Vehicle-road cooperative scheduling method and system between bus and station

Abstract

The invention discloses a vehicle-road cooperative scheduling method and system between a bus and a station, and the method comprises the steps: collecting data in real time through vehicle-road cooperative equipment at a bus side and a station side, obtaining the collected data, carrying out the preprocessing and fusion of the collected data, and obtaining a fused data set; constructing a safety early warning model of the bus and the station based on the fused data set, and identifying safety risks and generating safety early warning information by using the safety early warning model; according to the invention, efficient, safe and intelligent monitoring and scheduling functions are provided for urban public transportation, so that the travel experience of passengers can be improved, the bus operation cost can be reduced, and the sustainable development of the urban transportation can be promoted; and meanwhile, potential safety hazards and operation bottlenecks can be automatically identified through fusion of video data and radar data, so that safety early warning and scheduling optimization suggestions are generated, the safety of passengers and vehicles is ensured, the bus service efficiency and quality are improved, and the system is suitable for being widely popularized and used.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a vehicle-road collaborative scheduling method and system between a bus and a platform. Background Art

[0002] The public transportation system includes multiple modes, such as rail transportation, road transportation, water transportation and air transportation. These different modes of transportation together constitute a multi-level and extensive public transportation network to meet the travel needs of different regions and people. Public transportation is not only an important link connecting various areas of the city, but also an effective means to alleviate traffic congestion and reduce environmental pollution. In order to improve the efficiency and attractiveness of public transportation, many cities have taken measures such as setting up bus lanes, implementing regional rail commuting and promoting new energy buses. At the same time, some cities have further improved the service level and passenger volume of public transportation through intelligent technology.

[0003] At present, with the acceleration of urbanization and the continuous growth of public transportation demand, the operation efficiency and safety of the bus system have received increasing attention; the existing bus scheduling methods mostly rely on manual experience and limited real-time data, which makes it difficult for bus scheduling decisions to cope with complex and changeable traffic environments, and the current information island phenomenon between buses, platforms and vehicle-road cooperative equipment is serious. At the same time, the acquisition of road condition information and road surface status is delayed. Not only is there a lack of unified monitoring data management and security big data analysis platform, but also the equipment lacks real-time monitoring, and the bus scheduling decision lacks scientific basis and real-time performance, resulting in low bus punctuality and high average vehicle delays, which in turn cannot improve the road network operation efficiency, and the visual blind spots of buses and platforms are prone to safety hazards, which makes the safety of passengers unable to be guaranteed; therefore, it is necessary to design a vehicle-road cooperative scheduling method and system between buses and platforms. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art, in order to better and effectively solve the existing bus scheduling methods, which mostly rely on manual experience and limited real-time data, resulting in bus scheduling decisions being difficult to cope with complex and changeable traffic environments, and the current information island phenomenon between buses, platforms and vehicle-road cooperative equipment is serious, and the road condition information and road surface status are acquired with lags, not only lacking a unified monitoring data management and security big data analysis platform, but also resulting in a lack of real-time monitoring of equipment, a lack of scientific basis and real-time performance of bus scheduling decisions, resulting in low bus punctuality rate and high average vehicle delay, and thus unable to improve the road network operation efficiency, and the visual blind spots of buses and platforms are prone to potential safety hazards, which makes it impossible to ensure the safety of passengers. A vehicle-road cooperative scheduling method and system between buses and platforms is provided, which realizes the monitoring and scheduling functions of providing efficient, safe and intelligent urban public transportation, which not only helps to improve passengers' travel experience and reduce bus operating costs, but also promotes the sustainable development of urban transportation, and at the same time, by integrating video data with radar data, it can automatically identify safety hazards and operational bottlenecks to generate safety warnings and scheduling optimization suggestions, thereby ensuring the safety of passengers and vehicles and improving the efficiency and quality of bus services.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: A vehicle-road coordinated dispatching method between a bus and a platform comprises the following steps: Step A, using the vehicle-road cooperative equipment on the bus side and the platform side to collect data in real time and obtain collected data; Step B, preprocessing and fusing the collected data to obtain a fused data set; Step C: construct a safety warning model for buses and stations based on the fused data set, and then use the safety warning model to identify safety risks and generate safety warning information; Step D: Building a remote monitoring platform based on the safety warning information to remotely monitor the hardware equipment on the bus side and the platform side; In step E, safety warning information is issued to passengers on the bus and platform sides, and then the dispatch center outputs and executes dispatch optimization decisions based on the safety warning information and real-time data to complete the vehicle-road collaborative dispatch operation.

[0006] The aforementioned vehicle-road collaborative scheduling method between a bus and a platform, step A, uses the vehicle-road collaborative equipment on the bus side and the platform side to collect data in real time to obtain collected data, wherein the collected data includes video surveillance data, GIS map data, passenger flow analysis data, equipment status monitoring data, driving road congestion data, and bus basic data. The platform side uses high-definition cameras, millimeter-wave radars, edge computing units and temperature sensing devices to collect roadside data and receive on-board data transmitted by the bus side's on-board OBU. The data collected on the platform side includes platform side waiting passenger flow data, video surveillance data, radar point cloud data, traffic data around the platform, road condition information data, non-motorized vehicle lane pedestrian and vehicle data, weather temperature data and platform water and snow data. The road condition information data includes platform road congestion, road traffic flow, platform pavement conditions and construction conditions. The video surveillance data includes platform motor vehicle lane video surveillance data and non-motor vehicle lane video surveillance data. The bus basic data includes vehicle number, license plate number, driver information and vehicle equipment information.

[0007] In the aforementioned vehicle-road coordinated dispatching method between a bus and a station, step B pre-processes and fuses the collected data to obtain a fused data set. The specific steps are as follows: Step B1, preprocessing the collected data and obtaining preprocessed data, wherein the preprocessing specifically includes data cleaning, data conversion, data standardization, data compression and encryption, and data integration and storage; Step B2, performing data fusion on the preprocessed data and obtaining a fused data set, wherein the specific fusion contents include spatiotemporal data fusion, convolutional neural network data fusion and radar multimodal data fusion.

[0008] In the aforementioned vehicle-road coordinated dispatching method between a bus and a station, the specific steps of data preprocessing in step B1 are as follows: Step B11, clean the collected data and obtain the cleaned data. The specific steps are as follows: Step B111, backing up the video surveillance data, GIS map data, passenger flow analysis data, equipment status monitoring data, road congestion data, bus basic data and platform passenger flow data, video surveillance data, radar point cloud data, platform surrounding traffic data, road condition information data, non-motorized vehicle lane pedestrian and vehicle data, weather temperature data, platform water and snow data in the collected data; Step B112, deduplication processing is performed on the collected data, specifically, using a set to identify and delete identical data records, wherein records with different timestamps but the same other information are deduplicated by setting a threshold. Specifically, the timestamp of data record i is set to T i , the timestamp of data record j is T jand the threshold is ΔT1, if |T i -T j | ≤ ΔT1, then data record i and data record j are determined to be duplicate data. If |T i -T j | >ΔT1, then data record i and data record j are determined to be non-duplicate data; Step B113, performing missing value processing on the collected data, specifically filling the numerical data with the mean value and filling the categorical data with the mark; Step B114, detect and process outliers on the collected data, specifically using the IQR method to identify outliers. The IQR method sets a threshold ΔT2 based on business logic to determine outliers and delete the determined outliers. The IQR is specifically the difference between the third quartile Q3 and the first quartile Q1 in the data, as shown in formula (1). IQR = Q3 − Q1 (1) Among them, the abnormal values ​​are less than Q1 – ΔT2×IQR and greater than Q3 + ΔT2×IQR; Step B12, convert and standardize the cleaned data to obtain data in a unified format. The specific steps are as follows: Step B121, converting the cleaned data, specifically, unifying the data source date, time and format of different platforms and devices in the cleaned data into a standard date and time format, wherein the text data is converted using a coding conversion method; Step B122, standardize the cleaned data. Specifically, the numerical data is standardized by using the min-max standardization method, and the categorical data is standardized by using the label encoding method. The specific min-max standardization method is shown in formula (2): (2) in, is the min-max standard value, is the original value, is the minimum value, is the maximum value, is the coefficient; Step B123, discretizing the cleaned data, specifically using an equal-width binning method to discretize the continuous data; Step B13, compress and encrypt the data after format unification to obtain compressed and encrypted data. The specific steps are as follows: Step B131, compressing the format-unified data to obtain compressed data, wherein the video surveillance data is compressed using a video compression standard, and the data information is compressed using a GZIP compression algorithm; Step B132, encrypt the compressed data, specifically using AES+RSA hybrid encryption algorithm to encrypt the data, the specific steps are as follows: Step B1321, using the RSA algorithm to generate a public key and a private key, the public key is used to encrypt data, and the private key is used to decrypt data; Step B1322, the sender generates an original AES key, and then uses the original AES key to symmetrically encrypt the compressed data to obtain symmetrically encrypted data. Then, the sender uses the RSA public key to asymmetrically encrypt the original AES key to obtain an encrypted AES key, and then sends the encrypted AES key and the symmetrically encrypted data to the receiver. Step B1323, the receiver receives the encrypted AES key and the symmetrically encrypted data, uses the RSA private key to decrypt the AES key and restore the original AES key, and then uses the original AES key to decrypt the symmetrically encrypted data to restore the compressed data; Step B14, integrating and storing the compressed and encrypted data. The specific steps are as follows: Step B141, integrating the compressed and encrypted data into logically interrelated data sets according to the data types, wherein the logically interrelated data sets include video surveillance data sets, GIS map data sets, passenger flow analysis data sets, equipment condition monitoring data sets, traffic condition data sets, radar point cloud and vehicle flow data sets, road condition information and non-motorized lane data sets, and weather and environment data sets; Step B142, selecting a database to store the integrated data, wherein the database is divided into a relational database MySQL and a non-relational database MongoDB.

