Highway natural disaster early warning method and system

Abstract

The invention provides an expressway natural disaster early warning method and system, and the method comprises the steps: obtaining multi-dimensional expressway data of a set road section, and forming a real-time data set of the corresponding road section based on the multi-dimensional expressway data; wherein the multi-dimensional expressway data comprises historical disaster and risk data, video data, pavement maintenance monitoring data and meteorological data; inputting the real-time data set into a pre-trained highway natural disaster early warning model, and outputting and obtaining natural disaster early warning information of a corresponding road section; and sending a user notification message generated based on the natural disaster early warning information to a user touch device, so that a user receives the user notification message. According to the invention, natural disaster early warning can be quickly given out to users on the expressway, so that the safety of vehicles driving at a high speed is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of natural disaster early warning, and in particular to a natural disaster early warning method and system for expressways. Background Art

[0002] In recent years, natural disasters have occurred frequently, posing unprecedented challenges to human society and economic development. In the specific scenario of highways, given the high speed of vehicles (average speed of 100 kilometers per hour), if a natural disaster occurs (covering types such as collapse, landslides, mudslides, subsidence and collapse in geological disasters, and water damage in floods and droughts), it is difficult for vehicles to quickly respond effectively within 5 kilometers of the disaster. Specifically, within a range of 2 meters to 1 kilometer near the location of the natural disaster, vehicle drivers are unable to immediately perform reasonable driving operations due to limited reaction time; in the area of ​​1 to 5 kilometers near the disaster site, vehicles cannot obtain relevant information about the disaster in a timely manner, making it difficult to prepare for response in advance.

[0003] At present, although some navigation software can calculate the congestion status of a road section based on the fusion of signaling data from mobile communication operators and GPS positioning data, and can send advance notifications to vehicles outside a range of 5 kilometers for early warning, this method can only reflect the congestion situation, and the software's response speed is at the minute level.

[0004] The existing technologies for monitoring and warning natural disasters on highways often rely on traditional monitoring methods, such as regular manual inspections; and when possible signs of disasters are detected, multiple links of information transmission and confirmation are required before a warning can be issued. The existing technologies have the following defects:

[0005] (1) Low timeliness: Traditional monitoring methods have a large time interval and cannot achieve real-time monitoring of disasters. In addition, there is a lag in data collection and transmission. For example, geological disasters such as collapse and landslides may not be discovered by inspection personnel until the disaster has occurred and caused a certain impact. During this period, vehicles are still driving at high speed without knowing it, which greatly increases the risk of accidents.

[0006] (2) Slow response: The current early warning mechanism is cumbersome and inefficient in processing and analyzing disaster data, which makes it impossible for vehicles to quickly obtain effective response instructions when facing imminent danger. Take water damage in floods and droughts as an example. When the water level suddenly rises or the road is flooded, existing technologies may take several minutes or even longer to complete data processing and decision-making. However, when a vehicle is driving at a speed of 100 km / h on the highway, the vehicle has traveled a considerable distance by the time the decision is made, increasing the possibility of encountering danger.

[0007] Therefore, there is an urgent need for a fast-response highway natural disaster early warning method to reduce the risk of highway traffic and improve the operation efficiency of highways. Summary of the invention

[0008] In view of this, an embodiment of the present invention provides a method and system for early warning of natural disasters on a highway, which can respond quickly to early warning of natural disasters on a highway to ensure the safety of high-speed driving.

[0009] One aspect of the present invention provides a method for early warning of natural disasters on a highway, the method comprising the following steps:

[0010] Acquire multi-dimensional highway data of a set section, and form a real-time data set of the corresponding section based on the multi-dimensional highway data; wherein the multi-dimensional highway data includes historical disaster and risk data, video data, pavement maintenance monitoring data and meteorological data;

[0011] The real-time data set is input into the pre-trained highway natural disaster warning model, and the natural disaster warning information of the corresponding road section is output;

[0012] Sending a user notification message generated based on the natural disaster warning information to the user's access device, so that the user receives the user notification message;

[0013] Among them, the pre-trained highway natural disaster warning model is obtained in the following way:

[0014] Obtain historical data sets, input the historical data sets into the initial early warning model, and obtain a highway natural disaster early warning model through multiple rounds of iterative training;

[0015] Among them, the historical data set is obtained based on multi-dimensional highway data and warning labels within a set historical period. The warning labels and natural disaster warning information both include disaster warning signs used to indicate whether a disaster has occurred and the disaster warning level.

[0016] In some embodiments of the present invention, the multi-dimensional highway data further includes: optical cable transmission status data and vehicle traffic data; wherein the vehicle traffic data is determined based on communication identification between the vehicle and the highway gantry;

[0017] Meteorological data includes meteorological monitoring data and meteorological forecast data. Meteorological data includes the following field information: meteorological monitoring area, meteorological type and meteorological type level.

[0018] In some embodiments of the present invention, the historical disaster and risk data include historical disaster data and risk point data, the historical disaster data include the location of the disaster point and the characteristic information of the disaster loss, and the risk point data include the location of the risk point and the characteristic information of historical disasters; wherein the characteristic information of historical disasters includes one or more of the following field information: terrain characteristics, the frequency of occurrence of historical disasters, the time of occurrence of historical disasters, the type of historical disasters, the scale of historical disasters, the degree of harm of historical disasters, the objects of harm of historical disasters, the high-speed disconnection of historical disasters and the treatment of historical disasters, wherein the type of historical disasters is collapse, landslide, mud-rock flow, subsidence and collapse, water damage or high slope;

[0019] The pavement maintenance monitoring data includes the pavement maintenance monitoring location, the pavement maintenance monitoring type and the monitoring value of each pavement maintenance monitoring type, wherein the pavement maintenance monitoring type includes one or more of displacement monitoring, disaster inducing factor monitoring, protective facility monitoring and short-term monitoring;

[0020] The optical cable transmission status data includes the communication optical cable number and the optical cable transmission information, wherein the optical cable transmission information is measured by one or more of the following performance indicators: signal strength, transmission rate and bit error rate;

[0021] Vehicle traffic data includes the highway gantry number, the time when the vehicle passes through the highway gantry, and the vehicle model; and

[0022] The video data includes the camera number and surveillance video image data.

[0023] In some embodiments of the present invention, forming a real-time data set of a corresponding road section based on multi-dimensional highway data includes:

[0024] The feature field extraction model is used to extract specific field information from historical disaster and risk data to form risk and disaster vectors; the video feature extraction neural network model is used to extract feature data from the monitoring video image data of the video data to form a video vector; the pavement maintenance vector is determined based on the monitoring values ​​of each pavement maintenance monitoring type; the communication optical cable vector is obtained based on the optical cable transmission information in the optical cable transmission status data; the meteorological vector is obtained based on the level of each meteorological type; the gantry vector is formed based on the vehicle type and the time when the vehicle passes the gantry;

[0025] Based on the set road section, the risk and disaster vector, video vector, road maintenance vector, communication optical cable vector, meteorological vector and gantry vector are integrated to obtain the natural disaster vector corresponding to the set road section.

