A method and system for managing people and vehicles at airport crossings

Abstract

The present invention provides a method and system for managing people and vehicles at airport crossings, relating to the field of intelligent transportation systems. The method comprises: identifying vehicle license plate information through image processing technology and comparing facial features of people to obtain traffic data of people and vehicles at the crossing; constructing an LSTM neural network based on the traffic data of people and vehicles at the crossing, and analyzing the temporal variation trend of traffic flow at the crossing using the LSTM neural network to predict peak hours; mining the association rules between people and vehicle traffic behaviors through the Apriori algorithm to obtain traffic preferences of vehicle types; and clustering people and vehicles to identify different traffic patterns and group characteristics. The present invention solves the problems existing in existing airport crossing people and vehicle management systems, improves traffic efficiency, enhances safety, optimizes resource allocation, enhances user experience, and realizes intelligent processing at airports.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation systems, and in particular to a method and system for managing people and vehicles at airport crossings. Background Art

[0002] Traditional management methods for people and vehicles at airport crossings rely primarily on manual monitoring, simple counting devices, and fixed traffic rules. While these methods have been effective in maintaining normal order at crossings to a certain extent, they have gradually exposed numerous shortcomings as airport traffic continues to grow in volume and complexity. Manual monitoring is not only time-consuming and labor-intensive, but also susceptible to human factors such as fatigue and lack of concentration, resulting in poor monitoring effectiveness and difficulty in accurately and timely understanding of crossing traffic flow and abnormal behavior. Traditional counting devices are typically limited to simple vehicle and passenger counting and are unable to identify and analyze key information such as license plates and facial features, making them inadequate for modern airport crossing management. Fixed traffic rules often cannot be dynamically adjusted to reflect real-time traffic flow and demand. This leads to low crossing efficiency, long waiting times, and even traffic congestion during peak hours or under special circumstances. The lack of effective abnormal behavior detection mechanisms makes potential safety hazards such as illegal intrusions and unusual parking difficult to detect and address in a timely manner, posing a threat to the safe operation of airports.

[0003] Traditional management methods are unable to accurately predict peak traffic flow periods, making it difficult to prepare in advance. This leads to increased pressure on intersections and low traffic efficiency during peak hours. Due to a lack of analysis of different traffic patterns and group characteristics, airport management departments often lack specificity when formulating traffic strategies, resulting in irrational resource allocation and serious waste of resources. In the event of an emergency, traditional management methods are unable to respond quickly and cannot adjust scheduling plans in a timely manner, resulting in weak emergency response capabilities and difficulty in effectively addressing emergencies. Due to the lack of clear route instructions and real-time traffic information updates, people and vehicles often feel confused and inconvenienced when passing through intersections, resulting in a poor user experience. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for managing people and vehicles at airport crossings, so as to solve the problems existing in the existing management systems for people and vehicles at airport crossings, improve traffic efficiency, enhance safety, optimize resource allocation, improve user experience, and realize intelligent processing of airports.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0006] In a first aspect, a method for managing people and vehicles at an airport crossing is provided, the method comprising:

[0007] By using image processing technology to identify vehicle license plate information and compare facial features of people, traffic data of people and vehicles at the intersection can be obtained;

[0008] Based on the traffic data of people and vehicles at the road crossing, an LSTM neural network is constructed. The LSTM neural network is then used to analyze the temporal trend of traffic flow at the road crossing to predict peak hours.

[0009] Through the Apriori algorithm, the association rules between people and vehicles' traffic behaviors are mined to obtain the traffic preferences of vehicle types;

[0010] Cluster people and vehicles to identify different traffic patterns and group characteristics;

[0011] The One-Class SVM algorithm is used to detect abnormal traffic behavior of people and vehicles and obtain abnormal behavior characteristics;

[0012] Dispatch personnel and vehicles to optimize travel routes based on predicted peak hours, vehicle type preferences, travel patterns, group characteristics, and abnormal behavior characteristics;

[0013] According to the optimized route, electronic display screens and voice prompt systems are set up to provide traffic guidance and instructions for people and vehicles, so as to achieve the management of people and vehicles at the airport crossing.

[0014] Furthermore, image processing technology is used to identify vehicle license plate information and compare facial features of people to obtain traffic data of people and vehicles at the intersection, including:

[0015] High-definition cameras are installed at airport crossings to capture images of people and vehicles passing through the crossings. These cameras should have high resolution and clear imaging capabilities to ensure the accuracy of subsequent image processing.

[0016] Use image processing technology to perform license plate recognition on the collected vehicle images, count the number of vehicles passing and the time, and obtain the recognition results;

[0017] Extract and compare facial features of the collected person images to obtain facial feature comparison results;

[0018] Based on the results of vehicle license plate recognition and facial feature comparison, the personnel and vehicle flow data at the crossing are counted, including the number of vehicles and people passing through the crossing per unit time.

[0019] Furthermore, based on the traffic data of people and vehicles at the road crossing, an LSTM neural network is constructed. The LSTM neural network is used to analyze the temporal trend of traffic flow at the road crossing to predict peak hours, including:

[0020] Based on the traffic data of personnel and vehicles at the road crossing, the foundation of the LSTM neural network is constructed;

[0021] The LSTM neural network is trained using historical training data. During the training process, appropriate hyperparameters are set and the network parameters are adjusted using the backpropagation algorithm to minimize the error between the predicted value and the true value, resulting in a trained LSTM neural network. The hyperparameters include the learning rate, batch size, and number of training rounds.

[0022] The new traffic flow data is predicted through the trained LSTM neural network to obtain the predicted peak hours of traffic flow.

[0023] Furthermore, the Apriori algorithm is used to mine the association rules between people and vehicles’ traffic behaviors, and the traffic preferences of vehicle types are obtained, including:

[0024] Traffic data of people and vehicles is used to construct transaction item sets for the Apriori algorithm. Each transaction can be represented as a combination of the traffic behaviors of people and vehicles.

[0025] By iteratively generating candidate item sets, obtaining support, and pruning infrequent item sets, all frequent item sets that meet the minimum support threshold are obtained, and association rules are screened out;

[0026] The confidence of each rule is calculated to obtain the traffic preference of vehicle type.

