Chain-type anti-intrusion intelligent early warning system for highway reconstruction and extension construction area

Abstract

The invention discloses a chain type anti-intrusion intelligent early warning system for a highway reconstruction and extension construction area. The chain type anti-intrusion intelligent early warning system comprises a sensor network which is responsible for collecting multi-dimensional data in real time; the data processing and analysis center is used for storing and processing the collected data through a cloud computing platform, determining which features are crucial to the early warning effect in combination with the actual situation and domain knowledge of the highway reconstruction and extension construction area, and automatically searching an optimal feature subset by adopting a recursive feature elimination method to reduce the risk of key feature omission; the intelligent early warning and decision support module is used for simulating a construction area scene, generating synthetic data, training and optimizing an early warning model according to the synthetic data, triggering an early warning mechanism through the early warning model based on an analysis result, sending early warning information to a driver, construction personnel and related departments, and providing emergency disposal suggestions at the same time; and the communication and Internet of Things module is used for realizing real-time data transmission between the sensor network and the cloud computing platform as well as between the sensor network and the terminal equipment by using a communication technology.

Description

Technical Field

[0001] The present invention relates to the technical field of road construction warning, and particularly to a chain anti-intrusion intelligent warning system for highway reconstruction and expansion construction areas. Background Art

[0002] Highway reconstruction and expansion projects usually adopt the method of "constructing while maintaining traffic", which makes traffic organization complex and changeable, and the construction and operation interfere with each other seriously. Especially in the accidents of passing vehicles accidentally entering the construction operation control area, the situation is particularly severe. These accidents will not only cause serious casualties and property losses, but also have a negative impact on the construction progress and traffic fluency. Therefore, how to effectively prevent passing vehicles from accidentally entering the construction operation control area has become a major problem in highway reconstruction and expansion projects. The chain anti-intrusion intelligent warning system can monitor and warn potential dangerous vehicles in real time, providing valuable avoidance time for construction personnel, thus effectively reducing the accident rate in the construction area. Through intelligent warning and guidance, the system can prompt vehicles that have strayed into the construction area to change lanes or decelerate in time, reducing the impact on the normal traffic flow and ensuring the unobstructed flow of the highway.

[0003] In the prior art, in the highway reconstruction and expansion construction area, due to the complex environment, the sensor network may capture a large number of data features, such as vehicle speed, vehicle type, vehicle position, construction area length, construction area width, weather conditions, visibility, etc. However, not all features contribute significantly to the warning effect. Some redundant features, such as the detailed dimensions of the construction area (when the dimensions change little), may have little impact on the warning effect, but will increase the computational complexity and reduce the system efficiency. Therefore, a chain anti-intrusion intelligent warning system for highway reconstruction and expansion construction areas is proposed. Summary of the Invention

[0004] The purpose of the present invention is to solve the problems existing in the prior art that in the highway reconstruction and expansion construction area, due to the complex environment, the sensor network may capture a large number of data features, such as vehicle speed, vehicle type, vehicle position, construction area length, construction area width, weather conditions, visibility, etc. However, not all features contribute significantly to the warning effect. Some redundant features, such as the detailed dimensions of the construction area (when the dimensions change little), may have little impact on the warning effect, but will increase the computational complexity and reduce the system efficiency, and to propose a chain anti-intrusion intelligent warning system for highway reconstruction and expansion construction areas.

[0005] In order to achieve the above purpose, the present invention adopts the following technical scheme:

[0006] A chain anti-intrusion intelligent warning system for highway reconstruction and expansion construction areas, comprising:

[0007] Sensor network: Sensors such as radars and cameras deployed along the highway are responsible for collecting multi-dimensional data such as road environment and vehicle dynamics in real time;

[0008] Data processing and analysis center: The collected data is stored and processed through a cloud computing platform. Combining the actual situation and domain knowledge of the highway reconstruction and expansion construction area, determine which features (such as vehicle speed, vehicle type, relative position relationship between the vehicle and the construction area, weather conditions, etc.) are crucial for the warning effect. Adopt a recursive feature elimination method to automatically search for the optimal feature subset, reduce the risk of missing key features, and thus improve the accuracy and efficiency of the warning system;

[0009] Intelligent warning and decision support module: By simulating the construction area scenario, generate synthetic data to supplement the deficiency of real data. Train and optimize the warning model based on the synthetic data. Based on the analysis results, trigger the warning mechanism through the warning model, and send warning information to drivers, construction workers and relevant departments through various means such as LED displays, sounds, and lights. At the same time, provide emergency disposal suggestions to assist relevant departments in quickly responding to emergencies;

[0010] Communication and Internet of Things module: Utilize communication technologies such as NB-IoT and 5G to achieve real-time data transmission between the sensor network, the cloud computing platform and terminal devices, ensuring the immediacy and accuracy of warning information.

