Intelligent early warning method and system for urban flow of traffic checkpoint based on artificial intelligence
By monitoring video data at traffic checkpoints and analyzing it using convolutional neural networks and long short-term memory networks, the problems of lagging traffic flow monitoring and inaccurate prediction in traditional traffic management have been solved. This has enabled intelligent early warning and efficient decision-making in traffic management, reducing congestion and improving urban traffic flow.
Patent Information
- Application Number
- CN202510807862.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-17
- Publication Date
- 2025-10-31
AI Technical Summary
Traditional traffic management methods suffer from lagging traffic flow monitoring, inaccurate forecasting, and slow response, leading to increased traffic congestion. Existing methods are also unresponsive to sudden surges in traffic flow or extreme weather conditions.
By monitoring video data of traffic checkpoints for both outbound and inbound traffic, and using convolutional neural networks and long short-term memory networks for image analysis, the system identifies traffic flow, vehicle type, and license plate information, predicts future traffic flow, and provides intelligent early warnings.
It improves the accuracy and timeliness of traffic flow forecasting, enables early warnings, supports traffic management decisions, reduces traffic congestion, improves road efficiency, and enhances the intelligence and responsiveness of urban traffic management.
Smart Images
Figure CN120877504A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent transportation, and in particular to an intelligent early warning method and system for traffic flow entering the city at traffic checkpoints based on artificial intelligence. Background Technology
[0002] With the acceleration of urbanization and the continuous growth of vehicle ownership, urban transportation systems are facing increasingly serious congestion problems. Traditional traffic monitoring methods, such as fixed cameras, traffic sensors, and traffic flow detection equipment, can provide real-time data, but due to slow data processing speeds and a lack of intelligent analysis, they cannot promptly capture changes in traffic flow trends or accurately predict the timing and location of traffic congestion. This makes it difficult for traffic management departments to make timely and effective decisions in the face of traffic peaks, accidents, or emergencies, leading to exacerbated congestion, frequent traffic accidents, and even affecting the overall operational efficiency of the city.
[0003] Furthermore, most existing traffic flow prediction methods rely on simple statistical models or empirical formulas, lacking the ability to deeply analyze complex and ever-changing urban traffic conditions. Although some advanced traffic management systems have introduced machine learning and artificial intelligence technologies, due to incomplete data acquisition and insufficient accuracy of prediction algorithms, they often cannot accurately cope with traffic fluctuations under different conditions. In particular, under the influence of uncertain factors such as sudden traffic surges or extreme weather, the traffic prediction error is large, resulting in a slow response. Summary of the Invention
[0004] The purpose of this invention is to provide an intelligent early warning method and system for traffic flow entering the city based on artificial intelligence, in order to solve the technical problems of lagging traffic monitoring, inaccurate prediction, and slow response in traditional traffic management methods, including:
[0005] In a first aspect, the present invention provides an intelligent early warning method for traffic flow entering the city at traffic checkpoints based on artificial intelligence, comprising: monitoring and acquiring outbound video monitoring data sequences and inbound video monitoring data sequences at traffic checkpoints within a preset historical time zone; performing image analysis on the outbound video monitoring data sequences to obtain an outbound traffic flow sequence, an outbound vehicle type percentage sequence, and an outbound vehicle source percentage sequence; performing image analysis on the inbound video monitoring data sequences to obtain an inbound traffic flow sequence, an inbound vehicle type percentage sequence, and an inbound vehicle source percentage sequence; analyzing the outbound traffic flow sequence, outbound vehicle type percentage sequence, outbound vehicle source percentage sequence, inbound traffic flow sequence, inbound vehicle type percentage sequence, and inbound vehicle source percentage sequence to obtain a predicted new inbound traffic flow for a preset future time period; and issuing a traffic warning if the predicted new inbound traffic flow exceeds a preset traffic flow threshold.
[0006] Preferably, the intelligent early warning method for traffic flow entering the city based on artificial intelligence further includes: constructing a vehicle feature recognition channel based on a convolutional neural network, wherein the vehicle feature recognition channel includes a traffic flow recognition branch, a vehicle type recognition branch, and a license plate recognition branch; using the traffic flow recognition branch, the vehicle type recognition branch, and the license plate recognition branch, image analysis is performed on the outbound video monitoring data sequence, and the outbound traffic flow sequence, the outbound vehicle type proportion sequence, and the outbound vehicle source proportion sequence are output.
[0007] Preferably, the intelligent early warning method for traffic flow entering the city based on artificial intelligence further includes: collecting sample video monitoring datasets and sample traffic flow sets according to the historical traffic flow monitoring logs of the traffic checkpoints; using the sample video monitoring datasets as input and the sample traffic flow sets as supervision, training a convolutional neural network until convergence to obtain a traffic flow recognition branch.
