Safety monitoring intelligent question answering system based on AI large model

Abstract

The invention discloses a safety monitoring intelligent question-answering system based on an AI large model, which relates to the field of intelligent safety monitoring and comprises a client, a natural language processing unit, an acquisition unit, a storage unit, a safety monitoring model library, a matching unit, a knowledge graph unit, a question-answering matching model and a display unit. According to the system, the safety monitoring result of the corresponding monitoring object can be automatically matched according to the question of the user, the interaction is simple, the use is easy, and the corresponding safety monitoring result is easy to obtain without professional operation and professional technical background.

Description

Technical Field

[0001] The present invention relates to the field of intelligent security monitoring, and in particular, to an intelligent question-answering system for security monitoring based on an AI big model. Background Art

[0002] Traditional security monitoring systems install various monitoring terminals on the monitored objects to collect monitoring data, and then process them to obtain monitoring results. Traditional security monitoring systems are relatively simple in terms of user interaction. Users often obtain information through fixed format inputs. They lack flexible question-answering and interactive mechanisms and are not easy to use. In addition, the data displayed by traditional security monitoring systems requires professionals to understand, and non-professionals cannot ask or query to obtain the corresponding monitoring results. Summary of the invention

[0003] To solve the above problems, the purpose of the present invention is to provide a security monitoring intelligent question and answer system based on an AI big model, which can automatically match the security monitoring results of the corresponding monitoring objects according to the user's questions. The interaction is simple and easy to use, and the corresponding security monitoring results can be easily obtained without professional operation and professional technical background.

[0004] To achieve the above-mentioned purpose of the invention, the present invention provides a security monitoring intelligent question-answering system based on an AI big model, the system comprising:

[0005] The client is used for user login and obtaining the user's question information after the user logs in;

[0006] A natural language processing unit, used to analyze and process the question information to obtain a question processing result;

[0007] A collection unit, used for collecting monitoring data of the monitored object;

[0008] A storage unit, used for storing monitoring data;

[0009] A security monitoring model library is used to store multiple security monitoring models. The security monitoring models are used to perform security analysis and processing based on corresponding monitoring data to obtain security monitoring results.

[0010] A matching unit, the matching unit is used to match the corresponding monitoring data and the security monitoring model based on the question processing result, and obtain the security monitoring data corresponding to the question information based on the matched monitoring data and the security monitoring model;

[0011] A knowledge graph unit, used to establish a knowledge graph library based on the monitoring data and the safety monitoring data;

[0012] A question-answer matching model, used to match an answer result matching the question information from the knowledge graph library based on the question processing result;

[0013] A display unit is used to display the answer result to the user.

[0014] Among them, the traditional security monitoring system displays various monitoring data and alarm data or early warning data on the display, which requires professional technicians to interpret so that non-professionals can understand it, and even professionals need to input fixed-format query statements or instructions to obtain corresponding monitoring results, making the monitoring results of the traditional security monitoring system complex and difficult to operate. The present invention combines the security monitoring system with the AI ​​big model, uses the AI ​​big model to obtain corresponding question information based on the user's questions, and then uses the question signal to combine with the security monitoring system to obtain the security monitoring results, and then returns the security monitoring results to the user. The user only needs to ask questions, without the need for professional data understanding and interpretation, and does not need to enter complex and professional instructions and format statements for query, and can quickly and conveniently obtain security monitoring results.

[0015] Furthermore, matching corresponding monitoring data and security monitoring models based on the question processing results specifically includes:

[0016] Performing semantic analysis on the question processing result to obtain question processing semantic information;

[0017] Performing feature extraction on the question processing semantic information to obtain a feature processing result;

[0018] Inputting the feature processing result into the first matching model for matching to obtain corresponding monitoring data;

[0019] Inputting the feature processing result into a second matching model for matching to obtain a corresponding safety monitoring model;

[0020] The first matching model is obtained as follows:

[0021] Constructing first training data, the first training data includes a plurality of first-class sample data, the first-class sample data includes monitoring data and corresponding feature information;

[0022] Training the matching model based on the first training data to obtain a first matching model;

[0023] The second matching model is obtained as follows:

[0024] Constructing second training data, the second training data includes a plurality of second-category sample data, the second-category sample data includes a security monitoring model and corresponding feature information;

[0025] The matching model is trained based on the second training data to obtain a second matching model.

[0026] Furthermore, the client also includes a voice processing unit, which is used to collect and obtain the user's voice question information; pre-process the voice question information to obtain first voice information; perform spectrum analysis on the first voice information to obtain human voice information; and perform voiceprint recognition on the human voice information to obtain the user's question information.

[0027] Furthermore, the system also includes a statistical unit, a pre-processing unit and a cache unit;

[0028] The statistical unit is used to: record the historical question information of the user, count the historical question information, obtain the historical question times of the corresponding question information, sort the historical question information in descending order based on the historical question times, and obtain the commonly used question information based on the top N historical question information in the sorting, where N is an integer greater than 1;

[0029] The preprocessing unit is used to obtain the answer result corresponding to the commonly used question information based on the monitoring data and the commonly used question information at each preset time interval after the user logs in to the client;

[0030] A cache unit, the cache unit is used to store answer results corresponding to commonly used question information;

[0031] The client determines whether the user's question information is commonly used question information, and if so, matches and obtains the corresponding answer result from the cache unit; if not, the user's question information is sent to the natural language processing unit for processing.