[0011] The above technical solution further includes:

[0012] Furthermore, the multi-dimensional data includes road environment data, vehicle dynamic data, construction area data, and other relevant data. The road environment data includes road surface conditions, weather conditions, and traffic flow. Road surface conditions: Sensors can monitor the smoothness, slipperiness, damage, etc. of the road surface. This data helps to evaluate the driving safety of the road and timely warn of potential road surface risks. Weather conditions: By connecting with meteorological sensors or other meteorological systems, the system can obtain real-time weather information such as visibility, rainfall, wind speed, etc. This data is crucial for judging driving conditions, adjusting construction plans, etc. Traffic flow: Using devices such as cameras or radars, the system can monitor the traffic flow, vehicle speed, and lane occupancy in real time, so as to evaluate the congestion degree of the road and provide a decision-making basis for traffic guidance. The vehicle dynamic data includes vehicle position and speed, vehicle type and size, and driver behavior. Vehicle position and speed: Radars and cameras can accurately capture the position and speed information of vehicles. This data is of great significance for judging whether a vehicle has entered the construction area, whether there are violations such as speeding, etc. Vehicle type and size: Through image recognition technology, the system can identify the type and size of vehicles. This is very helpful for evaluating the impact of vehicles on the construction area and formulating targeted warning strategies. Driver behavior: Cameras can also monitor driver behavior, such as whether the driver is fatigued or distracted. This data helps to early warn of potential driving risks. The construction area data includes the boundary of the construction area and the status of construction activities. Boundary of the construction area: The system needs to master the boundary information of the construction area in real time to ensure that the warning system can accurately judge whether a vehicle has entered the construction area. Status of construction activities: Understanding the status of construction activities (such as whether high-risk operations such as excavation and hoisting are in progress) helps the system adjust the warning strategy and improve the accuracy and timeliness of warnings. The other relevant data includes emergency event information and historical data. Emergency event information: The system needs to connect with traffic management departments or other emergency service departments to obtain real-time emergency event information such as traffic accidents and road closures, so as to timely adjust the warning strategy. Historical data: By analyzing historical data, information such as the change trend of traffic flow and accident-prone areas can be revealed, providing more accurate warning support for the system.

[0013] Furthermore, the data processing and analysis center includes a data acquisition module, a data storage module, a data processing module, a data analysis unit, and a feature selection and optimization module. The data acquisition module is responsible for collecting data such as road environment and vehicle dynamics in real time from the sensor network. The data storage module stores the collected raw data, as well as the intermediate data and result data generated by subsequent processing and analysis. The data processing module performs preprocessing operations such as cleaning, formatting, and normalizing the stored data. The data analysis module mines and analyzes the preprocessed data to identify potential intrusion risks, and transmits the analysis results to the intelligent early warning and decision support module after analysis. The feature selection and optimization module combines the actual situation of the highway reconstruction and expansion construction area and domain knowledge to determine which features are crucial for the early warning effect, automatically searches for the optimal feature subset, and reduces the risk of missing key features. The feature selection and optimization module closely cooperates with the data analysis module, adjusts the feature subset according to the analysis results, and feeds back the optimized feature set to the data analysis module for further analysis.

[0014] Furthermore, the data analysis module uses LSTM to model the long-term dependence relationship of vehicle driving data and predicts the driving state of the vehicle in the next period of time. Then, GAN is used to perform anomaly detection on the prediction results to identify abnormal vehicle behaviors that do not conform to the normal driving state, that is, to identify potential intrusion risks, including the following steps:

[0015] Data preprocessing: Convert the original vehicle driving data into a format suitable for LSTM input, and perform cleaning and normalization processing;

[0016] LSTM model construction and training: Use LSTM to capture the long-term dependence relationship in vehicle driving data and predict the driving state of the vehicle in the next period of time; Model definition: Define the LSTM network structure, including the input layer, LSTM layer (multiple layers can be used), fully connected layer, and output layer; Input layer: Receive the preprocessed vehicle driving data sequence; LSTM layer: Capture the long-term dependence relationship in the data; Fully connected layer: Convert the output of the LSTM layer into a prediction result; Output layer: Output the predicted vehicle driving state, such as speed, position, etc.; Model training: Input the preprocessed data into the LSTM model for training, and update the model parameters through the backpropagation algorithm.

