Smart traffic terminal configuration method, system, device and program product

Abstract

The invention provides a smart traffic terminal configuration method, system and device and a program product. The smart traffic terminal configuration method comprises the steps of obtaining a natural language-based traffic terminal configuration text input by a user; identifying a configuration intention of the user based on the traffic terminal configuration text; standard configuration parameters matched with the traffic terminal configuration text are extracted from a standard configuration parameter set corresponding to the configuration intention of the user and serve as configuration parameters to be executed; and determining a configuration identification result based on the configuration intention of the user and the to-be-executed configuration parameters, and executing configuration on the intelligent traffic terminal based on the configuration identification result. According to the method, automatic configuration of the intelligent traffic terminal is realized through semantic recognition of the natural language text, the configuration method is simple, interference of configuration parameters of other configuration intentions is avoided by recognizing the configuration intention of the user and performing parameter matching in the standard configuration parameter set corresponding to the configuration intention, and the matching accuracy and efficiency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a smart traffic terminal configuration method, system, device and program product. BACKGROUND

[0002] With the rapid development of social economy and technology, the steady advancement of national smart city, as an important infrastructure of smart city, the smart traffic system has been widely applied in various road scenes such as urban roads, highways, urban ring lines, etc. Relying on visual perception and video analysis technology, the system can realize real-time perception of vehicle driving state, road environment and traffic participant behavior, and accurately identify traffic violations such as red light running, reverse driving, overspeed, etc. Through real-time traffic flow monitoring and response, the smart traffic system realizes the optimization of main road signals and the balance of traffic flow, and significantly improves the overall operation efficiency and management level. Overall, the smart traffic system provides a solid technical support for the precise control and safety protection of urban and highway networks with its multi-dimensional capabilities such as autonomous detection, tracking and identification, and becomes an important pillar of modern urban governance and transportation management. However, in practical application, there are certain limitations in system interaction experience, maintenance convenience and high professional barriers, which need to be further optimized and broken through.

[0003] Firstly, the traffic security industry has not yet formed a unified industry standard, and most suppliers use proprietary systems, interfaces and protocols, resulting in poor device interoperability, weak compatibility, and users being easily "locked" in a single supplier system. Secondly, the configuration process is complex, requiring in-depth learning and understanding of industry rules and configuration schemes, etc. Professional knowledge, making it difficult for ordinary maintenance personnel to quickly get started and locate problems, with high learning costs and serious dependence on professional engineer support. In addition, the further expansion of overseas markets promotes the export of smart traffic products, but due to language and cultural differences, etc. The deployment and maintenance costs are greatly increased, and the differences in network environment, IT infrastructure, etc. Also make it difficult for remote debugging to be stable and reliable to reproduce and locate problems, and the communication and cooperation efficiency is low.

[0004] With the large-scale deployment and complex development of smart traffic systems, current smart traffic terminal configuration and parameter tuning still highly rely on manual operation, which has the following problems:

[0005] (1) Smart traffic terminal configuration is complicated and highly dependent on professional knowledge, with low operation efficiency;

[0006] (2) Traditional graphical or script-based configuration methods cannot meet the needs of rapid deployment and multi-scene self-adaptation. SUMMARY

[0007] To address the problems in the existing technology, the purpose of this application is to provide a smart transportation terminal configuration method, system, device, and program product. This method achieves automatic configuration of smart transportation terminals through semantic recognition of natural language text, with a simple configuration method and improved matching accuracy and efficiency.

[0008] The first aspect of this application provides a method for configuring a smart transportation terminal, including the following steps:

[0009] Obtain the traffic terminal configuration text based on natural language input from the user;

[0010] Based on the configuration text recognition of transportation terminals, the user's configuration intent is identified.

[0011] Extract the standard configuration parameters that match the traffic terminal configuration text from the standard configuration parameter set corresponding to the user's configuration intent, and use them as the configuration parameters to be executed;

[0012] The configuration identification result is determined based on the user's configuration intent and the configuration parameters to be executed, and the configuration is executed on the intelligent transportation terminal based on the configuration identification result.

[0013] In some embodiments, recognizing a user's configuration intent based on the configuration text of a transportation terminal includes the following steps:

[0014] The traffic terminal configuration text is segmented into words to obtain the first candidate word set corresponding to the traffic terminal configuration text.

[0015] The first candidate word set is corrected based on the standard words in the pre-built knowledge base to obtain the second candidate word set after error correction.

[0016] Semantic matching is performed based on the second candidate word set and the intent dictionary in the knowledge base to identify the user's configuration intent.

[0017] In some embodiments, error correction processing is performed on the first candidate word set based on standard words in a pre-built knowledge base, including:

[0018] Calculate the character-level similarity between each first candidate word in the first candidate word set and each standard word in the pre-built knowledge base;

[0019] Calculate the semantic similarity between each first candidate word and each standard word;

[0020] Calculate the similarity between each first candidate word and each standard word in the context window word set, and use it as the context similarity between the first candidate word and the standard word;

[0021] The similarity between each first candidate word and each standard word is calculated based on character-level similarity, semantic similarity, and contextual similarity.

[0022] If the similarity between a first candidate word and a standard word is greater than a preset similarity threshold, then the first candidate word is replaced with the corresponding standard word and used as the second candidate word in the second candidate word set.

[0023] In some embodiments, the intelligent transportation terminal configuration method further includes the following steps:

[0024] Obtain user feedback on the configuration recognition results. The feedback information can be either confirmation that the recognition is accurate or confirmation that the recognition is incorrect.

[0025] Based on the first candidate word, the standard word, and the corresponding feedback information that were corrected and replaced during the error correction process, adjust the weights of the first candidate word and the standard word in the knowledge base, and / or adjust the confidence of the mapping relationship between the first candidate word and the standard word in the knowledge base.

[0026] In some embodiments, the intelligent transportation terminal configuration method further includes the following steps:

[0027] Historical configuration record samples are periodically acquired. Each historical configuration record sample includes the first candidate word, standard word, and user feedback information on the configuration recognition result, which is either confirmation of accurate recognition or confirmation of incorrect recognition.

[0028] Incremental training of the word vector mapping model is performed based on historical configuration record samples to adjust the word vector representation of the first candidate word.

[0029] In some embodiments, the intelligent transportation terminal configuration method further includes the following steps:

[0030] Acquire road images captured by the camera equipment corresponding to the intelligent transportation terminal;

[0031] Based on the identification of key elements in road images, the geometric features corresponding to the key elements are obtained;

[0032] Extracting configuration parameters from the traffic terminal configuration text includes the following steps:

[0033] If the geometric features corresponding to the key element match the user's configuration intent, then the configuration parameters are extracted from the geometric features corresponding to the key element and the traffic terminal configuration text.

[0034] In some embodiments, configuring the intelligent transportation terminal based on the user's configuration intent and the configuration parameters to be executed includes the following steps:

[0035] Based on the key parameter categories corresponding to the configuration intent, determine whether the configuration parameters to be executed already include all key parameter categories corresponding to the user's configuration intent;

[0036] If not, determine the missing parameter category and send a missing parameter supplementation request, including the missing parameter category, to the user.

[0037] A second aspect of this application provides a smart transportation terminal configuration system for implementing the smart transportation terminal configuration method of the first aspect. The system includes:

[0038] The text acquisition module is used to acquire traffic terminal configuration text based on natural language input by the user;

[0039] The intent recognition module is used to recognize the user's configuration intent based on the configuration text of the transportation terminal;

[0040] The parameter extraction module is used to extract standard configuration parameters that match the traffic terminal configuration text from the standard configuration parameter set corresponding to the user's configuration intent, and use them as configuration parameters to be executed.

[0041] The configuration execution module is used to perform configuration on the intelligent transportation terminal based on the user's configuration intent and the configuration parameters to be executed.

