Zero-contact traffic monitoring method, system and equipment and computer medium

Abstract

The invention discloses a traffic zero-contact monitoring method, system and device and a computer medium, and relates to the technical field of intelligent traffic monitoring and digital twinning, and the method comprises the steps: obtaining multi-source target data and monitoring demand information of a target traffic scene; processing the multi-source target data to generate a target structured semantic representation; analyzing the monitoring demand information to obtain target constraint information for constraining the monitoring interest attribute; obtaining target priori knowledge of the target traffic scene; processing the target structured semantic representation, the target constraint information and the target priori knowledge through a pre-trained large language model to generate target update information of the target monitoring attribute; and updating the digital twin state of the target traffic scene based on the target update information. According to the invention, the target updating information of the target monitoring attribute is automatically generated to update the digital twin state of the target traffic scene, and the response efficiency and scene adaptability of zero-contact monitoring are improved.

Description

Technical Field

[0001] This application relates to the field of intelligent traffic monitoring and digital twin technology, and more specifically, to a zero-contact traffic monitoring method, system, device, and computer medium. Background Technology

[0002] Currently, to achieve refined perception and dynamic management of traffic conditions, traffic monitoring systems typically need to continuously acquire and analyze the spatial location, movement status, and road operation indicators of traffic participants to provide a basis for subsequent decision-making. The emergence of digital twin technology offers new hope for solving traffic problems.

[0003] Digital twin technology models real-world transportation systems into virtual simulation environments, enabling the simulation and optimization of traffic flow, traffic planning, and traffic control. However, traffic monitoring and twin update processes largely rely on manual configuration of monitoring attributes and manual writing and maintenance of interest attribute calculation / simulation logic. This makes it difficult to automatically adjust attribute acquisition and update processes when monitoring needs change dynamically, resulting in high operation and maintenance costs, slow response times, and insufficient adaptability to different scenarios.

[0004] In conclusion, improving the response efficiency and scenario adaptability of traffic monitoring is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide a zero-contact traffic monitoring method, which can, to some extent, solve the technical problem of how to improve the response efficiency and scene adaptability of traffic monitoring. This application also provides a zero-contact traffic monitoring system, electronic equipment, and computer-readable storage medium.

[0006] To achieve the above objectives, this application provides the following technical solution: Firstly, this application provides a zero-contact traffic monitoring method, including: Acquire multi-source target data and monitoring requirement information for the target traffic scenario; The multi-source target data is processed to generate a structured semantic representation of the target; The monitoring requirement information is parsed to obtain target constraint information used to constrain the monitoring interest attributes; Obtain the target prior knowledge of the target traffic scenario; By using a pre-trained large language model, the target structured semantic representation, the target constraint information, and the target prior knowledge are processed to generate target update information for the target monitoring attributes. Based on the target update information, the digital twin state of the target traffic scenario is updated.

[0007] On the other hand, acquiring multi-source target data for the target traffic scenario includes: Acquire multi-source target data of the target traffic scene, wherein the multi-source target data includes one or more of the following: surveillance camera video data, roadside perception data, vehicle sensor data, signal control status data, V2X communication data, and high-precision map data.

[0008] On the other hand, the multi-source target data is processed to generate a structured semantic representation of the target, including: The multi-source target data is spatiotemporally aligned to obtain multi-source initial data; The initial multi-source data is cleaned to obtain cleaned multi-source data; Attribute extraction is performed on the multi-source cleaned data to obtain the basic attribute set of traffic participants; The basic attribute set is transformed to obtain the target structured semantic representation; The basic attribute set includes the identifier, category, timestamp, spatial location, speed, orientation, and confidence level of traffic participants; the structured semantic representation includes one or more of the following: object entity table, attribute table, relation table, event log, and spatiotemporal index.

[0009] On the other hand, the monitoring requirement information is parsed to obtain target constraint information for constraining monitoring interest attributes, including: The monitoring requirement information is parsed to obtain target constraint information, which includes a candidate set of interest attributes, attribute priority, update frequency, time delay constraint, monitoring object range, and spatiotemporal constraint.