[0027] Furthermore, people and vehicles are clustered to identify different traffic patterns and group characteristics, including:

[0028] Preprocess the traffic data of people and vehicles at the airport crossing and extract the features of the preprocessed data;

[0029] The similarity or distance between the features of the data is used to divide the data points into different clusters, each cluster represents a common pattern or group feature, and the clustering results are obtained;

[0030] By calculating the intra-cluster compactness and inter-cluster separation of the clustering results, different traffic patterns and group characteristics are identified.

[0031] Furthermore, the One-Class SVM algorithm is used to detect abnormal traffic behavior of people and vehicles and obtain abnormal behavior characteristics, including:

[0032] Collect personnel and vehicle flow data at airport intersections and pre-process the data;

[0033] The One-Class SVM algorithm is trained based on preprocessed normal traffic behavior data. During training, the algorithm learns the characteristic distribution of normal traffic behavior and constructs a hypersphere that encloses the normal data points. The boundary of the hypersphere represents the dividing line between normal and abnormal traffic behavior.

[0034] The traffic behavior data of people and vehicles to be detected are passed through the trained One-ClassSVM algorithm to determine whether each data point is inside the hypersphere. If the data point is outside the hypersphere, it is regarded as abnormal traffic behavior, so as to detect abnormal traffic behavior of people and vehicles and obtain abnormal behavior characteristics.

[0035] Furthermore, based on the predicted peak hours, traffic preferences of vehicle types, traffic patterns, group characteristics, and abnormal behavior characteristics, personnel and vehicles are dispatched to optimize traffic routes, including:

[0036] Integrate peak hour predictions, vehicle type preferences, traffic patterns and group characteristics, and abnormal behavior characteristics to obtain integrated data;

[0037] Based on the results of data integration and analysis, preliminary personnel and vehicle scheduling plans are developed. This includes arranging additional crossing staff or adjusting vehicle travel times based on peak hour forecasts, optimizing the travel order of different vehicle types or allocating dedicated lanes based on their travel preferences, and developing targeted travel strategies based on travel patterns and group characteristics.

[0038] According to the preliminary scheduling plan, the preliminary scheduling plan is adjusted according to the abnormal behavior characteristics to obtain the scheduling plan;

[0039] Optimize traffic routes by simulating and evaluating scheduling plans.

[0040] In a second aspect, an airport crossing personnel and vehicle management system includes:

[0041] The acquisition module is used to identify vehicle license plate information through image processing technology and compare facial features of people to obtain traffic data of people and vehicles at the intersection;

[0042] The recognition module is used to construct an LSTM neural network based on the traffic flow data of people and vehicles at the road crossing. The LSTM neural network is used to analyze the temporal trend of traffic flow at the road crossing to predict peak hours. The Apriori algorithm is used to discover the association rules between the traffic behavior of people and vehicles and obtain the traffic preferences of vehicle types. People and vehicles are clustered to identify different traffic patterns and group characteristics.

[0043] The optimization module is used to detect abnormal traffic behavior of people and vehicles in traffic data through the One-Class SVM algorithm and obtain abnormal behavior characteristics; it dispatches people and vehicles and optimizes traffic routes based on predicted peak hours, traffic preferences of vehicle types, traffic patterns, group characteristics and abnormal behavior characteristics; and it sets up electronic display screens and voice prompt systems based on the optimized routes to guide and instruct people and vehicles to achieve people and vehicle management at airport crossings.

[0044] According to a third aspect, a computing device includes:

[0045] one or more processors;

[0046] The storage device is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method.

[0047] In a fourth aspect, a computer-readable storage medium stores a program, which implements the method when executed by a processor.

[0048] The above solution of the present invention includes at least the following beneficial effects:

[0049] Through image processing technology, vehicle license plate information and facial features are accurately identified, and traffic data at road crossings is quickly counted. LSTM neural networks are used to predict peak hours, helping airport management departments prepare in advance and rationally arrange crossing staff and vehicle travel times, effectively alleviating traffic pressure and improving traffic efficiency. The Apriori algorithm is used to mine the association rules of traffic behavior and understand the traffic preferences of vehicle types, providing a basis for optimizing traffic sequences and allocating dedicated lanes, reducing vehicle waiting times, and improving crossing capacity.

[0050] By detecting abnormal traffic behavior, potential safety hazards can be discovered and addressed promptly. Scheduling plans can be adjusted based on the characteristics of abnormal behavior, improving the airport's emergency response capabilities, effectively responding to emergencies, and ensuring the safety of personnel and vehicles. By identifying different traffic patterns and group characteristics, data support can be provided for airport management departments to formulate targeted traffic strategies, optimize resource allocation, and improve resource utilization efficiency. Based on predicted peak hours and traffic preferences, the passage time of crossing staff and vehicles can be reasonably arranged to avoid resource waste and reduce operating costs. Electronic display screens and voice prompt systems are installed to provide clear route instructions, real-time traffic information updates, and safety prompts to help people and vehicles pass through the crossing quickly and accurately. In an emergency, the electronic display screens and voice prompt systems can quickly provide evacuation instructions to ensure the safe evacuation of personnel and vehicles, and enhance user satisfaction with airport services. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 The present invention provides a flowchart of a method for managing people and vehicles at an airport crossing.

[0052] Figure 2 This is a schematic diagram of an airport crossing personnel and vehicle management system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0054] like Figure 1 As shown, an embodiment of the present invention provides a method for managing people and vehicles at an airport intersection, the method comprising the following steps:

[0055] Step 11: Use image processing technology to identify vehicle license plate information and compare facial features of people to obtain traffic data of people and vehicles at the intersection;

[0056] Step 12: Based on the traffic flow data of people and vehicles at the road crossing, an LSTM neural network is constructed, and the time variation trend of the traffic flow at the road crossing is analyzed using the LSTM neural network to predict the peak period;

[0057] Step 13: Using the Apriori algorithm, the association rules between people and vehicles’ traffic behaviors are mined to obtain the traffic preferences of vehicle types.

