Government business scene model training method, control method, device and equipment

Abstract

The invention provides a government business scene model training method, control method, device and equipment, and the training method comprises the steps: obtaining a data set which comprises an unlabeled data set and a labeled small sample data set related to a target government business scene; carrying out unsupervised training on a computer vision CV large model by utilizing the unlabeled data set, carrying out iterative optimization adjustment on parameters and an algorithm of the CV large model according to a training result and an actual demand of a target governance business scene, and obtaining an intermediate CV large model when an optimization adjustment result reaches a minimum reconstruction error; performing optimization training on the intermediate CV large model by using the labeled small sample data set; and knowledge distillation is carried out on the trained CV large model to obtain a small model corresponding to the target treatment scene. According to the method, fine tuning training and knowledge distillation are carried out on the CV large model, so that the model can efficiently identify the thematic scene, the development cost is reduced, and the identification accuracy and generalization ability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital government construction, and in particular to a training method for a government business scenario model, a control method, device and electronic equipment for government business scenario governance. Background Art

[0002] With the rapid development of computer vision (CV) technology, smart city governance solutions based on CV large models are gradually emerging. These governance solutions usually use deep learning technology to train high-performance computer vision CV large models on a large amount of visual data to achieve accurate recognition and analysis of complex urban scenes. The current application capabilities of smart city governance are often used to solve specific scenario problems using a single or limited number of algorithm models, or to use large model capabilities to retrieve and discover generalized non-specific scenario problems.

[0003] In the related technologies, the business model based on the CV big model requires highly customized development for problems and business needs in specific scenarios. It has problems such as long development cycle, high cost, and difficulty in quickly adapting to new scenarios. At the same time, highly customized scenario results are difficult to reuse, which greatly increases the development cost. For new scenarios, the generalization ability of the big model is limited, the application method of the ability is not well matched with business needs, the user operation is complicated, and the experience is poor.

[0004] Therefore, based on the convenience and actual usage needs of common business scenarios in digital government, how to improve the efficiency of using large models is a problem that needs to be solved. Summary of the invention

[0005] The present invention provides a training method for a government business scenario model, a control method, device and electronic device for government business scenario governance, so as to at least solve the problems in the related art that the CV large model needs to be customized for specific scenarios and actual business needs, the cost is high, the generalization ability of the large model is limited, the matching degree between the ability application method and business needs is low, the user operation is complex, the experience is poor, and the efficiency of the large model is reduced. The technical solution of the present invention is as follows:

[0006] According to a first aspect of an embodiment of the present invention, a method for training a government business scenario model is provided, comprising:

[0007] Acquire a data set, wherein the data set includes: an unlabeled data set and a labeled small sample data set related to a target governance business scenario;

[0008] The unlabeled data set is used to perform unsupervised training on the computer vision CV large model, and the parameters and algorithms of the CV large model are iteratively optimized and adjusted according to the training results and the actual needs of the target governance business scenario. When the optimization and adjustment results achieve the minimum reconstruction error, an intermediate CV large model is obtained;

[0009] Utilizing the labeled small sample data set to optimize and train the intermediate CV large model to obtain a trained CV large model;

[0010] The trained large CV model is subjected to knowledge distillation to obtain a small model corresponding to the target governance scenario.

[0011] Optionally, the method further includes:

[0012] Integrate the corresponding small model of the target governance scenario into the preset scenario supermarket;

[0013] The integrated small models of the target governance scenario are encapsulated, and a matching user interface is set for each small model, so that the user can select the small model corresponding to the target governance scenario through the user interface.

[0014] Optionally, the method further includes:

[0015] Preprocessing the acquired unlabeled data set to obtain a processed unlabeled data set;

[0016] The unsupervised training of the computer vision CV large model using the unlabeled data set specifically includes: unsupervised training of the computer vision CV large model using the processed unlabeled data set, and iteratively optimizing and adjusting the parameters and algorithms of the CV large model according to the training results and the actual needs of the target governance business scenario, and obtaining an intermediate CV large model when the optimization and adjustment result achieves the goal of minimizing the reconstruction error.

[0017] Optionally, before acquiring the data set, the method further includes:

[0018] Obtaining city governance requirements, including data requirements, governance objectives, and user role requirements under different government governance business scenarios;

[0019] According to the data requirements of the government in different target governance business scenarios, determine the selection of corresponding surveillance camera points and preset urban video perception layout planning suggestions, where the layout planning suggestions include: camera selection, installation location and coverage;

[0020] Based on the processing speed and resource allocation of the computer vision CV large model, determine the application mode of each target governance business scenario and the algorithm that matches the corresponding application mode;

[0021] After determining the algorithm that matches the application mode of each target governance business scenario, corresponding scenario-specific functions are set based on each target governance business scenario to quickly deploy and adapt to different urban governance needs. The scenario-specific functions include: scenario-specific algorithms, scene camera screening, decision analysis report generation, and problem rectification record tracking.

[0022] Optionally, the processing speed and resource allocation of the computer vision CV large model are used to determine the application mode of each target governance business scenario and the algorithm matching the corresponding application mode, including:

[0023] Determine the application mode of multiple target governance business scenarios in the city based on the urban governance needs; and based on the processing speed and resource allocation, data statistics and visualization display of the computer vision CV large model, set the video artificial intelligence AI-perceived alarm event push method, and determine the algorithm that matches the application mode of each target governance business scenario based on the decision analysis of the AI ​​scenario.

[0024] Optionally, the method further includes:

[0025] System testing is carried out on the set video artificial intelligence AI perception and AI scene topics, as well as the corresponding application modes, to ensure the stable operation of the system in various target governance scenarios, wherein the system testing includes: functional testing, performance testing and user experience testing.

[0026] According to a second aspect of an embodiment of the present invention, a control method for government business scenario governance is provided, including:

[0027] In response to the user's operation instruction, the small model of the target governance scene and the corresponding scene parameters are selected from the scene supermarket through the user interface. The small model of the target governance scene is a small model obtained by training the computer vision CV large model and performing knowledge distillation on the trained CV large model;

[0028] Acquire the video perception data of the city to be detected for the target governance scenario;

[0029] Based on the selected small model of the target governance scenario and the corresponding scenario parameters, intelligent analysis is performed on the video perception data of the city to be detected to obtain a decision analysis report on the governance results of the corresponding target governance scenario;

[0030] When the decision analysis report does not meet the urban governance requirements of the corresponding scenario, an alarm event of the corresponding scenario is generated and pushed;

[0031] The decision analysis report of the target governance scenario governance results and the corresponding alarm events are displayed visually.

[0032] Optionally, the method further includes:

[0033] When the decision analysis report meets the urban governance needs of the corresponding scenario, regularly obtain the use evaluation results of the small model of the target governance scenario integrated in the scenario supermarket, as well as user feedback data;

[0034] The parameters of the small model of the target governance scenario integrated in the scenario supermarket are adjusted and optimized by using the usage evaluation results and the user feedback data.

