Traffic accident warning system and method based on intelligent warning signs

Abstract

This application relates to the field of traffic accident warnings. Specifically, it discloses a traffic accident warning system and method based on intelligent warning signs. First, it obtains the road traffic images collected by cameras installed in areas prone to road traffic accidents and the vehicle passing speed values within a predetermined time period collected by sensors. Then, using deep learning technology, it performs feature extraction and correlation analysis on the two. Finally, through a generator, it generates traffic accident warning sign warning prompts to more accurately identify potential traffic accident risk areas and help reduce the occurrence of secondary traffic accidents.

Description

Technical Field

[0001] This application relates to the field of traffic accident warning, and more specifically, to a traffic accident warning system and method based on an intelligent warning sign. Background Art

[0002] A secondary traffic accident refers to a subsequent accident caused by the negligence of the parties or rescue personnel on the basis of an original traffic accident, which may be triggered due to reasons such as failure to warn in time, improper warning measures or poor effects. Compared with the first accident, secondary accidents usually cause more serious casualties and property losses, increasing the difficulty of rescue work. Therefore, it is urgent to take measures to prevent such accidents.

[0003] Currently, the warning measures for traffic accidents mainly rely on the parties placing on-vehicle folding warning signs for prompt. However, due to possible casualties or panic of the parties after the accident, it is impossible to give early warnings in time, and the on-vehicle folding warning signs are small in area, short in placement distance and low in height, resulting in poor warning effects.

[0004] Therefore, there is a need for a traffic accident warning system and method based on an intelligent warning sign. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide a traffic accident warning system and method based on an intelligent warning sign, which first obtain the road traffic images collected by cameras installed in areas prone to road traffic accidents and the vehicle passing speed values collected by sensors during a predetermined period of time, then use deep learning technology to perform feature extraction and correlation analysis on the two, and finally use a generator to generate traffic accident warning sign warning prompts to more accurately identify potential traffic accident risk areas and help reduce the occurrence of secondary traffic accidents.

[0006] According to one aspect of this application, there is provided a traffic accident warning system based on an intelligent warning sign, which includes:

[0007] A traffic accident warning data acquisition module for obtaining the road traffic images collected by cameras installed in areas prone to road traffic accidents and the vehicle passing speed values collected by sensors during a predetermined period of time;

[0008] A traffic accident warning data extraction module for extracting an optimized road traffic correlation feature vector and a vehicle passing speed feature vector from the road traffic images collected by cameras installed in areas prone to road traffic accidents and the vehicle passing speed values collected by sensors during a predetermined period of time;

[0009] A warning sign warning prompt generation module, configured to generate a traffic accident warning sign warning prompt based on the optimized road traffic association feature vector and the vehicle passing speed feature vector.

[0010] According to another aspect of the present application, there is provided a traffic accident warning method based on an intelligent warning sign, which includes:

[0011] Obtain the road traffic images collected by cameras installed in areas prone to road traffic accidents and the vehicle passing speed values within a predetermined period collected by sensors;

[0012] Extract an optimized road traffic association feature vector and a vehicle passing speed feature vector from the road traffic images collected by cameras installed in areas prone to road traffic accidents and the vehicle passing speed values within a predetermined period collected by sensors;

[0013] Generate a traffic accident warning sign warning prompt based on the optimized road traffic association feature vector and the vehicle passing speed feature vector.

[0014] Compared with the prior art, a traffic accident warning system and method based on an intelligent warning sign provided by the present application first obtain the road traffic images collected by cameras installed in areas prone to road traffic accidents and the vehicle passing speed values within a predetermined period collected by sensors, then use deep learning technology to perform feature extraction and correlation analysis on the two, and finally use a generator to generate a traffic accident warning sign warning prompt to more accurately identify potential traffic accident risk areas and help reduce the occurrence of secondary traffic accidents. Description of the Drawings

[0015] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 It is a block diagram of a traffic accident warning system based on an intelligent warning sign according to an embodiment of the present application.

[0017] Figure 2 It is a block diagram of a traffic accident warning data extraction module in a traffic accident warning system based on an intelligent warning sign according to an embodiment of the present application.

[0018] Figure 3 It is a block diagram of a road traffic feature aggregation unit in a traffic accident warning system based on an intelligent warning sign according to an embodiment of the present application.

[0019] Figure 4 It is a flowchart of a traffic accident warning method based on an intelligent warning sign according to an embodiment of the present application.

[0020] Figure 5 It is a block diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0021] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0022] Figure 1 It is a block diagram of a traffic accident warning system based on an intelligent warning sign according to an embodiment of the present application. As Figure 1 shown, the traffic accident warning system 100 based on an intelligent warning sign according to an embodiment of the present application includes: a traffic accident warning data acquisition module 110, configured to acquire road traffic images collected by cameras installed in areas prone to road traffic accidents and vehicle passing speed values for a predetermined period collected by sensors; a traffic accident warning data extraction module 120, configured to extract an optimized road traffic correlation feature vector and a vehicle passing speed feature vector from the road traffic images collected by the cameras installed in areas prone to road traffic accidents and the vehicle passing speed values for a predetermined period collected by the sensors; a warning sign warning prompt generation module 130, configured to generate a traffic accident warning sign warning prompt based on the optimized road traffic correlation feature vector and the vehicle passing speed feature vector.

