Machine learning-based urban traffic accident influence range zoning early warning and management and control system
By using edge AI and convolutional neural network to identify traffic accidents in the impact range of urban traffic accidents, and combining nginx smooth weight polling algorithm to detect congestion situations, the existing system solves real-time data processing and traffic events in real-time system The problem of insufficient decision-making efficiency achieves fast and accurate traffic incident response and control, reducing the impact of traffic accidents.
Patent Information
- Application Number
- CN202410817134.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-24
- Publication Date
- 2024-11-08
AI Technical Summary
The existing urban traffic accident impact area zoning warning and control systems are inefficient in real-time data processing and traffic event decision-making, making it difficult to respond quickly and effectively prevent traffic accidents.
Design a partition warning and control system for urban traffic accident impact scope based on machine learning, including edge computing module, event detection result output module, traffic accident impact scope analysis module, data storage module and user terminal module, and use edge AI acceleration card to real-time Analyze the video stream data, identify traffic accidents through convolutional neural networks, combine the nginx smooth weight polling algorithm to detect traffic congestion situations, and conduct real-time traffic control through traffic agents.
It realizes rapid acquisition of traffic event information, improves the scientificity and accuracy of event processing, and can quickly identify the time and space location of accidents, reduces the impact range of traffic events, and prevents congestion.
Smart Images

Figure CN118918705A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data analysis technology, and more specifically, to a machine learning-based urban traffic accident impact area zoning warning and control system. Background Art
[0002] As the process of urban transportation modernization and intelligence accelerates, the urban traffic accident impact area zoning warning and control system based on machine learning is an important part of the traffic management system. The system can timely identify traffic accident situations and risk levels by accurately and timely collecting various traffic information, so as to call relevant emergency plans and conduct zoning control around traffic accidents.
[0003] The urban traffic environment changes in real time. By collecting relevant data through monitoring technology, the image information is processed in real time to obtain the accident type, the results are output and decisions are made, that is, the real-time road information, road condition information, and signal control are adjusted accordingly, and the surrounding traffic conditions are controlled in time. Vehicles can be controlled, which can effectively prevent the occurrence of subsequent traffic accidents. Summary of the invention
[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a city traffic accident impact range zoning warning and control system based on machine learning to solve the problems raised in the above-mentioned background technology.
[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a machine learning-based urban traffic accident impact range zoning warning and control system, comprising an edge computing module, an event detection result output module, a traffic accident impact range analysis module, a data storage module, and a user terminal module;
[0006] Edge computing module: includes the front-end video stream input submodule and the edge video AI algorithm loading submodule. It collects traffic scene video streams in real time through traffic cameras, uses edge devices and is equipped with AI acceleration cards to load and execute traffic accident recognition models, analyzes video images in real time, and detects traffic incidents based on the output results of the recognition model.
[0007] Event detection result output module: including a warning signal generation unit and a warning signal output unit;
[0008] Traffic accident impact range analysis module: Use NGI NX smooth weighted polling algorithm to perform image analysis, detect and perceive road traffic congestion;
[0009] Data storage module: stores the video stream data collected by the front-end video stream input submodule, the result data processed by the edge video AI algorithm loading submodule, and the warning signal data generated by the event detection result output module;
[0010] User terminal module: used to receive the warning signal generated by the event detection result output module, display it to the user, and perform traffic control based on the traffic agent.
[0011] In a preferred embodiment, the front-end video stream input submodule is used to obtain the real-time video stream of the urban road and transmit it to the edge video AI algorithm loading submodule. The specific steps are as follows:
[0012] Step A1, collecting the precise location of moving targets and event traffic information through high-precision maps, further includes the following steps:
[0013] Step A101, obtaining high-precision map data: obtaining high-precision map data from a map service provider, including road elements, road facilities, and surrounding building information;
[0014] Step A102, map data association: associate various data objects in the high-precision map with the road network to ensure high-precision location information of the target and traffic event data objects;
[0015] Step A2, converting the collected high-precision position information data into video data, extracting the real-time and accurate motion coordinates of the target in the video screen, and transmitting the acquired real-time video stream data to the edge video AI algorithm loading submodule, further includes the following steps:
[0016] Step A201, data association: matching the location information and the video frame through the unique identification of the timestamp, and associating the collected high-precision location information data with the video data;
[0017] Step A202, video processing: In the edge video AI algorithm loading module, computer vision technology, including target detection and target tracking algorithms, is used to identify the target object in the video frame and track the position information of the target object in the continuous video frames;
[0018] Step A203, real-time target extraction: after identifying the target object in the video frame according to the result of data association, the real-time accurate motion coordinates of the target in the video screen are extracted, including the position, speed and direction information of the vehicle;
[0019] Step A204, data conversion and integration: Integrate the extracted real-time precise motion coordinates with the video data to generate new video data, including the real-time motion trajectory, speed, and acceleration information of the target object, and associate the target features with the system's continuous time and the target's location information to obtain the coordinates and trajectory of the vehicle in the complete road.