[0010] On the other hand, before processing the target structured semantic representation, the target constraint information, and the target prior knowledge using a pre-trained large language model, the process also includes: Acquire the training structured semantic representation, training constraint information, and training prior knowledge of the training traffic scenario; Determine the complete set of candidate interest attributes for the training traffic scenario; Based on the attribute weights, attribute costs, and contributions to monitoring needs of the complete set of candidate interest attributes, a set of candidate interest attributes is determined. Based on the large language model, the conditional probability of the set of candidate interest attributes under the training structured semantic representation and the training constraint information is determined; Based on the prior training knowledge and the conditional probability, the training monitoring attributes are determined; Generate training update information for the training monitoring attributes; Based on the training structured semantic representation, the training constraint information, and the training prior knowledge, the large language model is trained to obtain a trained large language model.

[0011] On the other hand, generating training update information for the training monitoring attributes includes: Based on the training monitoring attributes, a structured update instruction is generated, which includes one or more of the following: object addition, object deletion, object status update, attribute field update, relationship edge update, and event log writing. Based on a preset update template, the structured update instructions are converted into training update information that can be executed by the twin.

[0012] On the other hand, after updating the digital twin state of the target traffic scene based on the target update information, the process also includes: Obtain the observation results of the observed attributes in the target traffic scenario, wherein the observed attributes include traffic flow and lane occupancy rate; Obtain the inference results of the observed attributes in the digital twin; Generate the error between the inference result and the observation result; If the error exceeds a preset threshold, the monitoring update strategy will be downgraded. Degradation includes reducing the update frequency, reducing the number of updated fields, or updating only one or more key attributes.

[0013] Secondly, a zero-contact traffic monitoring system is provided, including: The target scene acquisition module is used to acquire multi-source target data and monitoring requirement information of the target traffic scene; The target data processing module is used to process the multi-source target data and generate a structured semantic representation of the target. The parsing module is used to parse the monitoring requirement information to obtain target constraint information for constraining the monitoring interest attributes; The target prior knowledge acquisition module is used to acquire the target prior knowledge of the target traffic scenario; The target update information generation module is used to process the target structured semantic representation, the target constraint information and the target prior knowledge through a pre-trained large language model to generate target update information of the target monitoring attributes; The update module is used to update the digital twin state of the target traffic scenario based on the target update information.

[0014] Thirdly, an electronic device is provided, comprising: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of any of the above-described zero-contact traffic monitoring methods.

[0015] Fourthly, a computer-readable storage medium is provided, wherein a computer program is stored therein, and when executed by a processor, the computer program implements the steps of any of the above-described zero-contact traffic monitoring methods.

[0016] This application provides a zero-contact traffic monitoring method, which involves acquiring multi-source target data and monitoring requirement information of a target traffic scene; processing the multi-source target data to generate a structured semantic representation of the target; parsing the monitoring requirement information to obtain target constraint information used to constrain monitoring interest attributes; acquiring prior knowledge of the target traffic scene; processing the target structured semantic representation, target constraint information, and target prior knowledge through a pre-trained large language model to generate target update information of the target monitoring attributes; and updating the digital twin state of the target traffic scene based on the target update information. In this application, because multi-dimensional target data can reflect the information of the target traffic scene from multiple perspectives, the processed target structured semantic representation can reflect the characteristics of the target traffic scene in detail, providing a good data foundation for subsequent monitoring updates. Furthermore, monitoring requirements need to be converted into target constraint information that constrains monitoring interest attributes, avoiding disordered and large-scale attribute filtering and ensuring the accuracy of attribute updates. Finally, based on prior knowledge constraints and large language model reasoning, target update information for target monitoring attributes is automatically generated to update the digital twin state of the target traffic scene. The entire process requires no manual configuration or maintenance, reducing operation and maintenance costs, improving the response efficiency of zero-contact monitoring, and can dynamically adapt to different monitoring needs, demonstrating strong scenario adaptability. The traffic zero-monitoring system, electronic device, and computer-readable storage medium provided in this application also solve the corresponding technical problems. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a zero-contact traffic monitoring method provided in this application embodiment; Figure 2 This application provides a schematic diagram of the structure of a zero-contact traffic monitoring system as an embodiment of the present application. Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application; Figure 4 This is another structural schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Please see Figure 1 , Figure 1 A flowchart of a zero-contact traffic monitoring method provided in an embodiment of this application.