[0009] In the aforementioned vehicle-road coordinated dispatching method between a bus and a station, the specific steps of data fusion in step B2 are as follows: Step B21, performing spatiotemporal data fusion on the preprocessed data and obtaining spatiotemporal fused data, wherein the spatiotemporal data fusion is used to make the preprocessed data accurate to milliseconds in time and centimeters in space, thereby supporting coordinated safety monitoring and intelligent analysis-assisted scheduling of buses and stations. The specific steps are as follows: Step B211, using the Network Time Protocol (NTP) to synchronize the timestamps of the pre-processed data, and then correcting the timestamps of different devices to ensure consistency on the time axis; Step B212, the pre-processed data is subjected to spatial coordinate transformation to conform to the unified geographic information system (GIS) standard. The specific spatial coordinate transformation formula is shown in formula (3): Xnew = Xold*cos(α2) - Yold*sin(α2) + translation X, Ynew= Xold*sin(α2) + Yold*cos(α2) + translation Y(3) Among them, α2 is the rotation angle, Xold and Yold are the original coordinates, and Xnew and Ynew are the transformed coordinates; Step B22, performing convolutional neural network data fusion on the preprocessed data and obtaining convolutional neural network fused data, wherein the convolutional neural network data fusion specifically uses the convolutional neural network deep learning model to extract and analyze features of image data and video data to support coordinated safety monitoring, intelligent analysis, situation awareness and auxiliary scheduling of buses and platforms. The specific steps are as follows: Step B221, adjust and normalize the image data and video data to obtain the adjusted and normalized data. The specific steps are as follows: Step B2211, resizing the image data and the video data and obtaining resized image and video data, wherein the resizing specifically refers to adjusting the image size and the video frame to a uniform size; Step B2212, normalize the resized image and video data to obtain normalized image and video data, as shown in formula (4), (4) in, is the normalized pixel value, is the original pixel value, and are the minimum and maximum pixel values, respectively. is the adjustment coefficient; Step B2213, performing image enhancement on the pixel values ​​of the normalized image video data, wherein the specific content of the image enhancement is image rotation, image flipping or image cropping; The image rotation is used to enable the convolutional neural network deep learning model to learn the features of the object at different angles. Specifically, the image rotation uses a neighbor interpolation method to maintain the image boundary information, and then limits the rotation angle range according to different scene characteristics; The image flipping is used to enable the convolutional neural network deep learning model to learn the features of the object in the mirror direction. The image flipping is specifically divided into a horizontal flipping mode and a vertical flipping mode. The horizontal flipping mode specifically flips the image along the vertical central axis and generates a horizontally flipped image. The vertical flipping mode specifically flips the image along the horizontal central axis and generates a vertically flipped image. The image cropping is used to make the convolutional neural network deep learning model focus on the visual focus area in the image. The image cropping is specifically to crop the image using the rule of thirds, wherein the rule of thirds is to divide the photo into three parts in the vertical and horizontal directions, and then crop the four points where the lines intersect to form the visual focus; Step B222, based on the adjusted and normalized data, use the convolutional neural network deep learning model to extract and analyze the features of the image data and video data and obtain the convolutional neural network fused data. The specific steps are as follows: Step B2221, constructing a convolutional neural network deep learning model, wherein the convolutional neural network deep learning model is used to identify and extract features in images and videos, the convolutional neural network deep learning model comprises a convolutional layer, a pooling layer and a fully connected layer, and the convolutional neural network deep learning model specifically uses back propagation and gradient descent algorithms to optimize model parameters, wherein the gradient descent algorithm is shown in formula (5), (5) Among them, θ is the model parameter, J(θ) is the loss function, is the learning rate; Step B2222, the features extracted by the convolutional neural network deep learning model are fused with the sensor data. Specifically, the features extracted by the convolutional neural network deep learning model are combined with the sensor data using a feature cascade method to obtain the fused features of the convolutional neural network, and then the support vector machine SVM is used to classify the fused features of the convolutional neural network. The specific steps are as follows: Step B22221, using a feature cascade method to combine the features extracted by the convolutional neural network deep learning model with the sensor data and obtain convolutional neural network fusion features, specifically combining features from different data sources together through the feature cascade method, wherein the feature vector extracted by the convolutional neural network deep learning model is connected to the feature vector of the sensor data to form a long feature vector; Step B22222, using support vector machine SVM to classify the features after convolutional neural network fusion, specifically using support vector machine SVM to find the optimal hyperplane to classify the features after convolutional neural network fusion, the support vector machine SVM uses Lagrange multiplier method to transform it into a dual problem and uses sequential minimum optimization SMO algorithm to solve it, as shown in formula (6), (6) in, is the hyperplane normal vector, b is the intercept, C is the regularization parameter, x i is the slack variable, y i is the sample label, is the sample characteristics; Step B23, performing radar-vision multimodal data fusion on the pre-processed data and obtaining radar-vision multimodal fusion data, wherein the radar-vision multimodal data fusion is used to integrate point cloud data and visual data for comprehensive environmental perception to support coordinated safety monitoring of buses and stations. The specific steps are as follows: Step B231, using data synchronization technology to keep the radar point cloud data and visual data consistent in time and space, specifically using the Hungarian algorithm to match the radar detection target with the object in the visual image, wherein the matching basis includes distance features, speed features, shape features and size features, and the matching process uses a cost matrix to find the minimum match so as to correspond the radar target to the visual object, and each element in the matching cost is the matching cost between a radar target and a visual object; Step B232, using a Kalman filter data fusion algorithm to fuse information provided by different sensors, wherein the Kalman filter data fusion algorithm is used to combine radar distance, speed information and visual image features when performing state measurement; Step B24, using the spatiotemporal fused data, the convolutional neural network fused data and the radar multimodal fused data to perform target detection, target tracking and target classification and obtain a fused data set, the target detection specifically uses the deep learning-based target detection framework Faster R-CNN to detect and locate the target to obtain the detection target data, the target tracking specifically uses the multi-target tracking algorithm MOT to continuously monitor the movement of the target in time and space to obtain the tracking target data, and the target classification specifically classifies the detection target data and the tracking target data to distinguish between pedestrians, electric vehicles and cars.