[0023] In the above-mentioned traffic accident warning system 100 based on intelligent warning signs, the traffic accident warning data acquisition module 110 is used to acquire the road traffic images collected by cameras installed in areas prone to road traffic accidents and the vehicle passing speed values collected by sensors within a predetermined time period. It should be understood that, compared with the first accident, secondary accidents often cause greater casualties and property losses, bringing greater challenges to rescue work. Therefore, there is an urgent need to take measures to prevent the occurrence of such accidents. Currently, the warning measures during traffic accidents mainly rely on the parties to place on-vehicle folding warning signs to alert other drivers. However, since the parties may be injured or in a panicked state after the accident, they are unable to place these warning signs in a timely manner. In addition, the area of on-vehicle folding warning signs is small, the placement distance is short, and the height is low, resulting in poor warning effects. In the technical solution of this application, by acquiring the road traffic images collected by cameras installed in areas prone to road traffic accidents and the vehicle passing speed values collected by sensors within a predetermined time period, and combining deep learning technology, a warning prompt for traffic accident warning signs is generated to provide more timely, obvious, and effective warning information, helping to reduce the incidence of secondary accidents and ensuring the safety and smoothness of road traffic.

[0024] Specifically, the cameras in areas prone to road traffic accidents capture real-time road conditions, providing valuable visual data, including vehicle driving trajectories, road conditions, etc. At the same time, the vehicle passing speed values collected by sensors provide specific data on vehicle operating speeds, which can be used to evaluate important information such as traffic flow density and vehicle operating status. By comprehensively utilizing the images collected by cameras and the speed data collected by sensors, the system can comprehensively understand the road traffic situation, including important information such as vehicle density, speed distribution, and traffic flow. This comprehensive data acquisition method provides the system with multi-dimensional and multi-angle traffic information to more accurately identify potential traffic risks, issue warning prompts in a timely manner, improve traffic safety, reduce the likelihood of traffic accidents, and ensure the safety of road users.

[0025] In the above-mentioned traffic accident warning system 100 based on intelligent warning signs, the traffic accident warning data extraction module 120 is used to extract an optimized road traffic correlation feature vector and a vehicle passing speed feature vector from the road traffic images collected by cameras installed in areas prone to road traffic accidents and the vehicle passing speed values collected by sensors within a predetermined time period. It should be understood that extracting an optimized road traffic correlation feature vector and a vehicle passing speed feature vector from camera and sensor data provides key data support for the intelligent warning sign system, enabling it to accurately and timely generate traffic accident warning prompts, thereby improving traffic safety and efficiency.

[0026] Figure 2The block diagram of the traffic accident warning data extraction module in the traffic accident warning system based on an intelligent warning sign according to an embodiment of the present application. As Figure 2 shown, in a specific embodiment of the present application, the traffic accident warning data extraction module 120 includes: a road traffic attention feature encoding unit 121, configured to perform attention feature encoding on the road traffic images collected by the cameras installed in the areas prone to road traffic accidents to obtain a vehicle trajectory tracking enhanced feature map; a road traffic local feature extraction unit 122, configured to perform local feature extraction on the road traffic images collected by the cameras installed in the areas prone to road traffic accidents to obtain a road traffic Canny edge feature map; a road traffic feature aggregation unit 123, configured to perform feature aggregation on the vehicle trajectory tracking enhanced feature map and the road traffic Canny edge feature map to obtain the optimized road traffic association feature vector; a vehicle passing speed feature extraction unit 124, configured to perform feature extraction on the vehicle passing speed values in a predetermined time period collected by the sensor to obtain the vehicle passing speed feature vector.

[0027] It should be understood that attention feature encoding can help the system identify key information and important areas in road traffic images, such as accident-prone locations like intersections and crosswalks. By paying special attention to these areas, the system can better capture the vehicle movement trajectory and accurately track the driving path of the vehicle. This attention mechanism can improve the system's attention to important areas, thereby enhancing the effect of vehicle trajectory tracking. Among them, attention feature encoding can also help the system filter and reduce noise and interference information in the image, and concentrate on processing key areas and vehicles. By performing attention weighting on road traffic images, the system can effectively extract features related to vehicle trajectories, reduce unnecessary interference, and improve the accuracy and stability of vehicle trajectory tracking.

[0028] Furthermore, in areas prone to road traffic accidents, there are often complex traffic conditions and road structures. Local feature extraction can help the system better understand these local details, so as to accurately capture important information such as the position and direction of vehicles. Here, Canny edge detection is a classic edge detection algorithm that can effectively identify edge information in images. By extracting the Canny edge feature map, the system can more accurately capture the edge information in road traffic images, and then achieve more precise vehicle detection and trajectory analysis. Among them, local feature extraction can help the system focus on local areas in road traffic images, improve the attention to specific areas, and clearly express the edge structure in road traffic images, which is helpful for accurately positioning vehicle edges and road lines, thereby achieving precise vehicle detection and trajectory analysis.