[0042] A third aspect of this application also provides a smart transportation terminal configuration device, comprising:

[0043] processor;

[0044] A memory in which executable instructions of the processor are stored;

[0045] The processor is configured to execute the steps of the intelligent transportation terminal configuration method described above by executing the executable instructions.

[0046] The fourth aspect of this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described intelligent transportation terminal configuration method.

[0047] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application.

[0048] The intelligent transportation terminal configuration method, system, equipment, and program products of this application have the following beneficial effects:

[0049] By employing the intelligent transportation terminal configuration method of this application, after obtaining the user's input of traffic terminal configuration text based on natural language, the user's configuration intent is automatically obtained based on natural language text recognition. Standard configuration parameters matching the traffic terminal configuration text are extracted from the set of standard configuration parameters defined by the configuration intent and used as the configuration parameters to be executed. This allows for the configuration of the intelligent transportation terminal based on the automatically recognized user configuration intent and the configuration parameters to be executed. Using this intelligent transportation terminal configuration method, users can directly configure the intelligent transportation terminal by inputting natural language text. The configuration method is simple, which helps improve configuration and maintenance efficiency. The low operational threshold reduces the professional skill requirements for staff. This intelligent transportation terminal configuration method improves matching accuracy and efficiency by matching parameters within the set of standard configuration parameters corresponding to the configuration intent, avoiding interference from configuration parameters of other configuration intents. This improves the accuracy of intelligent transportation terminal configuration and avoids users repeatedly adjusting configuration parameters, further improving the configuration efficiency of intelligent transportation terminals. Furthermore, by matching parameters within the set of standard configuration parameters corresponding to the configuration intent, this application also avoids the problem of interference from configuration parameters with the same name in multiple configuration modules, achieving accurate keyword extraction through configuration intent-isolated retrieval of configuration parameters. Attached Figure Description

[0050] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0051] Figure 1 This is a flowchart of a smart transportation terminal configuration method according to an embodiment of this application;

[0052] Figure 2 This is a detailed flowchart of a smart transportation terminal configuration method according to an embodiment of this application;

[0053] Figure 3 This is a schematic diagram of some modules of a local knowledge base according to an embodiment of this application;

[0054] Figure 4 This is a schematic diagram of the configuration interaction page of a smart transportation terminal according to an embodiment of this application;

[0055] Figure 5 This is a schematic diagram of the structure of an intelligent transportation terminal configuration system according to an embodiment of this application;

[0056] Figure 6 This is a schematic diagram of the structure of a smart transportation terminal configuration device according to an embodiment of this application. Detailed Implementation

[0057] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0058] Furthermore, the accompanying drawings are merely illustrative of this application and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices. Although the terms "first" or "second," etc., are used in this specification to denote certain features, these are only for indicating function and not as a limitation on the number or importance of specific features.

[0059] The flowchart shown in the attached diagram is merely an illustrative example and does not necessarily include all steps. For example, some steps may be broken down, while others may be combined or partially combined. Therefore, the actual execution order may change depending on the specific circumstances.

[0060] like Figure 1 As shown in the figure, this application provides a method for configuring a smart transportation terminal, including the following steps:

[0061] S100: Obtain the traffic terminal configuration text based on natural language input by the user;

[0062] Specifically, a smart transportation terminal configuration interaction page is provided (such as...) Figure 4 As shown in the image, this page includes a text input box where users can enter natural language text to initiate the configuration process of the intelligent transportation terminal.

[0063] Intelligent transportation terminals refer to embedded special-purpose devices that integrate traffic perception, violation capture, and local intelligent processing capabilities. They use embedded terminals as hardware carriers, have built-in intelligent control algorithms, and can also collaborate with back-end management platforms. They are the core unit for on-site intelligent processing in intelligent transportation systems. Intelligent transportation terminals include, for example, embedded processors, cameras, radars, user interaction modules, and communication modules.

[0064] Traffic terminal configuration text refers to the natural language text input by the user that expresses the content the user wants to configure. It includes the category of the content the user wants to configure, the specific parameters the user wants to configure, etc. For example, if the user wants to configure red light violation photography, they can input the traffic terminal configuration text as "help me generate red light violation capture for one lane". Or, if the user wants to configure three lanes to capture lane violations, they can input the traffic terminal configuration text as "help me generate lane violation capture for one to three lanes".

[0065] S200: Recognizes the user's configuration intent based on the configuration text of the transportation terminal;

[0066] The user's configuration intent refers to the specific category of content that the user needs to configure. For example, in this embodiment, the user's configuration intent can be divided into various configuration intents such as scene configuration, capture configuration, device configuration, and detection configuration. Scene configuration is used to configure scene information of traffic terminal equipment, such as the number of lanes, lane direction, and lane attributes. Capture configuration is used to configure various specific capture types. Device configuration is used to configure basic device information of traffic terminal equipment, such as device name, device number, camera height, and communication protocol of communication module. Detection configuration is used to configure various detection types, such as motion detection, vehicle front detection, vehicle rear detection, and traffic light detection.

[0067] In this embodiment, the user's configuration intent is identified based on the traffic terminal configuration text. Keywords can be extracted from the traffic terminal configuration text as candidate words. The candidate words are compared with the standard words of various pre-stored configuration intents. The configuration intent corresponding to the standard word that matches the candidate word is taken as the identified configuration intent. When matching candidate words with standard words, similarity can be calculated, vector distance can be calculated, or other word matching methods can be used. For example, if the traffic terminal configuration text input by the user is "help me generate a red light capture for one lane", the candidate word "red light" is extracted. It has a high similarity to the standard word "run a red light" in the capture configuration intent. Therefore, the configuration intent is identified as capture configuration.

[0068] S300: Extract standard configuration parameters that match the traffic terminal configuration text from the standard configuration parameter set corresponding to the user's configuration intent, and use them as configuration parameters to be executed;

[0069] Configuration parameters are the specific configuration contents of each configuration. For example, for scene configuration, its configuration parameters may include lane attributes, lane direction, number of lanes, etc., and for capture configuration, its configuration parameters may include maximum speed limit, red light violation, etc.

[0070] The standard configuration parameter set is a collection of standard configuration parameters corresponding to each preset configuration intent. The standard configuration parameter set can be configuration parameters defined in advance by staff in the configuration system. The configuration system maintains the standard configuration parameter set corresponding to each configuration intent. For example, the configuration intent "scene configuration" corresponds to standard configuration parameters such as "lane", "direction", and "capture type". The standard configuration parameters can be determined according to the intent dictionary of each configuration intent. The configuration system uses the configuration intent as an index to summarize all the standard terms associated with the configuration intent and build an intent dictionary for configuration intent recognition and configuration parameter parsing. The standard terms corresponding to each configuration intent in the intent dictionary are each used as a standard configuration parameter. All the standard terms of each configuration intent constitute the standard configuration parameter set corresponding to the configuration intent.

[0071] In this embodiment, standard configuration parameters that match the traffic terminal configuration text are extracted from the standard configuration parameter set corresponding to the user's configuration intent. For example, this includes: performing word segmentation on the traffic terminal configuration text to obtain multiple words to be matched, matching the words to be matched with each standard configuration parameter in the standard configuration parameter set corresponding to the user's configuration intent (e.g., by calculating similarity, vector distance or other word matching methods), and finding the standard configuration parameters that match the words in the traffic terminal configuration text, which are then used as the configuration parameters to be executed.

[0072] S400: Determines the configuration identification result based on the user's configuration intent and the configuration parameters to be executed, and performs configuration on the intelligent transportation terminal based on the configuration identification result;

[0073] The configuration recognition result indicates the specific configuration content that needs to be executed for the intelligent transportation terminal. For example, if the user's configuration intent is scene configuration, and the configuration parameters are: number of lanes 3, lane attributes from left to right are left turn lane, straight lane, and right turn lane, then the corresponding configuration recognition result is to configure three lanes in the intelligent transportation terminal, with the lane attributes from left to right being left turn lane, straight lane, and right turn lane. If the user's configuration intent is capture configuration, and the configuration parameter is: running a red light, then the corresponding configuration recognition result is to further add a red light capture function to the lanes already configured in the intelligent transportation terminal.