[0032] Furthermore, the safety monitoring model library stores a bridge pavement icing safety monitoring model, and the bridge pavement icing safety monitoring model is used to perform bridge pavement icing safety analysis and processing based on corresponding monitoring data to obtain a bridge pavement icing safety monitoring result.

[0033] Furthermore, the acquisition unit includes:

[0034] Environmental collection module, used to collect environmental data around the bridge;

[0035] An image acquisition unit, used for acquiring bridge pavement image data;

[0036] A reflection collection unit, wherein the reflection collection unit includes a transmitting end and a collecting end distributed on both sides of the bridge pavement, the transmitting end is used to transmit detection light to the bridge pavement, and the collecting end is used to receive the detection light reflected from the bridge pavement to obtain light detection data;

[0037] The surface icing safety monitoring model includes:

[0038] An environmental safety monitoring model, used to obtain a bridge pavement environmental safety monitoring result based on the bridge surrounding environment data;

[0039] An image safety monitoring module, used to obtain bridge pavement image safety monitoring results based on bridge pavement image data;

[0040] A detection safety monitoring module is used to obtain bridge pavement detection safety monitoring results based on light detection data;

[0041] The output module is used to obtain the bridge pavement icing safety monitoring results based on the analysis of the beam pavement environment safety monitoring results, the bridge pavement image safety monitoring results and the bridge pavement detection safety monitoring results.

[0042] Furthermore, the bridge pavement icing safety monitoring results are obtained by analyzing the beam pavement environment safety monitoring results, the bridge pavement image safety monitoring results and the bridge pavement detection safety monitoring results, specifically including:

[0043] Obtained the first safety monitoring score based on the beam pavement environmental safety monitoring results;

[0044] Obtain the second safety monitoring score based on the bridge pavement image safety monitoring results;

[0045] Obtain the third safety monitoring score based on the bridge pavement detection safety monitoring results;

[0046] The total safety monitoring score is obtained based on the first to third safety monitoring scores, and the bridge pavement icing safety monitoring results are obtained based on the total safety monitoring score.

[0047] Further:

[0048] If the bridge pavement environmental safety monitoring result is that the bridge pavement is frozen, the first safety monitoring score is a; if the bridge pavement environmental safety monitoring result is that the bridge pavement is not frozen, the first safety monitoring score is b;

[0049] If the bridge pavement image safety monitoring result shows that the bridge pavement is frozen, the second safety monitoring score is c; if the bridge pavement environmental safety monitoring result shows that the bridge pavement is not frozen, the second safety monitoring score is d;

[0050] If the bridge pavement detection safety monitoring result is that the bridge pavement is frozen, the third safety monitoring score is e; if the bridge pavement detection safety monitoring result is that the bridge pavement is not frozen, the third safety monitoring score is f;

[0051] Total safety monitoring score = (a or b) + (c or d) + (e or f); where a <b,c<d,e<f;

[0052] If the total safety monitoring score is within the first preset range, the bridge pavement icing safety monitoring result is safe; if the total safety monitoring score is not within the first preset range, the bridge pavement icing safety monitoring result is unsafe.

[0053] Further:

[0054] The environmental safety monitoring model is obtained as follows:

[0055] A bridge model is established and placed in a test box. The structure and material of the bridge model are the same as those of the bridge. A plurality of vehicle models that can move back and forth are arranged on the bridge model. A temperature control system and an airflow control system are arranged in the test box. Temperature collection equipment and humidity collection equipment are installed on both sides of the bridge of the bridge model.

[0056] Conduct multiple tests in the test chamber, and record the test data of each test. The test content includes: starting the vehicle model to move back and forth, starting the temperature control system and the airflow control system, and the temperature control system controls the ambient temperature in the test chamber to drop from the initial temperature until the bridge deck of the bridge model freezes; the airflow control system is used to simulate and control the airflow velocity around the bridge model in the test chamber based on the meteorological data of the area where the monitored bridge is located; the test data includes: the ambient temperature data and ambient humidity data corresponding to the bridge deck of the bridge model when ice forms;

[0057] Training the prediction model based on the test data to obtain the environmental safety monitoring model;

[0058] The image security monitoring module is obtained as follows:

[0059] Collect images of the bridge pavement at various time periods when the pavement is not frozen and has no water accumulation to obtain a first image set;

[0060] The second image set is obtained by collecting images of the bridge road surface at various time periods when the bridge road surface is not frozen and has accumulated water;

[0061] The third image set is obtained by collecting images of the bridge road surface at various time periods when the bridge road surface is frozen;

[0062] A first data set is constructed based on the first to third image sets, the first data set is annotated to obtain a first training set, and a model is trained based on the first training set to obtain the image safety monitoring module.

[0063] Furthermore, the detection safety monitoring module is obtained in the following manner:

[0064] Collecting detection light reflected from the bridge pavement at each time period when the bridge pavement is not frozen, to obtain a first detection light signal;

[0065] Collecting detection light reflected from the bridge pavement at each time period when the bridge pavement is frozen, to obtain a second detection light signal;

[0066] A second data set is constructed based on the first detection light signal and the second detection light signal, the second data set is labeled to obtain a second training set, and the detection safety monitoring module is obtained based on a training model of the second training set.