[0020] In a preferred embodiment, the edge video AI algorithm loading submodule is used to use an edge AI acceleration card to perform real-time analysis on the video stream to detect whether a traffic accident has occurred, including a video acquisition unit, an image processing unit, an algorithm loading unit and a result output unit. The specific steps are as follows:
[0021] Step B1, video acquisition unit: using an edge AI acceleration card to receive real-time urban road video stream data from a front-end video stream input module;
[0022] Step B2, image processing unit: preprocessing the collected video images, the preprocessing including image normalization and resizing operations, and extracting key information in the video, including image data of traffic accidents and congestion conditions;
[0023] Step B3, algorithm loading unit: loading the machine learning algorithm into the image processing unit to identify the type, location and impact range of the traffic accident;
[0024] Step B4, result output unit: Generate a corresponding warning signal according to the recognition result of the algorithm loading unit, and output it to the user terminal module.
[0025] In a preferred embodiment, in the algorithm loading unit of step B3, a machine learning algorithm is loaded into the image processing unit, a traffic accident recognition model is constructed using a convolutional neural network algorithm, and features are extracted and classified on the image data, and the classification results of the image are output, further comprising the following steps:
[0026] Step B301, constructing a convolutional neural network: constructing a typical convolutional neural network model, including a convolutional layer, a pooling layer, a fully connected layer and an output layer;
[0027] Step B302, feature extraction: using the image data of traffic accidents and congestion extracted by the image processing unit as training samples, and marking whether there are parking events, vehicle accidents, congestion events, spilled objects and reflective cones, lane change events, and emergency lane occupation events in the images, and using them as supervision information for the training model;
[0028] Step B303, classification prediction: Use convolutional neural network to train the labeled image data to learn the characteristics of traffic accidents and congestion in the image, extract the image features through convolution layer and pooling layer, and pass the features to the fully connected layer, use Softmax function to classify in the output layer, and output the classification result of the image, which is expressed as: Z L =W L ×A L-1 +b L , Among them, W L and bL are the weight matrix and bias term of the output layer, L represents the last layer of the network, is the output of the model, indicating the probability distribution of the image belonging to each category, Z L represents the weighted input of the output layer, A L-1 Represents the output of the previous layer.
[0029] In a preferred embodiment, the event detection result output module includes a warning signal generating unit and a warning signal output unit, and the specific steps are as follows:
[0030] Step C1, warning signal generation unit: Generate corresponding warning signals, including sound, light, and text message, according to the detection results of the edge video AI algorithm loading module;
[0031] Step C2, warning signal output unit: output the warning signal generated by the warning signal generation unit to the user terminal module, so that the user terminal module can take corresponding management and control measures, including traffic light control and traffic induction.
[0032] In a preferred embodiment, the traffic accident impact range analysis module uses the nginx smooth weighted polling algorithm to perform image analysis, detect and perceive the road traffic congestion situation, and the specific steps are as follows:
[0033] Step D1, initializing weight values: extracting traffic accident features, including collision, vehicle speed and number of parking lots, through an image processing unit, and assigning initial weight values {3, 2, 1} to the three features of collision, vehicle speed and number of parking lots respectively;
[0034] Step D2, traffic accident feature selection: using ngi nx smooth weighted polling algorithm, add the current value of each feature to the weight of each feature for calculation, select the feature value with the largest current value as the selected feature, and update the current value of the feature;
[0035] Step D3, adjust weight: according to the selected feature value, adjust the weight of the feature. When a feature is selected, its weight is reduced; when it is not selected, its weight is increased to ensure smooth weight polling;
[0036] Step D4, perform multiple rounds of selection: Repeat steps D3 and D4 through multiple rounds of selection until the current value returns to {0,0,0} after 6 rounds of selection;
[0037] Step D5, loop operation: loop the above operation to meet the smoothing and weight-based requirements.