[0058] Step 14: Cluster people and vehicles to identify different traffic patterns and group characteristics;

[0059] Step 15: Using the One-Class SVM algorithm, detect abnormal traffic behavior of people and vehicles and obtain abnormal behavior features;

[0060] Step 16: Dispatch personnel and vehicles to optimize travel routes based on predicted peak hours, vehicle type preferences, travel patterns, group characteristics, and abnormal behavior characteristics;

[0061] Step 17: Based on the optimized route, an electronic display screen and a voice prompt system are set up to provide traffic guidance and instructions for people and vehicles, so as to achieve the management of people and vehicles at the airport crossing.

[0062] In this embodiment of the present invention, image processing technology can accurately identify vehicle license plates and facial features. By comparing facial features, illegal intruders or suspicious individuals can be promptly detected and dealt with, improving crossing safety. An LSTM neural network can capture temporal trends in traffic flow and accurately predict peak hours. This prediction allows for the rational scheduling of crossing staff and vehicle traffic times, optimizing resource allocation and improving crossing efficiency. The Apriori algorithm can identify correlations between human and vehicle traffic behavior, helping managers gain a deeper understanding of traffic patterns. Based on vehicle preferences, targeted traffic strategies can be developed, such as optimizing traffic order for different vehicle types or allocating dedicated lanes, improving crossing efficiency. Cluster analysis can identify different traffic patterns and group characteristics, helping managers discover underlying traffic patterns. Based on these patterns and group characteristics, personalized management strategies can be developed, such as providing customized services or preferential treatment for specific groups. The One-Class SVM algorithm can promptly detect abnormal traffic behavior of people and vehicles. Detecting abnormal traffic behavior helps to promptly identify and address safety hazards, ensuring the safe operation of crossings. Comprehensive scheduling and optimized routes can significantly improve crossing efficiency, reduce congestion and wait times. Optimized routes and scheduling strategies enhance user experience, making it easier for people and vehicles to pass through crossings. Electronic displays and voice prompts provide clear guidance and instructions for people and vehicles, reducing the risk of getting lost or mistaking their route. Timely prompts and guidance can improve crossing safety and reduce the risk of traffic accidents.

[0063] In a preferred embodiment of the present invention, the above step 11 may include:

[0064] Step 111: High-definition cameras are installed at the airport crossing to capture images of people and vehicles passing through the crossing. These cameras should have high resolution and clear imaging capabilities to ensure the accuracy of subsequent image processing.

[0065] Step 112: using image processing technology to perform license plate recognition on the collected vehicle images, counting the number and time of vehicle passage, and obtaining recognition results;

[0066] Step 113: extract and compare facial features of the collected person images to obtain a facial feature comparison result.

[0067] Step 114 , based on the vehicle license plate recognition results and the facial feature comparison results of the people, the traffic flow data of people and vehicles at the crossing is counted, including calculating the number of vehicles and people passing through the crossing per unit time.

[0068] In an embodiment of the present invention, high-definition cameras can capture clearer images, helping to identify potential security threats such as illegally carried items and suspicious individuals. High resolution and clear imaging ensure the accuracy of subsequent image processing. License plate recognition can accurately count the number and time of vehicle traffic, helping airport traffic management departments optimize traffic flow and reduce congestion. In the event of a security incident, license plate recognition results can be used to quickly trace the relevant vehicles, improving emergency response speed. Long-term vehicle traffic data can provide valuable reference for airport planning and operations, such as predicting peak hours and adjusting parking resources. Facial feature extraction and comparison can be used to verify personnel identities, ensuring that only authorized personnel can enter specific areas and improving airport security. Traffic flow data can be used to analyze the busyness of airport crossings. Based on this traffic flow data, resources such as personnel allocation and equipment layout can be rationally adjusted to improve crossing efficiency.

[0069] In the present invention, the specific steps include:

[0070] Step 111: Select a camera with high resolution and clear imaging capabilities to ensure that it can capture details of vehicle license plates and people's faces. Consider the width, length and traffic flow of the crossing, reasonably arrange the number of cameras to ensure that it can fully capture images of people and vehicles passing through the crossing, set the camera's acquisition frequency, and adjust the camera's exposure time, gain and other parameters.

[0071] Step 112 performs pre-processing operations such as denoising and enhancement on the collected vehicle images, converts the images into a format and resolution suitable for license plate recognition, uses image processing algorithms such as edge detection and morphological operations to locate the position of the vehicle license plate in the image, uses optical character recognition technology to recognize the segmented license plate image, extracts the character information in the license plate, compares it with a known license plate database, confirms the license plate number, records the passing time of each vehicle, and counts the number of vehicles passing through the intersection per unit time based on the timestamp.

[0072] Step 113 performs preprocessing operations such as denoising and enhancement on the collected person image, locates the face position in the image through a Haar cascade classifier, segments the face area from the image, extracts feature information of the face area using a local binary pattern, and converts the extracted feature information into a feature vector or feature descriptor; compares the extracted feature vector with a known facial feature database, calculates the similarity or distance between the feature vectors, determines whether they are the same person, and records the comparison results in the database.

[0073] Step 114 , extract relevant information from the vehicle license plate recognition results and the person facial feature comparison results, group the vehicles and people according to their passing time, and calculate the number of vehicles and people passing through the intersection per unit time.

[0074] In a preferred embodiment of the present invention, the above step 12 may include:

[0075] Step 121, constructing the foundation of the LSTM neural network based on the traffic data of people and vehicles at the road crossing;

[0076] Step 122: Training the LSTM neural network using historical training data. During the training process, appropriate hyperparameters are set and the network parameters are adjusted using a backpropagation algorithm to minimize the error between the predicted value and the true value, thereby obtaining a trained LSTM neural network. The hyperparameters include a learning rate, a batch size, and a number of training rounds.

[0077] In step 123, the new traffic flow data is predicted by the trained LSTM neural network to obtain the predicted peak period of traffic flow.