[0026] In some embodiments of the present invention, the feature field extraction model and the initial warning model are neural network models built based on a multi-layer perceptron, and the video feature extraction neural network model is a residual convolutional neural network model;

[0027] Vehicle traffic data, road maintenance monitoring data and meteorological data are obtained through the Hypertext Transfer Protocol (HTTPS), and video data is obtained through the Rapid Spanning Tree Protocol (RSTP). Historical disaster and risk data are obtained through the Java Database Connection (JDBC) protocol. Optical cable transmission status data is obtained through the Message Queue Telemetry Transmission (MQTT) protocol.

[0028] Before determining the pavement maintenance vector, the method further includes: normalizing the acquired pavement maintenance monitoring data.

[0029] In some embodiments of the present invention, the method further comprises:

[0030] Perform data cleaning on the acquired multi-dimensional highway data before forming a real-time data set; and / or

[0031] During the set update period, the initial warning model is retrained using the updated historical data set to obtain an updated highway natural disaster warning model.

[0032] In some embodiments of the present invention, the user notification message includes:

[0033] an operation instruction sent to a user access device within a first range of a set road section, and natural disaster warning information sent to users within a second range of the set road section, wherein the first range of the set road section is smaller than the second range of the set road section;

[0034] Before sending the user notification message to the user access device, the method further includes:

[0035] The order in which user notification messages are sent is determined based on the set vehicle model priority, wherein the lower the vehicle model priority, the later the order in which user notification messages are sent.

[0036] In some embodiments of the present invention, sending a user notification message generated based on natural disaster warning information to a user access device includes:

[0037] Send a user notification message generated based on natural disaster warning information to the message reaching adaptation module, so as to push the user notification message to the user reaching device by utilizing the communication protocol between the message reaching adaptation module and the user reaching device.

[0038] In some embodiments of the present invention, the user access device includes an in-vehicle device;

[0039] Pushing user notification messages to the user access device using the communication protocol between the message access adaptation module and the user access device includes:

[0040] Based on the pre-stored position correspondence between the road section and the highway gantry, the message reaching adaptation module is used to push the user notification message to the highway gantry within the set communication range, and then the communication protocol between the highway gantry and the vehicle-mounted equipment is used to send the user notification message to the user.

[0041] Another aspect of the present invention provides a highway natural disaster warning system, including a processor, a memory, and a computer program / instructions stored in the memory, wherein the processor is used to execute the computer program / instructions, and when the computer program / instructions are executed, the device implements the steps of the method described in any of the above embodiments.

[0042] The highway natural disaster warning method and system proposed in the present invention can obtain multi-dimensional highway data of a set section including historical disaster and risk data, video data and pavement maintenance monitoring data from a multi-source system, and obtain natural disaster warning information of the corresponding section according to a pre-trained highway natural disaster warning model, so that users can obtain warning information in time through user access devices, so as to be able to respond quickly and adjust the driving route to reduce the travel risk of the highway.

[0043] Additional advantages, purposes, and features of the present invention will be described in part in the following description, and will become apparent to those skilled in the art after studying the following, or may be learned from the practice of the present invention. The purposes and other advantages of the present invention may be achieved and obtained by the structures specifically indicated in the specification and the accompanying drawings.

[0044] Those skilled in the art will appreciate that the objectives and advantages that can be achieved with the present invention are not limited to the above specific description, and the above and other objectives that can be achieved by the present invention will be more clearly understood from the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of the present application, and do not constitute a limitation of the present invention. In the drawings:

[0046] Figure 1 The figure is a flow chart of a natural disaster early warning method for expressways in one embodiment of the present invention.

[0047] Figure 2 The figure is a flow chart of a natural disaster early warning method for highways in another embodiment of the present invention.

[0048] Figure 3 Schematic diagram of a natural disaster early warning system for highways in one embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0050] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or processing steps closely related to the solutions according to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.

[0051] It should be emphasized that the term “include / comprises” when used herein refers to the presence of features, elements, steps or components, but does not exclude the presence or addition of one or more other features, elements, steps or components.

[0052] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the accompanying drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

[0053] In the existing technology, the monitoring and early warning of natural disasters on highways often rely on traditional monitoring methods, and when possible signs of disasters are detected, it often requires information transmission and confirmation through multiple links before a warning can be finally issued, which has the disadvantages of slow response speed.

[0054] Based on this, the present invention considers real-time access to multi-source data such as video data from cameras along the highway, road maintenance monitoring data, and historical damage and risk data for historical natural disasters on the highway, and inputs the collected multi-source data into the highway natural disaster warning model, so as to quickly output the natural disaster warning information of the corresponding road section. Furthermore, the present application can use message notification mechanisms such as broadcasting to push user notification messages generated based on natural disaster warning information to user-accessible devices such as variable information boards, navigation software, vehicle computers, and vehicle-mounted terminals on the highway in real time, quickly, and comprehensively, so that users can respond in a timely manner according to the received messages to improve the safety of highway traffic.

[0055] Figure 1 FIG. 1 is a flow chart of a natural disaster early warning method for a highway according to an embodiment of the present invention. Figure 1 As shown, the method includes steps S110-S130.

[0056] Step S110: acquiring multi-dimensional highway data of a set section, and forming a real-time data set of the corresponding section based on the multi-dimensional highway data.

[0057] The disaster data collected after the disaster has a direct impact on the accuracy of future natural disaster predictions, and the highway risk point is the location where a secondary disaster may occur when a natural disaster occurs on the highway (the highway disaster point is the location where a natural disaster has occurred). Incorporating the data collected from the highway risk point and the disaster point into the influencing factors of highway natural disasters can more comprehensively consider the highway disaster warning problem. Therefore, this application can select historical disaster and risk data to show the impact of historical natural disasters on the current highway. Video data may include camera numbers and surveillance video image data captured by surveillance cameras along the set road section. Video data can intuitively reflect the relevant information of natural disasters and the impact of disasters on highway vehicles. Pavement maintenance monitoring data refers to relevant data obtained through highway monitoring during highway inspections, which has a potential impact on road traffic conditions before and after natural disasters. Moreover, extreme weather is also very likely to cause natural disasters. Therefore, the multidimensional highway data in this application may include historical disaster and risk data, video data, pavement maintenance monitoring data, and meteorological data.