[0029] Furthermore, by aggregating the enhanced vehicle trajectory tracking feature map and the road traffic Canny edge feature map, the system can comprehensively utilize vehicle trajectory information and road edge features to achieve more comprehensive and accurate road traffic analysis and accident warning. Here, feature aggregation can comprehensively utilize different types of information to enrich the feature representation. Among them, the enhanced vehicle trajectory tracking feature map contains key information such as vehicle movement trajectories, while the road traffic Canny edge feature map reflects road structure and edge information. By aggregating these two features, the system can obtain a richer and more comprehensive road traffic feature representation and improve the comprehensive utilization efficiency of traffic data.

[0030] Specifically, by extracting features from the vehicle passing speed values collected by the sensor within a predetermined time period, the system can obtain the speed information of the vehicle at different time points, including data such as average speed, maximum speed, and minimum speed, so as to comprehensively understand the driving state and speed distribution of the vehicle.

[0031] In a specific embodiment of the present application, the road traffic attention feature encoding unit 121 includes: passing the road traffic image collected by the camera installed in the accident-prone area of the road through the vehicle trajectory tracking pixel enhancer to obtain the vehicle trajectory tracking pixel enhanced image; passing the vehicle trajectory tracking pixel enhanced image through the vehicle trajectory tracking enhanced feature encoder based on the spatial attention mechanism to obtain the vehicle trajectory tracking enhanced feature map.

[0032] It should be understood that the vehicle trajectory tracking pixel enhancer can help the system accurately capture the vehicle movement trajectory in the road traffic image. In the accident-prone area of the road, the accurate tracking of the vehicle movement trajectory is of great significance for accident prevention. In complex traffic scenarios, the road traffic image may have problems such as blurring and noise, which affect the accurate identification of the vehicle trajectory by the system. Through the pixel enhancer, the system can enhance the key parts of the image, improve the clarity of the image, reduce interference, and provide more reliable data support for vehicle trajectory tracking. Specifically, the road traffic image collected by the camera installed in the accident-prone area of the road is input into the high-definition image generator based on the generative adversarial network, and the vehicle trajectory tracking pixel enhanced image is generated by the vehicle trajectory tracking pixel enhancer based on the generative adversarial network through deconvolution coding.

[0033] Furthermore, processing the vehicle trajectory tracking pixel-enhanced image through a vehicle trajectory tracking enhanced feature encoder based on a spatial attention mechanism can improve the accuracy and efficiency of vehicle trajectory tracking. Here, the vehicle trajectory tracking enhanced feature encoder based on a spatial attention mechanism can help the system better understand the movement trajectory of the vehicle on the road. By performing feature encoding on the vehicle trajectory tracking pixel-enhanced image, the system can focus on key areas and the vehicle movement path, thereby improving the accuracy and stability of vehicle trajectory tracking. Specifically, use the convolutional encoding part of the vehicle trajectory tracking enhanced feature encoder based on a spatial attention mechanism to perform deep convolutional encoding on the vehicle trajectory tracking pixel-enhanced image to obtain an initial convolutional feature map; input the initial convolutional feature map into the spatial attention part of the vehicle trajectory tracking enhanced feature encoder based on a spatial attention mechanism to obtain a spatial attention map; pass the spatial attention map through the Softmax activation function to obtain a spatial attention feature map; calculate the element-wise multiplication of the spatial attention feature map and the initial convolutional feature map to obtain the vehicle trajectory tracking enhanced feature map.

[0034] In a specific embodiment of the present application, the road traffic local feature extraction unit 122 includes: constructing the road traffic images collected by the cameras set in the areas prone to road traffic accidents into a plurality of road traffic local images; passing the plurality of road traffic local images through a road traffic local feature extractor based on Canny edge detection to obtain the road traffic Canny edge feature map.

[0035] It should be understood that considering that in complex traffic scenarios, a single road traffic image may contain a large amount of information, resulting in difficult system processing and high computational complexity. After decomposing the image into multiple local images, the system can process each local image in parallel, improving the processing efficiency, reducing errors and confusion at the same time, improving the accuracy of accident detection, and thus realizing more refined feature extraction and analysis. Among them, each local image can be regarded as an independent data sample. The system can extract features for each local image and combine global information for comprehensive analysis, so as to more comprehensively understand the road traffic situation and provide a more accurate basis for accident detection and warning. Constructing the road traffic image into a plurality of road traffic local images helps the system to observe the traffic situation on the road more carefully. By decomposing the overall road traffic image into multiple local images, the system can focus on the details of each local area, capture important information such as vehicle driving trajectories and traffic signal states, and thus improve the system's perception ability of the road traffic situation.