[0008] Preferably, the intelligent early warning method for traffic flow entering the city based on artificial intelligence further includes: collecting sample video monitoring datasets, sample vehicle type sets, and sample license plate information sets according to the historical traffic flow monitoring logs of the traffic checkpoints; training a convolutional neural network until convergence using the sample video monitoring dataset as input and the sample vehicle type set as supervision to obtain a vehicle type recognition branch; and training a convolutional neural network until convergence using the sample video monitoring dataset as input and the sample license plate information set as supervision to obtain a license plate recognition branch.
[0009] Preferably, the intelligent early warning method for traffic flow entering the city based on artificial intelligence further includes: using the vehicle flow recognition branch to perform image analysis on the outbound video monitoring data sequence and outputting the outbound vehicle flow sequence; using the vehicle type recognition branch to perform image analysis on the outbound video monitoring data sequence and outputting the vehicle type dataset sequence, calculating the outbound vehicle type proportion sequence; and using the license plate recognition branch to perform image analysis on the outbound video monitoring data sequence and outputting the license plate information set sequence, analyzing to obtain the outbound vehicle source proportion sequence.
[0010] Preferably, the intelligent early warning method for traffic flow entering the city based on artificial intelligence further includes: using the traffic flow recognition branch, vehicle type recognition branch and license plate recognition branch to perform image analysis on the entry video monitoring data sequence to obtain the entry traffic flow sequence, the entry vehicle type proportion sequence and the entry vehicle source proportion sequence.
[0011] Preferably, the intelligent early warning method for traffic flow entering the city based on artificial intelligence further includes: using a long short-term memory network to predict the outbound traffic flow for a preset future period based on the outbound traffic flow sequence, the outbound vehicle type proportion sequence, and the outbound vehicle source proportion sequence, and outputting the predicted outbound traffic flow; using a long short-term memory network to predict the inbound traffic flow for a preset future period based on the inbound traffic flow sequence, the inbound vehicle type proportion sequence, and the inbound vehicle source proportion sequence, and outputting the predicted inbound traffic flow; and calculating the predicted new inbound traffic flow based on the predicted outbound traffic flow and the predicted inbound traffic flow.
[0012] Secondly, the present invention also provides an intelligent early warning system for inbound traffic flow at traffic checkpoints based on artificial intelligence, used to execute an intelligent early warning method for inbound traffic flow at traffic checkpoints based on artificial intelligence as described in the first aspect, comprising: a video monitoring data acquisition module, used to monitor and acquire outbound video monitoring data sequences and inbound video monitoring data sequences at traffic checkpoints within a preset historical time zone; an outbound video analysis module, used to perform image analysis on the outbound video monitoring data sequences to acquire outbound traffic flow sequences, outbound vehicle type proportion sequences, and outbound vehicle source proportion sequences; an inbound video analysis module, used to perform image analysis on the inbound video monitoring data sequences to acquire inbound traffic flow sequences, inbound vehicle type proportion sequences, and inbound vehicle source proportion sequences; a new traffic flow prediction module, used to analyze and obtain a predicted new inbound traffic flow for a preset future time period based on the outbound traffic flow sequences, outbound vehicle type proportion sequences, outbound vehicle source proportion sequences, inbound traffic flow sequences, inbound vehicle type proportion sequences, and inbound vehicle source proportion sequences; and a traffic warning module, used to issue a traffic warning if the predicted new inbound traffic flow exceeds a preset traffic threshold.
[0013] The embodiments of the present invention have the following advantages:
[0014] By monitoring and acquiring outbound and inbound video monitoring data sequences from traffic checkpoints within a preset historical time zone, the system then performs image analysis on the outbound video monitoring data sequences to obtain outbound traffic flow sequences, outbound vehicle type percentage sequences, and outbound vehicle source percentage sequences. Further image analysis is performed on the inbound video monitoring data sequences to obtain inbound traffic flow sequences, inbound vehicle type percentage sequences, and inbound vehicle source percentage sequences. Based on these sequences, a predicted increase in inbound traffic flow for a preset future time period is obtained. Finally, if the predicted increase in inbound traffic flow exceeds a preset flow threshold, a traffic warning is issued. In other words, by using artificial intelligence and image analysis technology, the accuracy and timeliness of traffic flow prediction are improved, enabling early warnings and supporting traffic management decisions, thereby effectively reducing traffic congestion, improving road efficiency, and enhancing the intelligence and responsiveness of urban traffic management. Attached Figure Description
[0015] Figure 1 This is a flowchart illustrating the steps of an intelligent early warning method for traffic flow entering the city based on artificial intelligence, according to the present invention.
[0016] Figure 2 This is a schematic diagram of the structure of an intelligent early warning system for traffic flow entering the city based on artificial intelligence, according to the present invention.