[0010] In the aforementioned vehicle-road coordinated dispatching method between a bus and a platform, step C constructs a safety warning model for the bus and the platform based on the fused data set, and then uses the safety warning model to identify safety risks and generate safety warning information. The specific steps are as follows: Step C1, segmenting the fused data set and obtaining segmented data, specifically, dividing the fused data into a training set, a validation set, and a test set. Specifically, 70% of the data sources are randomly selected from the fused data as the training set, and 15% of the data sources are randomly selected from the data after the training set is extracted as the validation set, and the remaining 15% of the data sources are used as the test set; Step C2, feature selection is performed on the segmented data to extract model learning features, specifically, passenger flow density features, vehicle speed features, vehicle status features and weather condition features are selected from the training set, and then principal component analysis is performed on the passenger flow density features and vehicle speed features. The specific steps are as follows: Step C21, analyzing the correlation between each feature and the safety risk, and then using the correlation coefficient analysis method to select the passenger flow density feature, vehicle speed feature, vehicle state feature and weather condition feature; Step C22, using principal component analysis PCA to reduce the dimension of passenger flow density features and vehicle speed features and obtain features after dimension reduction, Step C23, standardizing and normalizing the passenger flow density features, vehicle speed features, vehicle state features, weather condition features, and dimensionality-reduced features so that all features are at the same scale, thereby obtaining model learning features; Step C3, building a safety warning model for buses and platforms based on model learning features. The specific steps are as follows: Step C31, constructing a safety warning model for buses and platforms according to the safety properties and data characteristics of buses and platforms, wherein the network structure of the safety warning model for buses and platforms includes an input layer, multiple convolutional layers, a pooling layer, a fully connected layer and an output layer, wherein the input layer is used to receive model learning features, the convolutional layer and the pooling layer are used for feature extraction and feature dimensionality reduction respectively, the fully connected layer is used to perform high-level abstract operations, and the output layer is used to output warning results; Step C32, training the safety warning model, specifically using the training set data to train the safety warning model, and then optimizing the performance of the safety warning model by adjusting the learning rate, convolution kernel size and activation function; Step C33, verifying the safety warning model, specifically using the verification set data to verify the accuracy index, recall index and F1 score index of the safety warning model, and then verifying the classification performance of the safety warning model by drawing a confusion matrix, and then verifying the binary classification ability of the safety warning model by using the receiver operating characteristic curve ROC and the area under the curve AUC; Step C4, using the security warning model to identify security risks and generate security warning information, the specific steps are as follows: Step C41, using the safety early warning model to identify safety risks, specifically using the verified safety early warning model of buses and platforms to identify safety hazards and determine whether there are safety risks; Step C42, generating security warning information, specifically converting the determined security risk into security warning information, wherein the specific content of the security warning information includes risk type, risk level and recommended measures.

[0011] In the aforementioned vehicle-road coordinated dispatching method between a bus and a platform, step D is to build a remote monitoring platform according to the safety warning information to remotely monitor the hardware equipment on the bus side and the platform side. The specific steps are as follows: Step D1, receiving security warning information, and then processing the security warning information and obtaining processed security warning information. The specific steps are as follows: Step D11, receiving security warning information, wherein the specific content of the security warning information includes device type, risk type, risk level, occurrence location and timestamp; Step D12, processing the security warning information and obtaining the processed security warning information, the specific steps are as follows: Step D121, removing duplicate data, redundant field data and invalid data in the security warning information, and then verifying the rationality of the timestamp and the validity of the location information in the security warning information; Step D122, classifying the security warning information according to the urgency and importance, wherein the risk categories include high risk, medium risk and low risk, and then prioritizing the security warning information according to the risk level factor, the impact scope factor and the time urgency factor; Step D2, build a remote monitoring platform, and then use the remote monitoring platform to handle faults and provide feedback based on the processed safety warning information. The specific steps are as follows: Step D21, construct a remote monitoring platform, the remote monitoring platform can receive, display and process security warning information in real time, and the remote monitoring platform can also support data visualization, equipment control and alarm notification. The specific steps are as follows: Step D211, displaying the received safety warning information in real time on the remote monitoring platform, and then judging whether it is necessary to trigger a public transportation safety warning according to the set safety threshold; Step D212: If a public transportation safety warning is triggered, the remote monitoring platform will notify the relevant staff of the public transportation safety warning information in real time by SMS, APP push and display on the monitoring center screen. After receiving the public transportation safety warning information, the relevant staff will take corresponding emergency measures to ensure the safety of passengers and vehicles. The corresponding emergency measures include adjusting the driving route, evacuating passengers and emergency stopping. Step D22, using the remote monitoring platform to perform fault processing and feedback based on the processed safety warning information, the specific steps are as follows: Step D221, using the remote monitoring platform to identify the fault type and equipment location based on the processed safety warning information, and then using remote monitoring technology to remotely diagnose and troubleshoot the faulty equipment and determine the cause of the fault and solution; Step D222, repair the faults that can be repaired remotely using the remote monitoring module, and automatically trigger an alarm notification to notify maintenance personnel to go to the site for repair if the faults need to be handled on site; Step D223, feed back the fault handling results to the remote monitoring platform in real time and update the equipment status information, then record the fault handling process and fault handling results and form a fault handling report.

[0012] In the aforementioned vehicle-road collaborative dispatching method between a bus and a platform, in step E, safety warning information is issued to passengers on the bus side and the platform side, and then the dispatching center outputs and executes the dispatching optimization decision based on the safety warning information and real-time data to complete the vehicle-road collaborative dispatching operation. The specific steps are as follows: Step E1, issuing safety warning information to passengers on the bus side and the platform side, specifically, using the on-board display screen channel, the platform display screen channel, the broadcast channel and the mobile phone APP channel to issue the safety warning information to the platform side, and then transmit it to the on-board OBU through the platform collaborative equipment and perform forward collision warning, emergency braking warning, lane change warning, blind spot warning, emergency vehicle reminder and abnormal vehicle reminder on the bus side; Step E2: The dispatch center outputs and executes dispatch optimization decisions based on safety warning information and real-time data. The dispatch optimization decisions include adjusting departure intervals, changing driving routes, and adding temporary shifts.

[0013] A vehicle-road cooperative dispatching system between a bus and a platform comprises a data acquisition module, a data processing and fusion module, a safety warning information generation module, a remote monitoring module and a dispatching output execution module, wherein the data acquisition module is used to collect data in real time using the vehicle-road cooperative equipment on the bus side and the platform side and obtain the collected data; the data processing and fusion module is used to preprocess and fuse the collected data and obtain a fused data set; the safety warning information generation module is used to construct a safety warning model for the bus and the platform based on the fused data set, and then use the safety warning model to identify safety risks and generate safety warning information; the remote monitoring module is used to construct a remote monitoring platform based on the safety warning information to remotely monitor the hardware equipment on the bus side and the platform side; the dispatching output execution module issues safety warning information to passengers on the bus side and the platform side, and then outputs and executes dispatching optimization decisions based on the safety warning information and real-time data through the dispatching center.

[0014] The aforementioned vehicle-road coordinated dispatching system between buses and platforms, the specific contents of preprocessing in the data processing and fusion module include data cleaning, data conversion, data standardization, data compression and encryption, and data integration and storage; the specific contents of fusion in the data processing and fusion module include spatiotemporal data fusion, convolutional neural network data fusion, and radar multimodal data fusion.

[0015] The beneficial effects of the present invention are as follows: a vehicle-road cooperative scheduling method and system between a bus and a platform of the present invention first utilizes the vehicle-road cooperative equipment on the bus side and the platform side to collect data in real time and obtain collected data, then pre-processes and fuses the collected data to obtain a fused data set, then builds a safety warning model for the bus and the platform based on the fused data set, then uses the safety warning model to identify safety risks and generate safety warning information, then builds a remote monitoring platform based on the safety warning information to remotely monitor the hardware equipment on the bus side and the platform side, then issues safety warning information to passengers on the bus side and the platform side, and then outputs and executes scheduling optimization decisions based on the safety warning information and real-time data through the dispatch center, thereby completing the vehicle-road cooperative scheduling operation; effectively realizing that the vehicle-road cooperative scheduling method and system between the bus and the platform have the advantages of providing services for urban public transportation Providing efficient, safe and intelligent monitoring and dispatching functions not only helps to improve passengers' travel experience and reduce bus operating costs, but also promotes the sustainable development of urban transportation. By deploying high-definition cameras, millimeter-wave radars and edge computing unit vehicle-road collaborative equipment on the platform side, bus data and real-time data on the platform side can be transmitted in real time, thereby significantly improving the accuracy and early warning capabilities of safety monitoring. After the real-time data is processed, a safety early warning platform is built for analysis, which can intuitively display the vehicle entry and exit operation trajectory, traffic flow and passenger distribution information, thereby improving the accuracy and intuitiveness of the analysis. At the same time, by integrating video data with radar data, it can automatically identify safety hazards and operational bottlenecks, thereby generating safety early warning warnings and dispatch optimization suggestions, thereby ensuring the safety of passengers and vehicles, which not only improves the efficiency and quality of bus services, but also ensures the stable and reliable operation of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of a vehicle-road cooperative scheduling method and system between a bus and a platform of the present invention. DETAILED DESCRIPTION

[0017] The present invention will be further described below in conjunction with the accompanying drawings.