[0067] One or more technical solutions provided by the present invention have at least the following technical effects or advantages:

[0068] The present invention can automatically match the security monitoring results of the corresponding monitoring objects according to the questions asked by the users, has simple interaction and is easy to use, and can easily obtain the corresponding security monitoring results without professional operation and professional technical background. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation on the embodiments of the present invention;

[0070] Figure 1 The figure is a schematic diagram of the composition of a security monitoring intelligent question-answering system based on an AI big model. DETAILED DESCRIPTION

[0071] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0072] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those within the scope of this description. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0073] Those skilled in the art should understand that, in the disclosure of the present invention, the orientation or position relationship indicated by the terms "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc. are based on the orientation or position relationship shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the above terms should not be understood as limiting the present invention.

[0074] It is to be understood that the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element may be one, while in another embodiment, the number of the element may be multiple, and the term "one" should not be understood as a limitation on the quantity.

[0075] Embodiment 1;

[0076] Please refer to Figure 1 , Figure 1 The figure is a schematic diagram of the composition of a security monitoring intelligent question-answering system based on an AI big model. The present invention provides a security monitoring intelligent question-answering system based on an AI big model, and the system includes:

[0077] The client is used for user login and obtaining the user's question information after the user logs in; the client can be an APP on a mobile smart device, such as a mobile phone smart APP, or an application software on a computer device, etc., and the embodiments of the present invention are not limited accordingly. The client can be connected to a mobile phone or a computer, and the user can input questions through an input device, such as a keyboard or a microphone;

[0078] The natural language processing unit is used to analyze and process the question information to obtain the question processing result; the natural language processing unit can use the existing natural language processing model to process and generate the machine language that the computer can understand, that is, the question processing result. Natural language processing is one of the core technologies of the intelligent question-answering system, involving text understanding, semantic analysis, named entity recognition and other aspects. Through NLP technology, the system can understand the questions raised by the user and convert them into a form that the computer can understand and process. In practical applications, the natural language processing model can be selected according to actual needs. The embodiments of the present invention do not make corresponding limitations, such as BERT model, GPT-2 model or RoBERTa model, etc., and existing AI software or platforms such as ChatGPT can also be used;

[0079] The acquisition unit is used to collect monitoring data of the monitored object; different monitoring objects and different monitoring purposes collect different monitoring data, and the corresponding acquisition units are also different. For example, structural safety monitoring mainly collects parameters such as vibration, stress, displacement, etc., and the corresponding acquisition units are vibration, stress, displacement and other sensors. For example, leakage monitoring collects corresponding data through leakage collectors or terminals. For example, geological disaster monitoring collects corresponding data through geological disaster monitoring data collectors. For example, water level monitoring collects corresponding water level data through water level sensors. For example, environmental monitoring data collection collects corresponding data through corresponding environmental monitoring collection equipment, such as temperature, humidity sensors and meteorological equipment. The present invention does not limit the specific implementation method of the acquisition unit, and can be adaptively adjusted according to the specific monitoring object and purpose to meet the corresponding monitoring purpose;

[0080] A storage unit is used to store monitoring data. The monitoring data can be stored through the storage unit to facilitate subsequent use and historical monitoring data query;

[0081] A safety monitoring model library is used to store multiple safety monitoring models. The safety monitoring models are used to perform safety analysis and processing based on corresponding monitoring data to obtain safety monitoring results. Different monitoring objects and purposes require different monitoring models. In order to enable the system to match the user's various questioning requirements, the safety monitoring model library of the present invention stores multiple safety monitoring models, such as dam safety monitoring models, bridge safety monitoring models, building safety monitoring models, landslide monitoring models, earthquake monitoring models, flood monitoring models, structural safety monitoring models, etc. These safety monitoring models can use existing monitoring models, or they can be monitoring models created and constructed according to actual conditions. The existing monitoring models and the models constructed by the present invention are stored in the safety monitoring model library to form a safety monitoring model library for subsequent use.

[0082] A matching unit, the matching unit is used to match the corresponding monitoring data and the safety monitoring model based on the question processing result, and obtain the safety monitoring data corresponding to the question information based on the matched monitoring data and the safety monitoring model; for example, if the user asks about the current safety status of the bridge, then the corresponding monitoring data corresponding to the bridge safety monitoring and the safety monitoring model matching the bridge safety monitoring are matched; if the user asks whether a landslide will occur at present, then the corresponding monitoring data corresponding to the landslide safety monitoring and the safety monitoring model matching the landslide safety monitoring are matched; the matching obtains the keyword features by analyzing the semantics, and then obtains the corresponding monitoring data type and the safety monitoring model type by feature matching;

[0083] The knowledge graph unit is used to establish a knowledge graph base based on the monitoring data and the security monitoring data; the knowledge graph is a knowledge base stored in a graph structure, which contains rich entity, attribute and relationship information. The intelligent question-answering system enables the machine to acquire a large amount of structured knowledge and apply it to the process of answering questions by establishing and maintaining the knowledge graph.