[0017] GAN Anomaly Detection: Use GAN to perform anomaly detection on the prediction results of LSTM, and identify abnormal vehicle behaviors that do not conform to the normal driving state; Generator Training: The generator learns to generate data similar to real vehicle driving data; Discriminator Training: The discriminator learns to distinguish between real data and data generated by the generator; Adversarial Training: The generator and the discriminator perform adversarial training until Nash equilibrium is reached; Anomaly Detection: Input the prediction results of LSTM into the trained GAN discriminator. If the discriminator considers the prediction results to be abnormal (i.e., significantly different from the real data distribution), it is marked as a potential intrusion risk;

[0018] Result Output and Warning: Output the anomaly detection results to the intelligent warning and decision support module to trigger the warning mechanism.

[0019] Furthermore, the generator and the discriminator perform adversarial training until Nash equilibrium is reached. Among them, the conditions for reaching Nash equilibrium include the optimality of the generator and the optimality of the discriminator. The optimality of the generator means that the generator has learned the distribution of real data and generates samples that are indistinguishable from real data. At this time, the discriminator cannot accurately distinguish between generated samples and real samples. The optimality of the discriminator means that the discriminator can accurately distinguish between generated samples and real samples. Given the strategy of the generator, the discriminator can no longer improve the accuracy by changing its own strategy. When the optimality of the generator and the optimality of the discriminator are satisfied simultaneously, GAN reaches Nash equilibrium.

[0020] Furthermore, the feature selection and optimization module combines the actual situation of the highway reconstruction and expansion construction area and domain knowledge. The features crucial for the warning effect include vehicle speed, vehicle type, the relative position relationship between the vehicle and the construction area, and weather conditions. Vehicle speed is one of the key factors affecting the warning effect. In the highway reconstruction and expansion construction area, excessive vehicle speed may cause the driver to react too late and increase the risk of intruding into the construction area. It is usually measured in meters per second (m / s) or kilometers per hour (km / h) to represent the driving speed of the vehicle near the construction area; Different types of vehicles (such as cars, trucks, buses, etc.) have differences in braking performance, maneuverability, etc., so their risks of intruding into the construction area are also different. They are usually identified through vehicle classification algorithms or vehicle type information in the database; The relative position relationship between the vehicle and the construction area, such as distance and direction, directly affects the triggering timing and accuracy of the warning system. The specific position information of the vehicle in the construction area can be obtained through technologies such as GPS positioning and radar detection, and its relative distance and direction from the construction area can be calculated; Bad weather (such as rainy days, foggy days, snowy days, etc.) will affect the driver's vision and the vehicle's maneuverability, increasing the risk of intruding into the construction area. The weather condition information is usually obtained through meteorological monitoring equipment or weather forecast data.

[0021] Furthermore, the feature selection and optimization module adopts the recursive feature elimination method to automatically search for the optimal feature subset and reduce the risk of missing key features, including the following steps:

[0022] Model training: Train a machine learning model using all features and evaluate its performance;

[0023] Feature ranking: Rank the features according to the importance evaluation metrics of the model (such as feature weights, coefficients, etc.);

[0024] Feature elimination: Remove the least important features and retrain the model;

[0025] Iteration: Repeat the above steps until the predetermined number of features is reached or the model performance no longer improves significantly.

[0026] Furthermore, the intelligent early warning and decision support module generates synthetic data by simulating the construction area scenario, and trains and optimizes the early warning model according to the synthetic data, including the following steps:

[0027] Determine simulation parameters: Determine the parameters of the construction area scenario to be simulated, such as lane width, vehicle speed, vehicle type, construction personnel location, construction equipment layout, etc. According to the actual situation of the highway reconstruction and expansion construction area, set reasonable parameter ranges and change rules;

[0028] Generate data using simulation software: Use traffic simulation software (such as SUMO, MATSim, etc.) or custom simulation programs to generate synthetic data of the construction area scenario according to the determined parameters. The synthetic data should include dynamic information such as vehicle driving trajectories, speeds, accelerations, etc., as well as static and dynamic position information of construction personnel and equipment;