[0036] Furthermore, Canny edge detection is a classic edge detection algorithm that can effectively identify edge information in images. By inputting local road traffic images into a feature extractor based on Canny edge detection, the system can quickly and accurately extract edge features in the images, including important information such as road lines and vehicle contours. The obtained Canny edge feature map of road traffic can help the system better distinguish different objects and structures in road traffic images. By extracting the Canny edge features of local road traffic images, the system can convert the images into a more representative and easily understandable feature representation form, thus better identifying road elements such as road signs, vehicles, and pedestrians.

[0037] Figure 3 It is a block diagram of a road traffic feature aggregation unit in a traffic accident warning system based on an intelligent warning sign according to an embodiment of the present application. As Figure 3 shown, in a specific embodiment of the present application, the road traffic feature aggregation unit 123 includes: a multi-channel road feature association sub-unit 1231 for associating the enhanced vehicle trajectory tracking feature map and the Canny edge feature map of road traffic to obtain a multi-channel road traffic input map; a road traffic association feature encoding sub-unit 1232 for passing the multi-channel road traffic input map through a road traffic convolutional neural network serving as a feature encoder to obtain a road traffic association feature vector; and a road traffic feature optimization sub-unit 1233 for performing depth-wise point-wise divergence domain feature compensation on the road traffic association feature vector based on the vehicle passing speed feature vector to obtain the optimized road traffic association feature vector.

[0038] It should be understood that by associating the enhanced vehicle trajectory tracking feature map and the Canny edge feature map of road traffic, the system can comprehensively consider vehicle movement trajectories and road structure features to more comprehensively describe road traffic conditions. Among them, the enhanced vehicle trajectory tracking feature map provides key information about the vehicle movement path, while the Canny edge feature map of road traffic reflects the road structure features, including road edges and traffic signs, etc. Associating these two feature maps can enable the system to consider more factors when analyzing traffic conditions and improve the accurate grasp of the road traffic state. The generation of the multi-channel road traffic input map can enhance the system's representation ability of road traffic images. This multi-channel input map combines information from different aspects, helps the system better understand road traffic images, and improves the perception ability of traffic conditions.

[0039] Furthermore, processing the multi-channel road traffic input image through a road traffic convolutional neural network acting as a feature encoder can extract key features in the road traffic image, further enhancing the system's understanding and analysis capabilities of traffic conditions. Here, the road traffic convolutional neural network acting as a feature encoder can effectively learn the abstract feature representation in the road traffic image. Through multi-layer convolutional and pooling operations, the neural network can gradually extract low-level features (such as edges and textures) and high-level semantic features (such as vehicles and pedestrians) in the image, thus realizing the feature extraction and representation of the road traffic image. Among them, the road traffic convolutional neural network can capture the spatial information and context associations in the road traffic image. The neural network will consider the spatial position relationship between pixels during the process of learning feature representation, so as to better understand the importance and relevance of different regions in the road traffic image, which helps to improve the accurate analysis of vehicle trajectories and traffic conditions. The road traffic correlation feature vector extracted by the road traffic convolutional neural network can provide a more compact and efficient representation for the system. These feature vectors contain the key information in the road traffic image, helping the system to better understand the road traffic situation and realize the timely warning and handling of traffic accidents. Specifically, each layer of the road traffic convolutional neural network acting as a feature encoder performs convolution processing, mean pooling processing based on the local feature matrix, and non-linear activation processing on the input data respectively during the forward pass of the layer to output the road traffic correlation feature vector by the last layer of the road traffic convolutional neural network acting as a feature encoder, where the input of the road traffic convolutional neural network acting as a feature encoder is the multi-channel road traffic input image.

[0040] Even further, the process of performing depth-wise point divergence domain feature compensation on the road traffic correlation feature vector based on the vehicle passing speed feature vector is crucial for improving the performance of the traffic accident warning system. Through this step, the road traffic correlation feature vector can be better optimized, enhancing the system's understanding and prediction capabilities of traffic conditions. Here, the vehicle passing speed feature vector provides important information about the vehicle's motion state. By considering the vehicle's speed information, the system can more accurately infer the vehicle's driving trajectory, lane-changing behavior, and possible dangerous situations, thus improving the system's prediction ability for the occurrence of traffic accidents. Among them, the depth-wise point divergence domain feature compensation can help the system better understand and utilize the road traffic correlation feature vector. By performing feature compensation in the depth-wise point divergence domain, the system can better capture the complex relationships and dependencies between features, improve the feature representation ability, and thus enhance the system's modeling ability of traffic conditions.

[0041] In a specific embodiment of the present application, the road traffic feature optimization subunit 1233 includes: calculating the position-wise subtraction of the road traffic correlation feature vector and the vehicle passing speed feature vector to obtain a difference feature vector; dividing the feature values at each position of the difference feature vector by two and then taking the absolute value, and then calculating the natural exponential function value to obtain an exponentialized feature vector; dividing a second hyperparameter by the feature values at each position of the exponentialized feature vector to obtain a second weighted coefficient vector; performing position-wise multiplication of the second weighted coefficient vector and the vehicle passing speed feature vector to obtain a weighted vehicle passing speed feature vector; performing position-wise multiplication of the first predetermined hyperparameter and the road traffic correlation feature vector to obtain a weighted road traffic correlation feature vector, and then performing position-wise addition of the weighted road traffic correlation feature vector and the weighted vehicle passing speed feature vector to obtain the optimized road traffic correlation feature vector.