[0074] By adopting the intelligent transportation terminal configuration method of this application, after obtaining the user's input of traffic terminal configuration text based on natural language in step S100, the user's configuration intent is automatically obtained based on natural language text recognition in step S200. In step S300, configuration parameters matching the traffic terminal configuration text are extracted from the standard configuration parameter set defined by the configuration intent and used as configuration parameters to be executed. Then, in step S400, the intelligent transportation terminal is configured based on the automatically recognized user configuration intent and the configuration parameters to be executed. By adopting this intelligent transportation terminal configuration method, users can directly configure the intelligent transportation terminal by inputting natural language text. The configuration method is simple, which helps to improve configuration and maintenance efficiency. The low operational threshold reduces the professional skill requirements for staff. This intelligent transportation terminal configuration method improves matching accuracy and efficiency by matching parameters in the standard configuration parameter set corresponding to the configuration intent, avoiding interference from configuration parameters of other configuration intents. This improves the accuracy of intelligent transportation terminal configuration, avoids repeated adjustments of configuration parameters by the user, and further improves the configuration efficiency of intelligent transportation terminals. Furthermore, this application avoids the problem of interference from configuration parameters with the same name in multiple configuration modules by matching parameters in the standard configuration parameter set corresponding to the configuration intent, and achieves accurate keyword extraction by isolating and retrieving configuration parameters through configuration intent.

[0075] Natural Language Processing (NLP) technology has demonstrated its value in various aspects of intelligent transportation management, such as traffic incident prediction and trip forecasting. For example, NLP interfaces can be used to provide explanatory queries for traffic violation rule determinations. However, most existing solutions focus on cloud-based data query and analysis, and have not yet achieved integrated control on embedded terminals. Furthermore, while NLP solutions based on large models significantly improve text processing capabilities and efficiency, they often rely on large amounts of industry-labeled data and high-performance computing, making them difficult to deploy in resource-constrained embedded environments, thus limiting functional scalability and cost-effectiveness. Therefore, the configuration of existing intelligent transportation terminals still suffers from the following problems: the natural language understanding module typically relies on large cloud models, making low-cost deployment on embedded terminals impossible.

[0076] This intelligent transportation terminal configuration method can be implemented through an intelligent transportation terminal configuration system. The intelligent transportation terminal configuration method and system of this application can be directly configured in the controller of the intelligent transportation terminal and can run independently in an embedded environment. This enables natural language recognition and automatic execution of configuration intentions within embedded transportation equipment, thereby significantly improving the intelligence level and deployment efficiency of the intelligent transportation system.

[0077] The scheme based on Named Entity Recognition (NER) can accurately extract domain-specific terms and configuration elements from natural language text, exhibiting good domain adaptability. This application uses entity recognition in natural language text to automatically identify and extract configuration intent and configuration parameters. The implementation method of this application will be described in detail below with reference to specific embodiments.

[0078] This application achieves accurate conversion from natural language to structured instructions through multi-level semantic processing. For example... Figure 2 As shown, in this embodiment, step S200: recognizing the user's configuration intent based on the traffic terminal configuration text includes the following steps:

[0079] S210: Perform word segmentation on the traffic terminal configuration text to obtain the first candidate word set corresponding to the traffic terminal configuration text;

[0080] In this embodiment, the word segmentation process further includes standardization and specialized word segmentation of the natural language text input by the user, ensuring the accuracy and stability of the configuration method in the semantic parsing stage in subsequent steps S220 and S300. The processing flow is as follows:

[0081] (1) Basic Standardization: Perform basic standardization operations on the traffic terminal configuration text input by the user, including:

[0082] Simplified / Traditional Chinese conversion: Ensures consistent input across regions;

[0083] Special character filtering: Removes noisy characters that are irrelevant to the configuration semantics;

[0084] Case normalization: Eliminates the impact of differences in English capitalization;

[0085] (2) Equivalent substitution: Before word segmentation, the Aho-Corasick automaton is used to perform equivalent substitution of synonyms / variants on the input text (e.g., "reverse driving" --> "reverse"). This step can effectively reduce ambiguity and improve the stability of subsequent intent recognition and parameter mapping; the Aho-Corasick automaton is an efficient multi-pattern string matching algorithm. By constructing a finite state machine of keyword dictionary through preprocessing, all target keywords can be matched in batches with a single scan of the text;

[0086] (3) Word segmentation: The cppjieba word segmentation engine is used as the basic word segmentation tool, combined with a proper noun dictionary automatically generated from the local knowledge base for fine-grained segmentation. This proper noun dictionary can be pre-built in the local knowledge base, which includes proper nouns in the field of intelligent transportation and other pre-set proper nouns. This step ensures that industry proper nouns are not incorrectly segmented by loading proper nouns in the field of intelligent transportation (e.g., "AI ultra-low light" is not segmented into "AI" and "ultra-low light"); the cppjieba word segmentation engine is a lightweight Chinese word segmentation tool that integrates algorithms such as dictionary matching and HMM hidden Markov model, supports accurate / full-mode word segmentation, balances word segmentation speed and accuracy, and adapts to the local semantic parsing needs of embedded terminals;

[0087] S220: Correct the first candidate word set based on the standard words in the pre-built knowledge base to obtain the corrected second candidate word set;

[0088] S230: Based on the second candidate word set and the intent dictionary in the knowledge base, perform semantic matching to identify the user's configuration intent.

[0089] In this embodiment, the pre-built knowledge base is a local knowledge base deployed on the intelligent transportation terminal. The local knowledge base, which is pre-built and stored on the intelligent transportation terminal, serves as the core of semantic parsing during intelligent transportation configuration. It employs a multi-level, structured management approach to organize and maintain professional terminology and configuration rules, achieving efficient and accurate semantic mapping. The construction and deployment process of the local knowledge base includes the following aspects:

[0090] (1) Ontology knowledge base data structure design

[0091] The local knowledge base is stored in a modular JSON format and designed with a two-layer structure to achieve accurate semantic mapping: the first layer is the functional layer, which describes the functional descriptions of different configuration modules (such as scene configuration, capture configuration, etc.); the second layer is the parameter definition layer, which describes the specific parameters, constraints and synonym mappings under each configuration module.

[0092] When the intelligent transportation terminal configuration method is running, the local knowledge base will be loaded into memory all at once to avoid frequent I / O operations and improve semantic parsing efficiency and processing response speed.

[0093] (2) Organization of ontology knowledge base content

[0094] Function Description: This section describes the overall function of each configuration module to facilitate module-level intent matching and output of matched information during semantic parsing. Each configuration module corresponds to a configuration intent, such as scene configuration module, capture configuration module, device configuration module, detection configuration module, etc.

[0095] Configuration parameters: Define the configuration parameters involved in each configuration module, such as "lane attributes" and "capture type".

[0096] Constraints: Specify the legal value range for each configuration parameter to ensure that the configuration operation conforms to the traffic business logic.

[0097] Synonym set: Summarizes common expressions for each configuration parameter, supporting the parsing and mapping of diverse natural language inputs.

[0098] Figure 3 The example illustrates part of the content structure of the scene configuration module and the capture configuration module. The scene configuration module includes two parts: a functional description and configuration parameters. Each configuration parameter further includes various information such as parameter name, constraints, and synonyms. The capture configuration module includes various capture types and corresponding configuration parameters for each type. These configuration parameters further include parameter name, minimum value, and maximum value. Figure 3 The architecture of the local knowledge base shown is merely an example and is not intended to limit the scope of protection of this application. In practical applications, the local knowledge base can adaptively adopt other architectures according to the needs of semantic analysis and recognition in the intelligent transportation configuration process.