[0035] Optionally, the method further includes:

[0036] The scene supermarket system is regularly maintained, and the daily maintenance includes at least one of data backup, system update and troubleshooting.

[0037] According to a third aspect of an embodiment of the present invention, there is provided a training device for a government business scenario model, comprising:

[0038] A first acquisition module is used to acquire a data set, wherein the data set includes: an unlabeled data set and a labeled small sample data set related to a target governance business scenario;

[0039] The first training module is used to perform unsupervised training on the computer vision CV large model using the unlabeled data set, and iteratively optimize and adjust the parameters and algorithms of the CV large model according to the training results and the actual needs of the target governance business scenario, and obtain the intermediate CV large model when the optimization and adjustment result achieves the minimum reconstruction error;

[0040] The second training module is used to optimize and train the intermediate CV large model using the labeled small sample data set to obtain a trained CV large model;

[0041] The knowledge distillation module is used to perform knowledge distillation on the trained CV large model to obtain a small model corresponding to the target governance scenario.

[0042] Optionally, the device further comprises:

[0043] An integration module, used to integrate the small model of the target governance scenario obtained by the knowledge distillation module into a preset scenario supermarket;

[0044] The encapsulation module is used to encapsulate the small models of the target governance scenario after integration by the integration module, and to set a matching user interface for each small model so that the user can select the small model corresponding to the target governance scenario through the user interface.

[0045] Optionally, the device further comprises:

[0046] A preprocessing module, used for preprocessing the acquired unlabeled data set to obtain a processed unlabeled data set;

[0047] The first training module is specifically used to perform unsupervised training on the computer vision CV large model using the processed unlabeled data set, and iteratively optimize and adjust the parameters and algorithms of the CV large model according to the training results and the actual needs of the target governance business scenario. When the optimization and adjustment results achieve the minimum reconstruction error, an intermediate CV large model is obtained.

[0048] Optionally, the device further comprises:

[0049] A second acquisition module is used to acquire city governance requirements before the first acquisition module acquires the data set, wherein the city governance requirements include: data requirements, governance objectives, and user role requirements under different government target governance business scenarios;

[0050] The point selection determination module is used to determine the corresponding surveillance camera point selection and preset urban video perception layout planning suggestions according to the data requirements of the government in different target governance business scenarios, wherein the layout planning suggestions include: camera selection, installation location and coverage;

[0051] The video perception module is used to determine the application mode of each target governance business scenario and the algorithm that matches the corresponding application mode based on the processing speed and resource allocation of the computer vision CV large model;

[0052] The scenario-specific module is used to set corresponding scenario-specific functions based on each target governance business scenario after determining the algorithm that matches the application mode of each target governance business scenario, so as to quickly deploy and adapt to different urban governance needs. The scenario-specific functions include: scenario-specific algorithms, scene camera screening, decision analysis report generation and problem rectification record tracking.

[0053] Optionally, the video perception module is specifically used to determine the application mode of multiple target governance business scenarios in the city according to the urban governance needs; and to set the alarm event push method of video artificial intelligence AI perception based on the processing speed and resource allocation, data statistics and visual display of the computer vision CV large model; and to determine the algorithm that matches the application mode of each target governance business scenario based on the decision analysis of the AI ​​scene topic, wherein the alarm event push method includes: platform display, topic subscription overview, SMS notification, at least one way to push the matching algorithm detection and recognition results, and the scenario decision analysis includes: customized recommendation and decision suggestion text setting for the corresponding scenario data.

[0054] Optionally, the device further comprises:

[0055] The testing module is used to perform system testing on the set video artificial intelligence AI perception and AI scene topics, as well as the corresponding application modes, to ensure the stable operation of the system in various target governance scenarios, wherein the system testing includes: functional testing, performance testing and user experience testing.

[0056] According to a fourth aspect of an embodiment of the present invention, a control device for government business scenario governance is provided, including:

[0057] A selection module is used to respond to the user's operation instruction and select the required small model of the target governance scene and the corresponding scene parameters from the scene supermarket through the user interface. The small model of the target governance scene is a small model obtained by training the computer vision CV large model and performing knowledge distillation on the trained CV large model;

[0058] A data acquisition module, used to acquire the video perception data of the city to be detected in the target governance scenario;

[0059] An intelligent analysis module, which is used to perform intelligent analysis on the video perception data of the city to be detected based on the selected small model of the target governance scenario and the corresponding scenario parameters, and obtain a decision analysis report on the governance results of the corresponding target governance scenario;

[0060] A generation module, used to generate and push an alarm event for a corresponding scenario when the urban governance requirements of the corresponding scenario are not met in the decision analysis report;

[0061] The display module is used to visualize the decision analysis report of the target governance scenario governance results and the corresponding alarm events.

[0062] Optionally, the device further comprises:

[0063] An evaluation result acquisition module is used to regularly obtain the use evaluation results of the small model of the target governance scenario integrated in the scenario supermarket and the user feedback data when the decision analysis report meets the urban governance needs of the corresponding scenario;

[0064] An adjustment and optimization module is used to adjust and optimize the parameters of the small model of the target governance scenario integrated in the scenario supermarket by using the usage evaluation results and the user feedback data.

[0065] Optionally, the device further comprises:

[0066] The maintenance module is used to regularly perform routine maintenance on the scene supermarket system, and the routine maintenance includes at least one of data backup, system update and troubleshooting.

[0067] According to a fifth aspect of an embodiment of the present invention, there is provided an electronic device, including:

[0068] processor;

[0069] a memory for storing instructions executable by the processor;

[0070] In which, the processor is configured to execute the instructions to implement the training method of the digital government business scenario model as described above or the control method of government business scenario governance as described above.

[0071] According to the sixth aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the training method of the digital government business scenario model as described above or the control method of government business scenario governance as described above.

[0072] According to the seventh aspect of an embodiment of the present invention, there is provided a computer program product, comprising a computer program or instructions, which, when executed by a processor of an electronic device, implements the training method of the digital government business scenario model as described above or the control method of government business scenario governance as described above.

[0073] The technical solution provided by the embodiments of the present invention brings at least the following beneficial effects:

[0074] In an embodiment of the present invention, a data set is obtained, and the data set includes: an unlabeled data set and a labeled small sample data set related to the target governance business scenario; the unlabeled data set is used to perform unsupervised training on the computer vision CV large model, and the parameters and algorithms of the CV large model are iteratively optimized and adjusted according to the training results and the actual needs of the target governance business scenario, and when the optimization and adjustment results reach the minimum reconstruction error, an intermediate CV large model is obtained; the labeled small sample data set is used to optimize the training of the intermediate CV large model to obtain a trained CV large model; the trained CV large model is subjected to knowledge distillation to obtain a small model corresponding to the target governance scenario. In other words, the embodiment of the present invention performs fine-tuning training and knowledge distillation on the computer vision large model, so that the model can extract the features of thematic scene data from massive data for efficient recognition, thereby reducing development costs, having powerful image processing and detection and recognition capabilities, and improving the generalization ability and accuracy of the large model.