[0058] In some embodiments of the present invention, historical disaster and risk data may include historical disaster data and risk point data of highways, that is, data on natural disasters on highways collected during and after the occurrence of historical disasters. Specifically, risk point data may include risk point locations and historical disaster feature information, and historical disaster feature information refers to data collected after a disaster occurs in a specific historical period, and may include one or more of the following field information: terrain features (concave banks, convex banks, or straight segments), the frequency of historical disasters, the time of occurrence of historical disasters, the type of historical disasters, the scale of historical disasters, the degree of harm caused by historical disasters (serious, relatively serious, general, or minor), the objects of harm caused by historical disasters (road sections, bridges, tunnel openings, roadbeds, or culverts, etc.), the ways of harm caused by historical disasters (impact, siltation, immersion, or flooding), the highway disconnection conditions caused by historical disasters (traffic interruption, half-width traffic, or two-way traffic), and the treatment of historical disasters. Historical disaster data may include the location of disaster points and disaster loss characteristics information. The disaster loss characteristics information may include basic data such as whether the road is interrupted, the time of disaster occurrence, road width and road type, as well as disaster facility loss information (which may include field information such as roadbed, bridge and tunnel facility types, damaged length, damaged quantity, and damaged amount);

[0059] Furthermore, the meteorological data may include the following field information: meteorological monitoring area, meteorological type and level of meteorological type. The pavement maintenance monitoring data may include pavement maintenance monitoring location, pavement maintenance monitoring type and monitoring values ​​of each pavement maintenance monitoring type, wherein the pavement maintenance monitoring type may include one or more of displacement monitoring, disaster inducing factor monitoring, protective facility monitoring and short-term monitoring, etc. Specifically, the monitoring value of displacement monitoring may be a numerical value representing the displacement of the pavement, the monitoring value of disaster inducing factor monitoring may be a set of numerical values ​​representing the intensity or existence of multiple inducing factors (such as factors that easily induce natural disasters such as pavement cracks and pavement pressure), the monitoring value of protective facility monitoring may be a set of coded numerical values ​​for the status of protective facilities, and short-term monitoring may monitor pavement conditions, bridge structures, tunnel facilities, slope stability, and traffic facilities and operating conditions, etc. The monitoring value of short-term monitoring may be a set of sequences.

[0060] The field information and pavement maintenance monitoring types included in the above-mentioned historical disaster and risk data are merely examples, and the present invention is not limited thereto.

[0061] As an example, the risk point types (i.e., historical disaster types) in the present invention may include collapse, subsidence, landslide, mudslide, water damage, high slope, etc. In addition, meteorological data may include meteorological monitoring data (obtained by meteorological monitoring stations) and meteorological forecast data (obtained based on meteorological monitoring data prediction), and the monitoring video image data in the video data may include monitoring videos and monitoring pictures.

[0062] In order to reflect the timeliness of disaster warning, the time required for the entire process from obtaining the multi-dimensional highway data of the set section to the user receiving the user notification message should be as short as possible, so as to achieve the purpose of timely warning of disasters. When obtaining historical disaster data in the multi-dimensional highway data, the specific historical time period corresponding to the historical disaster data can be set by itself (for example, the time period corresponding to the historical disaster data is before the time period corresponding to the video data), and the present invention does not specifically limit it. In addition, the disaster point location, risk point location and road maintenance monitoring location can be located on the set section.

[0063] In some embodiments of the present invention, considering that data transmission in communication optical cables is fast and the optical cables are buried underground, if the communication optical cables are interrupted, it is likely to indicate that a natural disaster has occurred on the highway. Therefore, the present application considers using the transmission status of the communication optical cables connected to the high-speed gantry equipment as a factor for early warning of natural disasters on highways; passing vehicles are the main objects of early warning of highway disasters, and the vehicle traffic data generated by vehicles traveling on highways is the basic data of highways. Introducing vehicle traffic data into early warning of natural disasters on highways is helpful for subsequent notification of users. Therefore, the multi-dimensional highway data in the present application may also include optical cable transmission status data and vehicle traffic data.

[0064] As an example, the optical cable transmission status data includes the communication optical cable number and the optical cable transmission information (also referred to as the optical cable transmission performance information), wherein the optical cable transmission information can be measured by one or more of the following performance indicators: signal strength, transmission rate and bit error rate; the vehicle passage data is determined based on the communication identification between the vehicle and the high-speed gantry, and the vehicle passage data may include the high-speed gantry number, the time for the vehicle to pass through the high-speed gantry (i.e., the time for the high-speed gantry to identify the passing vehicle) and the vehicle model.

[0065] The multi-dimensional highway data in this application may include historical highway collection data (historical disaster data), real-time highway monitoring data (risk point data, video data, road maintenance monitoring data, meteorological monitoring data, optical cable transmission status data and vehicle traffic data), and highway prediction data (meteorological forecast data), which can improve the accuracy of highway natural disaster warning from the time dimension.

[0066] In some embodiments of the present invention, vehicle traffic data, road maintenance monitoring data and meteorological data can be obtained through the HyperText Transfer Protocol Secure (HTTPS), video data can be obtained through the Rapid Spanning Tree Protocol (RSTP), historical disaster and risk data can be obtained through the Java Database Connectivity (JDBC) protocol, and optical cable transmission status data can be obtained through the Message Queuing Telemetry Transport (MQTT) protocol.

[0067] More specifically, the multi-source data obtained in this application can be divided into structured data and unstructured data. Structured data includes vehicle traffic data, historical disaster and risk data, road maintenance monitoring data, optical cable transmission status data and meteorological monitoring data, which can be obtained from the corresponding system through database JDBC interface protocol, HTTPS and MQTT protocol. Unstructured data includes video data obtained by cameras along the highway, which can be obtained through RSTP.

[0068] Furthermore, considering that the acquired multi-source data may not directly correspond to the highway section, the historical disaster and risk data of the set section can be obtained from the highway disaster emergency management system according to the correspondence between the risk point location and the highway section, and the correspondence between the disaster point location and the highway section; the video data of the set section can be obtained from the highway monitoring system according to the correspondence between the camera number and the highway section; the pavement maintenance monitoring data of the set section can be obtained from the highway maintenance management system according to the correspondence between the pavement maintenance monitoring position and the highway section; the optical cable transmission status data of the set section can be obtained from the high-speed optical cable communication system according to the correspondence between the communication optical cable number and the highway section; the meteorological monitoring data of the set section can be obtained from the meteorological monitoring system according to the correspondence between the meteorological monitoring area and the highway section; the vehicle traffic data of the set section can be obtained from the high-speed gantry management system according to the correspondence between the high-speed gantry and the highway section. Among them, the location of the disaster point, the location of the risk point, the camera number, the road maintenance monitoring location, the communication optical cable number, the meteorological monitoring area, and the correspondence between the high-speed gantry and the highway section can be pre-stored in a specific device.

[0069] In some embodiments of the present invention, before forming a real-time data set based on the acquired multi-dimensional highway data, the method further includes: performing data cleaning on the acquired multi-dimensional highway data. Specifically, after acquiring multi-source data (i.e., acquiring multi-dimensional highway data of a set section), data pre-cleaning can be performed based on the following rules: (1) processing missing values, duplicate values, and abnormal values ​​(including removing obviously erroneous or unreasonable data); (2) unifying data formats such as date, time, numbers, and text; (3) checking whether the data conforms to a specific matching relationship; and (4) unifying the character encoding of text data to avoid garbled characters.