[0099] The advantages of the local knowledge base in this application are: (1) Modular design: The multi-level structure enables independent management and expansion of different configuration modules, supporting the rapid addition of new configuration intentions and configuration parameters. (2) Accurate semantic mapping: The parameter constraints combined with the thesaurus ensure that the user's natural language input can be accurately mapped to the corresponding configuration operation. In addition, the configuration modules are isolated from each other and will not be affected by other configuration modules, which significantly improves the accuracy of configuration parameter parsing. Therefore, after the user's configuration intention is identified in step S200, the matching of configuration parameters in step S300 can be directly based on the set of configuration parameters in the configuration module corresponding to the configuration intention. (3) Efficient parsing: The structured configuration modules enable that after a certain configuration intention is identified, only the information under the configuration intention needs to be focused on without frequently traversing all configuration modules in the ontology knowledge base, which significantly improves the parsing efficiency and real-time response capability. (4) Maintainability and scalability: The operation and maintenance personnel can directly maintain and expand the local knowledge base by modifying the JSON file without modifying the source code extensively, which facilitates the iteration and functional upgrade of the intelligent transportation terminal configuration system.

[0100] Users inevitably make typos or provide inaccurate descriptions of the system's proprietary names when configuring intelligent transportation terminals, leading to incomplete terminology in the word segmentation results. To improve the accuracy and robustness of semantic recognition, this embodiment proposes a multi-stage fusion semantic error correction algorithm based on sliding splicing and similarity matching mechanisms. This algorithm combines character-level, semantic-level, and context-related-level fusion recognition to achieve multi-dimensional correction and adaptive optimization of the input text.

[0101] In this embodiment, step S220: performing error correction processing on the first candidate word set based on standard words in a pre-built knowledge base, including:

[0102] Calculate the character-level similarity between each first candidate word in the first candidate word set and each standard word in the pre-built knowledge base;

[0103] Calculate the semantic similarity between each first candidate word and each standard word;

[0104] Calculate the similarity between each first candidate word and each standard word in the context window word set, and use it as the context similarity between the first candidate word and the standard word;

[0105] The similarity between each first candidate word and each standard word is calculated based on character-level similarity, semantic similarity, and contextual similarity.

[0106] If the similarity between a first candidate word and a standard word is greater than a preset similarity threshold, then the first candidate word is replaced with the corresponding standard word and used as the second candidate word in the second candidate word set.

[0107] In this embodiment, character-level error correction is performed by calculating character-level similarity; the specific steps for calculating character-level similarity are as follows:

[0108] First, the traffic terminal configuration text is processed using an n-gram sliding window to generate a first candidate word set C = {c1, c2, ..., c...}. n The n-gram sliding window is a core method for sequence data preprocessing. Essentially, it uses a window of fixed length n, sliding along the sequence with a set step size, each time capturing n consecutive elements covered by the window to form an n-gram (n-tuple), ultimately decomposing the long sequence into multiple locally continuous subsequences. The edit distance between each first candidate word and each standard word (standard terminology) in the local knowledge base is calculated using EditDist. pinyin (c i ,s), calculate the pinyin distance score D based on the pinyin edit distance. pinyin (c i Character-level similarity (s) is calculated using the following formula:

[0109]

[0110] Among them, the Pinyin edit distance uses the weighted Levenshtein algorithm to measure the initial consonant similarity d. shengmu similarity of vowels d yunmu Pitch similarity d tone Assign weights ω1, ω2, and ω3 respectively, and EditDist pinyin (c 1i The formula for calculating ,s) is as follows:

[0111] EditDist pinyin (c i ,s)=ω1·d shengmu +ω2·d yunmu +ω3·d tone

[0112] Pinyin Distance Score D pinyin The higher the value, the higher the similarity between the pinyin.

[0113] In this embodiment, semantic-level error correction is achieved by calculating semantic similarity. Character-level similarity is used to filter standard words in the local knowledge base, selecting only those with character-level similarity greater than a first threshold for semantic similarity calculation. This embodiment introduces the FastText semantic embedding model to calculate the vector similarity between candidate words and standard words, which serves as the semantic similarity. The FastText model, based on sub-word embedding, can automatically handle words not found in the dictionary corpus and misspelled words, improving the accuracy of domain-specific word recognition. The word vector E(x) for each word x is defined as:

[0114]

[0115] Where G(x) is the set of n-gram subwords of word x, v + Let g be the embedding vector of the subword g.

[0116] First candidate word c i The semantic similarity S between a standard word s and a standard word s semantic (c i The formula for calculating ,s) is as follows:

[0117]

[0118] Where ||·|| represents the Euclidean length of the vector.

[0119] In this embodiment, to enhance contextual relevance, a weighted correction based on a context window is introduced. The core idea is that the closer a candidate word is semantically to other high-frequency words in the context, the higher its overall matching score should be. Character-level similarity can be used to filter standard words in the local knowledge base, selecting only those with character-level similarity greater than a first threshold for context similarity calculation. The context similarity between the first candidate word and the standard word is calculated as follows:

[0120]

[0121] Where W is the set of words in the context window near the candidate words in the user input text, f(w) is the weighting function, which is generally calculated by taking the frequency of the context words w appearing in the user input text, and μ is the balance factor, which controls the proportional weight between the semantic similarity between the candidate word and the standard word and the semantic similarity between the context and the target word. It is generally set to a value between 0 and 1, such as 0.2 to 0.6. If the influence of the context is to be weak, the value of μ is reduced; conversely, if the influence of the context is to be strong, the value of μ is increased. S cos (E(c i E(w)) is a candidate word c i Vector cosine similarity with context words. This mechanism can automatically identify and normalize misspellings, synonyms, and near-synonyms into standard terms.

[0122] In this embodiment, the similarity between each first candidate word and each standard word is calculated based on character-level similarity, semantic similarity, and contextual similarity, including calculating the mixed similarity S using the following formula. final :

[0123] S final =αD pinyin +βS semantic +γS content

[0124] Where α, β, and γ are the preset weights for character-level similarity, semantic similarity, and contextual similarity, respectively. When a first candidate word c 1i With a standard word S final When setting a preset similarity threshold, the first candidate word c i Replace it with the standard word to obtain the second candidate word c in the second candidate word set. 2i The preset similarity threshold can be set as needed, for example, to 0.7, 0.8, 0.9, etc. If a first candidate word c... i S with each standard term final If all similarity values ​​are less than or equal to the preset similarity threshold, then the first candidate word c is retained. i Without changing, the first candidate word c in the first candidate word set C is... iThe second candidate word q in the second candidate word set Q i .

[0125] By processing the user-input natural language text using the above method, the first candidate word requiring correction is automatically replaced with standard words (standard terms) from the knowledge base, and the word segmentation results are updated to obtain a second candidate word set for subsequent configuration intent recognition and configuration parameter recognition. This multi-stage fusion algorithm achieves adaptive error correction at the character, semantic, and context levels, significantly improving the accuracy and intelligence of natural language recognition.

[0126] like Figure 2 As shown, in step S230, semantic matching is performed based on the second candidate word set and the intent dictionary in the knowledge base to identify the user's configuration intent. In this embodiment, the BM25 algorithm is used to perform semantic matching between the second candidate word set and the intent dictionary to determine the user's configuration intent. BM25 is a document-query relevance ranking algorithm in the field of information retrieval. Specifically, the intelligent transportation terminal configuration system reads all the terms configured under each configuration intent module in the local knowledge base. For example, in the "scene configuration module," the terms include parameter categories such as "lane attributes" and "capture type," as well as their constraints. Then, using the configuration intent as an index, all terms associated with the configuration intent are summarized to construct an intent dictionary, which is used for configuration intent identification and configuration parameter parsing. When identifying the user's configuration intent, the BM25 score between all second candidate words in the second candidate word set and each configuration intent is calculated to determine the user's configuration intent.