[0038] In a preferred embodiment, the data storage module stores the video stream data collected by the front-end video stream input submodule, the result data processed by the edge video AI algorithm loading submodule, and the warning signal data generated by the event detection result output module. The specific steps are as follows:
[0039] Step E1, designing the database structure: designing the database tables, including the video stream data table, the processing result data table and the warning signal data table, and determining the fields, indexes and associations of each table;
[0040] Step E2, storing video stream data: storing the collected video stream data in the video stream data table according to the timestamp and camera number information, retaining the original data and related metadata information;
[0041] Step E3, storing result data: designing a data table of frame number, detection result, and confidence field according to the format of the processing result data, and storing the processed result data in the processing result table according to the data structure definition;
[0042] Step E4, storing warning signal data: designing the structure of the warning signal data table, including event type, location, time, and triggering reason fields, and storing the generated warning signal data in a special warning signal data table.
[0043] In a preferred embodiment, the user terminal module is used to receive the warning signal generated by the event detection result output module, display it to the user, and perform traffic control based on the traffic agent. The specific steps are as follows:
[0044] Step F1, user terminal module design: design the user terminal interface, including the warning signal display area and interactive control buttons, so that users can receive the warning signal generated by the event detection result output module, and develop the front-end application to realize the interaction and information display functions of the user interface;
[0045] Step F2, receiving and displaying: the user terminal module receives the warning signal from the event detection result output module, and displays the relevant information on the interface, including the accident location and the impact range;
[0046] Step F3, traffic control based on traffic agent: Integrate the traffic agent function in the user terminal module so that the user can interact with the traffic agent through the interface and issue instructions to the traffic agent, including requesting the dispatch of traffic lights, issuing work orders, and directing traffic diversion operations;
[0047] Step F4, the traffic agent performs traffic control tasks: after receiving the user's instructions, the traffic agent performs corresponding traffic control tasks according to preset rules and algorithms, including controlling traffic lights, directing intersection traffic, and coordinating emergency response to mitigate the impact of traffic accidents.
[0048] The beneficial effects of the present invention are: quickly obtaining traffic incident information, improving the scientific nature of incident handling, using videos with precise location information to quickly obtain the time and spatial location of the incident, determine the nature of the incident, judge its impact, and help traffic incident handling personnel scientifically arrange control measures and scope at the incident site, and control the impact of traffic incidents within the minimum range by carrying out upstream diversion and downstream guidance; the present invention can quickly and accurately extract road condition information, detect congestion patterns, and prevent congestion. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 is a system flow chart of the present invention;
[0050] Figure 2 This is the traffic accident feature weight distribution diagram. DETAILED DESCRIPTION
[0051] 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 those skilled in the art without creative work are within the scope of protection of this application.
[0052] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise clearly and specifically defined.
[0053] In the description of the present application, the term "for example" is used to mean "used as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid unnecessary details to obscure the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.
[0054] Example 1
[0055] This embodiment provides Figure 1 The system is a machine learning-based urban traffic accident impact range zoning warning and control system, which specifically includes an edge computing module, an event detection result output module, a traffic accident impact range analysis module, a data storage module, and a user terminal module;
[0056] Edge computing module: includes the front-end video stream input submodule and the edge video AI algorithm loading submodule. It collects traffic scene video streams in real time through traffic cameras, uses edge devices and is equipped with AI acceleration cards to load and execute traffic accident recognition models, analyzes video images in real time, and detects traffic incidents based on the output results of the recognition model.
[0057] Event detection result output module: including a warning signal generation unit and a warning signal output unit;
[0058] Traffic accident impact range analysis module: Use NGI NX smooth weighted polling algorithm to perform image analysis, detect and perceive road traffic congestion;
[0059] Data storage module: stores the video stream data collected by the front-end video stream input submodule, the result data processed by the edge video AI algorithm loading submodule, and the warning signal data generated by the event detection result output module;
[0060] User terminal module: used to receive the warning signal generated by the event detection result output module, display it to the user, and perform traffic control based on the traffic agent.
[0061] In this embodiment, the edge computing module specifically needs to be explained. The edge computing module includes a front-end video stream input submodule and an edge video AI algorithm loading submodule. The traffic scene video stream is collected in real time through a traffic camera, and an edge device is used with an AI acceleration card to load and execute a traffic accident recognition model, perform real-time analysis on the video image, and perform traffic event detection according to the output result of the recognition model.