[0078] In an embodiment of the present invention, an LSTM neural network is constructed using actual crossing personnel and vehicle flow data, making the model closer to the actual situation. The LSTM neural network is good at processing time series data and can capture the time dependence and trend in traffic flow data. The LSTM-based neural network structure is flexible and can be adjusted and expanded according to actual needs. By adjusting the network parameters through the backpropagation algorithm, the error between the predicted value and the true value can be minimized, thereby improving the prediction accuracy of the model. Setting appropriate hyperparameters can optimize the model training process and improve the convergence speed and stability of the model. By predicting the peak hours of traffic flow, the airport management department can make early warnings and plans in advance, such as increasing crossing staff, adjusting traffic lights, etc., to deal with possible traffic congestion. According to the prediction results, the airport can reasonably adjust resource allocation, such as increasing parking resources during peak hours, optimizing flight take-off and landing times, etc., to improve the airport's operational efficiency.

[0079] In the present invention, the specific steps include:

[0080] Step 121 , based on step 11 above, obtains traffic data for people and vehicles at the road crossing, including key information such as timestamps, number of people, number of vehicles, and vehicle type. The data is preprocessed to extract key features, such as time and traffic characteristics, and converted into a time series format, meaning each sample contains continuous traffic data for a period of time.

[0081] Conduct network structure design, including:

[0082] Input layer: Determine the dimension of the input layer, that is, the number of features at each time step;

[0083] LSTM layer: Design the number of LSTM layers and the number of neurons in each layer. Multi-layer LSTM can capture more complex temporal dependencies.

[0084] Output layer: Design the dimensions and activation function of the output layer based on the peak traffic flow period. For example, if the prediction is for traffic flow values ​​within a certain period of time, the output layer can be a fully connected layer that outputs continuous values; if the prediction is for the peak period identifier, the output layer can be a classification layer.

[0085] The weights and biases of the LSTM network are randomly initialized to ensure that the network has diversity at the beginning of training.

[0086] Step 122: Divide the preprocessed time series data into a training set and a validation set. The training set is used to train the network, and the validation set is used to evaluate network performance and adjust hyperparameters. Hyperparameters are set, including the learning rate, which controls the step size for updating network parameters. A large learning rate may prevent the network from convergence, while a small learning rate may lead to slow training. The batch size is the number of samples used in each training session, and the selection of the batch size affects training speed and memory usage. The number of training rounds is the number of times the entire training set is traversed. Too many training rounds may lead to overfitting, while too few training rounds may result in insufficient network training.

[0087] The training data set is fed into the LSTM network in batches. The hidden state and output are calculated for each time step. The loss function is calculated based on the mean squared error between the predicted and true values. The gradient of the loss function with respect to the network parameters is calculated using the backpropagation algorithm. The Adam optimizer is used to update the network parameters based on the gradient to minimize the loss function. This process is repeated until the preset number of training rounds is reached or the loss function converges below a certain threshold.

[0088] In step 123, the pre-processed new data is input into the trained LSTM network. The network will output the predicted traffic flow value or peak period identifier based on the input data. If the output is a continuous value, such as the traffic value in the future, the peak period can be determined by analyzing these values; if the output is a categorical identifier, such as peak period or non-peak period, it can be directly judged based on the identifier and the prediction result can be analyzed, such as calculating the error between the predicted value and the true value, drawing a prediction curve, etc.

[0089] In a preferred embodiment of the present invention, the above step 13 may include:

[0090] Step 131: Build a transaction item set for the Apriori algorithm based on the traffic data of people and vehicles. Each transaction can be represented as a combination of the traffic behaviors of people and vehicles.

[0091] Step 132, by iteratively generating candidate item sets, obtaining support, and pruning infrequent item sets, all frequent item sets that meet the minimum support threshold are obtained, and association rules are screened out;

[0092] Step 133 , calculate the confidence of each rule to obtain the traffic preference of the vehicle type.

[0093] In this embodiment of the present invention, passenger and vehicle traffic data is integrated into transaction item sets, making the data more structured. By constructing transaction item sets, potential associations between passenger and vehicle traffic behaviors can be revealed. Through the iterative process of the Apriori algorithm, frequent patterns in passenger and vehicle traffic behaviors—combinations of behaviors that frequently occur together—can be discovered. By setting a minimum support threshold, meaningful association rules can be screened out, avoiding the generation of excessive, useless rules. Understanding which passenger and vehicle traffic behaviors frequently occur together can help airports optimize resource allocation, such as adjusting intersection layouts and increasing traffic capacity during specific time periods. By calculating confidence, the strength of association rules can be quantified—that is, the probability that one behavior will also occur when another occurs. High-confidence association rules can reveal the traffic preferences of different vehicle types, such as certain vehicle types preferring to travel at specific time periods or intersections. Understanding vehicle traffic preferences helps airports develop more precise operational strategies, such as providing exclusive services for specific vehicle types and optimizing parking resource allocation.

[0094] In the present invention, the specific steps include:

[0095] Step 131 obtains passenger and vehicle traffic data at the airport crossing, including the identities of individuals and vehicle types passing through within each time window. The traffic within each time window is considered a transaction. The elements in a transaction are combinations of individual and vehicle traffic, for example, "{person A, car}" or "{person B, truck, person C}." If no individuals or vehicles pass through a time window, that time window does not constitute a transaction or can be considered an empty transaction. All transactions are integrated into a transaction database, with each transaction being a record. The elements in the transaction database are encoded or mapped to facilitate subsequent calculations and processing. For example, numbers or letters can be used to represent different types of individuals and vehicles.

[0096] In step 132, all single elements (i.e., 1-itemsets) are extracted from the transaction database as initial candidate itemsets. For example, if the transaction database contains elements such as "car," "truck," and "person A," the initial candidate itemsets are {car}, {truck}, {person A}, etc. The transaction database is traversed, and the frequency of each candidate itemset's occurrence in the transaction database, i.e., its support, is calculated. Support is defined as the ratio of the number of times a candidate itemset appears in the transaction database to the total number of transactions in the transaction database. A minimum support threshold is set to filter frequent itemsets. Only candidate itemsets with a support greater than or equal to the minimum support threshold are considered frequent itemsets. Based on these frequent itemsets, frequent itemsets are paired together to generate new candidate itemsets. For example, if {car} and {person A} are both frequent itemsets, the candidate itemset {car, person A} can be generated. According to the Apriori property, if an itemset is frequent, then all its non-empty subsets are also frequent. Candidate itemsets that cannot become frequent itemsets are pruned. For example, if a subset of {car, person A} (such as {car} or {person A}) is not frequent, then {car, person A} cannot be frequent either, so it can be pruned; repeat the support calculation and pruning process until no new frequent item sets can be generated; generate a new candidate item set.