[0029] Data verification and adjustment: Verify the generated synthetic data to ensure that it conforms to the actual situation and traffic rules of the highway reconstruction and expansion construction area. Adjust the simulation parameters and generation rules according to the verification results to improve the accuracy and usability of the synthetic data;

[0030] Data preprocessing: Preprocess the synthetic data, including data cleaning, format conversion, feature extraction, etc. Divide the processed data into a training set and a test set for training and validating the early warning model;

[0031] Train the early warning model: Use the training set data to train the early warning model so that it can identify potential hazards in the construction area scenario and trigger an early warning. During the training process, optimize the performance and accuracy of the model by adjusting hyperparameters such as model parameters and learning rates;

[0032] Model Validation and Optimization: Use the test set data to validate the trained early warning model, evaluate its early warning accuracy and robustness, and optimize the model according to the validation results, including adjusting the model structure, adding features, etc., to improve the early warning performance.

[0033] The present invention has the following beneficial effects:

[0034] In the present invention, by combining domain knowledge to determine key features, the key factors affecting the safety of the construction area can be accurately captured, thereby improving the accuracy of the early warning system. By using the recursive feature elimination method, it is possible to automatically search for and retain the features crucial for the early warning effect, reduce the interference of redundant information, and further improve the early warning accuracy. The streamlined feature subset can reduce the model complexity and calculation amount, thereby improving the response speed and operation efficiency of the early warning system. Synthetic data can simulate various construction conditions and traffic scenarios, enrich the content and diversity of the data set, and provide sufficient data support for the training of the early warning system. By introducing synthetic data, the early warning system can learn more traffic behavior patterns in different situations, thereby enhancing the generalization ability of the model in different scenarios. Brief Description of the Drawings

[0035] Figure 1 It is a system block diagram of a chain anti-intrusion intelligent early warning system for the construction area of highway reconstruction and expansion proposed by the present invention. Detailed Embodiments

[0036] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0037] Please refer to Figure 1 As shown, the present invention is a chain anti-intrusion intelligent early warning system for the construction area of highway reconstruction and expansion, including:

[0038] Sensor Network: Sensors such as radars and cameras deployed along the road are responsible for collecting multi-dimensional data such as road environment and vehicle dynamics in real time;

[0039] Data Processing and Analysis Center: The collected data is stored and processed through a cloud computing platform. Combining the actual situation and domain knowledge of the highway reconstruction and expansion construction area, it is determined which features (such as vehicle speed, vehicle type, relative position relationship between the vehicle and the construction area, weather conditions, etc.) are crucial for the early warning effect. The recursive feature elimination method is used to automatically search for the optimal feature subset, reducing the risk of missing key features, thereby improving the accuracy and efficiency of the early warning system;

[0040] Intelligent Early Warning and Decision Support Module: By simulating the construction area scenario, synthetic data is generated to supplement the deficiencies of real data. Based on the synthetic data, the early warning model is trained and optimized. Based on the analysis results, the early warning mechanism is triggered through the early warning model, and early warning information is sent to drivers, construction workers and relevant departments through various means such as LED displays, sounds, and lights. At the same time, emergency disposal suggestions are provided to assist relevant departments in quickly responding to emergencies;

[0041] Communication and Internet of Things Module: Using communication technologies such as NB-IoT and 5G, real-time data transmission between the sensor network, the cloud computing platform, and terminal devices is realized to ensure the immediacy and accuracy of early warning information.