[0062] Furthermore, the front-end video stream input submodule is used to obtain the real-time video stream of the urban road and transmit it to the edge video AI algorithm loading submodule. The specific steps are as follows:
[0063] Step A1, collecting the precise location of moving targets and event traffic information through high-precision maps, further includes the following steps:
[0064] Step A101, obtaining high-precision map data: obtaining high-precision map data from a map service provider, including road elements, road facilities, and surrounding building information;
[0065] Step A102, map data association: associate various data objects in the high-precision map with the road network to ensure high-precision location information of the target and traffic event data objects;
[0066] Step A2, converting the collected high-precision position information data into video data, extracting the real-time and accurate motion coordinates of the target in the video screen, and transmitting the acquired real-time video stream data to the edge video AI algorithm loading submodule, further includes the following steps:
[0067] Step A201, data association: matching the location information and the video frame through the unique identification of the timestamp, and associating the collected high-precision location information data with the video data;
[0068] Step A202, video processing: In the edge video AI algorithm loading module, computer vision technology, including target detection and target tracking algorithms, is used to identify the target object in the video frame and track the position information of the target object in the continuous video frames;
[0069] Step A203, real-time target extraction: after identifying the target object in the video frame according to the result of data association, the real-time accurate motion coordinates of the target in the video screen are extracted, including the position, speed and direction information of the vehicle;
[0070] Step A204, data conversion and integration: Integrate the extracted real-time precise motion coordinates with the video data to generate new video data, including the real-time motion trajectory, speed, and acceleration information of the target object, and associate the target features with the system's continuous time and the location information of the target to obtain the coordinates and trajectory of the vehicle in the complete road;
[0071] Furthermore, the edge video AI algorithm loading submodule is used to use an edge AI acceleration card to perform real-time analysis on the video stream to detect whether a traffic accident has occurred, including a video acquisition unit, an image processing unit, an algorithm loading unit and a result output unit. The specific steps are as follows:
[0072] Step B1, video acquisition unit: using an edge AI acceleration card to receive real-time urban road video stream data from a front-end video stream input module;
[0073] Step B2, image processing unit: preprocessing the collected video images, the preprocessing including image normalization and resizing operations, and extracting key information in the video, including image data of traffic accidents and congestion conditions;
[0074] Step B3, algorithm loading unit: loading the machine learning algorithm into the image processing unit to identify the type, location and impact range of the traffic accident;
[0075] Step B4, result output unit: generates a corresponding warning signal according to the recognition result of the algorithm loading unit, and outputs it to the user terminal module;
[0076] Among them, in the algorithm loading unit of step B3, the machine learning algorithm is loaded into the image processing unit, and the traffic accident recognition model is constructed using the convolutional neural network algorithm, by extracting and classifying the image data, and outputting the classification result of the image, further comprising the following steps:
[0077] Step B301, constructing a convolutional neural network: constructing a typical convolutional neural network model, including a convolutional layer, a pooling layer, a fully connected layer and an output layer;
[0078] Step B302, feature extraction: using the image data of traffic accidents and congestion extracted by the image processing unit as training samples, and marking whether there are parking events, vehicle accidents, congestion events, spilled objects and reflective cones, lane change events, and emergency lane occupation events in the images, and using them as supervision information for the training model;
[0079] Step B303, classification prediction: Use convolutional neural network to train the labeled image data to learn the characteristics of traffic accidents and congestion in the image, extract the image features through convolution layer and pooling layer, and pass the features to the fully connected layer, use Softmax function to classify in the output layer, and output the classification result of the image, which is expressed as: Z L =W L ×A L-1 +b L , Among them, W L and b L are the weight matrix and bias term of the output layer, L represents the last layer of the network, is the output of the model, indicating the probability distribution of the image belonging to each category, Z L represents the weighted input of the output layer, A L-1 Represents the output of the previous layer.
[0080] In this embodiment, the convolution layer operation is specifically described. In each convolution layer, features are extracted from the input image through the convolution operation. The formula is expressed as Z I =W I ×a I-1 +b I , A I =g(Z I ), where Z I is the weighted input of the convolutional layer, W I is the convolution kernel, A I is the output feature map after being processed by the activation function g, A I-1 is the input feature map, b I is the bias term, g(Z I ) is the neuron weighted input Z I Output after being processed by the activation function g;
[0081] The pooling operation uses the maximum pooling operation to reduce the dimension of the feature map and extract the most significant features;
[0082] The cross entropy loss function is used to calculate the loss of the classification task, and the model parameters are updated through the back propagation algorithm and gradient descent optimization algorithm to minimize the loss function.