[0097] Generate association rules from frequent item sets. The form of association rules is "X→Y", where X and Y are subsets of frequent item sets, and , that is, X and Y have no common elements. Traverse all frequent item sets and try to generate all possible association rules. For example, from the frequent item set {car, person A}, we can generate association rules "{person A}→{car}" and "{car}→{person A}".

[0098] Step 133: For each association rule "X→Y", Calculate its confidence, where is the frequency with which X and Y appear simultaneously, is the frequency of X, The confidence factor is the probability that a transaction containing X also contains Y. The confidence factor is defined as the probability that a transaction containing X also contains Y. This is the ratio of support (X∪Y) to support (X). The confidence factor reflects the credibility or reliability of a rule. For example, if the confidence factor for the rule "{person A} → {car}" is high, it indicates that person A is more likely to ride in a car when traveling. A minimum confidence threshold is set to filter meaningful association rules. Only association rules with a confidence factor greater than or equal to the minimum confidence threshold are considered valid.

[0099] Analyze the selected valid association rules, focusing on rules related to vehicle type. For example, by analyzing rules of the type "{personnel type X} → {vehicle type Y}", we can understand the preferences of different personnel types for vehicle types.

[0100] In a preferred embodiment of the present invention, the above step 14 may include:

[0101] Step 141, pre-processing the traffic data of people and vehicles at the airport crossing and extracting features of the pre-processed data;

[0102] Step 142 , using similarity or distance measurement between data features to divide the data points into different clusters, each cluster representing a traffic pattern or group feature, to obtain a clustering result;

[0103] Step 143, pass Calculate the intra-cluster density of the clustering results by Inter-cluster separation, identifying different traffic patterns and group characteristics, among which, It is A cluster contains a group of data points with similar characteristics; It is clusters, It is a cluster A data point within It is a cluster A data point within is the degree of separation between clusters, is the intra-cluster compactness, is the total number of features, and They are in the The value of a feature.

[0104] In embodiments of the present invention, the preprocessing step can remove noise, missing values, and outliers from the data, making the data cleaner and more accurate. Furthermore, standardization can ensure that different features have the same scale. By extracting data features, the raw data can be converted into more representative feature vectors, reducing the data dimension and improving the processing efficiency of subsequent algorithms. Cluster analysis can identify different traffic patterns or group characteristics, such as traffic patterns during peak hours or traffic patterns of specific vehicle types. Clustering results can group similar data points together, facilitating data grouping and summarization, and helping to discover patterns and trends in the data. By calculating intra-cluster compactness and inter-cluster separation, the quality of the clustering results can be evaluated to ensure that data points within each cluster are highly similar, while data points between different clusters are highly different. The calculation of intra-cluster compactness and inter-cluster separation helps to more accurately identify different traffic patterns and group characteristics, avoiding misjudgments or omissions. Based on the identified traffic patterns and group characteristics, airports can optimize resource allocation, such as adjusting road crossing layout, increasing traffic capacity during specific periods, and optimizing parking resource allocation, thereby improving airport operational efficiency and service quality.

[0105] In the present invention, the specific steps include:

[0106] Step 141, obtain the traffic data of people and vehicles. The data should include key information such as timestamp, person identity, vehicle type, traffic direction, and traffic speed. Preprocess the data and extract the preprocessed data. Extract time features such as hours, minutes, day of the week, and whether it is a holiday from the timestamp. If the data contains traffic location information, the location coordinates or area code can be extracted as spatial features. Based on the traffic records of people and vehicles, extract behavioral features such as traffic frequency, average traffic time, and traffic speed. For vehicles, you can also extract features such as vehicle type and load; for people, you can extract features such as whether they carry luggage and the frequency of changes in traffic direction. Combine the above features to construct more complex features, such as "peak hour traffic speed" and "average traffic time for specific vehicle types" to capture richer traffic patterns.

[0107] Step 142, based on the characteristics of the data and the clustering requirements, select the K-means clustering algorithm and select an appropriate distance metric to calculate the similarity or distance between data points. Input the pre-processed data into the selected clustering algorithm and perform the clustering operation. The algorithm will divide the data points into different clusters based on the similarity or distance metric between the data points. Each cluster represents a traffic pattern or group feature, and is divided into different clusters by The silhouette coefficient is obtained, where is the silhouette coefficient, It is A cluster contains a group of data points with similar characteristics; It is clusters, It is a cluster A data point within It is a cluster A data point within is the total number of features, and They are in the The value of the feature, is a data point The cluster contains a group of data points with similar characteristics. ∣ is a cluster The number of data points in Except Other clusters outside is the total number of features, and The data points are and No. eigenvalues, is the eigenvalue index; the value range of the silhouette coefficient is The silhouette coefficient is used to evaluate the quality of clustering results.

[0108] Step 143: Calculate the average distance or similarity between data points within each cluster as a measure of intra-cluster compactness. The higher the intra-cluster compactness, the more similar the data points within the cluster, and the more distinct the traffic pattern or group characteristics. Calculate the average distance or similarity between different clusters as a measure of inter-cluster separation. The higher the inter-cluster separation, the greater the difference between the data points of different clusters, and the easier it is to distinguish the traffic pattern or group characteristics. Based on the calculation results of the intra-cluster compactness and inter-cluster separation, different traffic patterns and group characteristics are identified. For example, a cluster may represent a "group of cars traveling quickly during peak hours," with high intra-cluster compactness and large separation from other clusters; another cluster may represent a "group of trucks traveling slowly during non-peak hours," with different characteristic performances.

[0109] In a preferred embodiment of the present invention, the above step 15 may include:

[0110] Step 151, collecting personnel and vehicle flow data at the airport intersection and preprocessing the data;

[0111] Step 152: Train the One-Class SVM algorithm based on the pre-processed normal traffic behavior data. During the training process, the algorithm learns the characteristic distribution of normal traffic behavior and constructs a hypersphere that encloses the normal data points. The boundary of the hypersphere represents the boundary between normal and abnormal traffic behavior.