[0075] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] The drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present invention, and are used together with the specification to explain the principles of the present invention, and do not constitute an improper limitation of the present invention. In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0077] Figure 1 It is a flowchart of a training method for a government business scenario model provided by an embodiment of the present invention.

[0078] Figure 2 It is a schematic diagram of training based on Telesearch provided by an embodiment of the present invention.

[0079] Figure 3 A schematic diagram of a scene supermarket system provided in an embodiment of the present invention that integrates a video AI sense module and an AI scene topic module.

[0080] Figure 4 It is a flow chart of a control method for government business scenario governance provided by an embodiment of the present invention.

[0081] Figure 5It is an application example diagram of a control method for government business scenario governance provided by an embodiment of the present invention.

[0082] Figure 6 It is a block diagram of a training device for a government business scenario model provided by an embodiment of the present invention.

[0083] Figure 7 It is a block diagram of a control device for government business scenario governance provided by an embodiment of the present invention.

[0084] Figure 8 It is a block diagram of an electronic device provided by an embodiment of the present invention.

[0085] Fig. 9 It is a block diagram of a device for training a government business scenario model or controlling government business scenario governance provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0086] In order to enable ordinary persons in the art to better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0087] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0088] Technical terms:

[0089] Scenario Supermarket: Provides multiple commonly used urban governance scenarios for the digital government industry, and sets matching interfaces and functions for each scenario within the application platform.

[0090] Generalization: refers to the ability of an AI model to maintain good performance on data that has not been used and trained.

[0091] Cross-validation: A technique for evaluating statistical analysis methods that evaluates the generalization ability of a model by reusing data. The original dataset is divided into several smaller subsets, and the model is trained and validated on these subsets multiple times, each time using a different subset of the data as the training set and validation set.

[0092] Computer vision (CV) large models refer to large deep learning models used for computer vision tasks. They are usually implemented using deep learning algorithms such as convolutional neural networks (CNN). With the development of deep learning technology and the improvement of computing power, CV large models have achieved many important results in the field of computer vision and have been applied in many fields.

[0093] On the basis of understanding the above technical terms, please refer to the following embodiments.

[0094] Figure 1 is a flow chart of a method for training a government business scenario model provided by an embodiment of the present invention, such as Figure 1 As shown, the method comprises the following steps:

[0095] Step 101: Acquire a data set, where the data set includes: an unlabeled data set and a labeled small sample data set related to a target governance business scenario.

[0096] Step 102: Use the unlabeled data set to perform unsupervised training on the computer vision CV large model, and iteratively optimize and adjust the parameters and algorithms of the CV large model according to the training results and the actual needs of the target governance business scenario. When the optimization and adjustment results achieve the minimum reconstruction error, an intermediate CV large model is obtained.

[0097] Step 103: Utilize the labeled small sample data set to optimize and train the intermediate CV large model to obtain a trained CV large model.

[0098] Step 104: Perform knowledge distillation on the trained CV large model to obtain a small model corresponding to the target governance scenario.

[0099] The embodiments of the present invention aim to solve the problems of existing business models based on CV big models in terms of high development cost for highly customized scenarios, limited generalization ability of big models, low matching between ability application methods and business needs, and poor user operation experience. Unsupervised training is performed, and the parameters and algorithms of the CV big model are iteratively optimized and adjusted according to the training results and the actual needs of the target governance business scenarios. The intermediate CV big model is optimized and trained using the labeled small sample data set to obtain a trained CV big model. Knowledge distillation is performed on the trained CV big model to obtain a small model corresponding to the target governance scenario, thereby improving the flexibility and scalability of the system. While fully enriching the application mode of the big model, it deeply meets the needs of user scenarios, so that the difficulty of using the government platform is reduced while being able to respond to diverse governance challenges more conveniently and efficiently, thereby providing the government with a powerful intelligent governance tool and promoting the in-depth development of smart city construction.

[0100] The training method of the government business scenario model described in the present invention can be applied to terminals, servers, etc., without limitation. The terminal implementation device can be an electronic device such as a smart phone, a laptop computer, a tablet computer, a desktop computer, a personal digital assistant (PDA, Personal Digital Assistant) and a wearable device. The server can be an independent server or a server cluster, or a server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, intermediate services, domain name services, security services, content distribution networks, or a big data and artificial intelligence platform, etc., without limitation.

[0101] Combine the following Figure 1 , the specific implementation steps of a training method for a government business scenario model provided by an embodiment of the present invention are described in detail.

[0102] In step 101, a data set is obtained, where the data set includes: an unlabeled data set and a labeled small sample data set related to a target governance business scenario.

[0103] In this step, a variety of data sets related to the target governance business scenario are obtained, and the data sets may include: unlabeled data sets and labeled small sample data sets related to the target governance business scenario. For example, it may include vehicle information data, monitoring video streams, geographic location information, problem history records, etc. For another example, for the special governance scenario of electric vehicles, data sets can be obtained by deploying cameras in key areas of the city, such as major traffic arteries, residential entrances and exits, commercial blocks, etc., to detect and judge the video every 5 seconds for 24 hours and perform frame detection every second to obtain identification detection data. These data not only include the appearance features of electric vehicles and places, but also cover key information such as their behavior, status, relative position and related objects.

[0104] The specific acquisition method may be to obtain the data set of the corresponding scene from the local database, or to obtain the data set of the corresponding scene from the cloud, or to obtain the data set by combining the two, which is not limited in this embodiment.

[0105] Among them, the target governance business scenarios in this embodiment may include but are not limited to: special governance of electric vehicles, special governance of electric vehicles, special governance of garbage trucks, special governance of rivers, special governance of traffic lights, and other special scenarios with complex digital government usage needs and high frequency of use.

[0106] In step 102, the unlabeled data set is used to perform unsupervised training on the computer vision CV large model, and the parameters and algorithms of the CV large model are iteratively optimized and adjusted according to the training results and the actual needs of the target governance business scenario. When the optimization and adjustment results achieve the minimum reconstruction error, the intermediate CV large model is obtained.

[0107] In this embodiment, the computer vision CV large model can be trained based on Telesearch as an example, as shown in the following example: Figure 2 As shown, Figure 2 A schematic diagram of a training based on Telesearch provided in an embodiment of the present invention. However, in practice, it is not limited to this. Based on Telesearch, unsupervised training can be performed using unlabeled data, and the model can be initially usable by minimizing the reconstruction error, that is, a usable pre-trained model (i.e., an intermediate CV large model) is obtained, and the formula for minimizing the reconstruction error is as follows:

[0108]

[0109] Among them, x l is the original data, is the reconstructed data. recon represents the reconstruction error.