[0127] BM25 Intended Score BM25(Q,d) j The calculation formula is as follows:

[0128]

[0129] Where Q is the second candidate word set, Q = {q1, q2, ..., q} n-1 ,q n}(e.g., ["configuration", "three lanes", "running a red light", "not driving in the designated lane"]), D = {d1, d2, ..., d j-1 ,d j} is the intent dictionary. Furthermore, it should be noted that BM25 is a typical document query matching model; therefore, this implementation example directly concatenates the standard terms under each configured intent category according to the bag-of-words model to form an "intent category document." This document is not an article formed from natural language, but rather a textual representation of the standard terms contained in that intent category. f(q i ,d j ) is the string q i In configuration intent d jIn the context of word frequency, k1 = 1.2 is used to adjust word frequency saturation (the value of k1 is only an example and is not limited to the value shown here; for example, it can also be set to 1.1, 1.3, etc.), b = 0.75 is used to control the impact of document length (the value of b is only an example and is not limited to the value shown here; for example, it can also be set to 0.8, 0.9, etc.), avgdl represents the average length of documents across all intent categories, and N is the total number of configured intent categories. For containing the string q i Number of configuration intent categories.

[0130] Here is an example of an intent dictionary:

[0131] Scene Configuration: ["Left Turn Lane", "Right Turn Lane", "Running a Red Light", "Driving Without Following Lane Directions", "Crossing Solid Line", ...]

[0132] "Device Configuration": ["Device Name", "Device Number", "Camera Height", ...]

[0133] "Detection Configuration": ["Motion Detection", "Front End", "Rear End", "Traffic Light Detection", ...]

[0134] like Figure 2 As shown, after step S200, it is determined whether the user's configuration intent is matched based on the second candidate word set and the intent dictionary. If no match is found, the configuration intent is dynamically queried from the user through the intelligent transportation terminal configuration interaction page until a correct match is found. Once the user's configuration intent is determined, the parameter index corresponding to the configuration intent is activated through the intent dictionary, and configuration parameter (keyword) extraction is performed only in the parameter set of the current configuration intent module.

[0135] This embodiment uses a finite state machine (FSM) to control computation within a state. A finite state machine is a computational model with a finite number of states that can transition between different states based on input. After matching the configuration intent, it enters the state S_i corresponding to the user's configuration intent. Then, it continues to step S300.

[0136] In this embodiment, step S300: When extracting standard configuration parameters that match the traffic terminal configuration text from the standard configuration parameter set corresponding to the user's configuration intent as configuration parameters to be executed, combined with the finite state machine semantic state control mechanism, the intelligent transportation terminal configuration system only performs matching within the parameter space of the currently active configuration intent module (i.e., corresponding to the identified configuration intent), and defines a candidate parameter constraint function:

[0137] Q′={q i ∈Q|q i∈Domain(State FSM )}

[0138] Among them, Domain(State) FSM The `(`)` represents the semantic domain allowed by the current FSM state (i.e., the set of standard configuration parameters corresponding to the user's configuration intent, which can be determined according to the intent dictionary of each configuration intent; for example, the standard configuration parameters corresponding to the configuration intent "scene configuration" include "lane," "direction," and "capture type"). When extracting matching standard configuration parameters, each word obtained from the traffic terminal configuration text segmentation can be matched with each standard configuration parameter in the standard configuration parameter set (e.g., after calculating similarity, selecting standard configuration parameters with similarity greater than a certain threshold) to obtain the matching standard configuration parameters. For example, each second candidate word in the corrected second candidate word set can be matched with each standard configuration parameter to improve matching efficiency and accuracy. By performing configuration parameter matching and searching only within the semantic library corresponding to the user's configuration intent, cross-module interference is effectively avoided by limiting the matching scope, improving matching consistency and accuracy. Furthermore, when matching configuration parameters, the keyword matching algorithm is only executed in the relevant configuration module, reducing the computational load and improving matching efficiency. In addition, by limiting the matching scope, it is ensured that parameter parsing and the user's configuration meaning graph are highly consistent, realizing modular parsing and execution, and enhancing semantic relevance.

[0139] like Figure 2 As shown, in step S300, a multi-turn dialogue framework based on a finite state machine (FSM) maps each user interaction step to a predefined state to achieve the orderly advancement of the configuration task. In each state, the intelligent transportation terminal configuration system utilizes the BM25+ sliding window fuzzy matching engine to perform real-time intent recognition and parameter extraction of user input. By introducing FSM into the field of intelligent transportation semantic control, the semantic recognition results are mapped to the configuration flow state, realizing multi-turn interactions of parameter completion, intent confirmation, correction, and execution. Through the FSM state transition logic, the system can dynamically adjust the interaction path according to the user dialogue, thereby achieving full-chain closed-loop control of semantic understanding and configuration execution.

[0140] The specific process of matching configuration parameters is as follows:

[0141] (1) State set and transition

[0142] The set of configuration intent states maintained by the intelligent transportation terminal configuration system is, for example, S = {S 场景配置 ,S 抓拍配置 ,S 设备配置 ...}, a list of required key configuration parameters for each state binding (e.g., S...). 场景配置It needs to include lane number, capture type, configuration parameter dependencies (such as "running a red light" depending on lane attributes), and exit conditions (complete configuration parameters or user-initiated cancellation).

[0143] (2) Configuration parameter extraction active query mechanism

[0144] After recognizing the configuration intent, the intelligent transportation terminal configuration system enters the state set and selects the corresponding state S. i The system only focuses on the parameter list related to the current state for parameter parsing. If the parameters required for the current state are insufficient, it will actively query to supplement the parameters based on whether the missing parameters are mandatory. Simultaneously, the system maintains a text queue for each dialogue, and configuration parameters confirmed in the dialogue are automatically inherited to subsequent states (e.g., when the user confirms in S...). 场景配置 The system is configured with 3 lanes; this value will be used by default if lane number information is encountered in subsequent states.

[0145] Therefore, in this embodiment, step S400: configuring the intelligent transportation terminal based on the user's configuration intent and the configuration parameters to be executed includes the following steps:

[0146] Based on the key parameter categories corresponding to the configuration intent, determine whether the configuration parameters to be executed already include all key parameter categories corresponding to the user's configuration intent; the key parameter categories are the categories of the required parameter items corresponding to the user's configuration intent.

[0147] If so, proceed to step S400;

[0148] If not, determine the missing parameter category and send a missing parameter supplementation request, including the missing parameter category, to the user, i.e., actively ask for parameter supplementation. Continue until the parameters corresponding to the key parameter category are supplemented, then continue to execute step S400.

[0149] In this embodiment, step S400: determining the configuration identification result based on the user's configuration intent and the configuration parameters to be executed, and performing configuration on the intelligent transportation terminal based on the configuration identification result, includes the following steps:

[0150] S410: Determine the configuration identification result based on the user's configuration intent and the configuration parameters to be executed, and generate a configuration preview based on the configuration identification result; This configuration preview is provided to the user through the intelligent transportation terminal configuration page, and the user is asked to confirm whether the configuration content is correct;

[0151] Therefore, at key configuration nodes, the intelligent transportation terminal configuration system generates a confirmation request containing semantic summary and impact analysis (such as "Confirm to adjust the speed limit from '80km / h' to '120km / h'?"), which includes a configuration preview, and users can confirm the content for text confirmation or correction.

[0152] S420: Receives user feedback information, which confirms accurate identification or confirms incorrect identification.

[0153] If the user does not intend to change the configuration and the user's feedback confirms that the information is accurate, then continue to step S430: perform configuration on the intelligent transportation configuration terminal based on the configuration recognition result.

[0154] If the user changes the configuration intent, the system will re-enter the state S of the changed configuration intent. j Continue executing step S300 under the user's changed configuration intent.

[0155] If the user does not intend to change the configuration but changes the configuration parameters, step S430 can be continued based on the changed configuration parameters. If the changed configuration parameters are also natural language text that requires semantic recognition, then return to step S300 to extract the configuration parameters again.