[0112] In step 153, the traffic behavior data of people and vehicles to be detected are passed through the trained One-ClassSVM algorithm to determine whether each data point is located inside the hypersphere. If the data point is outside the hypersphere, it is regarded as abnormal traffic behavior, so as to detect abnormal traffic behavior of people and vehicles and obtain abnormal behavior characteristics.

[0113] In an embodiment of the present invention, data on passenger and vehicle traffic at airport crossings is collected to provide a comprehensive understanding of the crossing's traffic conditions. Data preprocessing, such as denoising, normalization, and missing value processing, improves data accuracy and consistency. The One-Class SVM algorithm, trained on normal traffic behavior data, learns the characteristic distribution of normal behavior, thereby constructing a hypersphere that encloses normal data points. The hypersphere's boundary represents the dividing line between normal and abnormal traffic behavior. This boundary effectively detects abnormal traffic behavior that deviates from normal patterns. For data points detected as abnormal, the specific manifestations of the abnormal behavior are understood. Through anomaly detection, airports can promptly identify and address potential safety hazards, such as illegal intrusions and abnormal parking, thereby enhancing the airport's security and prevention capabilities.

[0114] In the present invention, the specific steps include:

[0115] Step 151: Collect the passenger and vehicle flow data at the airport crossing and pre-process the data. Select key features: Based on the analysis requirements, extract the features that have a significant impact on traffic behavior, such as travel time, vehicle type, travel speed, whether it is peak hours, etc., and construct combined features, such as "peak hour travel speed" and "average travel time for specific vehicle types", to capture more complex traffic behavior patterns.

[0116] Step 152: Divide the preprocessed data into a training set and a validation set. The training set is used to train the One-ClassSVM algorithm, which is selected as the anomaly detection tool. Select the radial basis function and set the kernel function parameters, such as the γ value of the RBF kernel; set the penalty parameter ν, which controls the trade-off between the proportion of misclassified samples in the training set and the volume of the hypersphere. Use the training set data to train the One-ClassSVM algorithm. The algorithm will learn the characteristic distribution of normal traffic behavior and construct a hypersphere that can surround normal data points. The boundary of the hypersphere represents the dividing line between normal traffic behavior and abnormal traffic behavior. The trained One-ClassSVM model is evaluated using the validation set data.

[0117] In step 153, the traffic behavior data of people and vehicles to be tested is collected and preprocessed and extracted in the same way as the training data. The data is then fed into the trained One-Class SVM model to determine whether each data point lies within the hypersphere. If the data point lies within the hypersphere, it is considered normal traffic behavior; if it lies outside the hypersphere, it is considered abnormal traffic behavior. For data points detected as abnormal, relevant feature information is extracted, such as timestamp, vehicle type, speed, and degree of abnormality (i.e., the distance from the data point to the hypersphere boundary).

[0118] In a preferred embodiment of the present invention, the above step 16 may include:

[0119] Step 161 , integrating the peak hour prediction, vehicle type traffic preference, traffic pattern and group characteristics, and abnormal behavior characteristics to obtain integrated data;

[0120] Step 162: Based on the results of data integration and analysis, a preliminary personnel and vehicle dispatching plan is developed. This includes arranging additional crossing staff or adjusting vehicle travel times based on peak hour forecasts, optimizing the travel order of different vehicle types or allocating dedicated lanes based on their travel preferences, and developing targeted travel strategies based on travel patterns and group characteristics.

[0121] Step 163, adjusting the preliminary scheduling plan according to the abnormal behavior characteristics to obtain a scheduling plan;

[0122] Step 164 , optimize the traffic route by simulating and evaluating the scheduling plan.

[0123] In an embodiment of the present invention, by integrating multiple data types, including peak hour predictions, vehicle type preferences, travel patterns and group characteristics, and abnormal behavior characteristics, a comprehensive, multi-dimensional dataset can be formed. This integrated data can fully leverage the connections and complementarities between various data types, improve overall data utilization, and avoid the limitations of a single data source. Arranging additional crossing staff or adjusting vehicle travel times based on peak hour predictions can ensure sufficient resources to cope with traffic pressure during peak hours and improve traffic efficiency. Optimizing the travel order or allocating dedicated lanes for different vehicle types based on their travel preferences can reduce vehicle wait times and improve overall traffic efficiency. Developing targeted travel strategies based on travel patterns and group characteristics can better meet the needs of different groups and vehicle types, improving service quality and satisfaction. Adjusting scheduling plans based on abnormal behavior characteristics can promptly identify and address potential safety hazards, such as illegal intrusions and abnormal parking, ensuring the safety of airport crossings. The adjusted scheduling plans are more flexible and adaptable, better able to handle emergencies and abnormal events, and enhance the airport's emergency response capabilities. By simulating the implementation effect of the scheduling plan, the feasibility and effectiveness of the plan can be verified. By optimizing the traffic route based on the simulation results, the vehicle driving distance and waiting time can be reduced, and the traffic efficiency and service quality can be improved.

[0124] In the present invention, the specific steps include:

[0125] Step 161 collects the peak hour prediction results, vehicle type traffic preference analysis, traffic pattern and group feature identification results, and abnormal behavior feature data obtained from the previous analysis, and merges the standardized data according to timestamps or related features to form an integrated data set.

[0126] Step 162: Based on the peak hour prediction results, additional crossing staff are assigned to the predicted peak hours to accommodate increased passenger volume. Vehicle travel times are adjusted, for example, by encouraging off-peak travel or implementing staggered travel policies to distribute traffic pressure during peak hours. The order of travel for different vehicle types is optimized based on their preferences. For example, emergency vehicles can be given the highest priority, and dedicated lanes can be allocated to specific vehicle types, such as large trucks and buses, to improve their efficiency. Targeted traffic strategies are developed based on identified traffic patterns and group characteristics. For example, for passengers who frequently carry large amounts of luggage, dedicated baggage inspection lanes or additional assistance services can be provided. The number of crossings open and staff allocation can be flexibly adjusted based on changing traffic patterns during different time periods.