[0021] This application provides a zero-contact traffic monitoring method, which may include the following steps: Step S101: Obtain multi-source target data and monitoring requirement information for the target traffic scenario.

[0022] In practical applications, multi-source target data and monitoring requirement information for the target traffic scenario can be acquired first. Multi-source target data is used to characterize the corresponding information of the target traffic scenario through data collected from different sources. Monitoring requirement information is used to characterize the monitoring needs for the target traffic scenario. This monitoring requirement information can be configured by the user or automatically generated by the digital twin scenario. Furthermore, the target traffic scenario can be flexibly determined according to actual needs; for example, the target traffic scenario could be an intersection traffic scenario or a road traffic scenario, etc.

[0023] In an exemplary embodiment, the source of multi-source target data can be flexibly determined according to the application scenario. For example, in the process of acquiring multi-source target data of a target traffic scenario, the multi-source target data may include one or more of the following: surveillance camera video data, roadside perception data, vehicle-mounted sensor data, signal control status data, V2X (Vehicle-to-Everything) communication data, and high-precision map data.

[0024] Step S102: Process the multi-source target data to generate a structured semantic representation of the target.

[0025] In practical applications, in order to facilitate the autonomous processing of data by computers, it is also necessary to process multi-source target data to generate a target structured semantic representation that is easy for computers to process.

[0026] In an exemplary embodiment, during the processing of multi-source target data to generate a structured semantic representation of the target, spatiotemporal alignment can be performed on the multi-source target data to eliminate differences in sampling frequency, transmission delay, and coordinate system between different data sources, thereby obtaining initial multi-source data. Spatiotemporal alignment can be divided into time alignment and spatial alignment. During time alignment, a unified time grid can be set as Δt, and the observation of any data source at time t can be mapped to the aligned timestamp. , ,in, To align the start times, this time alignment ensures that data with different frequencies and delays fall on the same time base. During spatial alignment, the points in the coordinate system of the i-th data source can be aligned. Transform to a unified world coordinate system. ,in, Let be a rotation matrix. This is a translation vector used to eliminate differences in sensor mounting pose. express The values ​​are defined in a unified world coordinate system. The initial multi-source data is cleaned, including outlier removal, missing value completion, deduplication, conflict arbitration, and low-confidence filtering, resulting in cleaned multi-source data. Attribute extraction is performed on the cleaned multi-source data to obtain a set of basic attributes for traffic participants. This set includes the participant's identifier, category, timestamp, spatial location, speed, orientation, and confidence level. The state vector of object o at time t is defined as follows. , ,in, Indicates location, Indicates speed, Indicates direction, It represents the category; it transforms the basic attribute set to obtain the target structured semantic representation, which includes one or more of the following: object entity table, attribute table, relationship table, event log, and spatiotemporal index.

[0027] Step S103: Parse the monitoring requirement information to obtain target constraint information used to constrain the monitoring interest attributes.

[0028] In practical applications, the main content of monitoring requirement information is the corresponding information of monitoring attributes. To facilitate subsequent processing, the monitoring requirement information can be parsed to obtain the target constraint information used to constrain the monitoring interest attributes.

[0029] In an exemplary embodiment, the type of target constraint information can be flexibly determined according to the application scenario. For example, in the process of parsing monitoring requirement information to obtain target constraint information for constraining monitoring interest attributes, the monitoring requirement information can be parsed to obtain target constraint information, which includes a candidate set of interest attributes, attribute priority, update frequency, latency constraint, monitoring object range, spatiotemporal constraint, available data constraint, etc.

[0030] Step S104: Obtain the target prior knowledge of the target traffic scenario.

[0031] In practical applications, to avoid the monitoring attribute update range being too large and deviating from the monitoring needs of the target traffic scenario, it is also possible to obtain the target prior knowledge of the target traffic scenario so that the target prior knowledge can be used to constrain, verify and filter the candidate results of interest attributes output by the large language model.