[0083] In this embodiment, what needs to be explained specifically is the event detection result output module, which includes a warning signal generating unit and a warning signal output unit. The specific steps are as follows:
[0084] Step C1, warning signal generation unit: Generate corresponding warning signals, including sound, light, and text message, according to the detection results of the edge video AI algorithm loading module;
[0085] Step C2, warning signal output unit: output the warning signal generated by the warning signal generation unit to the user terminal module, and take corresponding control measures, including traffic light control and traffic guidance, further comprising the following steps:
[0086] Step C201: transmit the warning signal to the warning signal output unit, which is responsible for transmitting the generated warning signal to the user terminal module;
[0087] Step C201: After receiving the warning signal, the user terminal module analyzes the warning content and takes corresponding control measures; the control measures include adjusting traffic light control, issuing traffic guidance information, and guiding vehicles to detour to reduce the risk of traffic accidents and alleviate traffic congestion.
[0088] In this embodiment, the traffic accident impact range analysis module specifically needs to be explained. The traffic accident impact range analysis module uses the nginx smooth weighted polling algorithm to perform image analysis, detect and perceive the road traffic congestion situation, and the specific steps are as follows:
[0089] Step D1, initializing weight values: extract traffic accident features, including collision, vehicle speed, and number of parking lots, through the image processing unit, and assign initial weight values {3, 2, 1} to the three features of collision, vehicle speed, and number of parking lots, respectively, such as Figure 2 As shown;
[0090] Step D2, traffic accident feature selection: using ngi nx smooth weighted polling algorithm, add the current value of each feature to the weight of each feature for calculation, select the feature value with the largest current value as the selected feature, and update the current value of the feature;
[0091] Step D3, adjust weight: according to the selected feature value, adjust the weight of the feature. When a feature is selected, its weight is reduced; when it is not selected, its weight is increased to ensure smooth weight polling;
[0092] Step D4, perform multiple rounds of selection: repeat steps D3 and D4 through multiple rounds of selection until the current value returns to {0,0,0} after 6 rounds of selection;
[0093] Step D5, loop operation: loop the above operation to meet the smoothing and weight-based requirements.
[0094] In this embodiment, it is specifically necessary to explain the data storage module, which stores the video stream data collected by the front-end video stream input submodule, the result data processed by the edge video AI algorithm loading submodule, and the warning signal data generated by the event detection result output module. The specific steps are as follows:
[0095] Step E1, designing the database structure: designing the database tables, including the video stream data table, the processing result data table and the warning signal data table, and determining the fields, indexes and associations of each table;
[0096] Step E2, storing video stream data: storing the collected video stream data in the video stream data table according to the timestamp and camera number information, retaining the original data and related metadata information;
[0097] Step E3, storing result data: according to the format of the processing result data, design a data table of frame number, detection result, and confidence field, and store the processed result data in the processing result table according to the data structure definition;
[0098] Step E4, storing warning signal data: designing the structure of the warning signal data table, including the event type, location, time, and triggering reason fields, and storing the generated warning signal data in a special warning signal data table.
[0099] In this embodiment, the user terminal module specifically needs to be explained. The user terminal module is used to receive the warning signal generated by the event detection result output module, display it to the user, and perform traffic control based on the traffic agent. The specific steps are as follows:
[0100] Step F1, user terminal module design: design the user terminal interface, including the warning signal display area and interactive control buttons, so that the user can receive the warning signal generated by the event detection result output module, and develop the front-end application to realize the interaction and information display functions of the user interface;
[0101] Step F2, receiving and displaying: the user terminal module receives the warning signal from the event detection result output module, and displays the relevant information on the interface, including the accident location and the impact range;
[0102] Step F3, traffic control based on traffic agent: Integrate the traffic agent function in the user terminal module, so that the user can interact with the traffic agent through the interface, and issue instructions to the traffic agent, including requesting to dispatch traffic lights, issuing work orders, and directing traffic diversion operations. The traffic agent is a virtual intelligent agent based on artificial intelligence and natural language processing technology, which is designed to communicate and direct with the user through dialogue interaction to achieve traffic control, emergency response to traffic accidents and traffic diversion tasks;
[0103] Step F4, the traffic agent performs traffic control tasks: after receiving the user's instructions, the traffic agent performs corresponding traffic control tasks according to preset rules and algorithms, including controlling traffic lights, directing intersection traffic, and coordinating emergency response to mitigate the impact of traffic accidents.