[0127] Step 163 adjusts the initial dispatch plan based on the analysis of abnormal behavior. For example, if abnormal behaviors such as frequent vehicle congestion or passenger delays occur during a certain time period, the number of crossing staff at that time period can be increased or the vehicle traffic order can be adjusted to alleviate congestion. For serious abnormal behaviors such as malicious violations or sabotage, stricter control measures or enhanced security monitoring can be implemented.

[0128] Step 164 uses traffic simulation software or a mathematical model to simulate the traffic environment at the airport intersection. The adjusted scheduling plan is input into the simulation environment, and the intersection operation under different time periods and traffic flows is simulated. The simulation environment is run, and key data such as vehicle travel time, passenger waiting time, and intersection utilization are collected during the simulation. Based on the simulation evaluation results, the traffic route is optimized. For example, lane layout and arrangement can be adjusted, traffic signal timing can be optimized, or temporary routes can be set up to accommodate special circumstances. The optimized route and scheduling plan are simulated and evaluated again to verify the effectiveness of the improvements.

[0129] In a preferred embodiment of the present invention, the above step 17 may include:

[0130] Step 171 , planning the layout and functions of the electronic display screen and voice prompt system based on the optimized route;

[0131] Step 172, setting up an electronic display screen and voice prompt system according to the actual conditions of the optimized route and the airport crossing;

[0132] Step 173 , providing traffic guidance and instructions based on the electronic display screen and voice prompt system to manage people and vehicles at the airport crossing. The traffic guidance and instructions include route instructions, traffic information updates, safety tips, and evacuation instructions in emergency situations.

[0133] In embodiments of the present invention, by rationally planning the layout of electronic display screens and voice prompt systems, accurate and timely information can be conveyed to personnel and vehicles at airport crossings, improving information transmission efficiency. By designing the functions of the electronic display screens and voice prompt systems based on optimized routes, such as providing clear route instructions and real-time traffic information updates, the user experience (including passengers and drivers) can be significantly enhanced. Setting up the electronic display screens and voice prompt systems based on the actual conditions of the airport crossing ensures that the systems are compatible with the layout, traffic flow, and environmental conditions of the crossing, improving the system's practicality and effectiveness. Through precise setup and debugging, the stable operation of the electronic display screens and voice prompt systems can be ensured, minimizing the occurrence of faults and error messages, and improving system reliability. Clear route instructions and traffic information updates provided by the electronic display screens and voice prompt systems can guide personnel and vehicles along the optimal routes, reducing congestion and wait times, and optimizing the passage process. The traffic guidance and prompting functions provided by the system help airport management departments better understand the flow of personnel and vehicles at the crossing, improving management efficiency, such as enabling timely adjustments to crossing staff deployment and responding to emergencies.

[0134] In the present invention, the specific steps include:

[0135] Step 171: Based on the optimized route obtained through simulation and evaluation, identify areas requiring special attention or guidance, such as turns, confluences, divergence points, and emergency exits. Based on the characteristics of the optimized route and the actual conditions of the airport crossing, determine the layout principles for the electronic display screen and voice prompt system. Electronic display screens are set up at key nodes and areas requiring guidance, such as crossing entrances, exits, turns, and important intersections. Plan the display screen content, such as route instructions, traffic information updates, and safety reminders, and set different display modes as needed. Set up voice prompts in areas requiring voice guidance, such as near turns, emergency exits, and construction areas. Plan the content of voice prompts, such as route instructions, safety warnings, and emergency evacuation instructions, and set different prompt frequencies and methods as needed.

[0136] Step 172, conduct an on-site survey of the airport crossing to understand the actual conditions such as terrain, landform, and traffic flow, conduct specific measurements and positioning according to the layout plan, determine the installation location of the electronic display screen and voice prompt equipment, and install the electronic display screen and voice prompt equipment according to the layout plan and installation location.

[0137] Step 173 uses electronic display screens to update traffic information in real time, such as road conditions, weather, and flight information, to provide accurate traffic guidance for drivers and pedestrians. Route instructions and safety tips are played through a voice prompt system to guide drivers and pedestrians along optimized routes. Traffic information on the electronic display screens is adjusted promptly based on changes in traffic flow, such as congestion warnings and detour suggestions. When approaching dangerous areas or special road sections, safety tips and warnings, such as speed limit warnings, no parking warnings, and construction zone warnings, are issued through the electronic display screens and voice prompt system to remind drivers and pedestrians to be safe. In the event of an emergency, the emergency evacuation mode of the electronic display screens and voice prompt system is immediately activated, issuing evacuation instructions and escape route directions to guide drivers and pedestrians to evacuate the scene quickly and orderly. The system is upgraded and optimized based on the development needs of the airport crossing and changes in traffic conditions to improve the accuracy and effectiveness of traffic guidance and instructions.

[0138] like Figure 2 As shown, an embodiment of the present invention further provides an airport crossing person and vehicle management system 20, comprising:

[0139] The acquisition module 21 is used to identify vehicle license plate information through image processing technology and compare facial features of people to obtain traffic data of people and vehicles at the road crossing;

[0140] Identification module 22 is used to construct an LSTM neural network based on the traffic flow data of people and vehicles at the road crossing, and analyze the temporal trend of traffic flow at the road crossing using the LSTM neural network to predict peak hours; use the Apriori algorithm to mine the association rules between the traffic behavior of people and vehicles to obtain the traffic preferences of vehicle types; cluster people and vehicles to identify different traffic patterns and group characteristics;

[0141] The optimization module 23 is used to detect abnormal traffic behavior of people and vehicles in traffic data using the One-Class SVM algorithm and obtain abnormal behavior characteristics; dispatch people and vehicles and optimize traffic routes based on predicted peak hours, traffic preferences of vehicle types, traffic patterns, group characteristics, and abnormal behavior characteristics; and set up electronic display screens and voice prompt systems based on the optimized routes to provide traffic guidance and instructions for people and vehicles, thereby achieving personnel and vehicle management at airport crossings.