[0017] Explanation of reference numerals in the attached figures:
[0018] The system includes a video monitoring data acquisition module 11, an outbound video analysis module 12, an inbound video analysis module 13, a newly added traffic flow prediction module 14, and a traffic flow early warning module 15. Detailed Implementation
[0019] This invention provides an intelligent early warning method and system for traffic flow at city checkpoints based on artificial intelligence, solving the technical problems of lagging traffic monitoring, inaccurate prediction, and slow response in traditional traffic management methods. By utilizing artificial intelligence and image analysis technology, it improves the accuracy and timeliness of traffic flow prediction, enabling early warnings and supporting traffic management decisions. This effectively reduces traffic congestion, improves road efficiency, and enhances the intelligence and responsiveness of urban traffic management.
[0020] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. It should also be noted that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, not all of them.
[0021] Example 1, please refer to the appendix. Figure 1 This invention provides an intelligent early warning method for traffic flow entering the city at traffic checkpoints based on artificial intelligence, which is applied to an intelligent early warning system for traffic flow entering the city at traffic checkpoints based on artificial intelligence, and specifically includes the following steps:
[0022] S10: Monitor and acquire video monitoring data sequences of traffic checkpoints for leaving the city and entering the city within a preset historical time zone.
[0023] Specifically, by installing video surveillance equipment at traffic checkpoints, the dynamic changes in traffic flow in and out of the city are monitored and recorded in real time. Traffic checkpoints refer to key points or road intersections where vehicles enter and exit the city. These checkpoints are typically equipped with high-definition cameras, radar detection devices, or other monitoring equipment to capture information about passing vehicles. In this process, the monitoring system needs to collect video surveillance data according to a set historical time zone (e.g., a fixed time period, such as morning rush hour, evening rush hour, or holidays), generating detailed monitoring data sequences. The outbound video monitoring data sequence refers to the real-time video data of all vehicles leaving the city recorded by the monitoring system. The system periodically or continuously captures each vehicle leaving the city, including video images of vehicles passing through checkpoints. This video data is stored and organized into a data sequence in chronological order. Similarly, the inbound video monitoring data sequence refers to the video data of all vehicles entering the city. The monitoring images of these vehicles are also recorded, analyzed, and organized into a data sequence in chronological order.
[0024] By comprehensively monitoring and collecting video data of vehicles leaving and entering the city, we can provide original and reliable data support for subsequent traffic flow analysis, trend prediction and early warning mechanisms.
[0025] S20: Perform image analysis on the outbound video monitoring data sequence to obtain the outbound traffic flow sequence, the outbound vehicle type percentage sequence, and the outbound vehicle source percentage sequence.
[0026] Furthermore, step S20 of the present invention also includes:
[0027] S21: Construct a vehicle feature recognition channel based on a convolutional neural network, wherein the vehicle feature recognition channel includes a traffic flow recognition branch, a vehicle type recognition branch, and a license plate recognition branch.
[0028] Furthermore, step S21 of the present invention also includes:
[0029] S211: Collect sample video monitoring dataset and sample traffic flow set based on the historical traffic flow monitoring logs of traffic checkpoints; S212: Use the sample video monitoring dataset as input and the sample traffic flow set as supervision to train a convolutional neural network until convergence, and obtain the traffic flow recognition branch.
[0030] Specifically, the first step is to obtain data on outbound and inbound traffic flow from historical monitoring logs at traffic checkpoints. This data typically includes vehicle passage patterns over different time periods, such as hourly, daily, or specific time periods (e.g., peak hours, holidays). By analyzing historical traffic flow logs, key data points are extracted and combined with corresponding video monitoring data to form a sample video monitoring dataset and a sample traffic flow dataset. The sample video monitoring dataset is a training set composed of historical video surveillance data, including image data or video clips captured from monitoring cameras at checkpoints. These videos record vehicles passing through the checkpoints, and the video content needs to be processed into a format suitable for model input. This typically involves cutting the video frame by frame into images and performing preprocessing such as normalization or resizing to meet the input requirements of the neural network. The sample traffic flow dataset contains traffic flow information corresponding to the video data, usually expressed numerically, such as traffic flow per minute or hour. Traffic flow data can be obtained through traditional traffic counting methods or through post-processing analysis of the video surveillance data.
[0031] Next, using the sample video monitoring dataset as input and the sample traffic flow dataset as supervision, a convolutional neural network is trained. The structure of a convolutional neural network typically includes multiple convolutional layers (for extracting local features of the image), pooling layers (for reducing image dimensionality and computational cost), and fully connected layers (for integrating information and outputting prediction results). The network learns various features extracted from the video images through convolutional operations, such as the number of vehicles in the lanes, vehicle speed, and driving mode, thereby helping to calculate traffic flow. During training, the input video image first enters the network and is processed through convolutional layers, pooling layers, and fully connected layers, ultimately outputting the predicted traffic flow value. Next, based on the error between the predicted and actual values, the gradient of each parameter is calculated using the backpropagation algorithm, and the parameters (weights and biases) in the network are updated using the gradient descent algorithm. The loss function is defined as the sum of squares of the errors between the predicted traffic flow and the actual traffic flow. Through the backpropagation algorithm, the weights and biases in the network are gradually adjusted until the model can fit the training data well. This process is repeated until the network's loss function converges, that is, the error reaches the minimum level, and the model training is completed. At this point, the convolutional neural network can accurately predict the traffic flow on the given input video surveillance data, thus obtaining the traffic flow recognition branch.