[0156] Existing intelligent transportation terminal configuration methods lack semantic interaction and learning capabilities, making continuous optimization based on user habits impossible. To enhance the system's self-learning ability and long-term semantic adaptability, this application proposes a dynamic learning mechanism for a knowledge base based on interactive feedback and incremental learning. During semantic recognition and configuration execution, this mechanism automatically collects user interaction and multimodal perception data, and dynamically adjusts the knowledge base weights and semantic mapping accordingly, achieving self-evolution and continuous optimization of knowledge.

[0157] In this embodiment, the technical approach of the dynamic learning mechanism is as follows: the system continuously records interaction logs during runtime, including user input, system recognition results, user confirmation / correction content, and perceptual information. The background process periodically extracts features from the logs to form a feedback sample set. Based on these samples, the system dynamically updates the knowledge base weights and synonym mapping relationships, while simultaneously performing incremental training of a lightweight FastText model in the background, gradually optimizing the word vector distribution according to user language habits.

[0158] In this embodiment, the intelligent transportation terminal configuration method further includes the following steps:

[0159] Obtain user feedback on the configuration recognition results. The feedback information can be either confirmation that the recognition is accurate or confirmation that the recognition is incorrect.

[0160] Based on the first candidate word, the standard word, and the corresponding feedback information that were corrected and replaced during the error correction process, adjust the weights of the first candidate word and the standard word in the knowledge base, and / or adjust the confidence of the mapping relationship between the first candidate word and the standard word in the knowledge base.

[0161] The intelligent transportation terminal configuration system continuously records interaction logs during operation, including user-inputted terminal configuration text, system-recognized configuration results, user feedback (confirmation / correction content), and perception information (described in detail below). The background process of intelligent transportation terminal configuration periodically extracts features from the logs to form historical configuration record samples. Based on these historical configuration record samples, the system dynamically updates the knowledge base weights and synonym mapping relationships.

[0162] To enable the intelligent transportation terminal configuration system to automatically learn users' expression habits and gradually form a standardized semantic mapping, this application maintains two types of weights for each term in the knowledge base:

[0163] Standard word weight W std : Indicates the stability and credibility of standard terms in the field;

[0164] Candidate word weight W cand This indicates the activity level and misuse frequency of user-input words in the system;

[0165] During operation, the system dynamically adjusts the two types of weights based on user interaction feedback and calculates the confidence score ConfMap between candidate words and standard words.

[0166] The specific update rules for the knowledge base weights and the synonym mapping relationship are as follows:

[0167]

[0168] The parameters in the above formula are defined as follows:

[0169] t represents the time before the update, and t+1 represents the time after the update;

[0170] Confirm(x): The number of times the term x has been confirmed as correct by users in the recent period;

[0171] Reject(x): Indicates the number of times the term x has been corrected or rejected by the user;

[0172] R ctx : The average semantic similarity of a term to other keywords in the context (calculated from FastText vectors, reflecting the semantic consistency of the context);

[0173] μ1, μ2, μ3, μ4, μ5: These represent the weights of each item, controlling the degree of influence of each factor.

[0174] After each update, the system automatically determines and processes data based on the set thresholds using the following rules:

[0175]

[0176] Among them, T Add is a preset first weight threshold, and T Del is a preset second weight threshold, and T Map is a preset mapping confidence threshold.

[0177] In this embodiment, the main knowledge base refers to the core standard word set that is pre-constructed by the system and used for the intelligent transportation terminal configuration task. The main knowledge base consists of standard terms determined by manual collation or industry specifications, and is used to support key functions such as intention recognition, parameter parsing, and configuration verification. This standard word set has high semantic stability and domain generality, and is the basis for the system's semantic interpretation. The local knowledge base stores the standard words in the main knowledge base and other standard words. Other standard words can be determined according to, for example, the configuration parameter name, etc. The other standard words that do not enter the main knowledge base are used for error correction processing in step S220. The standard words in the main knowledge base are used as core standard words. If a candidate word matches a core standard word, the candidate word can be directly replaced by the corresponding core standard word before the error correction processing in step S220.

[0178] During the operation of the system, as user interaction data accumulates continuously, when the standard word weight Wstd(s) of a certain candidate word reaches the set threshold T Add after dynamic update, the system automatically incorporates this standard word into the main knowledge base, making it a standard term officially recognized by the system. The newly added standard words entering the main knowledge base have the same status as the initial standard words within the system.

[0179] For example, when the user inputs "running a red light" multiple times and is recognized and corrected by the system to "running a red light" and confirmed to be correct, the system performs the following operations:

[0180] Increase the weight of the standard word "running a red light"; [[ID=2,2]]

[0181] Decrease the weight of the candidate word "running a red light";

[0182] Enhance the mapping confidence ConfMap("running a red light" → "running a red light").

[0183] When the weight value of "running a red light" exceeds the threshold T Add , add "running a red light" to the main knowledge base; and / or, when this mapping confidence exceeds the threshold T Map , the system automatically establishes a fixed mapping in the synonym table: "running a red light": "running a red light".

[0184] After that, when the user inputs "running a red light" again, the system will preferentially and automatically normalize "running a red light" to the standard term "running a red light" before the character-based and semantic error correction in step S220, without the need for re-error correction, improving the text processing efficiency and accuracy.

[0185] This mechanism realizes a closed-loop process from semantic error correction -> weight evolution -> self-learning mapping, enabling the system to continuously optimize its language understanding ability and maintain the adaptive evolution of the local knowledge base.

[0186] Furthermore, in this embodiment, lightweight FastText model incremental training is periodically executed in the system background to gradually optimize the word vector distribution according to the user's language habits. FastText is a lightweight natural language processing model, and its core positioning is to efficiently solve word embedding learning and text classification tasks. The intelligent transportation terminal configuration method further includes the following steps:

[0187] Periodically obtain historical configuration record samples, and each historical configuration record sample includes a first candidate word, a standard word that has been corrected and replaced during the error correction process, and the user's feedback information on the configuration recognition result. The feedback information is information indicating correct recognition or incorrect recognition;

[0188] Perform incremental training of the word vector mapping model based on the historical configuration record samples to adjust the word vector representation of the first candidate word.

[0189] In this embodiment, the system periodically executes FastText embedding incremental training to fine-tune the vector representations of newly emerging or frequently corrected words, so as to make the vector expressions more accurate when natural language texts are processed with embedded vectors. For example, when the user repeatedly inputs "running a red light" and the system recognizes and corrects it to "running a red light" and confirms it is correct, when vectorizing "running a red light", the vector of "running a red light" can be fine-tuned to make it closer to the word vector of "running a red light". And if a candidate word is replaced with a standard word and then repeatedly modified by the user, the vector of the candidate word is fine-tuned to make it far from the word vector of the standard word. At the same time, in this embodiment, a hash check and atomic replacement mechanism is adopted to achieve hot updating of the knowledge base. The update process is asynchronous and does not affect the real-time semantic recognition service.

[0190] Therefore, this embodiment constructs a multi-stage fusion semantic error correction algorithm based on FastText, which combines the pinyin edit distance, semantic embedding similarity, FSM intention constraint, and context weighting mechanism to achieve joint error correction at the character level, semantic level, and intention level. This algorithm can automatically identify and correct out-of-vocabulary words (keywords not in the local knowledge base), typos, and dialectal expressions, significantly improving the robustness and accuracy of semantic recognition in the field of intelligent transportation.

[0191] This embodiment establishes a dynamic learning and self-evolving knowledge base. The system automatically records error correction logs and user feedback during operation, and dynamically adjusts the word frequency and intention mapping relationship based on an online recalibration mechanism. Through continuous learning, the system can continuously optimize the semantic recognition model and configuration logic, forming an adaptive growth mechanism for the transportation field.

[0192] To achieve adaptive mapping from the real environment to semantic configuration, this application further proposes a multimodal perception fusion mechanism, which includes a visual perception submodule, a semantic fusion submodule, and a rule triggering submodule. The visual perception submodule obtains perceptual information, the semantic fusion submodule fuses the perceptual information with information extracted from natural text, and the rule triggering submodule determines when the perceptual information is fused into the final configuration recognition result.