[0042] In one embodiment, the multi-dimensional data includes road environment data, vehicle dynamic data, construction area data, and other relevant data. The road environment data includes road surface conditions, weather conditions, and traffic flow. Road surface conditions: Sensors can monitor the smoothness, slipperiness, damage, etc. of the road surface. This data helps to evaluate the driving safety of the road and timely warn of potential road surface risks. Weather conditions: By connecting with meteorological sensors or other meteorological systems, the system can obtain real-time weather information such as visibility, rainfall, wind speed, etc. This data is crucial for judging driving conditions, adjusting construction plans, etc. Traffic flow: Using devices such as cameras or radars, the system can monitor the traffic flow, vehicle speed, and lane occupancy in real time, so as to evaluate the congestion degree of the road and provide a decision-making basis for traffic guidance. The vehicle dynamic data includes vehicle position and speed, vehicle type and size, and driver behavior. Vehicle position and speed: Radars and cameras can accurately capture the vehicle's position and speed information. This data is of great significance for judging whether a vehicle has entered the construction area, whether there are violations such as speeding, etc. Vehicle type and size: Through image recognition technology, the system can identify the vehicle type and size, which is very helpful for evaluating the impact of the vehicle on the construction area and formulating targeted warning strategies. Driver behavior: Cameras can also monitor the driver's behavior, such as whether the driver is fatigued or distracted. This data helps to early warn of potential driving risks. The construction area data includes the construction area boundary and the construction activity status. Construction area boundary: The system needs to know the boundary information of the construction area in real time to ensure that the warning system can accurately judge whether a vehicle has entered the construction area. Construction activity status: Understanding the status of construction activities (such as whether high-risk operations such as excavation and hoisting are in progress) helps the system adjust the warning strategy and improve the accuracy and timeliness of the warning. The other relevant data includes emergency event information and historical data. Emergency event information: The system needs to connect with traffic management departments or other emergency service departments to obtain real-time emergency event information such as traffic accidents and road closures, so as to timely adjust the warning strategy. Historical data: By analyzing historical data, information such as the change trend of traffic flow and accident-prone areas can be revealed, providing more accurate warning support for the system.

[0043] In one embodiment, the data processing and analysis center includes a data acquisition module, a data storage module, a data processing module, a data analysis unit, and a feature selection and optimization module. The data acquisition module is responsible for collecting data such as road environment and vehicle dynamics from the sensor network in real time. The data storage module stores the collected raw data, as well as the intermediate data and result data generated by subsequent processing and analysis. The data processing module performs preprocessing operations such as cleaning, formatting, and normalization on the stored data. The data analysis module mines and analyzes the preprocessed data to identify potential intrusion risks, and transmits the analysis results to the intelligent early warning and decision support module after analysis. The feature selection and optimization module combines the actual situation of the highway reconstruction and expansion construction area and domain knowledge to determine which features are crucial for the early warning effect, automatically searches for the optimal feature subset, and reduces the risk of missing key features. The feature selection and optimization module closely cooperates with the data analysis module, adjusts the feature subset according to the analysis results, and feeds back the optimized feature set to the data analysis module for further analysis.

[0044] In one embodiment, the data analysis module uses LSTM to model the long-term dependence relationship of vehicle driving data and predicts the driving state of the vehicle in the future for a period of time. Then, GAN is used to perform anomaly detection on the prediction results to identify abnormal vehicle behaviors that do not conform to the normal driving state, that is, to identify potential intrusion risks, including the following steps:

[0045] Data preprocessing: Convert the original vehicle driving data into a format suitable for LSTM input, and perform cleaning and normalization processing;

[0046] LSTM model construction and training: Use LSTM to capture the long-term dependence relationship in vehicle driving data and predict the driving state of the vehicle in the future for a period of time; Model definition: Define the LSTM network structure, including the input layer, LSTM layer (multiple layers are available), fully connected layer, and output layer; Input layer: Receive the preprocessed vehicle driving data sequence; LSTM layer: Capture the long-term dependence relationship in the data; Fully connected layer: Convert the output of the LSTM layer into a prediction result; Output layer: Output the predicted vehicle driving state, such as speed, position, etc.; Model training: Input the preprocessed data into the LSTM model for training, and update the model parameters through the backpropagation algorithm;

[0047] GAN Anomaly Detection: Use GAN to perform anomaly detection on the prediction results of LSTM, and identify abnormal vehicle behaviors that do not conform to the normal driving state; Generator Training: The generator learns to generate data similar to real vehicle driving data; Discriminator Training: The discriminator learns to distinguish between real data and data generated by the generator; Adversarial Training: The generator and the discriminator perform adversarial training until a Nash equilibrium is reached; Anomaly Detection: Input the prediction results of LSTM into the trained GAN discriminator. If the discriminator believes that the prediction result is abnormal (i.e., significantly different from the real data distribution), it is marked as a potential intrusion risk;

[0048] Result Output and Warning: Output the anomaly detection results to the intelligent warning and decision support module to trigger the warning mechanism.