[0032] Furthermore, step S21 of the present invention also includes:
[0033] S213: Based on the historical traffic flow monitoring logs of traffic checkpoints, collect sample video monitoring datasets, sample vehicle type sets, and sample license plate information sets; S214: Using the sample video monitoring dataset as input and the sample vehicle type set as supervision, train a convolutional neural network until convergence to obtain a vehicle type recognition branch; S215: Using the sample video monitoring dataset as input and the sample license plate information set as supervision, train a convolutional neural network until convergence to obtain a license plate recognition branch.
[0034] Specifically, firstly, based on historical traffic flow monitoring logs from traffic checkpoints, sample video monitoring datasets, sample vehicle type sets, and sample license plate information sets are collected. The sample video monitoring dataset includes historical video surveillance data, recording all vehicles passing through the checkpoint within different time periods. This video data is acquired through the traffic checkpoint monitoring equipment and is typically processed in a time-series manner to form image sequences that can be input into the neural network. The sample vehicle type set contains vehicle type labels corresponding to the video surveillance data. By labeling historical videos, the system assigns a vehicle type label, such as sedan, truck, or bus, to each frame or video segment based on vehicle appearance or vehicle recognition technology (e.g., body shape, size, tire features). This dataset serves as supervision data in the training of the convolutional neural network, guiding the model to learn how to identify different types of vehicles from video images. The sample license plate information set includes license plate number labels corresponding to the video surveillance data. License plate recognition technology is used to extract the license plate information of each vehicle and associate it with the corresponding video data. This dataset serves as supervision data for the license plate recognition model, used to train the network to recognize the license plate number of each vehicle.
[0035] Next, using the sample video monitoring dataset as input and the sample vehicle type set as supervision, a convolutional neural network (CNN) is trained. The CNN extracts image features, such as vehicle shape, size, wheels, and body structure, through multiple convolutional and pooling layers. Then, the network integrates these features through fully connected layers, ultimately outputting a vehicle type prediction. During training, the loss function is typically cross-entropy loss, used to measure the difference between the network's predicted vehicle type and the actual label. The loss is calculated using backpropagation, and optimization algorithms (such as Adam and SGD) are used to adjust the network weights. After each training iteration, the network parameters are adjusted using gradient descent until the loss function converges, and the model can accurately identify vehicle types in the video, thus obtaining the vehicle type recognition branch.
[0036] On the other hand, using the sample video monitoring dataset as input and the sample license plate information set as supervision, a convolutional neural network is trained. The input to the convolutional neural network comes from images or video frames in the sample video monitoring dataset. Each frame contains one or more vehicles, and the network's task is to identify the license plate of each vehicle. The output is the license plate number, which typically includes letters and numbers; therefore, the output can be a character sequence representing the complete license plate number. For license plate recognition, character-level cross-entropy loss is commonly used, which is particularly suitable for sequence prediction tasks. During training, the network performs backpropagation and parameter updates by comparing the predicted license plate with the actual license plate information. When training converges, the license plate recognition branch of the model is complete and can identify and extract license plate numbers from the video stream in real time.
[0037] S22: Using the traffic flow recognition branch, vehicle type recognition branch, and license plate recognition branch, perform image analysis on the outbound video monitoring data sequence, and output the outbound traffic flow sequence, the outbound vehicle type proportion sequence, and the outbound vehicle source proportion sequence.
[0038] Furthermore, step S22 of the present invention also includes:
[0039] S221: Perform image analysis on the outbound video monitoring data sequence using the traffic flow recognition branch, and output the outbound traffic flow sequence; S222: Perform image analysis on the outbound video monitoring data sequence using the vehicle type recognition branch, and output the vehicle type dataset sequence, and calculate the outbound vehicle type proportion sequence; S223: Perform image analysis on the outbound video monitoring data sequence using the license plate recognition branch, and output the license plate information set sequence, and analyze to obtain the outbound vehicle source proportion sequence.
[0040] Specifically, firstly, the traffic flow recognition branch performs image analysis on the outbound video monitoring data sequence. The traffic flow recognition branch uses a convolutional neural network to extract vehicle features from the images, such as the vehicle body and wheels, locate the vehicles, and count them. For each time period (e.g., per minute, per hour), the traffic flow recognition branch outputs the number of vehicles in that time period, i.e., the traffic flow, based on the results of the image analysis. The traffic flow data in these time periods are organized in chronological order to form an outbound traffic flow sequence. This sequence records in detail the changes in traffic flow at different time points, providing real-time data support for traffic management.