[0070] As an example, checking whether the data conforms to a specific matching relationship mainly includes: whether the monitoring location, road section and monitoring equipment match; whether the monitoring frequency of the monitoring equipment and the obtained monitoring data match; whether the speed, travel time and traffic volume of passing vehicles conform to the relevant relationship; whether the meteorological data matches the real-time road conditions monitored by the road section; whether the traffic volume matches the traffic accident data; whether the alarm signal is reasonable; and whether the expected traffic pattern matches the historical data; etc.

[0071] Step S120: input the real-time data set into the pre-trained highway natural disaster warning model, and output the natural disaster warning information of the corresponding road section.

[0072] In some embodiments of the present invention, the highway natural disaster warning method proposed in the present application further includes step S110, training the initial warning model to obtain a highway natural disaster warning model. The pre-trained highway natural disaster warning model is obtained by:

[0073] Obtain a historical data set, input the obtained historical data set into the initial warning model, and adjust the hyperparameters through multiple rounds of iterative training to obtain the optimal solution of the initial warning model, thereby obtaining a highway natural disaster warning model; wherein the historical data set is obtained based on multi-dimensional highway data and warning labels within a set historical period.

[0074] In this application, the execution order of step S110 and step S120 is not specific, and step S100 can be executed before step S120. Figure 2 The illustrated step S100 may be before the step S110 .

[0075] As an example, the multidimensional highway data in the historical data set may be the multidimensional highway data of any highway section within a set historical period, that is, the multidimensional highway data in the historical data set may be the multidimensional highway data of a set section within a set historical period, or may be the multidimensional highway data of other non-set sections within a set historical period. In this application, the multidimensional highway data within a set historical period are the historical disaster and risk data, video data, and pavement maintenance monitoring data of the same highway section within the set historical period.

[0076] If the multidimensional highway data forming the real-time data set includes historical disaster and risk data, video data, road maintenance monitoring data and meteorological monitoring data, then the multidimensional highway data forming the historical data set also includes historical disaster and risk data, video data, road maintenance monitoring data and meteorological monitoring data, that is, the multidimensional highway data forming the real-time data set is the same data type as the multidimensional highway data forming the historical data set. In addition, the historical data set and the real-time data set can use the same data form to represent the multidimensional highway data. For example, if the historical data set is represented in the form of a vector, the real-time data set is also represented in the form of a vector. In addition, the historical period corresponding to the multidimensional highway data in the historical data set is earlier than the period corresponding to the multidimensional highway data in the real-time data set.

[0077] Furthermore, in the present application, the specific information included in the natural disaster warning information can be determined according to the warning label, so the warning label and the natural disaster warning information can both include a disaster warning mark used to indicate whether a disaster has occurred and the disaster warning level. For example, the disaster warning mark can be customized to "0, 1, 2, 3, 4", "0" means no natural disaster, "1" means a natural disaster is about to occur, and the disaster warning level is level one (blue mark can be used), "2" means a natural disaster is about to occur, and the disaster warning level is level two (yellow mark can be used), "3" means a natural disaster is about to occur, and the disaster warning level is level three (orange mark can be used), and "4" means a natural disaster is about to occur, and the disaster warning level is level four (red mark can be used).

[0078] The disaster warning labels mentioned above are only examples, and users can customize their settings according to their needs. Moreover, the disaster warning labels in the warning labels of the historical data set can be determined based on historical disaster and risk data, or based on video data within a set historical period. The present invention does not specifically limit the method for determining the warning labels. In addition, the warning labels can also include other information such as disaster types, and the present invention is not limited to this.

[0079] The initial warning model can be a neural network model based on a multi-layer perceptron, or other types of warning models, such as support vector machines (SVM) and other models that can be used for prediction, but the present invention is not limited thereto. The iteration conditions set during the initial warning model training process can be set according to the model requirements, but the present invention is not limited thereto. For example, the iteration condition can be that the loss function reaches a set threshold, or the iteration condition can be a self-set number of iterations.

[0080] As an example, the highway natural disaster warning model trained in this application has not only the real-time warning function, but also the function of predicting the probability of natural disasters occurring in the future. The present invention does not specifically describe its training method, and it can be trained using existing technology.

[0081] In some embodiments of the present invention, considering the different speeds of various vehicle models, even if user notification messages are sent at the same time, some vehicles may not be able to respond in time. Therefore, before sending the user notification message to the user access device, the method further includes: determining the order of sending the user notification message based on the set vehicle model priority, wherein the lower the vehicle model priority, the later the order of sending the user notification message. The vehicle model priority can be set by itself, and the present invention does not specifically limit it. For example, the faster the speed of the vehicle model, the lower the priority.

[0082] Step S130: Sending a user notification message generated based on the natural disaster warning information to the user access device, so that the user receives the user notification message. This step may also be that if the result output by the natural disaster warning model of the expressway includes information that a disaster is about to occur on a set section of the road, then sending the user notification message generated based on the natural disaster warning information to the user access device; if the result output by the natural disaster warning model of the expressway includes information that there is no disaster on the set section of the road, then not sending the user notification message generated based on the natural disaster warning information to the user access device.

[0083] In the present application, user notification messages can be sent to users traveling on the highway within a set section. The communication range can also be customized according to the set section, and user notification messages generated based on natural disaster warning information of the set section can be sent to users within the customized communication range. The present invention does not limit the specific scope for sending user notification messages, and can be set according to the specific conditions of the highway.

[0084] In some embodiments of the present invention, the user notification message may include:

[0085] Operation instructions are sent to users within a first range of a set road section through a touch device, and natural disaster warning information is sent to users within a second range of a set road section, wherein the number of users receiving the operation instructions may not be greater than the number of users receiving the natural disaster warning information.

[0086] More specifically, since the closer the vehicle is to a natural disaster, the more difficult it is to respond quickly, the present application may consider including driving operation instructions in the user notification message sent to the user access device within the first range close to the set road section, so that the vehicle driver can respond quickly according to the operation instructions; the user notification message sent to the user access device within the second range far from the set road section includes natural disaster warning information, and since there is enough reaction time, the user can respond in time and take corresponding driving operations. Among them, the operation instructions can be pre-determined and stored in a specific device in combination with historical accidents and disaster situations.

[0087] As an example, the second range of the set section may include the first range of the set section (in this case, the first range of the set section is smaller than the second range of the set section), or the maximum boundary of the first range of the set section is the same as the minimum boundary of the second range of the set section. In addition, the sum of the second range of the set section and the first range of the set section may be referred to as the set communication range.

[0088] The specific content of the above user notification message is only an example. The present application may also send a user notification message including operation instructions and natural disaster warning information to both user access devices within the first range of the set road section and the second range of the set road section, or may send a user notification message including only operation instructions to user access devices within the first range of the set road section and the second range of the set road section. The present application does not limit the specific content of the user notification message.