[0049] In one embodiment, the generator and the discriminator perform adversarial training until a Nash equilibrium is reached. The conditions for reaching the Nash equilibrium include the optimality of the generator and the optimality of the discriminator. The optimality of the generator means that the generator has learned the distribution of real data and generates samples that are indistinguishable from real data. At this time, the discriminator cannot accurately distinguish between generated samples and real samples. The optimality of the discriminator means that the discriminator can accurately distinguish between generated samples and real samples, and given the generator's strategy, the discriminator can no longer improve its accuracy by changing its own strategy. When both the optimality of the generator and the optimality of the discriminator are satisfied, GAN reaches the Nash equilibrium.

[0050] In one embodiment, the feature selection and optimization module combines the actual situation of the highway reconstruction and expansion construction area and domain knowledge. The features crucial for the warning effect include vehicle speed, vehicle type, the relative position relationship between the vehicle and the construction area, and weather conditions. Vehicle speed is one of the key factors affecting the warning effect. In the highway reconstruction and expansion construction area, excessive vehicle speed may cause the driver to react too late and increase the risk of intrusion into the construction area. It is usually measured in meters per second (m / s) or kilometers per hour (km / h) to represent the driving speed of the vehicle near the construction area; Different types of vehicles (such as cars, trucks, buses, etc.) have differences in braking performance, maneuverability, etc., so their risks of intrusion into the construction area are also different. Usually, vehicle classification algorithms or vehicle type information in the database are used to identify them; The relative position relationship between the vehicle and the construction area, such as distance and direction, directly affects the triggering timing and accuracy of the warning system. The specific position information of the vehicle in the construction area can be obtained through technologies such as GPS positioning and radar detection, and its relative distance and direction from the construction area can be calculated; Bad weather (such as rainy days, foggy days, snowy days, etc.) will affect the driver's vision and the vehicle's maneuverability, increasing the risk of intrusion into the construction area. Usually, weather condition information is obtained through meteorological monitoring equipment or weather forecast data.

[0051] In one embodiment, the feature selection and optimization module adopts a recursive feature elimination method to automatically search for the optimal feature subset and reduce the risk of missing key features, including the following steps:

[0052] Model training: Train a machine learning model using all features and evaluate its performance;

[0053] Feature ranking: Rank the features according to the model's importance evaluation metrics (such as feature weights, coefficients, etc.);

[0054] Feature elimination: Remove the least important feature and retrain the model;

[0055] Iteration: Repeat the above steps until the predetermined number of features is reached or the model performance no longer improves significantly.

[0056] Suppose there is a dataset containing four features: vehicle speed, vehicle type, relative position relationship between the vehicle and the construction area, and weather conditions. Use a random forest as the machine learning model and apply the RFE method for feature selection.

[0057] Initial training: Train a random forest model using all four features and evaluate its performance.

[0058] Feature ranking: Rank the four features according to the random forest's feature importance evaluation metrics. Suppose the ranking result is: vehicle speed > relative position relationship between the vehicle and the construction area > vehicle type > weather conditions.

[0059] Feature elimination: Remove the least important feature "weather conditions" and retrain the random forest model.

[0060] Iteration: Continue to remove the next least important feature "vehicle type" and retrain the model. At this time, the feature set only contains "vehicle speed" and "relative position relationship between the vehicle and the construction area".

[0061] Stop condition: Suppose the stop condition we set is that the number of features does not exceed 2, or the model performance no longer improves significantly. In this case, the iteration can be stopped and the final feature subset can be selected.

[0062] In one embodiment, the intelligent early warning and decision support module generates synthetic data by simulating the construction area scenario, and trains and optimizes the early warning model according to the synthetic data, including the following steps:

[0063] Determine simulation parameters: Determine the parameters of the construction area scenario to be simulated, such as lane width, vehicle speed, vehicle type, positions of construction workers, layout of construction equipment, etc. According to the actual situation of the highway reconstruction and expansion construction area, set reasonable parameter ranges and change rules;

[0064] Generating data using simulation software: Utilize traffic simulation software (such as SUMO, MATSim, etc.) or custom simulation programs to generate synthetic data for the construction area scenario according to the determined parameters. The synthetic data should include dynamic information such as vehicle driving trajectories, speeds, accelerations, as well as static and dynamic position information of construction personnel and equipment.

[0065] Data verification and adjustment: Verify the generated synthetic data to ensure it conforms to the actual situation and traffic rules of the highway reconstruction and expansion construction area. Adjust the simulation parameters and generation rules according to the verification results to improve the accuracy and usability of the synthetic data.