[0084] The question-answer matching model is used to match the answer result that matches the question information from the knowledge graph library based on the question processing result; the question-answer matching model finds the most relevant question and gives the corresponding answer by calculating the matching degree between the user's question and the question in the knowledge base. The question-answer matching model can be based on traditional machine learning algorithms or deep learning methods such as recurrent neural networks and attention mechanisms. The embodiments of the present invention do not make corresponding limitations and can be selected or adjusted according to actual needs;

[0085] The display unit is used to display the answer result to the user. The display unit can display it through text messages or display screens.

[0086] Among them, in the embodiment of the present invention, the matching of corresponding monitoring data and security monitoring model based on the question processing result specifically includes:

[0087] Performing semantic analysis on the question processing result to obtain question processing semantic information;

[0088] Performing feature extraction on the question processing semantic information to obtain a feature processing result;

[0089] Inputting the feature processing result into the first matching model for matching to obtain corresponding monitoring data;

[0090] Inputting the feature processing result into a second matching model for matching to obtain a corresponding safety monitoring model;

[0091] The first matching model is obtained as follows:

[0092] Constructing first training data, the first training data includes a plurality of first-class sample data, the first-class sample data includes monitoring data and corresponding feature information;

[0093] Training the matching model based on the first training data to obtain a first matching model;

[0094] The second matching model is obtained as follows:

[0095] Constructing second training data, the second training data includes a plurality of second-category sample data, the second-category sample data includes a security monitoring model and corresponding feature information;

[0096] The matching model is trained based on the second training data to obtain a second matching model.

[0097] Through the above method, the user's question can be analyzed after the user asks the question, and the user's question information can be accurately obtained. Then, according to the user's feature processing result, the corresponding monitoring data can be obtained based on the accurate matching of the feature processing result and the first matching model, and the corresponding security monitoring model can be obtained by accurately matching the feature processing result and the second matching model. Through the above method, the user can ask questions and accurately obtain the corresponding monitoring data and security monitoring model, and realize security questioning and answering.

[0098] Among them, in the embodiment of the present invention, the applicant has found that in actual applications, the scene is usually noisy and there is a lot of interference. Users ask questions through mobile phones. The voice collected by the mobile phone contains a lot of interfering voices and non-user voices, which interfere with the extraction of user information of the user, making it impossible for the question and answer system to handle the problem well, and thus unable to accurately obtain the answer to the question the user wants. In order to solve this problem, the present invention has made corresponding improvements. The client also includes a voice processing unit, which is used to collect and obtain the user's voice question information; pre-process the voice question information to obtain first voice information; perform spectrum analysis on the first voice information to obtain human voice information; and perform voiceprint recognition on the human voice information to obtain the user's question information.

[0099] Among them, the audio signal is collected by the voice processing unit and pre-processed, including removing noise and interference, and then spectral analysis is performed. Through spectral analysis, the frequency range of human voice and noise is identified. Human voice is usually concentrated in the mid-frequency area, while noise may be distributed in a wider frequency band. Taking advantage of this feature, technical means such as filtering and noise reduction can be used to separate noise from the audio signal while retaining the human voice. Since the human voice can have the voice of other users, in order to accurately obtain the user's human voice, the user's voiceprint features can be extracted when the user registers, and then the user's voice can be matched through the voiceprint, and other voices can be eliminated to avoid interference, so as to ensure the accurate recognition and extraction of the user's questions.

[0100] Among them, in the embodiment of the present invention, there is a relatively important difference between the security monitoring intelligent question-and-answer system in the present invention and the traditional intelligent question-and-answer system. The traditional question-and-answer system uses historical data to establish a database, that is, the answer, and then matches to obtain the corresponding result. It does not use real-time data, while real-time security monitoring data is required in security monitoring, so as to reflect the current security status of the monitoring target. This results in the corresponding security monitoring model being run each time a question is asked, so that the user will wait for a long time between asking a question and obtaining a result. In order to solve the problem of long-term waiting for the answer to the question reducing the user experience, the system has been improved in design to obtain conventional answers in advance for quick reply and shorten the waiting time. The specific improvement method is: the system also includes a statistical unit, a preprocessing unit and a cache unit;

[0101] The statistical unit is used to: record the historical question information of the user, count the historical question information, obtain the historical question times of the corresponding question information, sort the historical question information in descending order based on the historical question times, and obtain the commonly used question information based on the top N historical question information in the sorting, where N is an integer greater than 1; the purpose of the statistical unit is to obtain several frequently used questions of the user by counting historical data;

[0102] The preprocessing unit is used to obtain answer results corresponding to the commonly used question information based on the monitoring data and the commonly used question information at each preset time period after the user logs in to the client; after the user logs in, the answers corresponding to the commonly used questions are generated according to the current monitoring data and the commonly used questions, and when the user is likely to ask commonly used questions, the corresponding answers can be quickly retrieved without running the corresponding security monitoring model, and quick answers can be achieved; the preset time period can be 10 minutes or 20 minutes or several minutes, etc., and can be adjusted according to actual needs. The embodiment of the present invention does not make specific adjustments. The purpose of designing the interval time period is to ensure the real-time nature of the data.