[0042] Specifically, in the technical solution of the present application, it is considered that both the road traffic correlation feature vector and the vehicle passing speed feature vector may contain information about vehicles and road traffic. Since these two feature vectors are extracted from different data sources, they may contain some identical information, resulting in information duplication. Moreover, the road traffic correlation feature vector and the vehicle passing speed feature vector may contain some related features. For example, the vehicle passing speed may be associated with the position and motion state of the vehicle on the road, and this information may also be reflected in the road traffic correlation feature vector, resulting in feature redundancy. Repeated or redundant features may reduce the performance of the model. These parts may introduce noise or interference, making it difficult for the model to accurately learn the true patterns of the data. Eliminating these parts can improve the generalization ability and accuracy of the model. Excessive repetition or redundant information may cause the model to overfit the training data. Overfitting makes the model perform well on the training data but poorly on new data. Eliminating redundant parts helps to avoid overfitting of the model. Eliminating repeated or redundant information can make the features more interpretable. If the feature vector contains a large amount of similar information, it may be difficult for the model to explain its decision-making process. Clear features can help to explain the behavior of the model. To solve this problem, in the technical solution of the present application, based on the vehicle passing speed feature vector, deep point-wise divergence domain feature compensation is performed on the road traffic correlation feature vector to obtain the optimized road traffic correlation feature vector.

[0043] Specifically, performing deep point-wise divergence domain feature compensation on the road traffic correlation feature vector based on the vehicle passing speed feature vector to obtain the optimized road traffic correlation feature vector includes: based on the vehicle passing speed feature vector, performing deep point-wise divergence domain feature compensation on the road traffic correlation feature vector according to the following formula to obtain the optimized road traffic correlation feature vector; ; where Denote the eigenvalue at the -th position of the road traffic associated feature vector, denote the eigenvalue at the -th position of the vehicle passing speed feature vector, and denote a first predetermined hyperparameter and a second predetermined hyperparameter, denote multiplication by position, denote addition by position, denote the eigenvalue at the -th position of the optimized road traffic associated feature vector.

[0044] In order to eliminate the repetitive or redundant parts in the road traffic associated feature vector and the vehicle passing speed feature vector, so as to improve the generalization ability of the road traffic associated feature vector, in the technical solution of this application, based on the vehicle passing speed feature vector, a depth-by-depth divergence domain feature compensation is performed on the road traffic associated feature vector. It simulates the depth of the position scattering response of the road traffic associated feature vector and the vehicle passing speed feature vector in the class divergence space by calculating the gradient of the variance between the positions of the road traffic associated feature vector and the vehicle passing speed feature vector, and uses the gradient of the variance to construct a depth correlation compensation factor to perform feature depth correlation compensation on a position-by-position basis for the road traffic associated feature vector. In this way, the repetitive or redundant parts in the road traffic associated feature vector and the vehicle passing speed feature vector are suppressed to improve the generalization ability of the road traffic associated feature vector.

[0045] In a specific embodiment of this application, the vehicle passing speed feature extraction unit 124 includes: constructing a vehicle passing speed input vector from the vehicle passing speed values collected by the sensor within a predetermined time period; passing the vehicle passing speed input vector through a vehicle passing speed time series encoder to obtain the vehicle passing speed feature vector.

[0046] It should be understood that constructing a vehicle passing speed input vector can integrate the vehicle speed information within a predetermined time period into a single vector, facilitating unified processing and analysis by the system. By combining the speed values of different vehicles into a vector form, the system can better capture the correlation between vehicles and the overall traffic flow situation, helping the system to more accurately judge the traffic state and predict potential traffic hazards. In addition, the construction of the vehicle passing speed input vector also provides the system with richer feature information. In addition to the vehicle speed itself, the vehicle passing speed vector can also contain other relevant information, such as vehicle type, lane position, etc. These information can provide a more comprehensive description of the traffic situation for the system, helping the system to more accurately identify traffic anomalies and perform early warning processing.

[0047] Furthermore, the vehicle can encode and model the time series information in the vehicle passing speed input vector through the speed time series encoder. By considering the temporal variation of the vehicle speed values, the system can capture the motion states and behavioral characteristics of the vehicle at different time points, such as acceleration, deceleration, parking, etc., so as to more comprehensively describe the vehicle's motion trajectory and driving pattern. The vehicle passing speed feature vector obtained through the vehicle passing speed time series encoder can extract higher-level feature representations. These feature vectors can contain richer and more abstract information, such as the trend of vehicle driving, the pattern of speed change, the interaction relationship between vehicles, etc. These information are crucial for the system to analyze and predict traffic conditions. Specifically, use the fully connected layer of the speed time series encoder to perform fully connected encoding on the vehicle passing speed input vector to extract the high-dimensional implicit features of the feature values at each position in the vehicle passing speed input vector; and, use the one-dimensional convolutional layer of the speed time series encoder to perform one-dimensional encoding on the vehicle passing speed input vector to extract the high-dimensional implicit correlation features of the correlation between the feature values at each position in the vehicle passing speed input vector.