[0018] like Figure 1 As shown, a vehicle-road coordinated scheduling method between a bus and a platform of the present invention comprises the following steps: Step A, using the vehicle-road cooperative equipment on the bus side and the platform side to collect data in real time to obtain collected data, wherein the collected data includes video surveillance data, GIS map data, passenger flow analysis data, equipment status monitoring data, driving road congestion data, and bus basic data. The platform side uses high-definition cameras, millimeter-wave radars, edge computing units and temperature sensing devices to collect roadside data and receive on-board data transmitted by the bus-mounted OBU. The data collected on the platform side includes platform side waiting passenger flow data, video surveillance data, radar point cloud data, traffic data around the platform, road condition information data, non-motorized vehicle lane pedestrian and vehicle data, weather temperature data and platform water and snow data. The road condition information data includes platform road congestion, road traffic flow, platform pavement conditions and construction conditions. The video surveillance data includes platform motor vehicle lane video surveillance data and non-motorized vehicle lane video surveillance data. The bus basic data includes vehicle number, license plate number, driver information and vehicle equipment information.

[0019] Step B, preprocessing and fusion of the collected data to obtain a fused data set. The specific steps are as follows: Step B1, preprocess the collected data and obtain preprocessed data, wherein the preprocessing specifically includes data cleaning, data conversion, data standardization, data compression and encryption, and data integration and storage. The specific steps are as follows: Step B11, clean the collected data and obtain the cleaned data. The specific steps are as follows: Step B111, backing up the video surveillance data, GIS map data, passenger flow analysis data, equipment status monitoring data, road congestion data, bus basic data and platform passenger flow data, video surveillance data, radar point cloud data, platform surrounding traffic data, road condition information data, non-motorized vehicle lane pedestrian and vehicle data, weather temperature data, platform water and snow data in the collected data; Step B112, deduplication processing is performed on the collected data, specifically, using a set to identify and delete identical data records, wherein records with different timestamps but the same other information are deduplicated by setting a threshold. Specifically, the timestamp of data record i is set to T i , the timestamp of data record j is T j and the threshold is ΔT1, if |T i -T j | ≤ ΔT1, then data record i and data record j are determined to be duplicate data. If |T i -T j | >ΔT1, then data record i and data record j are determined to be non-duplicate data; Step B113, performing missing value processing on the collected data, specifically filling the numerical data with the mean value and filling the categorical data with the mark; Step B114, detect and process outliers on the collected data, specifically using the IQR method to identify outliers. The IQR method sets a threshold ΔT2 based on business logic to determine outliers and delete the determined outliers. The IQR is specifically the difference between the third quartile Q3 and the first quartile Q1 in the data, as shown in formula (1). IQR = Q3 − Q1 (1) Among them, the abnormal values ​​are less than Q1 – ΔT2×IQR and greater than Q3 + ΔT2×IQR; Step B12, convert and standardize the cleaned data to obtain data in a unified format. The specific steps are as follows: Step B121, converting the cleaned data, specifically, unifying the data source date, time and format of different platforms and devices in the cleaned data into a standard date and time format, wherein the text data is converted using a coding conversion method; Step B122, standardize the cleaned data. Specifically, the numerical data is standardized by using the min-max standardization method, and the categorical data is standardized by using the label encoding method. The specific min-max standardization method is shown in formula (2): (2) in, is the min-max standard value, is the original value, is the minimum value, is the maximum value, is the coefficient; Step B123, discretizing the cleaned data, specifically using an equal-width binning method to discretize the continuous data; Step B13, compress and encrypt the data after format unification to obtain compressed and encrypted data. The specific steps are as follows: Step B131, compressing the format-unified data to obtain compressed data, wherein the video surveillance data is compressed using a video compression standard, and the data information is compressed using a GZIP compression algorithm; Step B132, encrypt the compressed data, specifically using AES+RSA hybrid encryption algorithm to encrypt the data, the specific steps are as follows: Step B1321, using the RSA algorithm to generate a public key and a private key, the public key is used to encrypt data, and the private key is used to decrypt data; Step B1322, the sender generates an original AES key, and then uses the original AES key to symmetrically encrypt the compressed data to obtain symmetrically encrypted data. Then, the sender uses the RSA public key to asymmetrically encrypt the original AES key to obtain an encrypted AES key, and then sends the encrypted AES key and the symmetrically encrypted data to the receiver. Step B1323, the receiver receives the encrypted AES key and the symmetrically encrypted data, uses the RSA private key to decrypt the AES key and restore the original AES key, and then uses the original AES key to decrypt the symmetrically encrypted data to restore the compressed data; Step B14, integrating and storing the compressed and encrypted data. The specific steps are as follows: Step B141, integrating the compressed and encrypted data into logically interrelated data sets according to the data types, wherein the logically interrelated data sets include video surveillance data sets, GIS map data sets, passenger flow analysis data sets, equipment condition monitoring data sets, traffic condition data sets, radar point cloud and vehicle flow data sets, road condition information and non-motorized lane data sets, and weather and environment data sets; Step B142, selecting a database to store the integrated data, wherein the database is divided into a relational database MySQL and a non-relational database MongoDB.