[0103] A cache unit, the cache unit is used to store answer results corresponding to commonly used question information;

[0104] The client determines whether the user's question information is commonly used question information, and if so, matches and obtains the corresponding answer result from the cache unit; if not, the user's question information is sent to the natural language processing unit for processing.

[0105] Among them, in the embodiment of the present invention, in practical applications, in addition to the safety of the bridge structure, the icing safety of the bridge is also an object that needs to be focused on. Since the bridge is affected by factors such as heat convection and heat conduction, its road surface is prone to ice. The specific reasons are:

[0106] Thermal convection: The bridge is suspended in the air and is completely exposed to the air, while only one side of the road is exposed to the air. The contact area between the bridge and the air is larger, so thermal convection will cause the surface heat to dissipate faster, resulting in a lower bridge surface temperature.

[0107] Heat conduction: Bridges are usually made of reinforced concrete, which has a thermal conductivity of 1.74, while ordinary pavements use asphalt concrete, which has a thermal conductivity of 1.05. Therefore, bridges lose heat faster and have lower temperatures.

[0108] Insulation effect of the earth: Although the air temperature can quickly drop below 0°C, the earth has an insulation effect and the surface temperature drops slowly. The deep soil still maintains a temperature above 0°C and transfers heat to the surface soil, making the road surface temperature relatively high.

[0109] The icing on the pavement of a bridge may cause serious skidding of vehicles, resulting in accidents and even crashing into the bridge and falling into the river. Therefore, it is necessary to monitor the icing on the pavement of the bridge accordingly. Therefore, the present invention designs a corresponding bridge pavement icing safety monitoring model. The bridge pavement icing safety monitoring model is stored in the safety monitoring model library. The bridge pavement icing safety monitoring model is used to perform bridge pavement icing safety analysis and processing based on corresponding monitoring data to obtain bridge pavement icing safety monitoring results.

[0110] Among them, unlike other existing safety monitoring models in the safety monitoring model library, the surface icing safety monitoring model in the present invention is obtained through the design improvement of the present invention, and the specific implementation method is:

[0111] The acquisition unit comprises:

[0112] The environment acquisition module is used to collect the environmental data around the bridge. Since the icing of the bridge is related to environmental factors, such as temperature and humidity, the environmental data around the bridge can be collected as training data so that the environmental factors can be used as a condition for judging icing. However, it is inaccurate and incomplete to monitor icing only based on environmental data such as temperature and humidity, because these temperature and humidity sensors are installed on the guardrails, roadbeds or road piles on both sides of the bridge pavement, and the object to be monitored is the pavement. The temperature and humidity of the pavement are different from the temperature and humidity of the environment on both sides. For example, when the pavement is paved with snow removal agents, the icing of the pavement requires different environmental and humidity conditions than those under normal conditions. For example, when the traffic volume on the road is large, the temperature of the pavement is higher than the ambient temperature on both sides of the road due to the friction of the tires and the emission temperature of the exhaust gas. Therefore, it is inaccurate to use the ambient temperature on both sides of the road to warn of icing of the bridge pavement. Therefore, in order to obtain a model that can accurately judge whether the bridge pavement is frozen, the present invention also designs the image and detection monitoring to complement each other, so as to achieve comprehensive and accurate safety monitoring of bridge pavement icing.

[0113] The image acquisition unit is used to collect image data of the bridge pavement. The purpose of collecting the image of the bridge pavement is that the applicant has found that the image of the bridge pavement before it freezes is different. Therefore, by collecting image data, it can be determined whether the bridge pavement is frozen. The bridge pavement will have the following changes after it freezes: Color change: After the pavement freezes, the color may become brighter because the ice layer will reflect more light. In the snow, the color of snow is usually white, while the ice layer may appear brighter white or transparent, so it can present different performances in the image; Texture change: Unfrozen pavement usually has rough textures, such as ruts, gravel, etc. After freezing, these textures will be covered by ice, and these changes and differences can be reflected in the image; Reflective properties: After freezing, the pavement will become smoother and the ability to reflect light will be enhanced, especially under the sunlight or car lights, there will be obvious reflection, and these changes and differences can be reflected in the image; Transparency change: The ice layer may make the pavement look more transparent, especially at night or in dim light, this sense of transparency will be more obvious, and these changes and differences can be reflected in the image;

[0114] A reflection collection unit, wherein the reflection collection unit includes a transmitting end and a collecting end distributed on both sides of the bridge pavement, the transmitting end is used to transmit detection light to the bridge pavement, and the collecting end is used to receive the detection light reflected from the bridge pavement to obtain light detection data; when the pavement is not frozen, the detection light generated by the transmitting end enters the collecting end after being reflected by the pavement, and after the pavement is frozen, the route of the reflected detection light is changed due to the influence of the ice, so that the collecting end collects detection light different from the previous one, so whether the pavement is frozen can also be judged by the reflection collection unit, and the reflection collection unit is designed to include the transmitting end and the collecting end distributed on both sides of the bridge pavement to transmit the detection light from both sides of the pavement, which can reduce the interference to the vehicles traveling on the pavement, and has high safety;

[0115] The surface icing safety monitoring model includes:

[0116] An environmental safety monitoring model, used to obtain a bridge pavement environmental safety monitoring result based on the bridge surrounding environment data;

[0117] An image safety monitoring module, used to obtain bridge pavement image safety monitoring results based on bridge pavement image data;

[0118] A detection safety monitoring module is used to obtain bridge pavement detection safety monitoring results based on light detection data;

[0119] The output module is used to obtain the bridge pavement icing safety monitoring results based on the analysis of the beam pavement environment safety monitoring results, the bridge pavement image safety monitoring results and the bridge pavement detection safety monitoring results.