[0110] In step 103, the intermediate CV large model is optimized and trained using the labeled small sample data set to obtain a trained CV large model.

[0111] In this step, based on the available pre-trained model trained in step 102, supervised training is performed using a labeled small sample data set to obtain a pre-trained model (i.e., a trained CV large model) to improve the retrieval and recognition capabilities of the model in thematic scenarios, that is, the output results of the available pre-trained model are compared with the labeled small sample data set, and the loss function (i.e., the loss value) is calculated based on the difference in the comparison, and the parameters of the model are adjusted with the loss function: the parameters of the model are updated through a back propagation mechanism based on the loss function, and after multiple iterations, until the loss value reaches a minimum value (i.e., the loss function is minimized) or the model converges, a trained pre-trained model (i.e., a trained CV large model) is obtained.

[0112] Among them, the expression of minimizing the loss function is shown in the following formula:

[0113]

[0114] Among them, y i is the true label, is the probability predicted by the model.

[0115] It should be noted that during the training process, the algorithm will learn based on the relationship between the characteristics of the data and the target variable, and continuously adjust the model parameters to optimize the prediction performance. Through technical means such as cross-validation and hyperparameter tuning, the generalization ability and accuracy of the model can be further improved.

[0116] The training of the CV large model in this embodiment is carried out on the basis of Star Telesearch. The training process mainly uses the unlabeled data of each scene for unsupervised training. After that, the small samples labeled with the corresponding scene are used for optimization training to improve the recognition efficiency of the large model heap target governance scene and the generalization performance of new scenes.

[0117] In step 103, knowledge distillation is performed on the trained large CV model to obtain a small model corresponding to the target governance scenario.

[0118] In this step, after obtaining the trained CV large model, knowledge distillation is performed on the trained CV large model to obtain a small model corresponding to the target governance scenario. The small model is a lightweight and dedicated small model, which can also be called an application-adapted real-time model. The final completed small model will be able to automatically analyze the image stream data obtained after video frame extraction, quickly identify the scene target behavior or objects, and provide timely and accurate decision-making support for government regulatory departments.

[0119] Among them, knowledge distillation is simply understood as a method of extracting knowledge from cumbersome models and compressing it into a single model so that it can be deployed in practical applications.

[0120] In an embodiment of the present invention, a data set is obtained, and the data set includes: an unlabeled data set and a labeled small sample data set related to the target governance business scenario; the unlabeled data set is used to perform unsupervised training on the computer vision CV large model, and the parameters and algorithms of the CV large model are iteratively optimized and adjusted according to the training results and the actual needs of the target governance business scenario, and when the optimization and adjustment results reach the minimum reconstruction error, an intermediate CV large model is obtained; the labeled small sample data set is used to optimize the training of the intermediate CV large model to obtain a trained CV large model; the trained CV large model is subjected to knowledge distillation to obtain a small model corresponding to the target governance scenario. In other words, the embodiment of the present invention performs fine-tuning training and knowledge distillation on the computer vision large model, so that the model can extract the features of thematic scene data from massive data for efficient recognition, reduces development costs, has powerful image processing and detection and recognition capabilities, improves the generalization ability and accuracy of the large model, and improves the availability of model applications.

[0121] Optionally, in another embodiment, based on the above embodiment, the method may further include: integrating the small model of the target governance scenario into a preset scenario supermarket; packaging the integrated small model of the target governance scenario, and setting a matching user interface for each small model, so that the user can select the small model of the target governance scenario through the user interface. The structure of the scenario supermarket is preset, including: a user interface, a backend service, and a database.

[0122] In this embodiment, the architecture of the scene supermarket is pre-designed according to user operation instructions, which may include a front-end user interface, back-end services, a database, etc., to ensure the availability and scalability of the scene supermarket system.

[0123] After that, the model is integrated and packaged. First, the trained and tuned models are integrated into the scenario supermarket. Secondly, the small models of the target governance scenario are packaged after integration, and a matching user interface is set for each small model so that users can select the small model corresponding to the target governance scenario through the user interface. In other words, the integrated small models are packaged into service modules that are easy to deploy and manage. The purpose is to facilitate the deployment of the trained CV large model to the application platform, so that users can choose to deploy it.

[0124] Among them, before setting a matching user interface for each small model, it is necessary to pre-select and design the user interface. This embodiment can be designed as a concise and intuitive front-end user interface (i.e., user operation interface), for example, a user interface designed with a clear navigation structure, intuitive operation buttons, and real-time feedback mechanisms. In this way, the user's learning cost can be reduced and the system's usability and user satisfaction can be improved. The design of the application mode will have clear prompts in terms of interaction and operation, which are specifically reflected in: 1) The function icon is simple and easy to understand; 2) The location of common functions is easy to see; 3) Visible content is displayed based on user needs and supports customization; 4) The operation process is simple, and there are fewer steps in viewing and alarm event processing; 5) The push text content is concise and supports modification.

[0125] Optionally, in another embodiment, based on the above embodiment, the method may also include: preprocessing the acquired unlabeled data set to obtain a processed unlabeled data set; using the unlabeled data set to perform unsupervised training on the computer vision CV large model, specifically including: using the processed unlabeled data set to perform unsupervised training on the computer vision CV large model, and iteratively optimizing and adjusting the parameters and algorithms of the CV large model according to the training results and the actual needs of the target governance business scenario, and obtaining an intermediate CV large model when the optimization and adjustment result achieves the minimum reconstruction error.

[0126] In this embodiment, the acquired data set may be pre-processed such as cleaning to ensure the quality, consistency and applicability of the data and improve the accuracy of model training. The model is trained and fine-tuned based on the pre-processed data set to optimize model performance.

[0127] Optionally, in another embodiment, based on the above embodiment, before acquiring the data set, the method may further include:

[0128] 1) Obtaining city governance needs, including data needs, governance goals, and user role needs under different government governance business scenarios;

[0129] In this step, before obtaining the needs of urban governance, it is necessary to first conduct an in-depth analysis of the specific needs of digital government services and clarify the application requirements in different scenarios. You can select multiple key application scenarios in smart cities with similar monitoring modes and high usage frequency, that is, they meet the characteristics of high versatility, large number of events, scattered event points, and clear scene characteristics. The scene type can be represented by the set J = {1, 2, ..., j}, for example: special governance of electric vehicles, special governance of muck trucks, river governance, and special governance of traffic lights. The selection of scene types is mainly determined through user visits, industry information feedback, competitive product results research, and relevant data release. Among them, the dispersion of event points is determined by comparing the latitude and longitude variances of different scenes in the survey. To determine, among which, represents the longitude equation, represents the latitude variance, and j represents the scene type. Whether the scene features are clear or not is determined based on the feature average of the comparison scene algorithm. The calculation formula for the feature average is:

[0130]

[0131] Among them, j represents the scene type and n represents the total number of scene features.