[0104] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0105] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0106] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0107] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0108] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1The steps for the functions specified in one or more boxes.
[0109] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0110] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A machine learning-based urban traffic accident impact area zoning warning and control system, characterized by: It includes edge computing module, event detection result output module, traffic accident impact range analysis module, data storage module, and user terminal module; Edge computing module: includes the front-end video stream input submodule and the edge video AI algorithm loading submodule. It collects traffic scene video streams in real time through traffic cameras, uses edge devices and is equipped with AI acceleration cards to load and execute traffic accident recognition models, analyzes video images in real time, and detects traffic incidents based on the output results of the recognition model. Event detection result output module: including a warning signal generation unit and a warning signal output unit; Traffic accident impact range analysis module: Use nginx smooth weighted polling algorithm to perform image analysis, detect and perceive road traffic congestion; Data storage module: stores the video stream data collected by the front-end video stream input submodule, the result data processed by the edge video AI algorithm loading submodule, and the warning signal data generated by the event detection result output module; User terminal module: used to receive the warning signal generated by the event detection result output module, display it to the user, and perform traffic control based on the traffic agent.
2. According to the machine learning-based urban traffic accident impact area zoning warning and control system of claim 1, it is characterized by: The front-end video stream input submodule is used to obtain the real-time video stream of the urban road and transmit it to the edge video AI algorithm loading submodule. The specific steps are as follows: Step A1: Collect the precise location of moving targets and event traffic information through high-precision maps; Step A2: Convert the collected high-precision location information data into video data, extract the real-time and accurate motion coordinates of the target in the video screen, and transmit the acquired real-time video stream data to the edge video AI algorithm loading submodule.
3. According to claim 2, a machine learning-based urban traffic accident impact area zoning warning and control system is characterized by: The step A2 converts the collected high-precision position information data into video data, extracts the real-time and accurate motion coordinates of the target in the video screen, and further includes the following steps: Step A201, data association: matching the location information and the video frame through the unique identification of the timestamp, and associating the collected high-precision location information data with the video data; Step A202, video processing: In the edge video AI algorithm loading module, computer vision technology, including target detection and target tracking algorithms, is used to identify the target object in the video frame and track the position information of the target object in the continuous video frames; Step A203, real-time target extraction: after identifying the target object in the video frame according to the result of data association, the real-time accurate motion coordinates of the target in the video screen are extracted, including the position, speed and direction information of the vehicle; Step A204, data conversion and integration: Integrate the extracted real-time precise motion coordinates with the video data to generate new video data, including the real-time motion trajectory, speed, and acceleration information of the target object, and associate the target features with the system's continuous time and the target's location information to obtain the coordinates and trajectory of the vehicle in the complete road.
4. According to claim 3, a machine learning-based urban traffic accident impact area zoning warning and control system is characterized by: The edge video AI algorithm loading submodule is used to use the edge AI acceleration card to perform real-time analysis on the video stream to detect whether a traffic accident has occurred, including a video acquisition unit, an image processing unit, an algorithm loading unit and a result output unit. The specific steps are as follows: Step B1, video acquisition unit: using an edge AI acceleration card to receive real-time urban road video stream data from a front-end video stream input module; Step B2, image processing unit: pre-processing the collected video images, and extracting image data of traffic accidents and congestion in the video; Step B3, algorithm loading unit: loading the machine learning algorithm into the image processing unit to identify the type, location and impact range of the traffic accident; Step B4, result output unit: Generate a corresponding warning signal according to the recognition result of the algorithm loading unit, and output it to the user terminal module.