[0020] Step B2, data fusion is performed on the preprocessed data to obtain a fused data set, wherein the fusion specifically includes spatiotemporal data fusion, convolutional neural network data fusion and radar multimodal data fusion. The specific steps are as follows: Step B21, performing spatiotemporal data fusion on the preprocessed data and obtaining spatiotemporal fused data, wherein the spatiotemporal data fusion is used to make the preprocessed data accurate to milliseconds in time and centimeters in space, thereby supporting coordinated safety monitoring and intelligent analysis-assisted scheduling of buses and stations. The specific steps are as follows: Step B211, using the Network Time Protocol (NTP) to synchronize the timestamps of the pre-processed data, and then correcting the timestamps of different devices to ensure consistency on the time axis; Step B212, the pre-processed data is subjected to spatial coordinate transformation to conform to the unified geographic information system (GIS) standard. The specific spatial coordinate transformation formula is shown in formula (3): Xnew = Xold*cos(α2) - Yold*sin(α2) + translation X, Ynew= Xold*sin(α2) + Yold*cos(α2) + translation Y(3) Among them, α2 is the rotation angle, Xold and Yold are the original coordinates, and Xnew and Ynew are the transformed coordinates; Step B22, performing convolutional neural network data fusion on the preprocessed data and obtaining convolutional neural network fused data, wherein the convolutional neural network data fusion specifically uses the convolutional neural network deep learning model to extract and analyze features of image data and video data to support coordinated safety monitoring, intelligent analysis, situation awareness and auxiliary scheduling of buses and platforms. The specific steps are as follows: Step B221, adjust and normalize the image data and video data to obtain the adjusted and normalized data. The specific steps are as follows: Step B2211, resizing the image data and the video data and obtaining resized image and video data, wherein the resizing specifically refers to adjusting the image size and the video frame to a uniform size; Step B2212, normalize the resized image and video data to obtain normalized image and video data, as shown in formula (4), (4) in, is the normalized pixel value, is the original pixel value, and are the minimum and maximum pixel values, respectively. is the adjustment coefficient; Step B2213, performing image enhancement on the pixel values ​​of the normalized image video data, wherein the specific content of the image enhancement is image rotation, image flipping or image cropping; The image rotation is used to enable the convolutional neural network deep learning model to learn the features of the object at different angles. Specifically, the image rotation uses a neighbor interpolation method to maintain the image boundary information, and then limits the rotation angle range according to different scene characteristics; The image flipping is used to enable the convolutional neural network deep learning model to learn the features of the object in the mirror direction. The image flipping is specifically divided into a horizontal flipping mode and a vertical flipping mode. The horizontal flipping mode specifically flips the image along the vertical central axis and generates a horizontally flipped image. The vertical flipping mode specifically flips the image along the horizontal central axis and generates a vertically flipped image. The image cropping is used to make the convolutional neural network deep learning model focus on the visual focus area in the image. The image cropping is specifically to crop the image using the rule of thirds, wherein the rule of thirds is to divide the photo into three parts in the vertical and horizontal directions, and then crop the four points where the lines intersect to form the visual focus; Step B222, based on the adjusted and normalized data, use the convolutional neural network deep learning model to extract and analyze the features of the image data and video data and obtain the convolutional neural network fused data. The specific steps are as follows: Step B2221, constructing a convolutional neural network deep learning model, wherein the convolutional neural network deep learning model is used to identify and extract features in images and videos, the convolutional neural network deep learning model comprises a convolutional layer, a pooling layer and a fully connected layer, and the convolutional neural network deep learning model specifically uses back propagation and gradient descent algorithms to optimize model parameters, wherein the gradient descent algorithm is shown in formula (5), (5) Among them, θ is the model parameter, J(θ) is the loss function, is the learning rate; Step B2222, the features extracted by the convolutional neural network deep learning model are fused with the sensor data. Specifically, the features extracted by the convolutional neural network deep learning model are combined with the sensor data using a feature cascade method to obtain the fused features of the convolutional neural network, and then the support vector machine SVM is used to classify the fused features of the convolutional neural network. The specific steps are as follows: Step B22221, using a feature cascade method to combine the features extracted by the convolutional neural network deep learning model with the sensor data and obtain convolutional neural network fusion features, specifically combining features from different data sources together through the feature cascade method, wherein the feature vector extracted by the convolutional neural network deep learning model is connected to the feature vector of the sensor data to form a long feature vector; Step B22222, using support vector machine SVM to classify the features after convolutional neural network fusion, specifically using support vector machine SVM to find the optimal hyperplane to classify the features after convolutional neural network fusion, the support vector machine SVM uses Lagrange multiplier method to transform it into a dual problem and uses sequential minimum optimization SMO algorithm to solve it, as shown in formula (6), (6) in, is the hyperplane normal vector, b is the intercept, C is the regularization parameter, x i is the slack variable, y i is the sample label, is the sample characteristics; Step B23, performing radar-vision multimodal data fusion on the pre-processed data and obtaining radar-vision multimodal fusion data, wherein the radar-vision multimodal data fusion is used to integrate point cloud data and visual data for comprehensive environmental perception to support coordinated safety monitoring of buses and stations. The specific steps are as follows: Step B231, using data synchronization technology to keep the radar point cloud data and visual data consistent in time and space, specifically using the Hungarian algorithm to match the radar detection target with the object in the visual image, wherein the matching basis includes distance features, speed features, shape features and size features, and the matching process uses a cost matrix to find the minimum match so as to correspond the radar target to the visual object, and each element in the matching cost is the matching cost between a radar target and a visual object; Step B232, using a Kalman filter data fusion algorithm to fuse information provided by different sensors, wherein the Kalman filter data fusion algorithm is used to combine radar distance, speed information and visual image features when performing state measurement; Step B24, using the spatiotemporal fused data, the convolutional neural network fused data and the radar multimodal fused data to perform target detection, target tracking and target classification and obtain a fused data set, the target detection specifically uses the deep learning-based target detection framework Faster R-CNN to detect and locate the target to obtain the detection target data, the target tracking specifically uses the multi-target tracking algorithm MOT to continuously monitor the movement of the target in time and space to obtain the tracking target data, and the target classification specifically classifies the detection target data and the tracking target data to distinguish between pedestrians, electric vehicles and cars.

[0021] Step C: construct a safety warning model for buses and stations based on the fused data set, and then use the safety warning model to identify safety risks and generate safety warning information. The specific steps are as follows: Step C1, segmenting the fused data set and obtaining segmented data, specifically, dividing the fused data into a training set, a validation set, and a test set. Specifically, 70% of the data sources are randomly selected from the fused data as the training set, and 15% of the data sources are randomly selected from the data after the training set is extracted as the validation set, and the remaining 15% of the data sources are used as the test set; Step C2, feature selection is performed on the segmented data to extract model learning features, specifically, passenger flow density features, vehicle speed features, vehicle status features and weather condition features are selected from the training set, and then principal component analysis is performed on the passenger flow density features and vehicle speed features. The specific steps are as follows: Step C21, analyzing the correlation between each feature and the safety risk, and then using the correlation coefficient analysis method to select the passenger flow density feature, vehicle speed feature, vehicle state feature and weather condition feature; Step C22, using principal component analysis PCA to reduce the dimension of passenger flow density features and vehicle speed features and obtain features after dimension reduction, Step C23, standardizing and normalizing the passenger flow density features, vehicle speed features, vehicle state features, weather condition features, and dimensionality-reduced features so that all features are at the same scale, thereby obtaining model learning features; Step C3, building a safety warning model for buses and platforms based on model learning features. The specific steps are as follows: Step C31, constructing a safety warning model for buses and platforms according to the safety properties and data characteristics of buses and platforms, wherein the network structure of the safety warning model for buses and platforms includes an input layer, multiple convolutional layers, a pooling layer, a fully connected layer and an output layer, wherein the input layer is used to receive model learning features, the convolutional layer and the pooling layer are used for feature extraction and feature dimensionality reduction respectively, the fully connected layer is used to perform high-level abstract operations, and the output layer is used to output warning results; Step C32, training the safety warning model, specifically using the training set data to train the safety warning model, and then optimizing the performance of the safety warning model by adjusting the learning rate, convolution kernel size and activation function; Step C33, verifying the safety warning model, specifically using the verification set data to verify the accuracy index, recall index and F1 score index of the safety warning model, and then verifying the classification performance of the safety warning model by drawing a confusion matrix, and then verifying the binary classification ability of the safety warning model by using the receiver operating characteristic curve ROC and the area under the curve AUC; Step C4, using the security warning model to identify security risks and generate security warning information, the specific steps are as follows: Step C41, using the safety early warning model to identify safety risks, specifically using the verified safety early warning model of buses and platforms to identify safety hazards and determine whether there are safety risks; Step C42, generating security warning information, specifically converting the determined security risk into security warning information, wherein the specific content of the security warning information includes risk type, risk level and recommended measures.

[0022] Step D: Based on the safety warning information, a remote monitoring platform is built to remotely monitor the hardware equipment on the bus side and the platform side. The specific steps are as follows: Step D1, receiving security warning information, and then processing the security warning information and obtaining processed security warning information. The specific steps are as follows: Step D11, receiving security warning information, wherein the specific content of the security warning information includes device type, risk type, risk level, occurrence location and timestamp; Step D12, processing the security warning information and obtaining the processed security warning information, the specific steps are as follows: Step D121, removing duplicate data, redundant field data and invalid data in the security warning information, and then verifying the rationality of the timestamp and the validity of the location information in the security warning information; Step D122, classifying the security warning information according to the urgency and importance, wherein the risk categories include high risk, medium risk and low risk, and then prioritizing the security warning information according to the risk level factor, the impact scope factor and the time urgency factor; Step D2, build a remote monitoring platform, and then use the remote monitoring platform to handle faults and provide feedback based on the processed safety warning information. The specific steps are as follows: Step D21, construct a remote monitoring platform, the remote monitoring platform can receive, display and process security warning information in real time, and the remote monitoring platform can also support data visualization, equipment control and alarm notification. The specific steps are as follows: Step D211, displaying the received safety warning information in real time on the remote monitoring platform, and then judging whether it is necessary to trigger a public transportation safety warning according to the set safety threshold; Step D212: If a public transportation safety warning is triggered, the remote monitoring platform will notify the relevant staff of the public transportation safety warning information in real time by SMS, APP push and display on the monitoring center screen. After receiving the public transportation safety warning information, the relevant staff will take corresponding emergency measures to ensure the safety of passengers and vehicles. The corresponding emergency measures include adjusting the driving route, evacuating passengers and emergency stopping. Step D22, using the remote monitoring platform to perform fault processing and feedback based on the processed safety warning information, the specific steps are as follows: Step D221, using the remote monitoring platform to identify the fault type and equipment location based on the processed safety warning information, and then using remote monitoring technology to remotely diagnose and troubleshoot the faulty equipment and determine the cause of the fault and solution; Step D222, repair the faults that can be repaired remotely using the remote monitoring module, and automatically trigger an alarm notification to notify maintenance personnel to go to the site for repair if the faults need to be handled on site; Step D223, feed back the fault handling results to the remote monitoring platform in real time and update the equipment status information, then record the fault handling process and fault handling results and form a fault handling report.