[0048] In the above traffic accident warning system 100 based on intelligent warning signs, the warning sign warning prompt generation module 130 is used to generate traffic accident warning sign warning prompts based on the optimized road traffic correlation feature vector and the vehicle passing speed feature vector. It should be understood that the optimized road traffic correlation feature vector contains the overall situation and feature information of road traffic, such as traffic flow, vehicle density, road topology, etc. By analyzing and optimizing these feature vectors, the system can more accurately grasp the overall situation of road traffic and identify potential traffic bottlenecks, congestion and other potential risk factors. The vehicle passing speed feature vector provides important information about the individual behavior of the vehicle, such as vehicle speed, acceleration, driving trajectory, etc. By combining the vehicle passing speed feature vector and the road traffic correlation feature vector, the system can more comprehensively analyze the vehicle's motion pattern and traffic state, and identify potential dangerous behaviors or traffic accident risks. By comprehensively analyzing the optimized road traffic correlation feature vector and the vehicle passing speed feature vector, the system can generate traffic accident warning sign warning prompts. These prompts can include information such as the potential occurrence location of traffic accidents, possible dangerous behaviors, and recommended traffic improvement measures, to help drivers and traffic management departments take timely measures to avoid the occurrence of traffic accidents or reduce accident losses.

[0049] In a specific embodiment of the present application, the warning sign warning prompt generation module 130 includes: fusing the optimized road traffic correlation feature vector and the vehicle passing speed feature vector to obtain a traffic warning sign warning generation feature vector; passing the traffic warning sign warning generation feature vector through a generator to generate a traffic accident warning sign warning prompt.

[0050] It should be understood that the optimized road traffic correlation feature vector includes information such as road conditions, traffic flow, and traffic signals, reflecting the overall situation of the road; while the vehicle passing speed feature vector records the real-time running state of the vehicle on the road, such as speed, acceleration, etc. Fusing these two feature vectors can comprehensively understand the traffic environment and vehicle driving conditions, so as to accurately generate a traffic warning sign warning feature vector for the reference of drivers and traffic management departments, helping them take timely measures to avoid traffic accidents.

[0051] Furthermore, by analyzing the feature vector generated by fusing the optimized road traffic correlation feature vector and the vehicle passing speed feature vector, the generator can identify potential traffic risks and dangerous situations, and then generate corresponding traffic accident warning sign warning prompts. The generator can predict the possible types and locations of traffic accidents based on the data information in the feature vector, such as traffic flow, vehicle speed, road conditions, etc., and issue warnings to drivers in a timely manner. This intelligent generator can help drivers understand the road conditions in a timely manner, improve the driver's perception ability of potential dangers, thereby reducing the occurrence of traffic accidents, driving the vehicle more alertly, avoiding dangerous driving behaviors, effectively reducing the incidence of traffic accidents, and ensuring the safety and smoothness of road traffic. This intelligent traffic warning system can not only improve the efficiency of traffic management, but also provide a safer traffic environment for drivers and pedestrians, promoting the sustainable development of the traffic system.

[0052] In summary, the embodiment of the present application first obtains the road traffic images collected by the cameras installed in the areas prone to road traffic accidents and the vehicle passing speed values collected by the sensors during a predetermined period of time, then uses deep learning technology to perform feature extraction and correlation analysis on the two, and finally generates a traffic accident warning sign warning prompt through a generator to more accurately identify potential traffic accident risk areas and help reduce the occurrence of secondary traffic accidents.

[0053] As described above, the traffic accident warning system 100 based on intelligent warning signs according to the embodiments of the present application can be implemented in various terminal devices. In one example, the traffic accident warning system 100 based on intelligent warning signs can be integrated into the terminal device as a software module and / or a hardware module. For example, the traffic accident warning system 100 based on intelligent warning signs can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the traffic accident warning system 100 based on intelligent warning signs can also be one of the many hardware modules of the terminal device.

[0054] Alternatively, in another example, the traffic accident warning system 100 based on intelligent warning signs and the terminal device can also be separate devices, and the traffic accident warning system 100 based on intelligent warning signs can be connected to the terminal device through a wired and / or wireless network, and transmit and interact information in accordance with a predefined data format.

[0055] Figure 4 FIG. is a flowchart of a traffic accident warning method based on intelligent warning signs according to the embodiments of the present application. As Figure 4 shown, the traffic accident warning method based on intelligent warning signs according to the embodiments of the present application includes: S110, obtaining a road traffic image collected by a camera installed in an area prone to road traffic accidents and a vehicle passing speed value for a predetermined period collected by a sensor; S120, extracting an optimized road traffic correlation feature vector and a vehicle passing speed feature vector from the road traffic image collected by the camera installed in the area prone to road traffic accidents and the vehicle passing speed value for the predetermined period collected by the sensor; S130, generating a warning prompt for a traffic accident warning sign based on the optimized road traffic correlation feature vector and the vehicle passing speed feature vector.