[0032] In an exemplary embodiment, prior knowledge may include one or more of the following: constraint rules related to interest attributes, statistical priors, or topological consistency constraints; and in the process of updating prior knowledge, in order to avoid excessive interference with prior knowledge and affecting the derivation results of the large language model, prior knowledge can be learned and updated based on a small number of real attribute samples, which can be extracted from historical monitoring data or manually labeled samples.

[0033] Step S105: Process the target structured semantic representation, target constraint information and target prior knowledge through a pre-trained large language model to generate target update information for target monitoring attributes.

[0034] In practical applications, after obtaining the target's structured semantic representation, target constraint information, and target prior knowledge, a pre-trained large language model can be used to process these elements to generate target update information for the target's monitoring attributes. , , t Indicates a time step. S This represents a structured semantic representation. R Represents constraint information, P This represents prior knowledge; the target update information is used to implement one or more of the following: adding objects, deleting objects, updating object status, updating attribute fields, updating relationship edges, and writing event logs. The target update information can be update code, etc., without specific limitations.

[0035] In an exemplary embodiment, before processing the target structured semantic representation, target constraint information, and target prior knowledge using a pre-trained large language model, a usable large language model needs to be trained. During this process, the training structured semantic representation, training constraint information, and training prior knowledge of the training traffic scene can be obtained, and the principle is the same as that of the corresponding target structured semantic representation, target constraint information, and target prior knowledge. The complete set of candidate interest attributes for the training traffic scene is determined, denoted as [equation missing]. , a Indicates interest attributes, m Indicates the number of the interest attribute; the feasibility screening of the candidate interest attributes meets the requirements. Consistency determination of the constraints shown. Denotes the j-th prior constraint, 1[ [] is an indicator function used to explicitly use prior knowledge as the reasoning boundary to eliminate unreasonable candidates and improve the reliability of reasoning. Based on this, attribute weights of the entire set of candidate interest attributes can be used. Attribute Cost and contribution to monitoring needs Determine the set of potential interest attributes , , As a trade-off coefficient, For interest attributes The cost is used to characterize the resource overhead of inferring this attribute under the current configuration; based on the large language model, the conditional probability of the set of candidate interest attributes is determined under the training structured semantic representation S and the training constraint information R. Based on prior training knowledge and conditional probabilities, determine training monitoring attributes, such as the set of prior constraints. Generate a joint score for each attribute. , The highest scorer was selected as the training monitoring attribute. , , To express the degree of violation of the j-th prior constraint, max(0, ) indicates that only the violation is punished, and λ is the punishment coefficient; generate training update information for training monitoring attributes; based on the training structured semantic representation, training constraint information and training prior knowledge, train the large language model to obtain the trained large language model. During the training process, a set of instruction samples for traffic monitoring tasks can be applied. This set of instruction samples contains at least one or more of the following pairs: "structured semantic input - interest attribute output" and "constraint condition - feasibility judgment".

[0036] In specific application scenarios, during the process of generating training update information for training monitoring attributes, structured update instructions can be generated based on the training monitoring attributes. The structured update instructions include one or more of the following: object addition, object deletion, object status update, attribute field update, relationship edge update, and event log writing. Based on a preset update template, structured update instructions are transformed into training update information that can be executed by the twin.

[0037] Step S106: Update the digital twin state of the target traffic scene based on the target update information.

[0038] In practical applications, after generating target update information for the target monitoring attributes, the digital twin state of the target traffic scene can be updated based on this information. , Indicates time t The status of digital twins in transportation U ( ) indicates that the update execution operator is performed. This represents the traffic digital twin state updated after time step Δt, which is used to update the corresponding information of the target monitoring attributes in the digital twin data of the target traffic scenario, thereby providing reliable support for traffic simulation, prediction and strategy evaluation.