5. According to claim 4, a machine learning-based urban traffic accident impact area zoning warning and control system is characterized by: In the algorithm loading unit of step B3, the machine learning algorithm is loaded into the image processing unit, and a traffic accident recognition model is constructed using a convolutional neural network algorithm, by extracting features and classifying the image data, and outputting the classification results of the image, further comprising the following steps: Step B301, constructing a convolutional neural network: constructing a typical convolutional neural network model, including a convolutional layer, a pooling layer, a fully connected layer and an output layer; Step B302, feature extraction: using the image data of traffic accidents and congestion extracted by the image processing unit as training samples, and marking whether there are parking events, vehicle accidents, congestion events, spilled objects and reflective cones, lane change events, and emergency lane occupation events in the images, and using them as supervision information for the training model; Step B303, classification prediction: Use convolutional neural network to train the labeled image data to learn the characteristics of traffic accidents and congestion in the image, extract the image features through convolution layer and pooling layer, and pass the features to the fully connected layer, use Softmax function to classify in the output layer, and the classification result of the output image is expressed as: Z L =W L ×A L-1 +b L , Among them, W L and b L are the weight matrix and bias term of the output layer, L represents the last layer of the network, is the output of the model, indicating the probability distribution of the image belonging to each category, Z L represents the weighted input of the output layer, A L-1 Represents the output of the previous layer.
6. The urban traffic accident impact area zoning warning and control system based on machine learning according to claim 5 is characterized by: The event detection result output module includes a warning signal generating unit and a warning signal output unit, and the specific steps are as follows: Step C1, warning signal generation unit: Generate corresponding warning signals, including sound, light, and text message, according to the detection results of the edge video AI algorithm loading module; Step C2, warning signal output unit: outputs the warning signal generated by the warning signal generation unit to the user terminal module.
7. The urban traffic accident impact area zoning warning and control system based on machine learning according to claim 6 is characterized by: The traffic accident impact range analysis module uses nginx smooth weighted polling algorithm to perform image analysis, detect and perceive road traffic congestion situation, and the specific steps are as follows: Step D1, initializing weight values: extracting traffic accident features, including collision, vehicle speed and number of parking lots, through an image processing unit, and assigning initial weight values {3, 2, 1} to the three features of collision, vehicle speed and number of parking lots respectively; Step D2, traffic accident feature selection: using nginx smooth weighted polling algorithm, add the current value of each feature to the weight of each feature for calculation, select the feature value with the largest current value as the selected feature, and update the current value of the feature; Step D3, adjust weight: adjust the weight of the feature according to the selected feature value. When a feature is selected, reduce its weight; When it is not selected, its weight is increased to ensure smooth weight polling; Step D4, perform multiple rounds of selection: repeat steps D3 and D4 through multiple rounds of selection until the current value returns to {0,0,0} after 6 rounds of selection; Step D5, loop operation: loop the above operation to meet the smoothing and weight-based requirements.
8. The urban traffic accident impact area zoning warning and control system based on machine learning according to claim 7 is characterized by: The data storage module stores the video stream data collected by the front-end video stream input submodule, the result data processed by the edge video AI algorithm loading submodule, and the warning signal data generated by the event detection result output module. The specific steps are as follows: Step E1, designing the database structure: designing the database tables, including the video stream data table, the processing result data table and the warning signal data table, and determining the fields, indexes and associations of each table; Step E2, storing video stream data: storing the collected video stream data in the video stream data table according to the timestamp and camera number information, and retaining the original data; Step E3, storing result data: according to the format of the processing result data, design a data table of frame number, detection result, and confidence field, and store the processed result data in the processing result table according to the data structure definition; Step E4, storing warning signal data: designing the structure of the warning signal data table, including the event type, location, time, and triggering reason fields, and storing the generated warning signal data in a special warning signal data table.
9. The urban traffic accident impact area zoning warning and control system based on machine learning according to claim 8 is characterized by: The user terminal module is used to receive the warning signal generated by the event detection result output module, display it to the user, and perform traffic control based on the traffic agent. The specific steps are as follows: Step F1, user terminal module design: design the user terminal interface, including the warning signal display area, interactive control buttons, and develop the front-end application to perform user interface interaction and information display; Step F2, receiving and displaying: receiving the warning signal from the event detection result output module, and displaying the accident location and impact range information on the interface; Step F3, traffic control based on traffic agent: Integrate the traffic agent function in the user terminal module, so that the user can interact with the traffic agent through the interface and issue instructions to the traffic agent, including requesting the dispatch of traffic lights, issuing work orders, and directing traffic diversion operations; Step F4, the traffic agent performs traffic control tasks: after receiving the user's instructions, the traffic agent performs traffic control tasks, including controlling traffic lights, directing intersection traffic, and coordinating emergency response.