[0132] 2) According to the data requirements of the government in different target governance business scenarios, determine the selection of corresponding surveillance camera points and preset urban video perception layout planning suggestions, where the layout planning suggestions include: camera selection, installation location and coverage;

[0133] In this step, the selection of surveillance camera locations is determined based on the government's different thematic scenario requirements, and cameras of multiple types, specifications, and access methods are centrally managed. At the same time, recommendations can be made on the layout planning of the urban video perception system, including camera selection, installation location, coverage, etc.

[0134] 3) Based on the processing speed and resource allocation of the computer vision CV large model, determine the application mode for each target governance business scenario and the algorithm that matches the application mode.

[0135] This step specifically includes: determining the application mode of multiple target governance business scenarios in the city according to the urban governance needs; and setting the video artificial intelligence AI-perceived alarm event push method based on the processing speed and resource allocation, data statistics and visual display of the computer vision CV large model; and determining the algorithm that matches the application mode of each target governance business scenario based on the decision analysis of the AI ​​scenario, wherein the alarm event push method includes: pushing the matching algorithm detection and recognition results through at least one of platform display, special subscription overview, and SMS notification, and the scenario decision analysis includes: customized recommendation and decision suggestion text setting for corresponding scenario data.

[0136] That is to say, in this step, after determining the scenario, it is first necessary to design a real-time application mode and an algorithm that matches the application mode for the large model processing speed and resource allocation issues, wherein the application mode can be an application mode for key application scenarios in smart cities. The algorithm can be efficiently applied to data processing, optimize resource scheduling strategies, and realize a rapid response mechanism for the model.

[0137] This embodiment uses this design to ensure that the large model can quickly process massive amounts of data and respond to urban management needs in real time. In an optional embodiment, the application mode can be matched by setting the alarm event push method of the video AI perception module and the decision analysis of the AI ​​scene topic. Among them, the alarm event push supports multiple channels such as platform display, topic subscription overview, and SMS notification to push matching algorithm detection and recognition results, and the decision analysis report of the AI ​​scene topic supports customized recommendations for key data and decision suggestion text settings.

[0138] 4) After determining the algorithm that matches the application mode of each target governance business scenario, set the corresponding scenario-specific functions based on each target governance business scenario to quickly deploy and adapt to different urban governance needs. The scenario-specific functions may include: scenario-specific algorithms, scene camera screening, decision analysis report generation, and problem rectification record tracking.

[0139] In this step, after completing the design of the algorithm that matches the application mode, a series of preset scenario-specific module functions are developed based on the defined application scenarios. These scenario-specific module functions will contain algorithms and tools for specific scenarios, realizing modular design for rapid deployment and adaptation to different urban management needs. Define the function set of scenario j as F j ={A,C j , Rj , P j , T j ,}, including algorithm training function A, scene camera screening function C j , decision analysis report function R j , Problem rectification record tracking function j , scene-specific functions T j Among them, only the function design of A is not customized for the scenario, and only differs in the small sample data used for fine-tuning training. j It represents new functional modules proposed by users for thematic scenarios in addition to the preset functions, which provides space for the iteration and upgrade of the product. Other functions will have customized differences in certain fields or input and output data.

[0140] In the embodiments of the present invention, by designing real-time application modes and thematic scenarios that can respond quickly, the camera screen content is retrieved and identified in real time, and alarms are pushed in real time. This mode ensures that the large model can quickly process massive data and respond to the government's monitoring needs in real time, while optimizing the server resource configuration and reducing the cost and time consumption of model adaptation to different scenarios. And the embodiments of the present invention pre-design and integrate a series of thematic modules for different service scenarios to assist in unified monitoring and management of urban governance, safe construction, water conservancy and environmental protection, traffic management and other directions, such as electric vehicle management, river management, traffic light management and other scenarios, greatly reducing the demand and cost of customized development. This modular design improves the implementation efficiency and flexibility of the project, enables the system to be quickly applied to specific scenarios, and improves the availability, frequency of use and convenience of the system.

[0141] Optionally, in another embodiment, based on the above embodiment, the method may also include: performing system testing on the set video artificial intelligence AI perception and AI scene topics, and corresponding application modes to ensure the stable operation of the system in various target governance scenarios, wherein the system testing includes: functional testing, performance testing, and user experience testing, etc.

[0142] In this step, the application modes of the set video artificial intelligence AI perception (i.e., video AI sense module) and scene topics (i.e., AI scene topic module) are integrated into an application platform (such as the application platform of the scene supermarket system, etc.) for comprehensive system testing. Its system testing can include: functional testing, performance testing, user experience testing, etc., to ensure that the system can run stably in various scenarios. During the system design and testing process, it is necessary to continuously optimize the server resource configuration to reduce the additional cost and time consumption caused by the adaptation of the model to different scenarios. Realize efficient use of resources and ensure the economic feasibility of the system. Specifically, Figure 3 As shown, Figure 3A schematic diagram of a scene supermarket system provided in an embodiment of the present invention that integrates a video AI sense module and an AI scene topic module.

[0143] like Figure 3 As shown, the scene supermarket system includes: a video AI sense module and an AI scene topic module, wherein the video AI sense module includes: data statistics, visual display (or visual map display) and alarm event push functions as an example, but in actual application, it is not limited to this. The AI ​​scene topic module includes: scene topic algorithm (or algorithm training), scene camera screening, decision analysis report generation and problem rectification record tracking functions as an example, of course, it is not limited to this, and can also include expanded content, such as scene topic functions of newly added application scenarios, etc.

[0144] See also Figure 4 , Figure 4 A flowchart of a control method for government business scenario governance provided by an embodiment of the present invention, the method comprising:

[0145] Step 401: In response to the user's operation instructions, the required small model of the target governance scenario and the corresponding scene parameters are selected from the scene supermarket through the user interface. The small model of the target governance scenario is based on training the computer vision CV large model and performing knowledge distillation on the trained CV large model.

[0146] In this step, the user can select the required small model of the target governance scenario and the corresponding scene parameters from the scene supermarket through the user interface, wherein the small model of the target governance scenario is a small model obtained by training the computer vision CV large model and performing knowledge distillation on the trained CV large model. The specific training process is detailed in the corresponding embodiment above and will not be repeated here.

[0147] Step 402: Obtain the city video perception data to be detected for the target governance scenario.

[0148] In this step, the video perception data of the city to be detected in the target governance scene can be obtained in real time or regularly through the video artificial intelligence AI perception on the scene supermarket. Of course, the video perception data of the city to be detected in the target governance scene can also be obtained by other means, which is not limited in this implementation.