[0039] In an exemplary embodiment, considering the possibility of errors in monitoring updates, the monitoring process can be proactively adjusted to ensure accuracy. For example, after updating the digital twin state of the target traffic scene based on the target update information, the observation results of the observed attributes in the target traffic scene can be obtained. The observed attributes include traffic flow q and lane occupancy rate o. Accordingly, the observation results are represented as follows: and , , ρ is the lane density, and v is the observed speed. Kocc This represents the occupancy rate normalization coefficient. Vmin This represents the lower limit constant of velocity. Kq Indicates the conversion factor for flow units; obtains the inference results of observed attributes in digital twins. and Error between the generated inference results and the observed results , If the error exceeds the preset threshold, the monitoring update strategy will be downgraded. Downgrading includes reducing the update frequency, reducing the number of updated fields, or updating only key attributes. Of course, in addition to downgrading the update strategy, re-inference or rollback update can also be triggered, which is not specifically limited here.

[0040] This application provides a zero-contact traffic monitoring method, which involves acquiring multi-source target data and monitoring requirement information of a target traffic scene; processing the multi-source target data to generate a structured semantic representation of the target; parsing the monitoring requirement information to obtain target constraint information used to constrain monitoring interest attributes; acquiring prior knowledge of the target traffic scene; processing the target structured semantic representation, target constraint information, and target prior knowledge through a pre-trained large language model to generate target update information of the target monitoring attributes; and updating the digital twin state of the target traffic scene based on the target update information. In this application, since multi-dimensional target data can reflect the information of the target traffic scene from multiple perspectives, the processed target structured semantic representation can reflect the characteristics of the target traffic scene in more detail, providing a good data foundation for subsequent monitoring updates. Furthermore, the monitoring requirements need to be converted into target constraint information that constrains the monitoring interest attributes, avoiding disordered and large-scale attribute filtering and ensuring the accuracy of attribute updates. Finally, based on prior knowledge constraints and large language model reasoning, target update information of target monitoring attributes is automatically generated to update the digital twin state of the target traffic scene. The entire process does not require manual configuration or maintenance, reducing operation and maintenance costs, improving the response efficiency of zero-contact monitoring, and can dynamically adapt to different monitoring requirements, demonstrating strong scene adaptability.

[0041] Please see Figure 2 , Figure 2 This is a schematic diagram of a zero-contact traffic monitoring system provided in an embodiment of this application.

[0042] This application provides a zero-contact traffic monitoring system, which may include: The target scene acquisition module 101 is used to acquire multi-source target data and monitoring requirement information of the target traffic scene; The target data processing module 102 is used to process multi-source target data and generate a structured semantic representation of the target. The parsing module 103 is used to parse the monitoring requirement information to obtain target constraint information for constraining the monitoring interest attributes; The target prior knowledge acquisition module 104 is used to acquire the target prior knowledge of the target traffic scenario. The target update information generation module 105 is used to process the target structured semantic representation, target constraint information and target prior knowledge through a pre-trained large language model to generate target update information of target monitoring attributes. The update module 106 is used to update the digital twin state of the target traffic scenario based on the target update information.

[0043] This application provides a zero-contact traffic monitoring system, wherein the target scene acquisition module may include: The multi-source target data acquisition unit is used to acquire multi-source target data of the target traffic scene. The multi-source target data includes one or more of the following: surveillance camera video data, roadside perception data, vehicle sensor data, signal control status data, V2X communication data, and high-precision map data.

[0044] This application provides a zero-contact traffic monitoring system, wherein the target data processing module may include: The spatiotemporal alignment unit is used to perform spatiotemporal alignment on multi-source target data to obtain multi-source initial data. The cleaning unit is used to clean the initial data from multiple sources to obtain cleaned data from multiple sources. The attribute extraction unit is used to extract attributes from multi-source cleaned data to obtain the basic attribute set of traffic participants; The transformation unit is used to transform the basic attribute set to obtain the target structured semantic representation; The basic attribute set includes the identifier, category, timestamp, spatial location, speed, orientation, and confidence level of traffic participants; the structured semantic representation includes one or more of the following: object entity table, attribute table, relation table, event log, and spatiotemporal index.

[0045] This application provides a zero-contact traffic monitoring system, wherein the parsing module may include: The parsing unit is used to parse the monitoring requirement information to obtain the target constraint information, which includes the candidate set of interest attributes, attribute priority, update frequency, time delay constraint, monitoring object range, and spatiotemporal constraint.