[0120] Among them, the surface icing safety monitoring model in the present invention includes an environmental safety monitoring model, an image safety monitoring module and a detection safety monitoring module, which can comprehensively evaluate whether the road surface is frozen from three perspectives: environment, image and detection, so that the safety monitoring of bridge road surface icing is accurate and comprehensive.

[0121] Among them, in the embodiment of the present invention, the bridge pavement icing safety monitoring result is obtained based on the analysis of the beam pavement environment safety monitoring result, the bridge pavement image safety monitoring result and the bridge pavement detection safety monitoring result, specifically including:

[0122] Obtained the first safety monitoring score based on the beam pavement environmental safety monitoring results;

[0123] Obtain the second safety monitoring score based on the bridge pavement image safety monitoring results;

[0124] Obtain the third safety monitoring score based on the bridge pavement detection safety monitoring results;

[0125] The total safety monitoring score is obtained based on the first to third safety monitoring scores, and the bridge pavement icing safety monitoring results are obtained based on the total safety monitoring score.

[0126] Among them, each monitoring result can obtain a corresponding safety monitoring score, and then the corresponding monitoring scores can be summarized to obtain the bridge pavement icing safety monitoring result. Accurate and comprehensive bridge pavement icing safety monitoring results can be obtained through the above method, and the three safety monitoring modules can make up for each other's respective defects and shortcomings.

[0127] Among them, in the embodiment of the present invention:

[0128] If the bridge pavement environmental safety monitoring result is that the bridge pavement is frozen, the first safety monitoring score is a; if the bridge pavement environmental safety monitoring result is that the bridge pavement is not frozen, the first safety monitoring score is b;

[0129] If the bridge pavement image safety monitoring result shows that the bridge pavement is frozen, the second safety monitoring score is c; if the bridge pavement environmental safety monitoring result shows that the bridge pavement is not frozen, the second safety monitoring score is d;

[0130] If the bridge pavement detection safety monitoring result is that the bridge pavement is frozen, the third safety monitoring score is e; if the bridge pavement detection safety monitoring result is that the bridge pavement is not frozen, the third safety monitoring score is f;

[0131] Total safety monitoring score = (a or b) + (c or d) + (e or f); where a <b,c<d,e<f;

[0132] If the total safety monitoring score is within the first preset range, the bridge pavement icing safety monitoring result is safe; if the total safety monitoring score is not within the first preset range, the bridge pavement icing safety monitoring result is unsafe.

[0133] In practical applications, the value of af can be adjusted according to actual needs, and the embodiment of the present invention does not make corresponding limitations.

[0134] Among them, in the embodiment of the present invention:

[0135] The environmental safety monitoring model is obtained as follows:

[0136] A bridge model is established and placed in a test box. The structure and material of the bridge model are the same as those of the bridge. A plurality of vehicle models that can move back and forth are arranged on the bridge model. A temperature control system and an airflow control system are arranged in the test box. Temperature acquisition equipment and humidity acquisition equipment are installed on both sides of the bridge of the bridge model. The actual conditions of the bridge pavement can be simulated by the vehicle model, so that the prediction capability of the environmental safety monitoring model can be more accurate. Therefore, the traveling vehicle can increase the temperature of the bridge pavement. The temperature control system and the airflow control system can make the environment around the bridge pavement of the bridge model closer to the real environment around the bridge pavement.

[0137] Conduct multiple tests in the test chamber, and record the test data of each test. The test content includes: starting the vehicle model to move back and forth, starting the temperature control system and the airflow control system, and the temperature control system controls the ambient temperature in the test chamber from the initial temperature until the bridge deck of the bridge model freezes; the airflow control system is used to simulate and control the airflow speed around the bridge model in the test chamber based on the meteorological data of the area where the monitored bridge is located. The simulation can match the environment with the real environment to ensure the accuracy of the monitoring results of the safety monitoring model; the test data includes: the corresponding ambient temperature data and ambient humidity data when the bridge deck of the bridge model freezes;

[0138] The environmental safety monitoring model is obtained by training the prediction model based on the test data; the prediction model can be an existing AI model, a deep learning model, a machine learning model, an intelligent model, etc., and the embodiments of the present invention do not make corresponding limitations.

[0139] The image security monitoring module is obtained as follows:

[0140] Collect images of the bridge pavement at various time periods when the pavement is not frozen and has no water accumulation to obtain a first image set;

[0141] The second image set is obtained by collecting images of the bridge road surface at various time periods when the bridge road surface is not frozen and has accumulated water;

[0142] The third image set is obtained by collecting images of the bridge road surface at various time periods when the bridge road surface is frozen;

[0143] A first data set is constructed based on the first to third image sets, the first data set is annotated to obtain a first training set, and a model is trained based on the first training set to obtain the image safety monitoring module.