[0149] Step 403: Based on the selected small model of the target governance scenario and the corresponding scenario parameters, the video perception data of the city to be detected is intelligently analyzed to obtain a decision analysis report on the governance results of the corresponding target governance scenario.

[0150] In this step, the selected small model of the target governance scenario and the corresponding parameters are deployed and implemented in the scene supermarket of urban management, and the video perception data of the city to be detected is intelligently analyzed to monitor the operation of the system and obtain a decision analysis report on the governance results of the corresponding target governance scenario.

[0151] In this step, the system can receive data from the urban video perception system in real time, perform intelligent analysis through small models deployed in corresponding scenarios, identify and warn of various urban governance issues, and support alarm record tracing and decision-making analysis report generation.

[0152] In other words, this embodiment can timely and accurately discover emergency events in urban governance through the real-time monitoring and early warning mechanism of customized scenarios, improve the response speed and quality of public services, and the system deeply fits the needs of government departments, has high real-time and accuracy, and can effectively improve the governance efficiency of urban management departments. Provide timely and effective decision support.

[0153] Step 404: When the urban governance requirements of the corresponding scenario are not met in the decision analysis report, an alarm event of the corresponding scenario is generated and pushed.

[0154] In this step, the system first determines whether the decision analysis report meets the urban governance needs of the corresponding scenario. If the decision analysis report does not meet the urban governance needs of the corresponding scenario, an alarm event of the corresponding scenario is generated and pushed to the user. The push method can be through platform display, topic subscription overview, SMS notification and other methods. Of course, it can also be pushed through other methods, which are not limited in this embodiment.

[0155] Step 405: Visually display the decision analysis report of the target governance scenario governance results and the corresponding alarm events.

[0156] Among them, the city status and governance effects can be visualized in the form of data, charts, maps, etc. At the same time, the user interface design of the embodiment of the present invention is concise, intuitive, and conforms to the user operation interface and process of the digital government scene, and adopts a user-friendly design concept, such as a clear navigation structure, intuitive operation buttons, and real-time feedback mechanism, etc. It is convenient for users to intuitively understand the governance situation, solves the user learning burden caused by complex system interfaces and operation processes, reduces the user's learning cost, and improves the system's ease of use and user satisfaction.

[0157] Optionally, in another embodiment, based on the above embodiment, the method may also include: when the decision analysis report meets the urban governance needs of the corresponding scenario, regularly obtaining the usage evaluation results of the small model of the target governance scenario integrated in the scenario supermarket, as well as user feedback data; using the usage evaluation results and the user feedback data to adjust and optimize the parameters of the small model of the target governance scenario integrated in the scenario supermarket.

[0158] In this embodiment, after selecting the required small model of the target governance scenario and the corresponding scenario parameters for deployment, a continuous iteration and improvement mechanism is established, and the system functions are continuously optimized and the system performance is improved according to user feedback and market changes to ensure that the system always meets the actual use needs of digital government affairs. This embodiment can provide the government with a powerful intelligent governance tool and promote the in-depth development of smart city construction.

[0159] Optionally, in another embodiment, based on the above embodiment, the method may further include: regularly performing routine maintenance on the scene supermarket system, and the routine maintenance includes: at least one of data backup, system update and troubleshooting.

[0160] The embodiment of the present invention performs daily maintenance on the scene supermarket system, including: data backup, system update, fault troubleshooting, etc.

[0161] Furthermore, it is also possible to track the latest CV technologies and urban governance concepts, upgrade the technology and expand the functions of the scene supermarket, and maintain the system's advancement and competitiveness.

[0162] See also Figure 5 , is a flowchart of an application example of a training application of a government business scenario model provided by an embodiment of the present invention. The method comprises:

[0163] Step 500: Urban governance needs research and analysis

[0164] This step requires in-depth research into the specific needs of urban governance, including data requirements, governance goals, user roles and their needs in different scenarios, and determining commonly used urban governance scenarios based on scientific analysis methods.

[0165] Step 501: Video perception system planning:

[0166] In this step, the selection of surveillance camera locations is determined based on the determined urban governance scenario requirements, and cameras of multiple types, specifications, and access methods are centrally managed. At the same time, suggestions can be made on the layout planning of the urban video perception system, including camera selection, installation location, coverage, etc.

[0167] Step 502: Design of real-time application mode for large models

[0168] After determining the urban governance scenario, we first need to design a real-time application mode and matching algorithm for the large model in view of the large model processing speed and resource allocation issues, so that the algorithm can be efficiently applied to data processing, optimize resource scheduling strategies, and implement a rapid response mechanism for the large model. This design ensures that the large model can quickly process massive amounts of data and respond to urban management needs in real time.

[0169] Step 503: Preset scenario theme module design

[0170] In this step, after completing the application model design, a series of preset scenario-specific module functions will be developed based on the defined application scenarios. These functions will include algorithms and tools for specific scenarios, and realize modular design so that they can be quickly deployed and adapted to different urban management needs.

[0171] The preset scene theme modules can be integrated into the scene supermarket to realize the modular development of the scene supermarket.

[0172] Step 504: Optimizing user interaction experience

[0173] In this step, after the backend design is completed, it is necessary to design a concise and intuitive front-end user interface, using user-friendly design concepts such as clear navigation structure, intuitive operation buttons and real-time feedback mechanism. This will reduce the user's learning cost and improve the system's ease of use and user satisfaction. The design of the application mode will have clear prompts in terms of interaction and operation, which are specifically reflected in: 1) The function icon is simple and easy to understand; 2) The location of common functions is easy to see; 3) Visible content is displayed based on user needs and supports customization; 4) The operation process is simple, with fewer steps for viewing and alarm event processing; 5) The push text content is concise and supports modification.

[0174] Step 505: System integration and testing

[0175] In this step, after completing the application design, it is necessary to integrate all preset modules and real-time application modes into a unified application platform and conduct comprehensive system testing. Testing includes: functional testing, performance testing, user experience testing, etc., to ensure that the system can run stably in various scenarios. During the system design and testing process, it is necessary to continuously optimize the server resource configuration to reduce the additional cost and time consumption caused by model adaptation to different scenarios. Achieve efficient use of resources and ensure the economic feasibility of the system.

[0176] Step 506: Application deployment of the scene mini-model

[0177] After ensuring that the system is stable and reliable, we began to deploy and implement small models of corresponding scenarios in actual urban management scenarios.

[0178] It should be noted that the small model for the corresponding scenario deployed and implemented is a small model obtained by training the large computer vision CV model and performing knowledge distillation on the trained large CV model. The specific training process is described above and will not be repeated here.

