / O) interface 1005 .
[0138] The present application provides a storage medium, in which computer-readable instructions are stored. When the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute: obtaining a first video stream; the first video stream is a video stream shot at a tunnel entrance; based on the first video stream, determining the category and trajectory of each vehicle in the tunnel; obtaining the tunnel vehicle distribution at different times according to each trajectory; determining a vehicle set corresponding to each position in the tunnel according to the tunnel vehicle distribution, and determining a tunnel risk index corresponding to each position in the tunnel at a corresponding time according to the category of each vehicle in the vehicle set and a first corresponding relationship; the first corresponding relationship is a corresponding relationship between a category and a vehicle risk index; and obtaining a fire risk distribution according to each tunnel risk index.
[0139] In one embodiment, when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the tunnel fire risk assessment method in any of the above embodiments.
[0140] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article 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, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0141] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.
[0142] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A tunnel fire risk assessment method, characterized in that: include: Get the first video stream; The first video stream is a video stream shot at the tunnel entrance; Based on the first video stream, determining the type and trajectory of each vehicle in the tunnel; According to each of the trajectories, the distribution of tunnel vehicles at different times is obtained; Determine a vehicle set corresponding to each position in the tunnel according to the tunnel vehicle distribution, and determine a tunnel risk index corresponding to each position in the tunnel at a corresponding time according to the category of each vehicle in the vehicle set and a first corresponding relationship; the first corresponding relationship is a corresponding relationship between the category and the vehicle risk index; According to each of the tunnel risk indicators, a fire risk distribution is obtained.
2. The tunnel fire risk assessment method according to claim 1, characterized in that: The determining, based on the first video stream, the category and trajectory of each vehicle in the tunnel comprises: Performing target recognition and tracking based on the first video stream, thereby assigning a unique identifier to each vehicle entering the tunnel, and obtaining a category and a trajectory before entering the tunnel corresponding to the unique identifier; According to the pre-entry trajectory, a predicted average speed is obtained; According to the predicted average speed, a predicted post-entry trajectory is obtained; The predicted post-entry trajectory is associated with the unique identifier of the corresponding pre-entry trajectory to obtain the category and trajectory of the vehicle in the tunnel.
3. The tunnel fire risk assessment method according to claim 2, characterized in that: The target recognition and tracking is performed based on the first video stream, thereby assigning a unique identifier to each vehicle entering the tunnel, and obtaining a category and a trajectory before entering the tunnel corresponding to the unique identifier, including: Inputting the current frame in the first video stream into the target detection model to obtain a recognition result corresponding to the current frame; the recognition result includes the categories and bounding box information of multiple vehicles to be matched; For any unfinished pre-entry trajectory, generating a predicted position corresponding to the current frame, and matching the predicted position with the bounding box information of each vehicle to be matched; If the bounding box information of the to-be-matched vehicle successfully matches the predicted position of a pre-entry trajectory, updating the pre-entry trajectory according to the bounding box information of the to-be-matched vehicle, and associating the unique identifier of the pre-entry trajectory with the to-be-matched vehicle; If the bounding box information of the to-be-matched vehicle does not have a successfully matched pre-entry track, a new track is created for the to-be-matched vehicle, and a new unique identifier is assigned.
4. The tunnel fire risk assessment method according to claim 2, characterized in that: Before performing target recognition and tracking based on the first video stream, the method includes: Performing a perspective transformation on each video frame in the first video stream to convert the perspective of each video frame into a top-down perspective; The step of obtaining a predicted average speed according to the pre-entry trajectory includes: Determine the pixel displacement between each two adjacent frames according to the pre-entry trajectory; Convert each of the pixel displacements into a physical displacement; According to each of the physical displacements and frame intervals, a plurality of first speeds are obtained; The predicted average speed is obtained according to each of the first speeds.
5. The tunnel fire risk assessment method according to claim 2, characterized in that: Also includes: Based on the second video stream, determine the identification track of each vehicle within the shooting range by using a cross-camera identification method, and associate the identification track with the existing unique identifier; The second video stream is a video stream shot at one or more different locations in the tunnel; The corresponding predicted post-entry trajectory is corrected using the corresponding identified trajectory.
6. The tunnel fire risk assessment method according to claim 1, characterized in that: The vehicle risk index includes vehicle fire load and / or vehicle heat release rate, and the tunnel risk index includes single-point fire load and / or single-point heat release rate; The fire risk distribution is obtained according to each of the tunnel risk indicators, including: A first coordinate system is formed with the distance from the tunnel entrance as the horizontal axis and the size of the tunnel risk index as the vertical axis; The tunnel risk index at the same time is plotted in the first coordinate system to obtain a first fire risk distribution at the corresponding time.
7. The tunnel fire risk assessment method according to claim 1, characterized in that: The vehicle risk index includes vehicle fire load, and the tunnel risk index includes single point fire load; The fire risk distribution is obtained according to each of the tunnel risk indicators, including: Adding all the single-point fire loads at the same time to obtain the total fire load at the corresponding time; The second coordinate system is formed with time as the horizontal axis and the size of the fire load as the vertical axis; The total fire load at a plurality of consecutive moments is plotted in the second coordinate system to obtain a second fire risk distribution.
8. The tunnel fire risk assessment method according to claim 1, characterized in that: Also includes: Determine the available evacuation points of the tunnel based on the actual fire occurrence point; For any non-empty position corresponding to the vehicle set, the closest available evacuation point is used as the target evacuation point, and the estimated evacuation time corresponding to the position is determined according to the distance between the target evacuation point and the position; Determining the maximum value among the estimated evacuation times as the required safe evacuation time; The evacuation risk is determined according to the required safe evacuation time and the preset available safe evacuation time.
9. The tunnel fire risk assessment method according to claim 8, characterized in that: Also includes: For any non-empty position corresponding to the vehicle set, the number of people to be evacuated corresponding to the position is determined according to the category of each vehicle in the corresponding vehicle set and the second corresponding relationship; the second corresponding relationship is the corresponding relationship between the category and the passenger capacity; According to the number of people to be evacuated and the estimated evacuation time corresponding to each position in the tunnel, the remaining number of people to be evacuated at each moment is determined and plotted in a third coordinate system with time as the horizontal axis and the remaining number of people to be evacuated as the vertical axis to obtain the evacuation situation distribution.
10. A tunnel fire risk assessment device, characterized in that: include: A data acquisition module, used for acquiring a first video stream; The first video stream is a video stream shot at the tunnel entrance; an identification module, configured to determine the type and trajectory of each vehicle in the tunnel based on the first video stream; A vehicle distribution determination module, used to obtain the tunnel vehicle distribution at different times according to each of the trajectories; a tunnel risk index determination module, configured to determine a vehicle set corresponding to each position in the tunnel according to the tunnel vehicle distribution, and determine a tunnel risk index corresponding to each position in the tunnel at a corresponding time according to the category of each vehicle in the vehicle set and a first corresponding relationship; the first corresponding relationship being a corresponding relationship between the category and the vehicle risk index; The fire risk distribution determination module is used to obtain the fire risk distribution according to each of the tunnel risk indicators.
11. A computer device, characterized in that: The method comprises one or more processors and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the tunnel fire risk assessment method according to any one of claims 1 to 9 are executed.
12. A storage medium, characterized in that: The storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the tunnel fire risk assessment method according to any one of claims 1 to 9.