[0089] Considering that existing early warning notification methods are relatively limited, they can usually only cover limited areas or specific equipment. For example, some early warning information may only be displayed through a few fixed electronic display screens along the highway, and the distribution density of these display screens is low, making it difficult to effectively remind vehicles that are far away. For example, for the early warning push of some navigation software, many vehicles may not receive timely and accurate disaster notifications due to software settings or users' failure to turn on relevant functions in time. In addition, existing notification methods often cannot cover all types of vehicle equipment. For example, some old car models may not be able to receive the latest early warning information, leaving some vehicles in an information blind spot in the face of disasters. This application can push early warning information comprehensively through multiple channels so that users can receive user notification messages in a timely manner.

[0090] As an example, the user access device in the present application is hardware or software that enables highway users to receive user notification messages, and may include one or more of highway variable information boards, navigation software, vehicle computers, vehicle radios, and vehicle-mounted OBU (On Board Unit) devices.

[0091] For example, if the user access device is a vehicle-mounted device, the user notification message can be sent directly to each vehicle-mounted device on the highway, or the user notification message can be sent to the gantry system, thereby sending the user notification message to the user based on the pre-stored position correspondence between the road section and the highway gantry and utilizing the communication between the highway gantry and the vehicle-mounted device.

[0092] In some embodiments of the present invention, the present application can send user notification messages to users through multiple channels. Considering that multiple user access devices are involved, a predetermined communication protocol is required between different devices to transmit user notification messages. Therefore, a message access adaptation module can be used to integrate various types of user access devices, thereby pushing user notification messages to multiple channels more comprehensively, so that passing vehicles can receive warning information in a timely manner.

[0093] Sending user notification messages generated based on natural disaster warning information to user access devices, including:

[0094] Send a user notification message generated based on natural disaster warning information to the message reaching adaptation module, so as to push the user notification message to the user reaching device by utilizing the communication protocol between the message reaching adaptation module and the user reaching device.

[0095] As an example, if the user access device is a variable information board on a highway, the message access adaptation module can be used to push user notification messages to the variable information boards on the highway within the set communication range; if the user access device is navigation software, the message access adaptation module can be used to push user notification messages to the navigation software operator, and the navigation software operator will subsequently push the obtained user notification messages to users within the set communication range; if the user access device is a vehicle computer, the message access adaptation module can be used to push user notification messages to the Internet of Vehicles, and the variable information boards on the highway within the set communication range of the Internet of Vehicles will subsequently push user notification messages; if the user access device is a vehicle-mounted device, the message access adaptation module can be used to push user notification messages to the highway gantries within the set communication range based on the pre-stored position correspondence between the road section and the highway gantry, and then the communication between the highway gantry and the vehicle-mounted device can be used to send the user notification message to the user.

[0096] In step S110, after acquiring the multi-dimensional highway data of the set section, a real-time data set of the set section can be formed based on the multi-dimensional highway data. The present invention does not limit the form of the real-time data set. For example, the real-time data set can be a collection of various values ​​or a vector generated based on the multi-dimensional highway data. Taking the formation of a natural disaster vector of the set section based on the multi-dimensional highway data as an example, the specific process is as follows:

[0097] In some embodiments of the present invention, forming a real-time data set of a corresponding road section based on multi-dimensional highway data includes:

[0098] The feature field extraction model is used to extract feature field information from historical disaster and risk data to form risk and disaster vectors; the video feature extraction neural network model is used to extract feature data from the monitoring video image data of the video data to form a video vector; the pavement maintenance vector is determined based on the monitoring values ​​of each pavement maintenance monitoring type; the communication optical cable vector is obtained based on the optical cable transmission information in the optical cable transmission status data; the meteorological vector is obtained based on the level of each meteorological type; the gantry vector is formed based on the vehicle type and the time when the vehicle passes the gantry;

[0099] Based on the fusion of risk and disaster vectors, video vectors, pavement maintenance vectors, communication optical cable vectors, meteorological vectors and gantry vectors on highway sections, the natural disaster vector corresponding to the set section is obtained.

[0100] As an example, the natural disaster vector can be obtained as follows:

[0101] (1) The pre-trained feature field extraction model can be used to extract feature field information from historical disaster and risk data, and the feature field value (N) can be obtained based on the extracted feature field information, thereby forming a risk and disaster vector according to the set road section (G) and the feature field value (N).

[0102] Among them, the feature field extraction model can be a neural network model based on a multi-layer perceptron. The feature field information can be part or all of the field information in the historical disaster information, and the feature field vector can be generated according to the corresponding arrangement order using the information corresponding to each feature field, such as {feature 1, feature 2, ..., feature n}, or the feature field value (N) can be obtained by assigning values ​​and weighted fusion.

[0103] The above-mentioned feature field information, feature field extraction model and feature field value generation method are only examples, and the present invention is not limited thereto.

[0104] (2) The pre-trained video feature extraction neural network model can be used to extract feature data from the video data, thereby forming a video vector by using the set road section (G), camera number (U) and video feature data (X)

[0105] The above-mentioned video vector generation method is only an example, and the video vector (C) can also be formed by only using the set road section (G) and the video feature data (X). The present invention does not specifically limit the generation method of the video vector. In addition, the video feature extraction neural network model can be a residual convolutional neural network model or other video feature extraction models, and the present invention is not limited thereto.

[0106] (3) The pavement maintenance vector can be formed by using the set road section (G), pavement maintenance monitoring position (P) and pavement maintenance monitoring value (Y)

[0107] As an example, the pavement maintenance monitoring position (P) may be the location of the entire monitoring area, or the position of any point in the monitoring area may be used to represent it. For example, the center point of the monitoring area may be used to represent the monitoring area position (P), but the present invention is not limited thereto.

[0108] The above-mentioned generation method of the pavement maintenance vector (D) is only an example, and it can also be generated by only setting the road section (G) and the pavement maintenance monitoring value (Y), and the present invention is not limited to this. In addition, the present application does not limit the generation method of the pavement maintenance monitoring value (Y), and the pavement maintenance monitoring value (Y) can be a one-dimensional vector formed by the monitoring values ​​of each pavement maintenance monitoring type, or it can be a value obtained by integrating the monitoring values ​​of each pavement maintenance monitoring type.

[0109] Furthermore, since the monitoring value of each pavement monitoring type is a value or a group of values, and the dimensions or value ranges of each monitoring value are quite different, it is necessary to normalize the acquired pavement maintenance monitoring data before using the pavement maintenance monitoring data to form a pavement maintenance vector, so as to ensure that the monitoring values ​​of each pavement maintenance monitoring type are relatively comparable. For example, the normalization formula can be:

[0110]

[0111] Among them, J i It represents the normalized value of the monitoring value of a certain pavement maintenance monitoring type, J represents the original monitoring value of this type of monitoring, min represents the minimum value among the original monitoring values ​​of this type of monitoring, and max represents the maximum value among the original monitoring values ​​of this type of monitoring.

[0112] The above-mentioned normalization formula is only an example, and the present invention is not limited thereto. Other normalization processing methods such as Z-Score normalization may also be used.