[0179] Step 507: Determine whether the decision analysis report of the governance result output by the small model meets the government governance requirements. If not, execute step 508 and return to step 504; if yes, execute step 509;

[0180] After the small model is deployed and implemented, the system operation is monitored to determine whether the decision analysis report of the governance results output by the small model meets the government governance needs. If not, execute step 508 to collect user feedback information and then return to step 504; if it is compliant, execute step 509 to continue iterating and improving, that is, regularly adjust and optimize the parameters of the small model for the corresponding scenario.

[0181] Step 508: Obtain user feedback information;

[0182] In this step, after obtaining the user feedback information, the process returns to step 504 .

[0183] Step 509: Continue to iterate and improve.

[0184] This step can specifically involve regularly adjusting and optimizing the parameters of the small model for the corresponding scenario. In other words, after the system is deployed, a continuous iteration and improvement mechanism is established. Based on user feedback and market changes, system functions are continuously optimized, system performance is improved, and the system is ensured to always meet the actual use needs of digital government affairs.

[0185] In this embodiment, the accuracy of the small model is evaluated, that is, by comparing the system recognition results with the manual labeling results, the accuracy of the model in electric vehicle detection, license plate recognition, and illegal behavior recognition is evaluated. The results show that the accuracy of the small model used in the present invention is above 90%, and all indicators reach or exceed the industry average.

[0186] This embodiment provides a small model with strong generalization ability for the corresponding scene, that is, the small model is applied to other similar scenes to evaluate its performance in the new scene. The results show that the model has good generalization ability, can quickly adapt to new environments and maintain a high recognition accuracy.

[0187] In this embodiment, the system performance evaluation of the deployed small model is relatively strong in real-time and stability. For real-time, the response time and data processing speed of the system under different network environments are tested. The results show that the system can process large-scale video data in real time and complete recognition and alarm generation in real time. For stability: the system is run for a long time and at a high frequency to observe whether it crashes, freezes, and other abnormal conditions occur. The results show that the system runs stably without obvious failures.

[0188] This embodiment achieves a high user experience evaluation and satisfaction, among which, user convenience: invite government officials with non-technical backgrounds to try the system and collect their feedback. The results show that the system operation interface is concise and easy to use; users can quickly master the use of the system and complete related tasks. High satisfaction, collect user satisfaction evaluations on the system through questionnaires and other methods. The results show that users are generally satisfied with the performance, stability and ease of use of the system; they believe that the system effectively improves the efficiency and accuracy of traffic monitoring.

[0189] For easier understanding, please also refer to the following application examples.

[0190] The embodiment of the present invention takes the digital government scenario supermarket system based on the CV large model as an example. Its application field is not limited to a single traffic monitoring and other field scenarios, but can be widely extended to multiple urban governance fields and specific technology usage scenarios. Below, this embodiment takes the four aspects of electric vehicle special governance, muck truck special governance, river governance and traffic light special governance as examples to explain in detail different usage scenarios:

[0191] First, special treatment of electric vehicles

[0192] In response to the problems of electric vehicles entering buildings and homes, illegal parking, illegal modification, illegal driving, illegal charging, etc., this special scenario integrates high-definition video surveillance, intelligent recognition algorithms and big data analysis technology to achieve real-time monitoring and early warning of electric vehicle behavior. The system can automatically identify electric vehicle violations and send alarms to relevant departments in real time to improve law enforcement efficiency and achieve efficient management of electric vehicles. The system supports the generation of decision-making analysis reports to provide a basis for electric vehicle management; it supports problem rectification records to achieve accurate tracing of alarm events.

[0193] Second, special treatment of muck trucks

[0194] As an indispensable part of urban construction, muck trucks have serious problems such as management compliance and dust pollution during transportation that seriously affect urban traffic and environmental governance. This special scenario uses high-definition video AI perception to monitor the implementation of regulations, loading status, and environmental protection measures for muck trucks in real time. The system can automatically analyze whether the muck truck's license plate top light spraying is compliant, whether the loading exceeds the standard, whether effective dust prevention measures are taken, etc. Once a violation is found, an alarm will be immediately sent to the management department, and the vehicle's location will be tracked to ensure that the violation is promptly investigated and dealt with. The system supports the generation of decision-making analysis reports to provide a basis for muck truck management; it supports problem rectification records to achieve accurate tracing of alarm events.

[0195] Third, river management

[0196] In response to problems such as river pollution, illegal fishing, illegal swimming, the spread of aquatic plants, and illegal construction in rivers, this special scenario uses technical means such as drone monitoring, high-definition video perception, and intelligent analysis platforms to achieve all-round monitoring of rivers. Drones regularly cruise the river and take high-definition images and videos. The system automatically analyzes the water quality of the river and illegal activities based on the color of the water surface and objects on the water surface. Once an abnormality is found, an alarm is immediately sent to the environmental protection department, and detailed information on the pollution area and pollution source is provided to provide strong support for governance work. The system supports the generation of decision-making analysis reports to provide a basis for river governance; it supports problem rectification records to achieve accurate tracing of alarm events.

[0197] Fourth, special treatment of traffic lights

[0198] Install high-definition cameras and sensors at key traffic intersections to achieve real-time monitoring of violations such as vehicles running red lights, crossing the line, giving way to pedestrians, driving in the wrong lane, and pedestrians running red lights and walking on the wrong side of the road. Secondly, use image recognition technology and intelligent algorithms to quickly process and analyze the collected video data, accurately identify and mark violations. Then upload the violation identification results and decision analysis reports to the traffic management center in real time as a basis and suggestion for subsequent penalties. This special scenario can intelligently identify violation targets and generate alarm events and reports based on monitoring data, improve the efficiency of identifying and handling violations, and ensure traffic order and safety. The system supports problem rectification records and achieves accurate tracing of alarm events.

[0199] To sum up, the thematic scenarios of this embodiment play an important role in key scenarios in the digital government field such as special governance of electric vehicles, special governance of garbage trucks, special governance of rivers, and special governance of traffic lights through highly customized and intelligent means, providing strong technical support and decision-making basis for the government, and promoting urban governance to develop in a more refined and intelligent direction.

[0200] The governance process for different scenarios includes the following steps:

[0201] 1) Demand research and analysis: Conduct in-depth research on the specific needs of urban governance, including data requirements, governance goals, user roles and their needs in different scenarios, and determine commonly used urban governance scenarios based on scientific analysis methods.

[0202] 2) Video perception system planning: Determine the selection of surveillance camera locations based on the needs of thematic scenarios, and centrally manage cameras of multiple types, specifications, and access methods. At the same time, provide suggestions on the layout planning of the city's video perception system, including camera selection, installation location, coverage, etc.

[0203] 3) Design of real-time application mode for large models: After determining the scenario, we first need to design a real-time application mode to match the algorithm for the processing speed and resource allocation of large models, so that the algorithm can be efficiently applied to data processing, optimize resource scheduling strategies, and implement a rapid response mechanism for the model. This design ensures that the large model can quickly process massive data and respond to urban management needs in real time.