[0193] In this embodiment, the intelligent transportation terminal configuration method further includes the following steps:

[0194] The visual perception submodule acquires road images captured by the camera equipment corresponding to the intelligent transportation terminal; the road images can be single or multiple images captured directly, or video frames from road videos;

[0195] The visual perception submodule identifies key elements in road images and obtains the corresponding geometric features of these key elements. The visual perception submodule further performs target detection and semantic segmentation algorithms on the road images to identify key elements in the road scene, including lane lines, traffic lights, and traffic signs. The system can automatically determine whether there are lane lines in the current image and output the number, direction, and position of the lane lines.

[0196] The semantic fusion submodule is used to structure the visual perception results from the visual perception submodule into parsable scene description information (e.g., number of lanes = 3, existence of double solid line boundaries, traffic light status = red light), and use it as part of the semantic input. This information, along with the user-input traffic terminal configuration text, participates in semantic parsing and intent recognition. When the user inputs statements such as "configure red light violation capture" or "configure lane crossing capture," the system can automatically select the corresponding rule template based on the perceived lane line and traffic light features.

[0197] The function of the rule triggering submodule is as follows: In the FSM state machine, if the geometric features corresponding to the key elements perceived by vision (such as lane line type or traffic light status) match the configuration intent, the system will automatically activate the corresponding violation rule parameters. For example, when a lane guidance sign is detected, if the user does not explicitly indicate in the input text whether the lane direction is "left turn", "straight" or "right turn", the system will use the automatically perceived features to complete the parameters, thereby improving configuration efficiency and accuracy.

[0198] Therefore, in this embodiment, extracting configuration parameters from the traffic terminal configuration text in step S300 includes the following steps:

[0199] If the geometric features corresponding to the key element match the user's configuration intent, then the configuration parameters are extracted from the geometric features corresponding to the key element and the traffic terminal configuration text.

[0200] Therefore, this embodiment introduces a visual perception module based on semantic control to automatically detect scene elements such as lane lines, traffic signs, and traffic lights, and then fuses and matches the perception results with semantic input. For example, when the system detects the number and direction of lanes, it automatically completes the lane attribute information in the scene configuration without requiring user input. This mechanism achieves adaptive linkage from real visual perception to semantic configuration. This embodiment implements dual-drive semantic and multimodal perception, combining semantic recognition with multimodal visual perception to achieve automatic linkage from visual features to configuration rules. Users can quickly complete the configuration without explicitly specifying lane type, direction, or sign status.

[0201] In summary, this embodiment constructs a smart transportation terminal configuration system that can run independently in an embedded environment and has semantic understanding and self-learning capabilities. Through local semantic parsing, finite state machine control, multimodal perception, and dynamic learning mechanisms, it realizes natural language recognition and automatic execution of configuration intentions in embedded transportation devices, achieving closed-loop control from natural language input to device configuration execution, thereby significantly improving the system's intelligence and deployment efficiency.

[0202] This embodiment provides a lightweight local semantic parsing framework, based on the FastText embedding model and the BM25 intent scoring algorithm, combined with the Aho-Corasick automaton to achieve synonym recognition and batch matching and replacement. This intelligent transportation terminal configuration method does not require large cloud model support and can run independently on an embedded processor, achieving real-time processing of semantic recognition, error correction, and intent extraction, significantly reducing system deployment and maintenance costs.

[0203] The following example illustrates the implementation process of the intelligent transportation terminal configuration method in this embodiment.

[0204] In this example, the traffic terminal configuration text input by the user in step S100 is: "Help me generate traffic rules for lanes one to three where people cross red lights without following the guides and crossing the lines."

[0205] In step S200, after word segmentation and error correction of the traffic terminal configuration text, the user's configuration intent is identified as "scenario configuration." In step S300, configuration parameters are extracted under the scenario configuration state, and it is determined whether the key configuration parameters are complete. The configuration intent and configuration parameters parsed through steps S200 and S300 are as follows:

[0206]

[0207]

[0208] Here, "intent" indicates that the configuration intent is "scene configuration" ("sceneConfig"), and the configuration parameters include the number of lanes and lane attributes, where lane attributes are missing.

[0209] Therefore, the system reports a missing required field [lane attribute] to the user. The user then enters: "From left to right, the lanes are left turn, straight, and right turn." The user's input is then parsed again to obtain the following:

[0210] {

[0211] "laneAttribution":[TURN_LEDT,GO_STRAGIHT,TURN_RIGHT]

[0212] }

[0213] Where laneAttribution represents the lane attribute, TIRN_LEDT represents the left-turn lane, GO_STRAGITH represents the straight-ahead lane, and TIRN_REDT represents the right-turn lane.

[0214] The system completes the recognition and obtains the configuration recognition result as "Scenario configuration required: three lanes, left turn, straight, right turn, running a red light, not following lane markings, crossing solid lines." Then, based on the configuration recognition result, it generates a configuration preview: "Do you want to configure the scenario: three lanes, left turn, straight, right turn, running a red light, not following lane markings, crossing solid lines?" and sends this information to the user.

[0215] User input: Confirm.

[0216] The system then configures the intelligent transportation terminal based on the configuration recognition results and sends a message to the user upon successful configuration.

[0217] like Figure 5 As shown in the figure, this application embodiment also provides a smart transportation terminal configuration system for implementing the smart transportation terminal configuration method of the first aspect. The system includes:

[0218] The text acquisition module M100 is used to acquire traffic terminal configuration text based on natural language input by the user;

[0219] The intent recognition module M200 is used to recognize the user's configuration intent based on the configuration text of the transportation terminal;

[0220] The parameter extraction module M300 is used to extract standard configuration parameters that match the traffic terminal configuration text from the standard configuration parameter set corresponding to the user's configuration intent, and use them as configuration parameters to be executed.

[0221] The configuration execution module M400 is used to perform configuration on the intelligent transportation terminal based on the user's configuration intent and the configuration parameters to be executed.

[0222] In the intelligent transportation terminal configuration system of this application, the functions of each module can be implemented using the specific implementation method of the intelligent transportation terminal configuration method described above, which will not be elaborated here.

[0223] By employing the intelligent transportation terminal configuration system of this application, after the text acquisition module M100 obtains the user's input of traffic terminal configuration text based on natural language, the intent recognition module M200 automatically obtains the user's configuration intent based on natural language text recognition. The parameter extraction module M300 extracts the configuration parameters that match the traffic terminal configuration text from the standard configuration parameter set defined by the configuration intent, which are then used as the configuration parameters to be executed. The configuration execution module M400 then configures the intelligent transportation terminal based on the automatically recognized user configuration intent and the configuration parameters to be executed. By adopting this intelligent transportation terminal configuration method, users can directly configure the intelligent transportation terminal by inputting natural language text. The configuration method is simple, which helps improve configuration and maintenance efficiency. The low operational threshold reduces the professional skill requirements for staff. This intelligent transportation terminal configuration method improves matching accuracy and efficiency by matching parameters in the standard configuration parameter set corresponding to the configuration intent, avoiding interference from configuration parameters of other configuration intents. This improves the accuracy of intelligent transportation terminal configuration, avoids repeated adjustments of configuration parameters by the user, and further improves the configuration efficiency of intelligent transportation terminals. Furthermore, this application avoids the problem of interference from configuration parameters with the same name in multiple configuration modules by matching parameters in the standard configuration parameter set corresponding to the configuration intent, and achieves accurate keyword extraction by isolating and retrieving configuration parameters through configuration intent.

[0224] This application embodiment also provides a smart transportation terminal configuration device, including a processor; a memory storing executable instructions of the processor; wherein the processor is configured to perform the steps of the smart transportation terminal configuration method by executing the executable instructions.

[0225] Those skilled in the art will understand that various aspects of this application can be implemented as systems, methods, or computer program products. Therefore, various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, collectively referred to herein as a "circuit," "module," or "platform."