[0023] Step E: Issue safety warning information to passengers on the bus and platform sides, and then the dispatch center will output dispatch optimization decisions and execute them based on the safety warning information and real-time data to complete the vehicle-road collaborative dispatch operation. The specific steps are as follows: Step E1, issuing safety warning information to passengers on the bus side and the platform side, specifically, using the on-board display screen channel, the platform display screen channel, the broadcast channel and the mobile phone APP channel to issue the safety warning information to the platform side, and then transmit it to the on-board OBU through the platform collaborative equipment and perform forward collision warning, emergency braking warning, lane change warning, blind spot warning, emergency vehicle reminder and abnormal vehicle reminder on the bus side; Step E2: The dispatch center outputs and executes dispatch optimization decisions based on safety warning information and real-time data. The dispatch optimization decisions include adjusting departure intervals, changing driving routes, and adding temporary shifts.

[0024] A vehicle-road cooperative dispatching system between a bus and a platform comprises a data acquisition module, a data processing and fusion module, a safety warning information generation module, a remote monitoring module and a dispatching output execution module, wherein the data acquisition module is used to collect data in real time using the vehicle-road cooperative equipment on the bus side and the platform side and obtain the collected data; the data processing and fusion module is used to preprocess and fuse the collected data and obtain a fused data set; the safety warning information generation module is used to construct a safety warning model for the bus and the platform based on the fused data set, and then use the safety warning model to identify safety risks and generate safety warning information; the remote monitoring module is used to construct a remote monitoring platform based on the safety warning information to remotely monitor the hardware equipment on the bus side and the platform side; the dispatching output execution module issues safety warning information to passengers on the bus side and the platform side, and then outputs and executes dispatching optimization decisions based on the safety warning information and real-time data through the dispatching center.

[0025] Specifically, the preprocessing contents in the data processing fusion module include data cleaning, data conversion, data standardization, data compression and encryption, and data integration and storage; the fusion contents in the data processing fusion module include spatiotemporal data fusion, convolutional neural network data fusion, and radar multimodal data fusion.

[0026] In summary, the vehicle-road cooperative scheduling method and system between a bus and a platform of the present invention first utilizes the vehicle-road cooperative equipment on the bus side and the platform side to collect data in real time and obtain the collected data, then pre-processes and fuses the collected data to obtain the fused data set, then builds a safety warning model for the bus and the platform based on the fused data set, and then uses the safety warning model to identify safety risks and generate safety warning information, and then builds a remote monitoring platform based on the safety warning information to remotely monitor the hardware equipment on the bus side and the platform side, and then issues safety warning information to passengers on the bus side and the platform side, and then outputs and executes scheduling optimization decisions based on the safety warning information and real-time data through the dispatch center, thereby completing the vehicle-road cooperative scheduling operation; effectively realizing that the vehicle-road cooperative scheduling method and system between the bus and the platform have the advantages of providing urban public transportation Efficient, safe and intelligent monitoring and dispatching functions not only help to improve passengers' travel experience and reduce bus operating costs, but also promote the sustainable development of urban transportation. By deploying high-definition cameras, millimeter-wave radars and edge computing unit vehicle-road collaborative equipment on the platform side, bus data and real-time data on the platform side can be transmitted in real time, thereby significantly improving the accuracy and early warning capabilities of safety monitoring. After real-time data is processed, a safety early warning platform is built for analysis, which can intuitively display the vehicle entry and exit operation trajectory, traffic flow and passenger distribution information, thereby improving the accuracy and intuitiveness of the analysis. At the same time, by integrating video data with radar data, it can automatically identify safety hazards and operational bottlenecks, thereby generating safety early warning warnings and dispatch optimization suggestions, thereby ensuring the safety of passengers and vehicles, which not only improves the efficiency and quality of bus services, but also ensures the stable and reliable operation of equipment.

[0027] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A vehicle-road coordinated dispatching method between a bus and a platform, characterized by: The following steps are included: Step A, using the vehicle-road cooperative equipment on the bus side and the platform side to collect data in real time and obtain collected data; Step B, preprocessing and fusing the collected data to obtain a fused data set; Step C: construct a safety warning model for buses and stations based on the fused data set, and then use the safety warning model to identify safety risks and generate safety warning information; Step D: Building a remote monitoring platform based on the safety warning information to remotely monitor the hardware equipment on the bus side and the platform side; In step E, safety warning information is issued to passengers on the bus and platform sides, and then the dispatch center outputs and executes dispatch optimization decisions based on the safety warning information and real-time data to complete the vehicle-road collaborative dispatch operation.

2. A vehicle-road coordinated dispatching method between a bus and a platform according to claim 1, characterized in that: Step A, using the vehicle-road cooperative equipment on the bus side and the platform side to collect data in real time to obtain collected data, wherein the collected data includes video surveillance data, GIS map data, passenger flow analysis data, equipment status monitoring data, driving road congestion data, and bus basic data. The platform side uses high-definition cameras, millimeter-wave radars, edge computing units and temperature sensing devices to collect roadside data and receive on-board data transmitted by the bus-mounted OBU. The data collected on the platform side includes platform side waiting passenger flow data, video surveillance data, radar point cloud data, traffic data around the platform, road condition information data, non-motorized vehicle lane pedestrian and vehicle data, weather temperature data and platform water and snow data. The road condition information data includes platform road congestion, road traffic flow, platform pavement conditions and construction conditions. The video surveillance data includes platform motor vehicle lane video surveillance data and non-motorized vehicle lane video surveillance data. The bus basic data includes vehicle number, license plate number, driver information and vehicle equipment information.

3. The vehicle-road coordinated dispatching method between a bus and a platform according to claim 2, characterized in that: Step B, preprocessing and fusion of the collected data to obtain a fused data set. The specific steps are as follows: Step B1, preprocessing the collected data and obtaining preprocessed data, wherein the preprocessing specifically includes data cleaning, data conversion, data standardization, data compression and encryption, and data integration and storage; Step B2, performing data fusion on the preprocessed data and obtaining a fused data set, wherein the specific fusion contents include spatiotemporal data fusion, convolutional neural network data fusion and radar multimodal data fusion.