[0041] Next, the vehicle type recognition branch performs image analysis on the outbound video monitoring data sequence. This branch processes each frame of the image using a convolutional neural network (CNN) to automatically identify the vehicle type. The CNN learns the vehicle's external features (such as body shape, size, number of wheels, etc.) to determine the type of each vehicle. After recognition, the network outputs the type of each vehicle in each frame, forming a vehicle type dataset sequence that records the type information of all vehicles in different time periods. Then, the system calculates the proportion of each vehicle type in each time period based on this data. Specifically, it counts the number of different types of vehicles in each time period and calculates their proportion in the total traffic flow. For example, if 70 out of 100 vehicles passing through in a certain time period are cars, 20 are trucks, and 10 are motorcycles, then the proportion of cars is calculated to be 70%, trucks 20%, and motorcycles 10%. Finally, an outbound vehicle type proportion sequence is obtained. This sequence records in detail the proportion of each type of vehicle in different time periods, providing traffic managers with key data on vehicle composition and helping to optimize traffic management strategies and improve road planning and scheduling.
[0042] On the other hand, the license plate recognition branch performs image analysis on the outbound video monitoring data sequence. First, the input outbound video monitoring data sequence consists of multiple consecutive video frames over multiple time periods. Each frame contains vehicles passing through checkpoints. The license plate recognition branch uses image processing technology to locate the license plate area in the image and uses a character recognition algorithm to extract the license plate number from the license plate area. Each time a license plate number is recognized, the system generates a license plate information set sequence. This sequence records the license plate information of all outbound vehicles. These license plate numbers typically contain area codes (such as city or province codes), which allows the vehicle's registration location or origin to be inferred from the license plate number. Next, by analyzing these license plate numbers, the origin region of each vehicle is determined. Then, the origin of all outbound vehicles is statistically analyzed, and the proportion of vehicles from each origin region to the total traffic flow is calculated, forming an outbound vehicle origin percentage sequence. For example, if 100 vehicles leave the city during a certain period, with 60 from city A, 30 from city B, and 10 from city C, then city A accounts for 60%, city B for 30%, and city C for 10%. Ultimately, the outbound vehicle origin percentage sequence provides data on the proportion of vehicles from different origin regions, which is invaluable for traffic management departments to understand the geographical distribution of traffic flow, analyze the inflow of vehicles from other regions, and optimize traffic control strategies between cities.
[0043] S30: Perform image analysis on the video monitoring data sequence of vehicles entering the city to obtain the vehicle flow sequence, the vehicle type proportion sequence, and the vehicle source proportion sequence.
[0044] Furthermore, step S30 of the present invention also includes:
[0045] S31: Using the traffic flow recognition branch, vehicle type recognition branch, and license plate recognition branch, perform image analysis on the entry video monitoring data sequence to obtain the entry traffic flow sequence, the entry vehicle type proportion sequence, and the entry vehicle source proportion sequence.
[0046] Specifically, the system utilizes the traffic flow recognition branch, vehicle type recognition branch, and license plate recognition branch to perform image analysis on the video monitoring data sequence of vehicles entering the city. Specifically, the traffic flow recognition branch identifies and counts all vehicles entering the city, extracting and organizing the number of vehicles in each frame into an entry traffic flow sequence, representing the number of vehicles passing through at different time periods. The vehicle type recognition branch identifies and classifies vehicle types (such as cars, trucks, buses, etc.), and after statistically analyzing the types of all entering vehicles, calculates and outputs an entry vehicle type percentage sequence, reflecting the proportion of different vehicle types in each time period. The license plate recognition branch identifies and extracts the license plate information of each vehicle. Whenever a license plate number is identified, the system organizes this information into an entry vehicle origin percentage sequence, representing the proportion of vehicles from different regions in each time period.
[0047] S40: Based on the outbound traffic flow sequence, outbound vehicle type proportion sequence, outbound vehicle source proportion sequence, inbound traffic flow sequence, inbound vehicle type proportion sequence, and inbound vehicle source proportion sequence, the predicted new inbound traffic flow for a preset future time period is obtained.
[0048] Furthermore, step S40 of the present invention further includes:
[0049] S41: Using a Long Short-Term Memory (LSTM) network, predict the outbound traffic flow for a preset future time period based on the outbound traffic flow sequence, the outbound vehicle type percentage sequence, and the outbound vehicle source percentage sequence, and output the predicted outbound traffic flow. S42: Using an LSTM network, predict the inbound traffic flow for a preset future time period based on the inbound traffic flow sequence, the inbound vehicle type percentage sequence, and the inbound vehicle source percentage sequence, and output the predicted inbound traffic flow. S43: Calculate the predicted new inbound traffic flow based on the predicted outbound traffic flow and the predicted inbound traffic flow.