[0113] (4) The transmission status of the communication optical cable can be measured by performance indicators such as signal strength, transmission rate and bit error rate. The corresponding values ​​of each indicator are integrated to obtain the optical cable transmission status information (O), so as to obtain the communication optical cable vector by using the set section (G) and the transmission status information (O) of the communication optical cable.

[0114] For example, multiple metrics can be fused in a weighted fusion manner. Assume that the weights of signal strength, transmission rate, and bit error rate are w s 、w r and w e , and w s +w r +w e =1, the values ​​of signal strength, transmission rate and bit error rate are Z s , Z r and Z e , then the calculation formula of the transmission status data (O) of the communication optical cable can be expressed as: O = w s ×Z s +w r ×Z r +w e ×Z e +b, where b represents the bias. In addition, the weight w corresponding to the measurement index of the communication optical cable transmission status s 、w r and w e It can be obtained based on specific data of the monitored indicators and a gradient descent algorithm, or by other methods, such as a pre-set method, but the present invention is not limited thereto.

[0115] The above-mentioned generation methods of the communication optical cable transmission state information (O) and the communication optical cable vector (E) are only examples, and the present invention does not specifically limit them. In addition, the measurement index of the communication optical cable transmission state is only an example, and other indicators such as bandwidth may also be used.

[0116] (5) Meteorological types (referring to meteorological disaster types or extreme weather types) in meteorological monitoring data may include heavy rain, strong winds, high temperatures, low temperatures, heavy fog, snowfall, and hail, etc. Each meteorological type can be divided into different levels based on monitoring data and classification standards, and different levels of each meteorological type can correspond to different set values.

[0117] Taking the meteorological types including heavy rain, strong wind, high temperature, low temperature and heavy fog as an example, the heavy rain level can be divided into level 1, level 2, level 3, level 4 and level 5, and the set values ​​are 1, 2, 3, 4 and 5 respectively; the strong wind level can be divided into level 1, level 2 and level 3, and the corresponding set values ​​are 1, 2 and 3 respectively; the high temperature level can be divided into level 1 (30-35℃) and level 2 (36-40℃), and the corresponding set values ​​are 1 and 2 respectively; the low temperature level can be divided into level 1 (0--5℃) and level 2 (-5--10℃), and the corresponding set values ​​are also 1 and 2 respectively; the heavy fog level can be divided into level 1 (visibility is 500-1000m), level 2 (visibility is 200-500m) and level 3 (visibility is less than 200m), and the corresponding set values ​​are 1, 2 and 3 respectively. That is, the higher the level of each meteorological type, the larger the corresponding set value.

[0118] In the present application, a weather vector can be obtained based on the level of each weather type. For example, a weather vector is formed based on a set road section (G) and a set value of the level of each weather type. The meteorological value (Q) can also be obtained from the meteorological monitoring data by weighted fusion, and a meteorological vector can be formed based on the set road section (G) and the meteorological value (Q).

[0119] As an example, the weights of various weather types (such as the rainstorm weight w rain , wind weight w wind , high temperature weight w high_temp , low temperature weight w low_temp and the fog weight w fog ) can be set by yourself, or obtained by other methods such as gradient descent algorithm, but the present invention is not limited thereto. rain +w wind +w high_temp +w low_temp +w fog =1.

[0120] The calculation formula of meteorological value (Q) can be expressed as:

[0121] Q=Zrain × rain +Z wind × wind +Z high_temp × high_temp +Z low_temp ×

[0122] w low_temp +Z fog × fog +b;

[0123] Where b represents bias, Z rain Indicates the set value corresponding to the rainstorm level, Z wind Indicates the setting value corresponding to the high wind level, Z high_temp Indicates the set value corresponding to the high temperature level, Z low_temp Indicates the set value corresponding to the low temperature level, Z fog Indicates the setting value corresponding to the heavy fog level.

[0124] As an example, the present application can adjust the impact of natural disasters on highways by adjusting the weights of various meteorological types, so the set values ​​set according to the meteorological type level can use values ​​of the same dimension. The above setting method of the set value is only an example, and other methods can also be used, such as obtaining the corresponding set value by normalizing the monitoring values ​​of various meteorological types, but the present invention is not limited thereto.

[0125] The above-mentioned method of obtaining meteorological values ​​by weighted fusion according to the set values ​​of each meteorological type and generating a meteorological vector (F) is only an example, and the present invention is not limited thereto.

[0126] (6) Considering that the driving speeds of vehicles of the same model are not much different, the present application can count the time interval distribution of vehicles of each model on a set road section to obtain vector data when the same model passes through the set road section.

[0127] In this application, a gantry vector (A) can be formed based on the vehicle type and the time when the vehicle passes the gantry. For example, the average and standard deviation of the time intervals of each vehicle type passing through the set road section can be calculated, and the gantry vector can be formed using the set road section (G), the vehicle type code (M) and the standard deviation σ of the time intervals of each vehicle type passing through the gantry section.

[0128] Among them, the vehicle model code (M) can be a number set according to the vehicle model, or it can be an existing code. Different vehicle model codes correspond to different vehicle models.

[0129] The above-mentioned method of generating the gantry vector in the present invention is only an example. Other vehicle traffic data can also be used to generate the gantry vector. For example, the gantry vector can be generated according to the average speed and average travel time of the vehicle. For example, It indicates the average speed of all vehicles passing through the portal section G1 during the set historical period. It represents the average time interval of the portal section G1 of all vehicles passing through in a set historical period, but the present invention is not limited thereto.

[0130] As an example, the vector type included in the natural disaster vector is determined according to the data type in the multidimensional highway data. For example, the risk and disaster vector, video vector, road maintenance vector, communication cable vector, meteorological vector and gantry vector can be integrated according to the highway section to form a natural disaster vector with the set section (G) as the primary key. At this point, the natural disaster vector corresponding to the historical data set can be expressed as

[0131] In some embodiments of the present invention, the set road section may be a long section. Therefore, in order to accurately predict the section where natural disasters may occur, the set road section can be divided into multiple adjacent gantry sections (referring to the section between adjacent highway gantries on the highway, which may also be referred to as gantry sections in this application), and the historical data set or real-time data set of each adjacent gantry section is input during the training or inference process, so as to obtain the natural disaster warning information of each adjacent gantry section through the highway natural disaster warning model.

[0132] For example, if the unit is the gantry section, the calculation method of the gantry vector (A) can be as follows:

[0133] Calculate the average time interval for all vehicles of type i to pass a certain gantry section The formula is as follows:

[0134]

[0135] Wherein, n represents the number of vehicles of the i-th type passing through adjacent gantry sections recorded in the vehicle traffic data within the set historical period, t1 is the time for a certain gantry in the gantry section for vehicle traffic, and t2 is the time for another gantry in the gantry section for vehicle traffic, that is, t2-t1 represents the time interval of a certain gantry section for vehicle traffic of the i-th type (both are positive values).

[0136] The standard deviation σ of the time interval for the i-th type of vehicle to pass a certain gantry section is calculated by the following formula i :

[0137]

[0138] Among them, σ i Indicates that the vehicle model code is M i Vehicle access gantry section G iThe standard deviation of the time interval.