4. The vehicle-road coordinated dispatching method between a bus and a platform according to claim 3, characterized in that: The specific steps of data preprocessing in step B1 are as follows: Step B11, clean the collected data and obtain the cleaned data. The specific steps are as follows: Step B111, backing up the video surveillance data, GIS map data, passenger flow analysis data, equipment status monitoring data, road congestion data, bus basic data and platform passenger flow data, video surveillance data, radar point cloud data, platform surrounding traffic data, road condition information data, non-motorized vehicle lane pedestrian and vehicle data, weather temperature data, platform water and snow data in the collected data; Step B112, deduplication processing is performed on the collected data, specifically, using a set to identify and delete identical data records, wherein records with different timestamps but the same other information are deduplicated by setting a threshold. Specifically, the timestamp of data record i is set to T i , the timestamp of data record j is T j and the threshold is ΔT1, if |T i -T j | ≤ ΔT1, then data record i and data record j are determined to be duplicate data. If |T i -T j | >ΔT1, then data record i and data record j are determined to be non-duplicate data; Step B113, performing missing value processing on the collected data, specifically filling the numerical data with the mean value and filling the categorical data with the mark; Step B114, detect and process outliers on the collected data, specifically using the IQR method to identify outliers. The IQR method sets a threshold ΔT2 based on business logic to determine outliers and delete the determined outliers. The IQR is specifically the difference between the third quartile Q3 and the first quartile Q1 in the data, as shown in formula (1). IQR = Q3 − Q1 (1) Among them, the abnormal values ​​are less than Q1 – ΔT2×IQR and greater than Q3 + ΔT2×IQR; Step B12, convert and standardize the cleaned data to obtain data in a unified format. The specific steps are as follows: Step B121, converting the cleaned data, specifically, unifying the data source date, time and format of different platforms and devices in the cleaned data into a standard date and time format, wherein the text data is converted using a coding conversion method; Step B122, standardize the cleaned data. Specifically, the numerical data is standardized by using the min-max standardization method, and the categorical data is standardized by using the label encoding method. The specific min-max standardization method is shown in formula (2): （2） in, is the min-max standard value, is the original value, is the minimum value, is the maximum value, is the coefficient; Step B123, discretizing the cleaned data, specifically using an equal-width binning method to discretize the continuous data; Step B13, compress and encrypt the data after format unification to obtain compressed and encrypted data. The specific steps are as follows: Step B131, compressing the format-unified data to obtain compressed data, wherein the video surveillance data is compressed using a video compression standard, and the data information is compressed using a GZIP compression algorithm; Step B132, encrypt the compressed data, specifically using AES+RSA hybrid encryption algorithm to encrypt the data, the specific steps are as follows: Step B1321, using the RSA algorithm to generate a public key and a private key, the public key is used to encrypt data, and the private key is used to decrypt data; Step B1322, the sender generates an original AES key, and then uses the original AES key to symmetrically encrypt the compressed data to obtain symmetrically encrypted data. Then, the sender uses the RSA public key to asymmetrically encrypt the original AES key to obtain an encrypted AES key, and then sends the encrypted AES key and the symmetrically encrypted data to the receiver. Step B1323, the receiver receives the encrypted AES key and the symmetrically encrypted data, uses the RSA private key to decrypt the AES key and restore the original AES key, and then uses the original AES key to decrypt the symmetrically encrypted data to restore the compressed data; Step B14, integrating and storing the compressed and encrypted data. The specific steps are as follows: Step B141, integrating the compressed and encrypted data into logically interrelated data sets according to the data types, wherein the logically interrelated data sets include video surveillance data sets, GIS map data sets, passenger flow analysis data sets, equipment condition monitoring data sets, traffic condition data sets, radar point cloud and vehicle flow data sets, road condition information and non-motorized lane data sets, and weather and environment data sets; Step B142, selecting a database to store the integrated data, wherein the database is divided into a relational database MySQL and a non-relational database MongoDB.

5. The vehicle-road coordinated dispatching method between a bus and a platform according to claim 4, characterized in that: The specific steps of data fusion in step B2 are as follows: Step B21, performing spatiotemporal data fusion on the preprocessed data and obtaining spatiotemporal fused data, wherein the spatiotemporal data fusion is used to make the preprocessed data accurate to milliseconds in time and centimeters in space, thereby supporting coordinated safety monitoring and intelligent analysis-assisted scheduling of buses and stations. The specific steps are as follows: Step B211, using the Network Time Protocol (NTP) to synchronize the timestamps of the pre-processed data, and then correcting the timestamps of different devices to ensure consistency on the time axis; Step B212, the pre-processed data is subjected to spatial coordinate transformation to conform to the unified geographic information system (GIS) standard. The specific spatial coordinate transformation formula is shown in formula (3): Xnew = Xold*cos(α2) - Yold*sin(α2) + translation X, Ynew= Xold*sin(α2) + Yold*cos(α2) + translation Y (3) Among them, α2 is the rotation angle, Xold and Yold are the original coordinates, and Xnew and Ynew are the transformed coordinates; Step B22, performing convolutional neural network data fusion on the preprocessed data and obtaining convolutional neural network fused data, wherein the convolutional neural network data fusion specifically uses the convolutional neural network deep learning model to extract and analyze features of image data and video data to support coordinated safety monitoring, intelligent analysis, situation awareness and auxiliary scheduling of buses and platforms. The specific steps are as follows: Step B221, adjust and normalize the image data and video data to obtain the adjusted and normalized data. The specific steps are as follows: Step B2211, resizing the image data and the video data and obtaining resized image and video data, wherein the resizing specifically refers to adjusting the image size and the video frame to a uniform size; Step B2212, normalize the resized image and video data to obtain normalized image and video data, as shown in formula (4), （4） in, is the normalized pixel value, is the original pixel value, and are the minimum and maximum pixel values, respectively. is the adjustment coefficient; Step B2213, performing image enhancement on the pixel values ​​of the normalized image video data, wherein the specific content of the image enhancement is image rotation, image flipping or image cropping; The image rotation is used to enable the convolutional neural network deep learning model to learn the features of the object at different angles. Specifically, the image rotation uses a neighbor interpolation method to maintain the image boundary information, and then limits the rotation angle range according to different scene characteristics; The image flipping is used to enable the convolutional neural network deep learning model to learn the features of the object in the mirror direction. The image flipping is specifically divided into a horizontal flipping mode and a vertical flipping mode. The horizontal flipping mode specifically flips the image along the vertical central axis and generates a horizontally flipped image. The vertical flipping mode specifically flips the image along the horizontal central axis and generates a vertically flipped image. The image cropping is used to make the convolutional neural network deep learning model focus on the visual focus area in the image. The image cropping is specifically to crop the image using the rule of thirds, wherein the rule of thirds is to divide the photo into three parts in the vertical and horizontal directions, and then crop the four points where the lines intersect to form the visual focus; Step B222, based on the adjusted and normalized data, use the convolutional neural network deep learning model to extract and analyze the features of the image data and video data and obtain the convolutional neural network fused data. The specific steps are as follows: Step B2221, constructing a convolutional neural network deep learning model, wherein the convolutional neural network deep learning model is used to identify and extract features in images and videos, the convolutional neural network deep learning model comprises a convolutional layer, a pooling layer and a fully connected layer, and the convolutional neural network deep learning model specifically uses back propagation and gradient descent algorithms to optimize model parameters, wherein the gradient descent algorithm is shown in formula (5), （5） Among them, θ is the model parameter, J(θ) is the loss function, is the learning rate; Step B2222, the features extracted by the convolutional neural network deep learning model are fused with the sensor data. Specifically, the features extracted by the convolutional neural network deep learning model are combined with the sensor data using a feature cascade method to obtain the fused features of the convolutional neural network, and then the support vector machine SVM is used to classify the fused features of the convolutional neural network. The specific steps are as follows: Step B22221, using a feature cascade method to combine the features extracted by the convolutional neural network deep learning model with the sensor data and obtain convolutional neural network fusion features, specifically combining features from different data sources together through the feature cascade method, wherein the feature vector extracted by the convolutional neural network deep learning model is connected to the feature vector of the sensor data to form a long feature vector; Step B22222, using support vector machine SVM to classify the features after convolutional neural network fusion, specifically using support vector machine SVM to find the optimal hyperplane to classify the features after convolutional neural network fusion, the support vector machine SVM uses Lagrange multiplier method to transform it into a dual problem and uses sequential minimum optimization SMO algorithm to solve it, as shown in formula (6), （6） in, is the hyperplane normal vector, b is the intercept, C is the regularization parameter, x i is the slack variable, y i is the sample label, is the sample characteristics; Step B23, performing radar-vision multimodal data fusion on the pre-processed data and obtaining radar-vision multimodal fusion data, wherein the radar-vision multimodal data fusion is used to integrate point cloud data and visual data for comprehensive environmental perception to support coordinated safety monitoring of buses and stations. The specific steps are as follows: Step B231, using data synchronization technology to keep the radar point cloud data and visual data consistent in time and space, specifically using the Hungarian algorithm to match the radar detection target with the object in the visual image, wherein the matching basis includes distance features, speed features, shape features and size features, and the matching process uses a cost matrix to find the minimum match so as to correspond the radar target to the visual object, and each element in the matching cost is the matching cost between a radar target and a visual object; Step B232, using a Kalman filter data fusion algorithm to fuse information provided by different sensors, wherein the Kalman filter data fusion algorithm is used to combine radar distance, speed information and visual image features when performing state measurement; Step B24, using the spatiotemporal fused data, the convolutional neural network fused data and the radar multimodal fused data to perform target detection, target tracking and target classification and obtain a fused data set, the target detection specifically uses the deep learning-based target detection framework Faster R-CNN to detect and locate the target to obtain the detection target data, the target tracking specifically uses the multi-target tracking algorithm MOT to continuously monitor the movement of the target in time and space to obtain the tracking target data, and the target classification specifically classifies the detection target data and the tracking target data to distinguish between pedestrians, electric vehicles and cars.