[0050] Specifically, Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network that excels at processing and predicting time-series data, capturing long-term dependencies and temporal changes within the data. In this step, LSTM networks are used to predict outbound traffic flow for future periods based on outbound traffic flow sequences, outbound vehicle type distribution sequences, and outbound vehicle source distribution sequences. Through its unique gating mechanism, LSTM networks can remember long-term dependencies in the input sequence while eliminating interference from irrelevant information. This allows LSTM networks to effectively identify historical traffic flow patterns, vehicle type distributions, and source distributions, thereby modeling the temporal evolution trend of outbound traffic flow. Based on historical data, LSTM networks predict outbound traffic flow for a future period. By capturing temporal relationships and patterns in the data, LSTM can predict potential future traffic flow trends based on the input historical data. Ultimately, the LSTM model outputs a predicted outbound traffic flow, i.e., the number of vehicles passing through the checkpoint in the preset future period. This prediction result is of great significance to traffic management departments, helping them to prepare for traffic scheduling and congestion prevention in advance.
[0051] On the other hand, using a Long Short-Term Memory (LSTM) network, the network predicts the inbound traffic flow for a predetermined future time period based on the inbound traffic flow sequence, the inbound vehicle type percentage sequence, and the inbound vehicle source percentage sequence. Based on historical data, the LSTM network predicts the inbound traffic flow for a future time period based on the input traffic flow sequence and other relevant features (such as vehicle type and source). The network applies the learned patterns to future time periods, thus outputting a predicted inbound traffic flow. Ultimately, the LSTM provides a predicted inbound traffic flow value for a future time period. This means that traffic management departments can know in advance the possible traffic flow during a certain period, helping them to make reasonable traffic scheduling, flow control, and congestion prevention measures.
[0052] The predicted new inbound traffic flow is calculated based on the predicted outbound traffic flow and the predicted inbound traffic flow. The new inbound traffic flow refers to the number of vehicles entering the city minus the number of vehicles leaving the city within a certain period, reflecting the new traffic flow entering the city during that period.
[0053] S50: If the predicted new traffic flow into the city exceeds the preset traffic flow threshold, a traffic warning will be issued.
[0054] Specifically, the system compares the predicted increase in inbound traffic volume with a pre-set threshold. When the predicted traffic volume exceeds the threshold, the system automatically issues an alert, notifying traffic management departments of potential traffic pressure or congestion. This traffic flow warning prompts management departments to take proactive measures, such as optimizing traffic lights, increasing road patrols, and adjusting public transportation, to alleviate upcoming traffic peaks and prevent traffic accidents or severe congestion. This early warning mechanism significantly improves the response speed of traffic management, reduces traffic problems caused by excessive traffic volume, and thus enhances the smoothness and efficiency of urban traffic.
[0055] In summary, the intelligent early warning method for traffic flow entering the city based on artificial intelligence provided by this invention has the following technical effects:
[0056] By monitoring and acquiring outbound and inbound video monitoring data sequences from traffic checkpoints within a preset historical time zone, the system then performs image analysis on the outbound video monitoring data sequences to obtain outbound traffic flow sequences, outbound vehicle type percentage sequences, and outbound vehicle source percentage sequences. Further image analysis is performed on the inbound video monitoring data sequences to obtain inbound traffic flow sequences, inbound vehicle type percentage sequences, and inbound vehicle source percentage sequences. Based on these sequences, a predicted increase in inbound traffic flow for a preset future time period is obtained. Finally, if the predicted increase in inbound traffic flow exceeds a preset flow threshold, a traffic warning is issued. In other words, by using artificial intelligence and image analysis technology, the accuracy and timeliness of traffic flow prediction are improved, enabling early warnings and supporting traffic management decisions, thereby effectively reducing traffic congestion, improving road efficiency, and enhancing the intelligence and responsiveness of urban traffic management.
[0057] Example 2: Based on the same inventive concept as the AI-based intelligent early warning method for traffic flow entering the city at traffic checkpoints in the foregoing examples, this invention also provides an AI-based intelligent early warning system for traffic flow entering the city at traffic checkpoints. Please refer to the appendix. Figure 2The system includes: a video monitoring data acquisition module 11, used to monitor and acquire outbound and inbound video monitoring data sequences of traffic checkpoints within a preset historical time zone; an outbound video analysis module 12, used to perform image analysis on the outbound video monitoring data sequences to obtain outbound traffic flow sequences, outbound vehicle type proportion sequences, and outbound vehicle source proportion sequences; an inbound video analysis module 13, used to perform image analysis on the inbound video monitoring data sequences to obtain inbound traffic flow sequences, inbound vehicle type proportion sequences, and inbound vehicle source proportion sequences; a new traffic flow prediction module 14, used to analyze the outbound traffic flow sequences, outbound vehicle type proportion sequences, outbound vehicle source proportion sequences, inbound traffic flow sequences, inbound vehicle type proportion sequences, and inbound vehicle source proportion sequences to obtain a predicted new inbound traffic flow for a preset future time period; and a traffic warning module 15, used to issue a traffic warning if the predicted new inbound traffic flow exceeds a preset traffic threshold.