[0144] Among them, the applicant discovered during actual use that water accumulation would also change the emission route of the detection light, causing the detection safety monitoring module to detect inaccurately when there is water accumulation. The image safety monitoring module obtained by taking into account the water accumulation situation in the above manner can make up for this defect. Similarly, the environmental safety monitoring model is based on the environmental monitoring data on both sides of the bridge pavement, which deviates from the actual environment of the bridge pavement. This deviation can be compensated by the image safety monitoring module and the detection safety monitoring module. The above three safety monitoring models can be combined to obtain accurate and comprehensive bridge pavement icing safety monitoring results, and the three safety monitoring modules can make up for each other's respective defects and deficiencies.

[0145] Among them, in the embodiment of the present invention, the detection safety monitoring module is obtained in the following manner:

[0146] Collecting detection light reflected from the bridge pavement at each time period when the bridge pavement is not frozen, to obtain a first detection light signal;

[0147] Collecting detection light reflected from the bridge pavement at each time period when the bridge pavement is frozen, to obtain a second detection light signal;

[0148] A second data set is constructed based on the first detection light signal and the second detection light signal, the second data set is labeled to obtain a second training set, and the detection safety monitoring module is obtained based on a training model of the second training set.

[0149] Among them, the bridge road surface itself is an asphalt road surface, which is bumpy and uneven. When it is frozen, the detection light has a fixed reflection path. After the road surface is frozen, the road surface becomes flat and changes the reflection path of the detection light. Therefore, it is possible to judge whether it is frozen by judging whether the reflected detection light path has changed. The transmitting end and the collecting end can be corresponding devices or components with light or signal transmitting and receiving functions, and the embodiments of the present invention do not make corresponding limitations.

[0150] Among them, the transmitting end and the collecting end distributed on both sides of the bridge pavement can be installed on both sides of the bridge pavement through corresponding installation or fixing equipment, such as being fixed on both sides of the pavement through a bracket, and then both the transmitting end and the receiving end are facing the middle of the pavement. Other fixing methods can also be used, such as directly fixing on the road blocks on both sides of the bridge, etc. The embodiments of the present invention do not make corresponding limitations and can be adjusted according to actual needs.

[0151] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0152] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A security monitoring intelligent question-answering system based on AI big model, characterized in that: The system comprises: The client is used for user login and obtaining the user's question information after the user logs in; A natural language processing unit, used to analyze and process the question information to obtain a question processing result; A collection unit, used for collecting monitoring data of the monitored object; A storage unit, used for storing monitoring data; A security monitoring model library is used to store multiple security monitoring models. The security monitoring models are used to perform security analysis and processing based on corresponding monitoring data to obtain security monitoring results. A matching unit, the matching unit is used to match the corresponding monitoring data and the security monitoring model based on the question processing result, and obtain the security monitoring data corresponding to the question information based on the matched monitoring data and the security monitoring model; A knowledge graph unit, used to establish a knowledge graph library based on the monitoring data and the safety monitoring data; A question-answer matching model, used to match an answer result matching the question information from the knowledge graph library based on the question processing result; A display unit is used to display the answer result to the user.

2. According to claim 1, a security monitoring intelligent question-answering system based on an AI big model is characterized in that: The matching of corresponding monitoring data and security monitoring model based on the question processing result specifically includes: Performing semantic analysis on the question processing result to obtain question processing semantic information; Performing feature extraction on the question processing semantic information to obtain a feature processing result; Inputting the feature processing result into the first matching model for matching to obtain corresponding monitoring data; Inputting the feature processing result into a second matching model for matching to obtain a corresponding safety monitoring model; The first matching model is obtained as follows: Constructing first training data, the first training data includes a plurality of first-class sample data, the first-class sample data includes monitoring data and corresponding feature information; Training the matching model based on the first training data to obtain a first matching model; The second matching model is obtained as follows: Constructing second training data, the second training data includes a plurality of second-category sample data, the second-category sample data includes a security monitoring model and corresponding feature information; The matching model is trained based on the second training data to obtain a second matching model.

3. According to claim 1, a security monitoring intelligent question-answering system based on an AI big model is characterized in that: The client also includes a voice processing unit, which is used to collect and obtain user voice question information; pre-process the voice question information to obtain first voice information; perform spectrum analysis on the first voice information to obtain human voice information; and perform voiceprint recognition on the human voice information to obtain the user's question information.

4. According to claim 1, a security monitoring intelligent question-answering system based on an AI big model is characterized in that: The system also includes a statistical unit, a pre-processing unit and a cache unit; The statistical unit is used to: record the historical question information of the user, count the historical question information, obtain the historical question times of the corresponding question information, sort the historical question information in descending order based on the historical question times, and obtain the commonly used question information based on the top N historical question information in the sorting, where N is an integer greater than 1; The preprocessing unit is used to obtain the answer result corresponding to the commonly used question information based on the monitoring data and the commonly used question information at each preset time interval after the user logs in to the client; A cache unit, the cache unit is used to store answer results corresponding to commonly used question information; The client determines whether the user's question information is commonly used question information, and if so, matches and obtains the corresponding answer result from the cache unit; if not, the user's question information is sent to the natural language processing unit for processing.