[0066] Data preprocessing: Preprocess the synthetic data, including data cleaning, format conversion, feature extraction, etc. Divide the processed data into a training set and a test set for training and validating the early warning model.

[0067] Training the early warning model: Use the training set data to train the early warning model so that it can identify potential hazards in the construction area scenario and trigger an early warning. During the training process, optimize the performance and accuracy of the model by adjusting hyperparameters such as model parameters and learning rates.

[0068] Model verification and optimization: Use the test set data to verify the trained early warning model, evaluate its early warning accuracy and robustness, and optimize the model according to the verification results, including adjusting the model structure, adding features, etc., to improve the early warning performance.

[0069] Simulating the construction area scenario: Set the lane width to 3.75 meters and the speed limit to 80 km / h. Set up a 50-meter early warning area in front of and behind the operation area, with a total early warning range of 100 meters. Deploy sensors and monitoring equipment in the early warning area to monitor the dynamic information of vehicles and construction personnel in real time.

[0070] Generating synthetic data: Use traffic simulation software to generate synthetic data under different conditions such as vehicle speeds, vehicle types, and construction personnel positions. Ensure that the synthetic data can cover various possible construction area scenarios and potential hazard situations.

[0071] Training the early warning model: Select SVM for model training. Use the synthetic data to train the early warning model so that it can identify potential hazards in the early warning area and trigger an early warning.

[0072] Triggering the early warning mechanism and emergency response: When a vehicle enters the early warning area at a speed exceeding the speed limit of 80 km / h, the early warning model identifies the potential hazard and triggers the early warning mechanism. Display the early warning message "Construction ahead, please slow down" on the LED display screen, and remind the driver through sound and light. At the same time, send the early warning information and emergency response suggestions to relevant departments and personnel for prompt response measures.

[0073] Although embodiments of the present invention have been shown and described, those of ordinary skill in the art will appreciate that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A chain-type anti-intrusion intelligent early warning system for highway reconstruction and expansion construction areas, characterized in that: include: Sensor network: responsible for collecting multi-dimensional data in real time; Data processing and analysis center: The collected data is stored and processed through the cloud computing platform. Combined with the actual situation and domain knowledge of the highway reconstruction and expansion construction area, it determines which features are crucial to the early warning effect. The recursive feature elimination method is used to automatically search for the optimal feature subset to reduce the risk of missing key features. Intelligent warning and decision support module: Generate synthetic data by simulating construction zone scenarios, train and optimize warning models based on synthetic data, trigger warning mechanisms through warning models based on analysis results, send warning information to drivers, construction workers and relevant departments, and provide emergency response suggestions; Communication and Internet of Things module: Use communication technology to achieve real-time data transmission between sensor networks and cloud computing platforms and terminal devices.

2. According to claim 1, a chain-type anti-intrusion intelligent early warning system for highway reconstruction and expansion construction areas is characterized in that: The multi-dimensional data includes road environment data, vehicle dynamic data, construction zone data and other related data. The road environment data includes road conditions, weather conditions and traffic flow. The vehicle dynamic data includes vehicle position and speed, vehicle type and size and driver behavior. The construction zone data includes construction area boundaries and construction activity status. The other related data includes emergency event information and historical data.

3. The chain-type anti-intrusion intelligent early warning system for highway reconstruction and expansion construction areas according to claim 1 is characterized in that: The data processing and analysis center includes a data acquisition module, a data storage module, a data processing module, a data analysis unit and a feature selection and optimization module. The data acquisition module is responsible for real-time data acquisition from the sensor network. The data storage module stores the collected raw data and the intermediate data and result data generated by subsequent processing and analysis. The data processing module performs preprocessing operations on the stored data. The data analysis module mines and analyzes the preprocessed data to identify potential intrusion risks, and transmits the analysis results to the intelligent early warning and decision support module after analysis. The feature selection and optimization module combines the actual situation and domain knowledge of the highway reconstruction and expansion construction area to determine which features are crucial to the early warning effect, automatically searches for the optimal feature subset, and reduces the risk of missing key features. The feature selection and optimization module works closely with the data analysis module to adjust the feature subset according to the analysis results, and feeds the optimized feature set back to the data analysis module for further analysis.