[0142] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for managing people and vehicles at an airport crossing, characterized in that: The method comprises: By using image processing technology to identify vehicle license plate information and compare facial features of people, traffic data of people and vehicles at the intersection can be obtained; Based on the traffic data of people and vehicles at the road crossing, an LSTM neural network is constructed. The LSTM neural network is then used to analyze the temporal trend of traffic flow at the road crossing to predict peak hours. The Apriori algorithm is used to mine association rules between the travel behaviors of people and vehicles, and to obtain the travel preferences of vehicle types. This includes: using traffic data of people and vehicles, constructing transaction itemsets for the Apriori algorithm, where each transaction can be represented as a combination of people and vehicle travel behaviors; by iteratively generating candidate itemsets, obtaining support, and pruning infrequent itemsets, all frequent itemsets that meet the minimum support threshold are obtained, and association rules are screened out; the confidence of each rule is calculated to obtain the travel preferences of vehicle types; Clustering people and vehicles to identify different traffic patterns and group characteristics, including: preprocessing the traffic data of people and vehicles at airport intersections and extracting features from the preprocessed data; dividing the data points into different clusters based on similarity or distance between the data features, with each cluster representing a traffic pattern or group characteristic, to obtain clustering results; and identifying different traffic patterns and group characteristics by calculating the intra-cluster compactness and inter-cluster separation of the clustering results. The One-Class SVM algorithm is used to detect abnormal traffic behavior of people and vehicles and obtain abnormal behavior characteristics; Dispatch personnel and vehicles to optimize travel routes based on predicted peak hours, vehicle type preferences, travel patterns, group characteristics, and abnormal behavior characteristics; According to the optimized route, electronic display screens and voice prompt systems are set up to provide traffic guidance and instructions for people and vehicles, so as to achieve the management of people and vehicles at the airport crossing.

2. The method for managing people and vehicles at an airport crossing according to claim 1, characterized in that: Image processing technology is used to identify vehicle license plate information and compare facial features of people to obtain traffic data of people and vehicles at the crossing, including: Install high-definition cameras at airport crossings to capture images of people and vehicles passing through the crossings; Use image processing technology to perform license plate recognition on the collected vehicle images, count the number of vehicles passing and the time, and obtain the recognition results; Extract and compare facial features of the collected person images to obtain facial feature comparison results; Based on the results of vehicle license plate recognition and facial feature comparison, the personnel and vehicle flow data at the crossing are counted, including the number of vehicles and people passing through the crossing per unit time.

3. The method for managing people and vehicles at an airport crossing according to claim 2, characterized in that: Based on the traffic data of people and vehicles at the road crossing, an LSTM neural network is constructed. The LSTM neural network is then used to analyze the temporal trend of traffic flow at the road crossing to predict peak hours, including: Based on the traffic data of personnel and vehicles at the road crossing, the foundation of the LSTM neural network is constructed; The LSTM neural network is trained using historical training data. During the training process, appropriate hyperparameters are set and the network parameters are adjusted using the backpropagation algorithm to minimize the error between the predicted value and the true value, resulting in a trained LSTM neural network. The hyperparameters include the learning rate, batch size, and number of training rounds. The new traffic flow data is predicted through the trained LSTM neural network to obtain the predicted peak hours of traffic flow.

4. The method for managing people and vehicles at an airport crossing according to claim 3, characterized in that: The One-Class SVM algorithm is used to detect abnormal traffic behavior of people and vehicles and obtain abnormal behavior characteristics, including: Collect personnel and vehicle flow data at airport intersections and pre-process the data; The One-Class SVM algorithm is trained based on preprocessed normal traffic behavior data. During training, the algorithm learns the characteristic distribution of normal traffic behavior and constructs a hypersphere that encloses the normal data points. The boundary of the hypersphere represents the dividing line between normal and abnormal traffic behavior. The traffic behavior data of people and vehicles to be detected are passed through the trained One-ClassSVM algorithm to determine whether each data point is located inside the hypersphere. If the data point is outside the hypersphere, it is regarded as abnormal traffic behavior, so as to detect abnormal traffic behavior of people and vehicles and obtain abnormal behavior characteristics.

5. The method for managing people and vehicles at an airport crossing according to claim 4, characterized in that: Based on the predicted peak hours, traffic preferences of vehicle types, traffic patterns, group characteristics, and abnormal behavior characteristics, personnel and vehicles are dispatched to optimize traffic routes, including: Integrate peak hour predictions, vehicle type preferences, traffic patterns and group characteristics, and abnormal behavior characteristics to obtain integrated data; Based on the results of data integration and analysis, preliminary personnel and vehicle scheduling plans are developed. This includes arranging additional crossing staff or adjusting vehicle travel times based on peak hour forecasts, optimizing the travel order of different vehicle types or allocating dedicated lanes based on their travel preferences, and developing targeted travel strategies based on travel patterns and group characteristics. According to the preliminary scheduling plan, the preliminary scheduling plan is adjusted according to the abnormal behavior characteristics to obtain the scheduling plan; Optimize traffic routes by simulating and evaluating scheduling plans.

6. An airport crossing personnel and vehicle management system, the system implementing the method according to any one of claims 1 to 5, characterized in that: include: The acquisition module is used to identify vehicle license plate information through image processing technology and compare facial features of people to obtain traffic data of people and vehicles at the intersection; The recognition module is used to build an LSTM neural network based on the traffic data of people and vehicles at the road crossing, and analyze the time variation trend of traffic flow at the road crossing based on the LSTM neural network to predict the peak period; Through the Apriori algorithm, the association rules between people and vehicles' traffic behaviors are mined to obtain traffic preferences of vehicle types; people and vehicles are clustered to identify different traffic patterns and group characteristics; The optimization module is used to detect abnormal traffic behavior of people and vehicles in traffic data through the One-Class SVM algorithm and obtain abnormal behavior characteristics; it dispatches people and vehicles and optimizes traffic routes based on predicted peak hours, traffic preferences of vehicle types, traffic patterns, group characteristics and abnormal behavior characteristics; and it sets up electronic display screens and voice prompt systems based on the optimized routes to guide and instruct people and vehicles to achieve people and vehicle management at airport crossings.

7. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, which implements the method according to any one of claims 1 to 5 when executed by a processor.

Images