5. According to claim 1, a security monitoring intelligent question-answering system based on an AI big model is characterized in that: The safety monitoring model library stores a bridge pavement icing safety monitoring model, and the bridge pavement icing safety monitoring model is used to perform bridge pavement icing safety analysis processing based on corresponding monitoring data to obtain a bridge pavement icing safety monitoring result.

6. According to claim 5, a security monitoring intelligent question-answering system based on an AI big model is characterized in that: The acquisition unit comprises: Environmental collection module, used to collect environmental data around the bridge; An image acquisition unit, used for acquiring bridge pavement image data; A reflection collection unit, wherein the reflection collection unit includes a transmitting end and a collecting end distributed on both sides of the bridge pavement, the transmitting end is used to transmit detection light to the bridge pavement, and the collecting end is used to receive the detection light reflected from the bridge pavement to obtain light detection data; The surface icing safety monitoring model includes: An environmental safety monitoring model, used to obtain a bridge pavement environmental safety monitoring result based on the bridge surrounding environment data; An image safety monitoring module, used to obtain bridge pavement image safety monitoring results based on bridge pavement image data; A detection safety monitoring module is used to obtain bridge pavement detection safety monitoring results based on light detection data; The output module is used to obtain the bridge pavement icing safety monitoring results based on the analysis of the beam pavement environment safety monitoring results, the bridge pavement image safety monitoring results and the bridge pavement detection safety monitoring results.

7. The intelligent question-answering system for security monitoring based on AI big model according to claim 6 is characterized in that: The bridge pavement icing safety monitoring results are obtained based on the analysis of the beam pavement environment safety monitoring results, the bridge pavement image safety monitoring results and the bridge pavement detection safety monitoring results, specifically including: Obtained the first safety monitoring score based on the beam pavement environmental safety monitoring results; Obtain the second safety monitoring score based on the bridge pavement image safety monitoring results; Obtain the third safety monitoring score based on the bridge pavement detection safety monitoring results; The total safety monitoring score is obtained based on the first to third safety monitoring scores, and the bridge pavement icing safety monitoring results are obtained based on the total safety monitoring score.

8. The security monitoring intelligent question-answering system based on AI big model according to claim 7 is characterized by: If the bridge pavement environmental safety monitoring result is that the bridge pavement is frozen, the first safety monitoring score is a; if the bridge pavement environmental safety monitoring result is that the bridge pavement is not frozen, the first safety monitoring score is b; If the bridge pavement image safety monitoring result shows that the bridge pavement is frozen, the second safety monitoring score is c; if the bridge pavement environmental safety monitoring result shows that the bridge pavement is not frozen, the second safety monitoring score is d; If the bridge pavement detection safety monitoring result is that the bridge pavement is frozen, the third safety monitoring score is e; if the bridge pavement detection safety monitoring result is that the bridge pavement is not frozen, the third safety monitoring score is f; Total safety monitoring score = (a or b) + (c or d) + (e or f); where a <b,c<d,e<f; If the total safety monitoring score is within the first preset range, the bridge pavement icing safety monitoring result is safe; if the total safety monitoring score is not within the first preset range, the bridge pavement icing safety monitoring result is unsafe.

9. The security monitoring intelligent question-answering system based on AI big model according to claim 6 is characterized by: The environmental safety monitoring model is obtained as follows: A bridge model is established and placed in a test box. The structure and material of the bridge model are the same as those of the bridge. A plurality of vehicle models that can move back and forth are arranged on the bridge model. A temperature control system and an airflow control system are arranged in the test box. Temperature collection equipment and humidity collection equipment are installed on both sides of the bridge of the bridge model. Conduct multiple tests in the test chamber, and record the test data of each test. The test content includes: starting the vehicle model to move back and forth, starting the temperature control system and the airflow control system, and the temperature control system controls the ambient temperature in the test chamber to drop from the initial temperature until the bridge deck of the bridge model freezes; the airflow control system is used to simulate and control the airflow velocity around the bridge model in the test chamber based on the meteorological data of the area where the monitored bridge is located; the test data includes: the ambient temperature data and ambient humidity data corresponding to the bridge deck of the bridge model when ice forms; Training the prediction model based on the test data to obtain the environmental safety monitoring model; The image security monitoring module is obtained as follows: Collect images of the bridge pavement at various time periods when the pavement is not frozen and has no water accumulation to obtain a first image set; The second image set is obtained by collecting images of the bridge road surface at various time periods when the bridge road surface is not frozen and has accumulated water; The third image set is obtained by collecting images of the bridge road surface at various time periods when the bridge road surface is frozen; A first data set is constructed based on the first to third image sets, the first data set is annotated to obtain a first training set, and a model is trained based on the first training set to obtain the image safety monitoring module.

10. The intelligent question-answering system for security monitoring based on AI big model according to claim 6 is characterized in that: The detection safety monitoring module is obtained in the following manner: Collecting detection light reflected from the bridge pavement at each time period when the bridge pavement is not frozen, to obtain a first detection light signal; Collecting detection light reflected from the bridge pavement at each time period when the bridge pavement is frozen, to obtain a second detection light signal; A second data set is constructed based on the first detection light signal and the second detection light signal, the second data set is labeled to obtain a second training set, and the detection safety monitoring module is obtained based on a training model of the second training set.

Images