6. A vehicle-road coordinated dispatching method between a bus and a platform according to claim 5, characterized in that: Step C: construct a safety warning model for buses and stations based on the fused data set, and then use the safety warning model to identify safety risks and generate safety warning information. The specific steps are as follows: Step C1, segmenting the fused data set and obtaining segmented data, specifically, dividing the fused data into a training set, a validation set, and a test set. Specifically, 70% of the data sources are randomly selected from the fused data as the training set, and 15% of the data sources are randomly selected from the data after the training set is extracted as the validation set, and the remaining 15% of the data sources are used as the test set; Step C2, feature selection is performed on the segmented data to extract model learning features, specifically, passenger flow density features, vehicle speed features, vehicle status features and weather condition features are selected from the training set, and then principal component analysis is performed on the passenger flow density features and vehicle speed features. The specific steps are as follows: Step C21, analyzing the correlation between each feature and the safety risk, and then using the correlation coefficient analysis method to select the passenger flow density feature, vehicle speed feature, vehicle state feature and weather condition feature; Step C22, using principal component analysis PCA to reduce the dimension of passenger flow density features and vehicle speed features and obtain features after dimension reduction, Step C23, standardizing and normalizing the passenger flow density features, vehicle speed features, vehicle state features, weather condition features, and dimensionality-reduced features so that all features are at the same scale, thereby obtaining model learning features; Step C3, building a safety warning model for buses and platforms based on model learning features. The specific steps are as follows: Step C31, constructing a safety warning model for buses and platforms according to the safety properties and data characteristics of buses and platforms, wherein the network structure of the safety warning model for buses and platforms includes an input layer, multiple convolutional layers, a pooling layer, a fully connected layer and an output layer, wherein the input layer is used to receive model learning features, the convolutional layer and the pooling layer are used for feature extraction and feature dimensionality reduction respectively, the fully connected layer is used to perform high-level abstract operations, and the output layer is used to output warning results; Step C32, training the safety warning model, specifically using the training set data to train the safety warning model, and then optimizing the performance of the safety warning model by adjusting the learning rate, convolution kernel size and activation function; Step C33, verifying the safety warning model, specifically using the verification set data to verify the accuracy index, recall index and F1 score index of the safety warning model, and then verifying the classification performance of the safety warning model by drawing a confusion matrix, and then verifying the binary classification ability of the safety warning model by using the receiver operating characteristic curve ROC and the area under the curve AUC; Step C4, using the security warning model to identify security risks and generate security warning information, the specific steps are as follows: Step C41, using the safety early warning model to identify safety risks, specifically using the verified safety early warning model of buses and platforms to identify safety hazards and determine whether there are safety risks; Step C42, generating security warning information, specifically converting the determined security risk into security warning information, wherein the specific content of the security warning information includes risk type, risk level and recommended measures.

7. The vehicle-road coordinated dispatching method between a bus and a platform according to claim 6, characterized in that: Step D: Based on the safety warning information, a remote monitoring platform is built to remotely monitor the hardware equipment on the bus side and the platform side. The specific steps are as follows: Step D1, receiving security warning information, and then processing the security warning information and obtaining processed security warning information. The specific steps are as follows: Step D11, receiving security warning information, wherein the specific content of the security warning information includes device type, risk type, risk level, occurrence location and timestamp; Step D12, processing the security warning information and obtaining the processed security warning information, the specific steps are as follows: Step D121, removing duplicate data, redundant field data and invalid data in the security warning information, and then verifying the rationality of the timestamp and the validity of the location information in the security warning information; Step D122, classifying the security warning information according to the urgency and importance, wherein the risk categories include high risk, medium risk and low risk, and then prioritizing the security warning information according to the risk level factor, the impact scope factor and the time urgency factor; Step D2, build a remote monitoring platform, and then use the remote monitoring platform to handle faults and provide feedback based on the processed safety warning information. The specific steps are as follows: Step D21, construct a remote monitoring platform, the remote monitoring platform can receive, display and process security warning information in real time, and the remote monitoring platform can also support data visualization, equipment control and alarm notification. The specific steps are as follows: Step D211, displaying the received safety warning information in real time on the remote monitoring platform, and then judging whether it is necessary to trigger a public transportation safety warning according to the set safety threshold; Step D212: If a public transportation safety warning is triggered, the remote monitoring platform will notify the relevant staff of the public transportation safety warning information in real time by SMS, APP push and display on the monitoring center screen. After receiving the public transportation safety warning information, the relevant staff will take corresponding emergency measures to ensure the safety of passengers and vehicles. The corresponding emergency measures include adjusting the driving route, evacuating passengers and emergency stopping. Step D22, using the remote monitoring platform to perform fault processing and feedback based on the processed safety warning information, the specific steps are as follows: Step D221, using the remote monitoring platform to identify the fault type and equipment location based on the processed safety warning information, and then using remote monitoring technology to remotely diagnose and troubleshoot the faulty equipment and determine the cause of the fault and solution; Step D222, repair the faults that can be repaired remotely using the remote monitoring module, and automatically trigger an alarm notification to notify maintenance personnel to go to the site for repair if the faults need to be handled on site; Step D223, feed back the fault handling results to the remote monitoring platform in real time and update the equipment status information, then record the fault handling process and fault handling results and form a fault handling report.

8. The vehicle-road coordinated dispatching method between a bus and a platform according to claim 7, characterized in that: Step E: Issue safety warning information to passengers on the bus and platform sides, and then the dispatch center will output dispatch optimization decisions and execute them based on the safety warning information and real-time data to complete the vehicle-road collaborative dispatch operation. The specific steps are as follows: Step E1, issuing safety warning information to passengers on the bus side and the platform side, specifically, using the on-board display screen channel, the platform display screen channel, the broadcast channel and the mobile phone APP channel to issue the safety warning information to the platform side, and then transmit it to the on-board OBU through the platform collaborative equipment and perform forward collision warning, emergency braking warning, lane change warning, blind spot warning, emergency vehicle reminder and abnormal vehicle reminder on the bus side; Step E2: The dispatch center outputs and executes dispatch optimization decisions based on safety warning information and real-time data. The dispatch optimization decisions include adjusting departure intervals, changing driving routes, and adding temporary shifts.

9. A vehicle-road collaborative dispatching system between a bus and a platform, wherein the specific vehicle-road collaborative dispatching process of the vehicle-road collaborative dispatching system is based on the vehicle-road collaborative dispatching method according to any one of claims 1 to 8, characterized in that: It includes a data acquisition module, a data processing and fusion module, a safety warning information generation module, a remote monitoring module and a dispatch output execution module. The data acquisition module is used to collect data in real time using the vehicle-road cooperative equipment on the bus side and the platform side and obtain the collected data; The data processing and fusion module is used to preprocess and fuse the collected data and obtain a fused data set; The safety warning information generation module is used to build a safety warning model for buses and platforms based on the fused data set, and then use the safety warning model to identify safety risks and generate safety warning information; The remote monitoring module is used to build a remote monitoring platform based on safety warning information to remotely monitor the hardware equipment on the bus side and the platform side; The dispatch output execution module issues safety warning information to passengers on the bus side and the platform side, and then outputs and executes dispatch optimization decisions based on the safety warning information and real-time data through the dispatch center.

10. The vehicle-road cooperative dispatching system between a bus and a platform according to claim 9, characterized in that: The specific contents of preprocessing in the data processing fusion module include data cleaning, data conversion, data standardization, data compression and encryption, and data integration and storage. The specific contents of fusion in the data processing fusion module include spatiotemporal data fusion, convolutional neural network data fusion, and radar multimodal data fusion.

Images