[0058] Furthermore, the AI-based intelligent early warning system for traffic flow entering the city at traffic checkpoints is also used to: construct a vehicle feature recognition channel based on a convolutional neural network, wherein the vehicle feature recognition channel includes a traffic flow recognition branch, a vehicle type recognition branch, and a license plate recognition branch; and use the traffic flow recognition branch, vehicle type recognition branch, and license plate recognition branch to perform image analysis on the outbound video monitoring data sequence, and output the outbound traffic flow sequence, the outbound vehicle type proportion sequence, and the outbound vehicle source proportion sequence.
[0059] Furthermore, the AI-based intelligent early warning system for traffic checkpoints entering the city is also used to: collect sample video monitoring datasets and sample traffic flow sets based on historical traffic flow monitoring logs of traffic checkpoints; use the sample video monitoring datasets as input and the sample traffic flow sets as supervision to train a convolutional neural network until convergence, thereby obtaining a traffic flow recognition branch.
[0060] Furthermore, the AI-based intelligent early warning system for traffic flow entering the city at traffic checkpoints is also used for: collecting sample video monitoring datasets, sample vehicle type sets, and sample license plate information sets based on historical traffic flow monitoring logs at traffic checkpoints; training a convolutional neural network to convergence using the sample video monitoring dataset as input and the sample vehicle type set as supervision to obtain a vehicle type recognition branch; and training a convolutional neural network to convergence using the sample video monitoring dataset as input and the sample license plate information set as supervision to obtain a license plate recognition branch.
[0061] Furthermore, the AI-based intelligent early warning system for traffic checkpoints entering the city is also used to: perform image analysis on the outbound video monitoring data sequence using the vehicle flow recognition branch, and output the outbound vehicle flow sequence; perform image analysis on the outbound video monitoring data sequence using the vehicle type recognition branch, and output the vehicle type dataset sequence, and calculate the outbound vehicle type proportion sequence; and perform image analysis on the outbound video monitoring data sequence using the license plate recognition branch, and output the license plate information set sequence, and analyze to obtain the outbound vehicle source proportion sequence.
[0062] Furthermore, the AI-based intelligent early warning system for traffic checkpoints entering the city is also used to: utilize the traffic flow recognition branch, vehicle type recognition branch, and license plate recognition branch to perform image analysis on the entry video monitoring data sequence to obtain the entry traffic flow sequence, the entry vehicle type proportion sequence, and the entry vehicle source proportion sequence.
[0063] Furthermore, the AI-based intelligent early warning system for traffic checkpoints entering the city is also used to: utilize a long short-term memory network to predict outbound traffic flow for a preset future period based on the outbound traffic flow sequence, the outbound vehicle type percentage sequence, and the outbound vehicle source percentage sequence, and output the predicted outbound traffic flow; utilize a long short-term memory network to predict inbound traffic flow for a preset future period based on the inbound traffic flow sequence, the inbound vehicle type percentage sequence, and the inbound vehicle source percentage sequence, and output the predicted inbound traffic flow; and calculate the predicted new inbound traffic flow based on the predicted outbound traffic flow and the predicted inbound traffic flow.
[0064] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. The AI-based intelligent early warning method and specific examples for urban traffic flow at traffic checkpoints described in Embodiment 1 are also applicable to the AI-based intelligent early warning system for urban traffic flow at traffic checkpoints in this embodiment. Through the foregoing detailed description of the AI-based intelligent early warning method for urban traffic flow at traffic checkpoints, those skilled in the art can clearly understand the AI-based intelligent early warning system for urban traffic flow at traffic checkpoints in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here. As for the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant details can be found in the method section.
[0065] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
[0066] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.
Claims
1. An intelligent early warning method for traffic flow entering the city at traffic checkpoints based on artificial intelligence, characterized in that the method... include: The system monitors and acquires video surveillance data sequences of traffic checkpoints for both outbound and inbound traffic within a preset historical time zone. Image analysis is performed on the outbound video monitoring data sequence to obtain the outbound traffic flow sequence, the outbound vehicle type percentage sequence, and the outbound vehicle source percentage sequence; Image analysis is performed on the video monitoring data sequence of vehicles entering the city to obtain the vehicle flow sequence, the vehicle type proportion sequence, and the vehicle source proportion sequence. Based on the outbound traffic flow sequence, outbound vehicle type percentage sequence, outbound vehicle source percentage sequence, inbound traffic flow sequence, inbound vehicle type percentage sequence, and inbound vehicle source percentage sequence, the predicted new inbound traffic flow for a preset future time period is obtained. If the predicted increase in the number of vehicles entering the city exceeds the preset traffic threshold, a traffic warning will be issued.
2. The intelligent early warning method for traffic flow entering the city based on artificial intelligence according to claim 1, characterized in that, Image analysis is performed on the outbound video monitoring data sequence to obtain the outbound traffic flow sequence, the outbound vehicle type percentage sequence, and the outbound vehicle source percentage sequence, including: A vehicle feature recognition channel is constructed based on a convolutional neural network, wherein the vehicle feature recognition channel includes a traffic flow recognition branch, a vehicle type recognition branch, and a license plate recognition branch; Using the traffic flow recognition branch, vehicle type recognition branch, and license plate recognition branch, image analysis is performed on the outbound video monitoring data sequence to output the outbound traffic flow sequence, the outbound vehicle type percentage sequence, and the outbound vehicle source percentage sequence.