[0056] Here, those skilled in the art can understand that the specific operations of each step in the above traffic accident warning method based on intelligent warning signs have been described in detail in the description of the traffic accident warning system based on intelligent warning signs above, and therefore, the repeated description thereof will be omitted. Figures 1 to 3 The following will describe an electronic device according to the embodiments of the present application with reference to

[0057] Next, reference will be made to Figure 5 to describe an electronic device according to the embodiments of the present application.

[0058] As Figure 5As shown, the electronic device 10 includes an input device 11, an input interface 12, a central processing unit 13, a memory 14, an output interface 15, an output device 16, and a bus 17. Among them, the input interface 12, the central processing unit 13, the memory 14, and the output interface 15 are interconnected through the bus 17. The input device 11 and the output device 16 are respectively connected to the bus 17 through the input interface 12 and the output interface 15, and then connected to other components of the electronic device 10.

[0059] Specifically, the input device 11 receives input information from the outside and transmits the input information to the central processing unit 13 through the input interface 12; the central processing unit 13 processes the input information based on computer-executable instructions stored in the memory 14 to generate output information, temporarily or permanently stores the output information in the memory 14, and then transmits the output information to the output device 16 through the output interface 15; the output device 16 outputs the output information to the outside of the electronic device 10 for user use.

[0060] In one embodiment, Figure 5 the electronic device 10 shown can be implemented as a network device, and the network device may include: a memory configured to store a program; a processor configured to run the program stored in the memory to execute any one of the tendering methods described in the above embodiments.

[0061] According to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments of the present application include a computer program product, which includes a computer program tangibly contained on a machine-readable medium, and the computer program includes program codes for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network, and / or installed from a removable storage medium.

[0062] Those of ordinary skill in the art will understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and appropriate combinations thereof. In the hardware implementation, the division between the functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, one physical component can have multiple functions, or one function or step can be executed by several physical components in cooperation. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, communication media typically contains computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0063] It should be understood that the above embodiments are merely exemplary embodiments adopted to illustrate the principles of the present application, but the present application is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present application, and these modifications and improvements are also regarded as the protection scope of the present application.

Claims

1. A traffic accident warning system based on intelligent warning signs, characterized in that: include: A traffic accident warning data acquisition module is used to acquire road traffic images collected by cameras installed in areas prone to road traffic accidents and vehicle passing speed values ​​collected by sensors in a predetermined time period; A traffic accident warning data extraction module, used to extract an optimized road traffic association feature vector and a vehicle passing speed feature vector from the road traffic image collected by the camera set up in the area prone to road traffic accidents and the vehicle passing speed value of the predetermined time period collected by the sensor; A warning sign warning prompt generation module is used to generate a traffic accident warning sign warning prompt based on the optimized road traffic association feature vector and the vehicle passing speed feature vector. Wherein, the traffic accident warning data extraction module includes: A road traffic attention feature encoding unit, used for performing attention feature encoding on the road traffic image collected by the camera set up in the area where road traffic accidents are prone to occur, so as to obtain a vehicle trajectory tracking enhanced feature map; A road traffic local feature extraction unit, used for extracting local features from the road traffic image collected by the camera set up in the area where road traffic accidents are prone to occur, so as to obtain a road traffic Canny edge feature map; A road traffic feature aggregation unit, used for performing feature aggregation on the vehicle trajectory tracking enhancement feature map and the road traffic Canny edge feature map to obtain the optimized road traffic association feature vector; A vehicle passing speed feature extraction unit, used for extracting features from the vehicle passing speed values ​​in a predetermined time period collected by the sensor to obtain the vehicle passing speed feature vector; Wherein, the road traffic feature aggregation unit includes: A multi-channel road feature association subunit, used for associating the vehicle trajectory tracking enhancement feature map with the road traffic Canny edge feature map to obtain a multi-channel road traffic input map; A road traffic associated feature encoding subunit, used for passing the multi-channel road traffic input graph through a road traffic convolutional neural network as a feature encoder to obtain a road traffic associated feature vector; A road traffic feature optimization subunit, configured to perform deep point-by-point divergence domain feature compensation on the road traffic associated feature vector based on the vehicle passing speed feature vector to obtain the optimized road traffic associated feature vector; Wherein, the road traffic characteristics optimization subunit includes: Calculating the position-based subtraction between the road traffic associated feature vector and the vehicle passing speed feature vector to obtain a difference feature vector; Dividing the eigenvalues ​​of each position of the phase difference eigenvector by two and taking the absolute value, and then calculating the natural exponential function value to obtain an exponential eigenvector; Dividing a second hyperparameter by each eigenvalue of the exponentially modified eigenvector to obtain a second weighting coefficient vector; Multiplying the second weighting coefficient vector by the vehicle passing speed feature vector according to position to obtain a weighted vehicle passing speed feature vector; The first predetermined hyperparameter is multiplied by the road traffic associated feature vector by position to obtain a weighted road traffic associated feature vector, which is then added to the weighted vehicle passing speed feature vector by position to obtain the optimized road traffic associated feature vector.