[0139] Furthermore, since the acquired multi-source data may not directly correspond to the road section, in the case of adjacent gantry sections as units, the multi-source data and adjacent gantry sections can be matched in the following ways:

[0140] Use historical disaster and risk data to determine the location of the risk point (the specific location can be represented by longitude and latitude or starting and ending point pile numbers), and match adjacent gantry sections based on the longitude and latitude of the risk point location and the disaster location; match adjacent gantry sections based on camera location information (such as camera number); match adjacent gantry sections through pavement maintenance monitoring location information; use the connection relationship between the high-speed gantry and the communication optical cable to determine the communication optical cable on the adjacent gantry section; and match adjacent gantry sections through meteorological monitoring areas.

[0141] The numbering of adjacent gantry sections can be generated based on the original numbering of the gantry, or can be set by oneself, but the present invention is not limited to this.

[0142] In a specific embodiment of the present invention, the process of using the residual convolutional neural network model as a video feature extraction neural network model to obtain video feature data is as follows:

[0143] Assume that the video stream input to the residual convolutional neural network can be expressed as Among them, C in Indicates the number of channels of input video stream data, H in Indicates the height of the input video stream data, W in Indicates the width of the input video stream data.

[0144] The input video stream data can be used in the first convolution operation of the residual convolutional neural network using the convolution kernel Get the output feature map Among them, C out1 Indicates the number of output channels of the first layer of convolution operation, k h1 Represents the height of the first layer of convolution kernel, k w1 Represents the width of the first layer convolution kernel, the step size is s1, and the padding is p1.

[0145] The formula for calculating the size of the output feature map can be:

[0146]

[0147] Among them, H out1 Represents the output height of the first convolution operation, W out1 Represents the output width of the first layer of convolution operation.

[0148] The calculation formula of the output feature map I1(x,y,z) can be:

[0149]

[0150] Furthermore, after the first layer of convolution operation, the output feature map I1 will be pooled. Assume that the pooling window size is k p1 , the step length is s p1 , then after the pooling operation, the feature map can be output

[0151] The size calculation formula of the output feature map I1′ can be:

[0152]

[0153] In addition, for the residual block, assuming the input is a l , after residual transformation f l The output after can be expressed as b l , the residual formula can be expressed as:

[0154] b l =f l (a l ,{w li})+a l ;

[0155] Among them, a l is a function composed of convolutional layers and activation functions, etc., w li is the corresponding weight parameter.

[0156] After the first layer of convolution and pooling operations, the feature map is calculated using the same steps as above using the subsequent convolution and pooling layers in the residual convolutional neural network. After multiple layers of convolution, pooling, and residual operations, the final feature map output by the residual convolutional neural network can be obtained.

[0157] Flatten the final feature map to get a one-dimensional vector And the obtained one-dimensional vector I ′ m is regarded as the video feature data (X) along the highway. ′ m The calculation process can be expressed as:

[0158] I ′ m =Flatten(I m )=[I m (0,0,0),I m (0,0,1),…,I m (C outm -1,H outm -1,Woutm -1)];

[0159] As an example, the convolutional layers and pooling layers in the residual convolutional neural network are alternately arranged, and the number of convolutional layers in the residual convolutional neural network may be 10 to 50.

[0160] The present application does not specifically limit the structure of the residual convolutional neural network, and the existing network structure can be used. In addition, the residual convolutional neural network model mentioned above is only an example, and the video feature extraction neural network model can also be other types of models, and the present invention is not limited thereto.

[0161] Furthermore, the highway natural disaster warning method proposed in the present application also includes: within a set update period, inputting the updated historical data set into the initial warning model for retraining, and updating the hyperparameters in the model through multiple rounds of iterative training to obtain an updated highway natural disaster warning model.

[0162] Corresponding to the above method, the present application proposes a highway natural disaster warning system, including a processor, a memory, and a computer program / instructions stored in the memory, wherein the processor is used to execute the computer program / instructions, and when the computer program / instructions are executed, the device can implement the method described in any of the above embodiments.

[0163] Specifically, Figure 3 As shown, the system may include: a data access module, a data processing module, a model building module and a data storage module. The description is as follows:

[0164] The data access module can obtain structured data such as vehicle traffic data, historical disaster and risk data, road maintenance monitoring data, optical cable transmission status data and meteorological monitoring data from various source systems through the database-type JDBC interface protocol, HTTPS and MQTT protocols; and obtain video data from the highway monitoring and management system through RSTP.

[0165] The data processing module can use a streaming computing engine (such as a streaming computing engine flink) to process the multi-dimensional highway data obtained in real time by the data access module, and input it into the highway natural disaster warning model in a timely manner to reduce data processing delays and improve response speed. The data processing module can also use the vehicle traffic data, historical disaster and risk data, video stream, pavement maintenance monitoring data, optical cable transmission status data and meteorological monitoring data accessed by the data access module to calculate the gantry vector, risk and disaster vector, video vector, pavement maintenance vector, communication optical cable vector and meteorological vector respectively, and fuse or partially obtain the natural disaster vector. In addition, the obtained gantry vector, risk and disaster vector, video vector, pavement maintenance vector, communication optical cable vector, meteorological vector and natural disaster vector and other vector data can be stored in the vector library in the data storage module.

[0166] The message delivery module is used for multi-channel disaster warning delivery, and pushes user notification messages generated based on natural disaster warning information in real time and quickly to the highway variable information boards, navigation software, vehicle computers, OBU and other equipment or software in the corresponding sections.

[0167] The data storage module is used to store all data in the natural disaster warning process of highways. To facilitate data management and storage, the data storage module can use a distributed database to store data. The data resource library stores the accessed and processed multi-dimensional highway data in a hierarchical strategy, which is divided into: original database (originaldb) → cleaning database (cleandb) → basic database (basedb) → shared database (sharedb), and problem database (errordb). Among them, the original database stores the original data obtained by the data access module from various source systems without data preprocessing; the cleaning database stores the data after data preprocessing; the problem database stores the problem data in the data cleaning process; the data stored in the basic database is the data obtained after the data in the cleaning database is integrated, which is the most complete; the data set in the shared database is organized according to the needs of different business systems. It is usually not recommended to use the original database and the problem database directly externally. The shared database gives priority to providing services to the outside world, and the basic database can provide services to the outside world.

[0168] The data storage module may also include a vector library and a model library constructed based on the network model. The model library stores models such as the initial warning model, multiple versions of highway natural disaster warning models, video feature extraction neural network models, feature field extraction models, and model parameter data.

[0169] As an example, the highway natural disaster warning system in the present application may also include a model building module for building and training multiple models required in the warning process, and for building an initial warning model, a video feature extraction neural network model, and a feature field extraction model, etc. Among them, the feature field extraction model and the initial warning model may be neural network models built based on a multi-layer perceptron, and the video feature extraction neural network model may be a residual convolutional neural network model.