4. The chain-type anti-intrusion intelligent early warning system for highway reconstruction and expansion construction areas according to claim 3 is characterized in that: The data analysis module uses LSTM to model the long-term dependency of vehicle driving data and predict the driving state of the vehicle in the future. Then, GAN is used to perform anomaly detection on the prediction results to identify abnormal vehicle behaviors that are inconsistent with normal driving states, that is, to identify potential intrusion risks, including the following steps: Data preprocessing: convert the original vehicle driving data into a format suitable for LSTM input, and perform cleaning and normalization; LSTM model construction and training: Use LSTM to capture long-term dependencies in vehicle driving data and predict the vehicle's driving status in the future; Model definition: Define the LSTM network structure, including input layer, LSTM layer, fully connected layer and output layer; Input layer: Receive the pre-processed vehicle driving data sequence; LSTM layer: Capture the long-term dependencies in the data; Fully connected layer: Convert the output of the LSTM layer into a prediction result; Output layer: Output the predicted vehicle driving status; Model training: Input the pre-processed data into the LSTM model for training, and update the model parameters through the back-propagation algorithm; GAN anomaly detection: GAN is used to detect anomalies in the prediction results of LSTM and identify abnormal vehicle behaviors that are inconsistent with normal driving status; Generator training: The generator learns to generate data similar to real vehicle driving data; Discriminator training: The discriminator learns to distinguish between real data and data generated by the generator; Adversarial training: The generator and the discriminator are trained adversarially until a Nash equilibrium is reached; Anomaly detection: The prediction results of LSTM are input into the trained GAN discriminator. If the discriminator considers the prediction result to be abnormal, it is marked as a potential intrusion risk; Result output and early warning: Output the anomaly detection results to the intelligent early warning and decision support module to trigger the early warning mechanism.

5. The chain-type anti-intrusion intelligent early warning system for highway reconstruction and expansion construction areas according to claim 4 is characterized in that: The generator and the discriminator are trained adversarially until a Nash equilibrium is reached, wherein the conditions for reaching the Nash equilibrium include an optimal generator and an optimal discriminator. The optimal generator indicates that the generator has learned the distribution of real data and generates samples that are difficult to distinguish from real data. At this time, the discriminator cannot accurately distinguish between the generated samples and the real samples. The optimal discriminator indicates that the discriminator accurately distinguishes between the generated samples and the real samples. Given the generator strategy, the discriminator can no longer improve the accuracy by changing its own strategy. When the generator is optimal and the discriminator is optimal at the same time, GAN reaches a Nash equilibrium.

6. The chain-type anti-intrusion intelligent early warning system for highway reconstruction and expansion construction areas according to claim 5 is characterized in that: The feature selection and optimization module combines the actual situation and domain knowledge of the highway reconstruction and expansion construction area. The features that are crucial to the warning effect include vehicle speed, vehicle type, the relative position relationship between the vehicle and the construction area, and weather conditions.

7. The chain-type anti-intrusion intelligent early warning system for highway reconstruction and expansion construction areas according to claim 5 is characterized in that: The feature selection and optimization module uses a recursive feature elimination method to automatically search for the optimal feature subset and reduce the risk of missing key features, including the following steps: Model training: Use all features to train a machine learning model and evaluate its performance; Feature sorting: Sort features according to the model’s importance evaluation index; Feature elimination: remove the least important features and retrain the model; Iteration: Repeat the above steps until the predetermined number of features is reached or the model performance no longer improves significantly.

8. The chain-type anti-intrusion intelligent early warning system for highway reconstruction and expansion construction areas according to claim 1 is characterized in that: The intelligent early warning and decision support module generates synthetic data by simulating construction zone scenes, and trains and optimizes the early warning model based on the synthetic data, including the following steps: Determine simulation parameters: Determine the simulated construction zone scene parameters, and set parameter ranges and change rules based on the actual conditions of the highway reconstruction and expansion construction zone; Generate data using simulation software: Use traffic simulation software or custom simulation programs to generate synthetic data of construction zone scenarios based on determined parameters; Data verification and adjustment: Verify the generated synthetic data and adjust the simulation parameters and generation rules based on the verification results; Data preprocessing: Preprocess the synthetic data and divide the processed data into training sets and test sets for training and verifying the early warning model; Training the warning model: Use the training set data to train the warning model so that it can identify potential hazards in the construction zone and trigger warnings. During the training process, the performance and accuracy of the model are optimized by adjusting the hyperparameters. Model verification and optimization: Use the test set data to verify the trained early warning model, evaluate its warning accuracy and robustness, and optimize the model based on the verification results.

Images