[0046] The zero-contact traffic monitoring system provided in this application embodiment may further include: The training data acquisition module is used by the target update information generation module to acquire the training structured semantic representation, training constraint information and training prior knowledge of the training traffic scenario before the target update information generation module processes the target structured semantic representation, target constraint information and target prior knowledge through the pre-trained large language model. The candidate interest determination module is used to determine the complete set of candidate interest attributes for the training traffic scene; The candidate interest determination module is used to determine the set of candidate interest attributes based on the attribute weights, attribute costs, and contributions to monitoring needs of the entire set of candidate interest attributes. The conditional probability determination module is used to determine the conditional probability of the set of candidate interest attributes under the training of structured semantic representation and training constraint information based on a large language model. The training monitoring attribute determination unit is used to determine training monitoring attributes based on prior training knowledge and conditional probabilities. The training update information generation module is used to generate training update information for training monitoring attributes. The training module is used to train a large language model based on training structured semantic representation, training constraint information, and training prior knowledge, so as to obtain a trained large language model.

[0047] This application provides a zero-contact traffic monitoring system, wherein the training update information generation module may include: The update instruction generation unit is used to generate structured update instructions based on the training monitoring attributes. The structured update instructions include one or more of the following: object addition, object deletion, object status update, attribute field update, relationship edge update, and event log writing. The training update information generation unit is used to convert structured update instructions into training update information that can be executed by the twin based on a preset update template.

[0048] The zero-contact traffic monitoring system provided in this application embodiment may further include: The observation result acquisition module is used to update the digital twin state of the target traffic scene based on the target update information, and then acquire the observation results of the observation attributes in the target traffic scene. The observation attributes include traffic flow and lane occupancy rate. The reasoning result acquisition module is used to acquire the reasoning results of observed attributes in the digital twin; The error generation module is used to generate the error between the inference results and the observation results; The adjustment module is used to downgrade the monitoring update strategy in response to errors exceeding a preset threshold. Degradation includes reducing the update frequency, reducing the number of updated fields, or updating only one or more key attributes.

[0049] This application also provides an electronic device and a computer-readable storage medium, both of which have the corresponding effects of the zero-contact traffic monitoring method provided in the embodiments of this application. Please refer to... Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0050] An electronic device provided in this application includes a memory 201 and a processor 202. The memory 201 stores a computer program, and the processor 202 executes the computer program to implement the steps of the zero-contact traffic monitoring method described in any of the above embodiments.

[0051] Please see Figure 4Another electronic device provided in this application embodiment may further include: an input port 203 connected to the processor 202 for transmitting commands input from the outside to the processor 202; a display unit 204 connected to the processor 202 for displaying the processing results of the processor 202 to the outside; and a communication module 205 connected to the processor 202 for enabling communication between the electronic device and the outside. The display unit 204 may be a display panel, a laser scanning display, etc.; the communication method adopted by the communication module 205 includes, but is not limited to, Mobile High-Definition Link (MHL), Universal Serial Bus (USB), High-Definition Multimedia Interface (HDMI), wireless connection: Wireless Fidelity (WiFi), Bluetooth communication technology, Bluetooth Low Energy communication technology, and communication technology based on IEEE 802.11s.

[0052] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the zero-contact traffic monitoring method described in any of the above embodiments.

[0053] The computer-readable storage media involved in this application include random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs (compact disc read-only memory), or any other form of storage media known in the art.

[0054] This application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the zero-contact traffic monitoring method described in any of the above embodiments.

[0055] For descriptions of relevant parts in the zero-contact traffic monitoring system, electronic device, and computer-readable storage medium provided in this application's embodiments, please refer to the detailed description of the corresponding parts in the zero-contact traffic monitoring system provided in this application's embodiments; they will not be repeated here. Furthermore, parts of the technical solutions provided in this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0056] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0057] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A zero-contact traffic monitoring method, characterized in that, include: Acquire multi-source target data and monitoring requirement information for the target traffic scenario; The multi-source target data is processed to generate a structured semantic representation of the target; The monitoring requirement information is parsed to obtain target constraint information used to constrain the monitoring interest attributes; Obtain the target prior knowledge of the target traffic scenario; By using a pre-trained large language model, the target structured semantic representation, the target constraint information, and the target prior knowledge are processed to generate target update information for the target monitoring attributes. Based on the target update information, the digital twin state of the target traffic scenario is updated.