2. The traffic accident warning system based on the intelligent warning sign according to claim 1 is characterized in that: The road traffic attention feature encoding unit comprises: The road traffic image collected by the camera set up in the area where road traffic accidents are prone to occur is passed through a vehicle trajectory tracking pixel enhancement generator to obtain a vehicle trajectory tracking pixel enhanced image; The vehicle trajectory tracking pixel enhanced image is passed through a vehicle trajectory tracking enhanced feature encoder based on a spatial attention mechanism to obtain the vehicle trajectory tracking enhanced feature map.

3. The traffic accident warning system based on the intelligent warning sign according to claim 2 is characterized in that: The road traffic local feature extraction unit comprises: Constructing the road traffic image collected by the camera installed in the area where road traffic accidents are prone to occur into a plurality of road traffic partial images; The multiple road traffic local images are passed through a road traffic local feature extractor based on Canny edge detection to obtain the road traffic Canny edge feature map.

4. The traffic accident warning system based on the intelligent warning sign according to claim 3 is characterized in that: The vehicle passes through a speed feature extraction unit, comprising: constructing the vehicle passing speed value of the predetermined time period collected by the sensor as a vehicle passing speed input vector; The vehicle passing speed input vector is passed through a vehicle passing speed timing encoder to obtain the vehicle passing speed feature vector.

5. The traffic accident warning system based on the intelligent warning sign according to claim 4 is characterized in that: The warning sign alarm prompt generation module includes: The optimized road traffic association feature vector and the vehicle passing speed feature vector are merged to obtain a traffic warning sign warning generation feature vector; The traffic warning sign warning generation feature vector is passed through a generator to generate a traffic accident warning sign warning prompt.

6. A traffic accident warning method based on intelligent warning signs, characterized in that: include: Acquire road traffic images collected by cameras installed in areas prone to road traffic accidents and vehicle passing speed values ​​collected by sensors in a predetermined time period; Extracting an optimized road traffic association feature vector and a vehicle passing speed feature vector from the road traffic image collected by the camera set up in the area prone to road traffic accidents and the vehicle passing speed value of the predetermined time period collected by the sensor; Based on the optimized road traffic associated feature vector and the vehicle passing speed feature vector, generating a traffic accident warning sign warning prompt; The step of extracting the optimized road traffic association feature vector and the vehicle passing speed feature vector from the road traffic image collected by the camera set up in the area prone to road traffic accidents and the vehicle passing speed value in the predetermined time period collected by the sensor includes: Performing attention feature encoding on the road traffic image collected by the camera installed in the area where road traffic accidents are prone to occur to obtain a vehicle trajectory tracking enhanced feature map; Extracting local features from the road traffic image collected by the camera located in the area where road traffic accidents are prone to occur, so as to obtain a road traffic Canny edge feature map; Performing feature aggregation on the vehicle trajectory tracking enhancement feature map and the road traffic Canny edge feature map to obtain the optimized road traffic association feature vector; Performing feature extraction on the vehicle passing speed value in the predetermined time period collected by the sensor to obtain the vehicle passing speed feature vector; The step of performing feature aggregation on the vehicle trajectory tracking enhancement feature map and the road traffic Canny edge feature map to obtain the optimized road traffic association feature vector includes: Associating the vehicle trajectory tracking enhancement feature map with the road traffic Canny edge feature map to obtain a multi-channel road traffic input map; Passing the multi-channel road traffic input graph through a road traffic convolutional neural network as a feature encoder to obtain a road traffic associated feature vector; Based on the vehicle passing speed feature vector, performing deep point-by-point divergence domain feature compensation on the road traffic associated feature vector to obtain the optimized road traffic associated feature vector; Wherein, based on the vehicle passing speed feature vector, performing deep point-by-point divergence domain feature compensation on the road traffic associated feature vector to obtain the optimized road traffic associated feature vector includes: Calculating the position-based subtraction between the road traffic associated feature vector and the vehicle passing speed feature vector to obtain a phase difference feature vector; Dividing the eigenvalues ​​of each position of the phase difference eigenvector by two and taking the absolute value, and then calculating the natural exponential function value to obtain an exponential eigenvector; Dividing a second hyperparameter by each eigenvalue of the exponentially modified eigenvector to obtain a second weighting coefficient vector; Multiplying the second weighting coefficient vector by the vehicle passing speed feature vector according to position to obtain a weighted vehicle passing speed feature vector; The first predetermined hyperparameter is multiplied by the road traffic associated feature vector by position to obtain a weighted road traffic associated feature vector, which is then added to the weighted vehicle passing speed feature vector by position to obtain the optimized road traffic associated feature vector.

Images