3. The intelligent early warning method for traffic flow entering the city based on artificial intelligence according to claim 2, characterized in that, Construct a traffic flow recognition branch, including: Based on the historical traffic flow monitoring logs of traffic checkpoints, sample video monitoring datasets and sample traffic flow datasets were collected. Using the sample video monitoring dataset as input and the sample traffic flow set as supervision, a convolutional neural network is trained until convergence to obtain the traffic flow recognition branch.
4. The intelligent early warning method for traffic flow entering the city based on artificial intelligence according to claim 2, characterized in that, Construct vehicle type recognition and license plate recognition branches, including: Based on the historical traffic flow monitoring logs of traffic checkpoints, sample video monitoring datasets, sample vehicle type datasets, and sample license plate information datasets were collected. Using the sample video monitoring dataset as input and the sample vehicle type set as supervision, a convolutional neural network is trained until convergence to obtain the vehicle type recognition branch. Using the sample video monitoring dataset as input and the sample license plate information set as supervision, a convolutional neural network is trained until convergence to obtain the license plate recognition branch.
5. The intelligent early warning method for traffic flow entering the city based on artificial intelligence according to claim 2, characterized in that, Using the traffic flow recognition branch, vehicle type recognition branch, and license plate recognition branch, image analysis is performed on the outbound video monitoring data sequence, outputting the outbound traffic flow sequence, the outbound vehicle type percentage sequence, and the outbound vehicle source percentage sequence, including: The outbound video monitoring data sequence is analyzed using the traffic flow recognition branch to output the outbound traffic flow sequence. The vehicle type recognition branch is used to perform image analysis on the outbound video monitoring data sequence, outputting a vehicle type dataset sequence, and calculating the outbound vehicle type proportion sequence. The license plate recognition branch is used to perform image analysis on the outbound video monitoring data sequence, outputting a license plate information set sequence, and the source proportion sequence of outbound vehicles is obtained through analysis.
6. The intelligent early warning method for traffic flow entering the city based on artificial intelligence according to claim 2, characterized in that, Using the traffic flow recognition branch, vehicle type recognition branch, and license plate recognition branch, image analysis is performed on the entry video monitoring data sequence to obtain the entry traffic flow sequence, the entry vehicle type proportion sequence, and the entry vehicle source proportion sequence.
7. The intelligent early warning method for traffic flow entering the city based on artificial intelligence according to claim 1, characterized in that, Based on the analysis of the outbound traffic flow sequence, the outbound vehicle type percentage sequence, the outbound vehicle source percentage sequence, the inbound traffic flow sequence, the inbound vehicle type percentage sequence, and the inbound vehicle source percentage sequence, the predicted new inbound traffic flow for a preset future time period is obtained, including: Using a long short-term memory network, outbound traffic flow is predicted for a preset future time period based on the outbound traffic flow sequence, the outbound vehicle type proportion sequence, and the outbound vehicle source proportion sequence, and the predicted outbound traffic flow is output. Using a long short-term memory network, the inbound traffic flow is predicted for a preset future time period based on the inbound vehicle flow sequence, the inbound vehicle type proportion sequence, and the inbound vehicle source proportion sequence, and the predicted inbound traffic flow is output. The predicted increase in inbound traffic volume is calculated based on the predicted outbound traffic volume and the predicted inbound traffic volume.
8. An intelligent early warning system for traffic flow entering the city based on artificial intelligence, characterized in that: The steps for implementing the AI-based intelligent early warning method for traffic flow entering the city at traffic checkpoints as described in any one of claims 1 to 7 include: The video monitoring data acquisition module is used to monitor and acquire outbound and inbound video monitoring data sequences of traffic checkpoints within a preset historical time zone. The outbound video analysis module is used to perform image analysis on the outbound video monitoring data sequence to obtain the outbound traffic flow sequence, the outbound vehicle type percentage sequence, and the outbound vehicle source percentage sequence. The city entry video analysis module is used to perform image analysis on the city entry video monitoring data sequence to obtain the city entry traffic flow sequence, the city entry vehicle type proportion sequence, and the city entry vehicle source proportion sequence. A new traffic flow prediction module is added, which is used to analyze the outbound traffic flow sequence, the outbound vehicle type proportion sequence, the outbound vehicle source proportion sequence, the inbound traffic flow sequence, the inbound vehicle type proportion sequence, and the inbound vehicle source proportion sequence to obtain the predicted new inbound traffic flow for a preset future period. The traffic flow warning module is used to issue a traffic flow warning if the predicted new traffic flow into the city exceeds a preset traffic flow threshold.