[0170] During the model training phase, the application uses the historical data of each source system within a set historical period and the corresponding warning labels to form a historical data set; during the real-time warning phase, the application can obtain multidimensional data from each source system in real time to generate a real-time data set.

[0171] The highway natural disaster early warning method and system proposed in this application have the following advantages:

[0172] (1) Intelligent warning: A natural disaster warning model for highways is generated based on the fusion calculation of multi-source data, which reduces the cost of manual inspections and the probability of misjudgment due to human factors.

[0173] (2) Rapid response: When the network is good, it is possible to obtain multi-source data in real time and quickly obtain data push results, so that vehicles near the disaster site have enough time to respond and reduce driving risks.

[0174] (3) Wide range of notifications: By integrating accessible devices along the highway and connecting with navigation software manufacturers, based on basic data such as vehicle gantries, notifications are quickly pushed to the corresponding highway variable information boards, navigation software, vehicle computers, vehicle-mounted terminals and other equipment or software, so as to reach users comprehensively and timely.

[0175] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0176] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0177] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0178] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A natural disaster early warning method for highways, characterized in that: The method comprises the following steps: Acquire multi-dimensional highway data of a set section, and form a real-time data set of the corresponding section based on the multi-dimensional highway data; wherein the multi-dimensional highway data includes historical disaster and risk data, video data, road maintenance monitoring data and meteorological data; Inputting the real-time data set into a pre-trained highway natural disaster warning model, and outputting natural disaster warning information for the corresponding road section; Sending a user notification message generated based on the natural disaster warning information to a user access device, so that the user receives the user notification message; The pre-trained highway natural disaster warning model is obtained in the following way: Acquire a historical data set, input the historical data set into an initial early warning model, and obtain the highway natural disaster early warning model through multiple rounds of iterative training; The historical data set is obtained based on multi-dimensional highway data and warning labels within a set historical period, and the warning labels and the natural disaster warning information both include disaster warning labels for indicating whether a disaster has occurred and the disaster warning level.

2. The method according to claim 1, characterized in that The multi-dimensional highway data also includes: optical cable transmission status data and vehicle traffic data; wherein the vehicle traffic data is determined based on communication identification between the vehicle and the highway gantry; The meteorological data includes meteorological monitoring data and meteorological forecast data, and the meteorological data includes the following field information: meteorological monitoring area, meteorological type and level of meteorological type.

3. The method according to claim 2, characterized in that The historical disaster and risk data include historical disaster data and risk point data, the historical disaster data include the location of the disaster point and the characteristic information of the disaster loss, and the risk point data include the location of the risk point and the characteristic information of historical disasters; wherein the characteristic information of historical disasters includes one or more of the following field information: terrain characteristics, the frequency of occurrence of historical disasters, the time of occurrence of historical disasters, the type of historical disasters, the scale of historical disasters, the degree of harm caused by historical disasters, the objects of harm caused by historical disasters, the high-speed disconnection status of historical disasters and the treatment of historical disasters, wherein the type of historical disasters is collapse, landslide, mud-rock flow, subsidence and collapse, water damage or high slope; The pavement maintenance monitoring data includes the pavement maintenance monitoring location, the pavement maintenance monitoring type and the monitoring value of each pavement maintenance monitoring type, wherein the pavement maintenance monitoring type includes one or more of displacement monitoring, disaster inducing factor monitoring, protective facility monitoring and short-term monitoring; The optical cable transmission status data includes a communication optical cable number and optical cable transmission information, wherein the optical cable transmission information is measured by one or more of the following performance indicators: signal strength, transmission rate, and bit error rate; The vehicle traffic data includes the high-speed gantry number, the time when the vehicle passes the high-speed gantry, and the vehicle model; and The video data includes camera number and monitoring video image data.

4. The method according to claim 3, characterized in that The forming of a real-time data set of a corresponding road section based on the multi-dimensional highway data includes: Utilize a feature field extraction model to extract specific field information from the historical disaster and risk data to form a risk and disaster vector; utilize a video feature extraction neural network model to extract feature data from the monitoring video image data of the video data to form a video vector; determine a pavement maintenance vector based on the monitoring values ​​of each pavement maintenance monitoring type; obtain a communication optical cable vector based on the optical cable transmission information in the optical cable transmission status data; obtain a meteorological vector based on the level of each meteorological type; and form a gantry vector based on the vehicle type and the time when the vehicle passes the gantry; Based on the set road section, the risk and disaster vector, video vector, road maintenance vector, communication optical cable vector, meteorological vector and gantry vector are integrated to obtain a natural disaster vector corresponding to the set road section.

5. The method according to claim 4, characterized in that The feature field extraction model and the initial warning model are neural network models constructed based on multi-layer perceptrons, and the video feature extraction neural network model is a residual convolutional neural network model; The vehicle traffic data, road maintenance monitoring data and meteorological data are obtained through the Hypertext Transfer Protocol (HTTPS), and the video data is obtained through the Rapid Spanning Tree Protocol (RSTP); the historical disaster and risk data are obtained through the Java Database Connection (JDBC) protocol; the optical cable transmission status data is obtained through the Message Queue Telemetry Transmission (MQTT) protocol; Before determining the pavement maintenance vector, the method further includes: normalizing the acquired pavement maintenance monitoring data.

6. The method according to claim 1, characterized in that The method further comprises: Before forming the real-time data set, performing data cleaning on the acquired multi-dimensional highway data; and / or During a set update period, the initial warning model is retrained using an updated historical data set to obtain an updated highway natural disaster warning model.

7. The method according to claim 1, characterized in that The user notification message includes: an operation instruction sent to a user access device within a first range of a set road section, and natural disaster warning information sent to users within a second range of the set road section, wherein the first range of the set road section is smaller than the second range of the set road section; Before sending the user notification message to the user access device, the method further includes: The order in which user notification messages are sent is determined based on the set vehicle model priority, wherein the lower the vehicle model priority, the later the order in which user notification messages are sent.

8. The method according to claim 1, characterized in that The sending a user notification message generated based on the natural disaster warning information to the user access device includes: A user notification message generated based on the natural disaster warning information is sent to a message reaching adaptation module, so as to push the user notification message to the user reaching device by utilizing the communication protocol between the message reaching adaptation module and the user reaching device.

9. The method according to claim 8, characterized in that The user access device includes a vehicle-mounted device; The using the communication protocol between the message reaching adaptation module and the user reaching device to push the user notification message to the user reaching device includes: Based on the pre-stored position correspondence between the road section and the high-speed gantry, the message reaching adaptation module is used to push the user notification message to the high-speed gantry within the set communication range, and then the user notification message is sent to the user by using the communication protocol between the high-speed gantry and the vehicle-mounted equipment.

10. A highway natural disaster early warning system, characterized in that: The device comprises a processor, a memory and a computer program / instruction stored in the memory, wherein the processor is used to execute the computer program / instruction, and when the computer program / instruction is executed, the device implements the steps of the method as claimed in any one of claims 1 to 9.

Images