[0226] The following reference Figure 6 To describe an electronic device 600 according to this embodiment of the present application.Figure 6 The electronic device 600 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this application. Figure 6 As shown, the electronic device 600 is presented in the form of a general-purpose computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different system components (including storage unit 620 and processing unit 610), a display unit 640, etc.

[0227] The storage unit stores program code that can be executed by the processing unit 610, causing the processing unit 610 to perform the steps described in the above-described intelligent transportation terminal configuration method section of this specification according to various exemplary embodiments of this application. For example, the processing unit 610 can perform actions such as... Figure 1 The steps are shown in the diagram. The storage unit 620 may include a readable medium in the form of volatile storage units, such as random access memory (RAM) 6201 and / or cache memory 6202, and may further include read-only memory (ROM) 6203. The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, including but not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. The bus 630 may represent one or more of several bus structures, including a storage unit bus or storage unit controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures. The electronic device 600 may also communicate with one or more external devices 700 (e.g., a keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with the electronic device 600, and / or with any device that enables the electronic device 600 to communicate with one or more other computing devices (e.g., a router, modem, etc.). This communication can be performed via input / output (I / O) interface 650. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via network adapter 660. Network adapter 660 can communicate with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0228] In the intelligent transportation terminal configuration device, when the program in the memory is executed by the processor, it implements the steps of the intelligent transportation terminal configuration method. Therefore, the device can also obtain the technical effects of the intelligent transportation terminal configuration method.

[0229] An exemplary embodiment of this application also provides a computer program product. The computer program product includes a computer program, which, when executed by a processor, implements the steps of the intelligent transportation terminal configuration method described above. In one embodiment, the computer program product can be a tangible product containing a computer program, such as a computer-readable storage medium storing the computer program. The readable storage medium can be a storage medium based on electrical, magnetic, optical, electromagnetic, infrared, or other signals, including but not limited to: random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory, hard disk drive (HDD), solid-state drive (SSD), etc. Exemplarily, the computer program product can be implemented as a non-volatile storage medium storing a computer program, such as read-only memory, NAND flash memory, etc. In one embodiment, the computer program product can be an intangible product containing a computer program. Exemplarily, the computer program product can be implemented as a virtual digital product, such as an executable file storing a computer program, an installation package, or other digital file. The code of the computer program can be written in one or more programming languages. Programming languages ​​include C, Java, C++, Python, etc. The program code can execute entirely on the user's computing device, or partially on the user's computing device, or as a standalone software package, or partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing devices can be connected to the user's computing device through any type of network, such as a local area network (LAN), a wide area network (WAN), etc., or they can be connected to external computing devices (e.g., through an internet connection provided by an operator).

[0230] Computer programs can be carried or transmitted via signals such as electrical, magnetic, optical, electromagnetic, and infrared rays. Electronic devices can convert signals carrying computer programs into digital signals, thereby running the computer programs. When a computer program runs on an electronic device, its code is used to cause the electronic device to execute (more specifically, the processor of the electronic device to execute) the method steps of various exemplary embodiments of this application, such as the steps of the intelligent transportation terminal configuration method described above.

[0231] When the computer program is executed by the processor, it implements the steps of the above-described intelligent transportation terminal configuration method. Therefore, the computer program product can also achieve the technical effects of the above-described intelligent transportation terminal configuration method.

[0232] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of this application and should not be construed as limiting the specific implementation of this application to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of this application, and all such modifications or substitutions should be considered within the scope of protection of this application.

Claims

1. A method for configuring a smart transportation terminal, characterized in that, Includes the following steps: Obtain the traffic terminal configuration text based on natural language input from the user; The user's configuration intent is identified based on the configuration text of the transportation terminal. Extract configuration parameters that match the traffic terminal configuration text from the standard configuration parameter set corresponding to the user's configuration intent, and use them as configuration parameters to be executed; Based on the user's configuration intent and the configuration parameters to be executed, the configuration identification result is determined, and the configuration is executed on the intelligent transportation terminal based on the configuration identification result.

2. The intelligent transportation terminal configuration method according to claim 1, characterized in that, Based on the configuration text of the transportation terminal, the user's configuration intent is recognized, including the following steps: The traffic terminal configuration text is segmented to obtain the first candidate word set corresponding to the traffic terminal configuration text; The first candidate word set is corrected based on the standard words in the pre-built knowledge base to obtain the second candidate word set after error correction. Semantic matching is performed based on the second candidate word set and the intent dictionary in the knowledge base to identify the user's configuration intent.

3. The intelligent transportation terminal configuration method according to claim 2, characterized in that, The step of correcting the first candidate word set based on standard words in a pre-built knowledge base includes: Calculate the character-level similarity between each first candidate word in the first candidate word set and each standard word in the pre-built knowledge base; Calculate the semantic similarity between each first candidate word and each standard word; Calculate the similarity between each first candidate word and each standard word in the context window word set, and use it as the context similarity between the first candidate word and the standard word; The similarity between each first candidate word and each standard word is calculated based on the character-level similarity, the semantic similarity, and the contextual similarity. If the similarity between the first candidate word and the standard word is greater than a preset similarity threshold, then the first candidate word is replaced with the corresponding standard word and used as the second candidate word in the second candidate word set.

4. The intelligent transportation terminal configuration method according to claim 2, characterized in that, The method further includes the following steps: Obtain user feedback on the configuration recognition result, wherein the feedback information is confirmation of accurate recognition or confirmation of incorrect recognition; Based on the first candidate word, the standard word, and the corresponding feedback information that were corrected and replaced during the error correction process, adjust the weights of the first candidate word and the standard word in the knowledge base, and / or adjust the confidence of the mapping relationship between the first candidate word and the standard word in the knowledge base.

5. The intelligent transportation terminal configuration method according to claim 2, characterized in that, The method further includes the following steps: Historical configuration record samples are periodically acquired. Each historical configuration record sample includes a first candidate word, a standard word, and user feedback information on the configuration recognition result, which is either confirmation of accurate recognition or confirmation of incorrect recognition. Incremental training of the word vector mapping model is performed based on the historical configuration record samples to adjust the word vector representation of the first candidate word.

6. The intelligent transportation terminal configuration method according to claim 1, characterized in that, The method further includes the following steps: Acquire road images captured by the camera device corresponding to the intelligent transportation terminal; Based on the road image, key elements are identified, and the geometric features corresponding to the key elements are obtained; Extracting configuration parameters from the traffic terminal configuration text includes the following steps: If the geometric features corresponding to the key element match the user's configuration intent, then the configuration parameters are extracted from the geometric features corresponding to the key element and the traffic terminal configuration text.

7. The intelligent transportation terminal configuration method according to claim 1, characterized in that, The configuration of the intelligent transportation terminal is performed based on the user's configuration intent and the configuration parameters to be executed, including the following steps: Based on the key parameter categories corresponding to the configuration intent, determine whether the configuration parameters to be executed already include all key parameter categories corresponding to the user's configuration intent; If not, determine the missing parameter category and send a missing parameter supplementation request, including the missing parameter category, to the user.

8. A smart transportation terminal configuration system, characterized in that, The system for implementing the intelligent transportation terminal configuration method according to any one of claims 1 to 7, the system comprising: The text acquisition module is used to acquire traffic terminal configuration text based on natural language input by the user; The intent recognition module is used to recognize the user's configuration intent based on the configuration text of the transportation terminal; The parameter extraction module is used to extract configuration parameters that match the traffic terminal configuration text from the standard configuration parameter set corresponding to the user's configuration intent, and use them as configuration parameters to be executed. The configuration execution module is used to perform configuration on the intelligent transportation terminal based on the user's configuration intent and the configuration parameters to be executed.

9. A smart transportation terminal configuration device, characterized in that, include: processor; A memory in which executable instructions of the processor are stored; The processor is configured to perform the steps of the intelligent transportation terminal configuration method according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the intelligent transportation terminal configuration method according to any one of claims 1 to 7.

Images