2. The method according to claim 1, characterized in that, Acquire multi-source target data for the target traffic scenario, including: Acquire multi-source target data of the target traffic scene, wherein the multi-source target data includes one or more of the following: surveillance camera video data, roadside perception data, vehicle sensor data, signal control status data, V2X communication data, and high-precision map data.

3. The method according to claim 2, characterized in that, The multi-source target data is processed to generate a structured semantic representation of the target, including: The multi-source target data is spatiotemporally aligned to obtain multi-source initial data; The initial multi-source data is cleaned to obtain cleaned multi-source data; Attribute extraction is performed on the multi-source cleaned data to obtain the basic attribute set of traffic participants; The basic attribute set is transformed to obtain the target structured semantic representation; The basic attribute set includes the identifier, category, timestamp, spatial location, speed, orientation, and confidence level of traffic participants; the structured semantic representation includes one or more of the following: object entity table, attribute table, relation table, event log, and spatiotemporal index.

4. The method according to claim 1, characterized in that, The monitoring requirement information is parsed to obtain target constraint information for constraining monitoring interest attributes, including: The monitoring requirement information is parsed to obtain target constraint information, which includes a candidate set of interest attributes, attribute priority, update frequency, time delay constraint, monitoring object range, and spatiotemporal constraint.

5. The method according to claim 1, characterized in that, Before processing the target structured semantic representation, the target constraint information, and the target prior knowledge using a pre-trained large language model, the process further includes: Acquire the training structured semantic representation, training constraint information, and training prior knowledge of the training traffic scenario; Determine the complete set of candidate interest attributes for the training traffic scenario; Based on the attribute weights, attribute costs, and contributions to monitoring needs of the complete set of candidate interest attributes, a set of candidate interest attributes is determined. Based on the large language model, the conditional probability of the set of candidate interest attributes under the training structured semantic representation and the training constraint information is determined; Based on the prior training knowledge and the conditional probability, the training monitoring attributes are determined; Generate training update information for the training monitoring attributes; Based on the training structured semantic representation, the training constraint information, and the training prior knowledge, the large language model is trained to obtain a trained large language model.

6. The method according to claim 5, characterized in that, Generating training update information for the training monitoring attributes includes: Based on the training monitoring attributes, a structured update instruction is generated, which includes one or more of the following: object addition, object deletion, object status update, attribute field update, relationship edge update, and event log writing. Based on a preset update template, the structured update instructions are converted into training update information that can be executed by the twin.

7. The method according to claim 1, characterized in that, After updating the digital twin state of the target traffic scene based on the target update information, the process further includes: Obtain the observation results of the observed attributes in the target traffic scenario, wherein the observed attributes include traffic flow and lane occupancy rate; Obtain the inference results of the observed attributes in the digital twin; Generate the error between the inference result and the observation result; If the error exceeds a preset threshold, the monitoring update strategy will be downgraded. Degradation includes reducing the update frequency, reducing the number of updated fields, or updating only one or more key attributes.

8. A zero-contact traffic monitoring system, characterized in that, include: The target scene acquisition module is used to acquire multi-source target data and monitoring requirement information of the target traffic scene; The target data processing module is used to process the multi-source target data and generate a structured semantic representation of the target. The parsing module is used to parse the monitoring requirement information to obtain target constraint information for constraining the monitoring interest attributes; The target prior knowledge acquisition module is used to acquire the target prior knowledge of the target traffic scenario; The target update information generation module is used to process the target structured semantic representation, the target constraint information and the target prior knowledge through a pre-trained large language model to generate target update information of the target monitoring attributes; The update module is used to update the digital twin state of the target traffic scenario based on the target update information.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the zero-contact traffic monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the zero-contact traffic monitoring method as described in any one of claims 1 to 7.

Images