[0204] 4) Preset scenario-specific module design: After completing the application model design, a series of preset scenario-specific module functions will be developed based on the defined application scenarios. These functions will contain algorithms and tools for specific scenarios, and realize modular design for rapid deployment and adaptation to different urban management needs.

[0205] 5) Data collection: Prepare a wide and diverse data set, perform preprocessing such as data cleaning and labeling to ensure data quality; then train and fine-tune the model to optimize model performance; then conduct a series of tests and evaluations to ensure that the model achieves the expected performance; finally, deploy the trained Telesearch model to the application platform.

[0206] 6) Model tuning: Based on the pre-training results and the actual needs of the subject scenario, the model is further tuned, including adjusting model parameters, optimizing algorithms, etc., to improve model performance.

[0207] 7) Scenario supermarket architecture design: Design the architecture of the scenario supermarket, including the front-end user interface, back-end services, database, etc., to ensure high availability and scalability of the system.

[0208] 8) Model integration and packaging: Integrate pre-trained and tuned models into the scenario supermarket and package them into service modules that are easy to deploy and manage.

[0209] 9) Effect evaluation and feedback: Regularly evaluate the model governance effect, collect user feedback, and adjust and optimize the model based on the evaluation results and feedback.

[0210] 10) System maintenance: Perform daily maintenance of the scene supermarket system, including data backup, system updates, troubleshooting, etc.

[0211] 11) Technology upgrade and product iteration: Track the latest CV technology and urban governance concepts, upgrade the technology and expand the functions of the scene supermarket, and maintain the system's advancement and competitiveness.

[0212] For ease of understanding, the embodiments of the present invention take the processes of special management of electric vehicles and river management as examples respectively. Other management scenarios are similar and will not be described in detail in this embodiment.

[0213] 1. The special treatment process for electric vehicles specifically includes:

[0214] 11) Data Collection

[0215] In the special treatment of electric vehicles, data collection is the primary link. Through the deployment of cameras in key areas of the city, such as traffic arteries, entrances and exits of residential areas, commercial blocks, etc., the video is tested and judged every 5 seconds for 24 hours, and the frame detection is performed every second to obtain identification detection data. These data not only include the appearance characteristics of electric vehicles and places, but also cover key information such as their behavior, status, relative position and related objects.

[0216] 12) Data preprocessing

[0217] The collected raw video data needs to go through preprocessing steps such as cleaning and labeling. The cleaning process mainly removes low-quality data such as invalid frames and blurred images to ensure the accuracy of subsequent analysis.

[0218] 13) Small sample pre-training

[0219] Using the multimodal generalization platform of the Telesearch platform, the CV large model is pre-trained locally with small samples. This step aims to use a small amount of labeled data to quickly optimize the model's ability to recognize electric vehicle-related features and improve the model's generalization performance in new scenarios. Through pre-training, the model can more accurately identify electric vehicles and their behavioral characteristics.

[0220] 14) Model training and optimization

[0221] On the basis of pre-training, we continuously improve the recognition accuracy and processing speed of the model in the special treatment of electric vehicles by adjusting model parameters, optimizing algorithms and other strategies. At the same time, we use cross-validation and other methods to evaluate the model performance to ensure the stability and reliability of the model in practical applications.

[0222] 15) Scenario supermarket construction and model deployment

[0223] The trained electric vehicle special governance model is integrated into the scenario supermarket for users to choose to deploy. The scenario supermarket provides an intuitive user interface and convenient operation process. Users can select the appropriate scenario model and configure relevant parameters according to actual needs. After deployment, the system can receive data from the city video AI perception system in real time and perform intelligent analysis through the deployed model.

[0224] 16) Real-time monitoring and effect evaluation

[0225] The system monitors the driving and parking of electric vehicles in real time, automatically identifies and triggers alarms for behaviors such as not wearing a helmet, illegal charging, and electric vehicles entering homes. The system also has data visualization functions, which display data overviews in the form of numbers, charts, maps, etc., so that users can intuitively understand the governance situation. By regularly evaluating the model governance effect and user feedback, the model performance and service quality are continuously optimized.

[0226] 2. The river management process specifically includes:

[0227] 21) River pollution monitoring

[0228] The water quality is monitored in real time along the river and in key watersheds through video analysis technology. The system sets a river pollution identification algorithm to automatically detect and identify pollution signs such as floating garbage, oil pollution, abnormal color changes, etc. Once pollution is found, the system will immediately trigger an alarm and push it to the responsible person of the relevant department through SMS, application platform, etc. At the same time, the system supports automatic generation of decision analysis reports and problem rectification tracing.

[0229] 22) Illegal swimming monitoring

[0230] Video surveillance points are set up in river areas where swimming is prohibited to conduct intelligent identification of surveillance images. The system can automatically detect and track the activities of people in the river in the designated area, especially identifying swimming behavior. Once an illegal swimmer is confirmed, the system will immediately generate an alarm message, automatically lock the swimmer's location, and push the alarm message to law enforcement personnel or management departments through the preset communication channels. At the same time, the system supports the automatic generation of decision analysis reports and problem rectification tracing.

[0231] 23) Illegal construction warning in river channel

[0232] The system uses drones and high-altitude cameras installed along the river to monitor both sides of the river, detect and identify buildings or other illegal constructions that meet the characteristics of illegal construction. Once suspected illegal construction is found, the system will automatically mark and generate alarm information, including the specific location, scale and time of the illegal construction, and push the information to the river management department for rapid response and on-site verification. The system will also automatically generate a decision analysis report and record the results of each illegal construction detection, form a problem rectification record, and track the progress and results of the handling of illegal construction.

[0233] In the embodiment of the present invention, the application of a multimodal CV model is realized, that is, a large-scale computer vision (CV) model trained by deep learning technology, which has powerful image processing and detection and recognition capabilities. Combining various types of data such as video, image, text, etc., the recognition accuracy and robustness of the model are improved through a multimodal fusion algorithm.

[0234] In the embodiment of the present invention, for the CV model with a large number of parameters, through fine-tuning training and knowledge distillation, it is possible to extract the features of thematic scene data from massive data and perform efficient recognition, has strong generalization ability and high accuracy, has higher usability than the application of other models, and reduces the manual burden.

[0235] The real-time monitoring and intelligent analysis system of the embodiment of the present invention, that is, the system can receive data from the urban video perception system in real time, perform intelligent analysis through the small models of the deployed corresponding scenes, identify and warn of various urban governance issues, and support alarm record tracing and decision analysis report generation. It realizes the real-time monitoring and early warning mechanism of customized scenes, can timely and accurately detect emergency events in urban governance, improves the response speed and quality of public services, and the system is deeply embedded in the use needs of government departments, has a high degree of real-time and accuracy, and can effectively improve the governance efficiency of urban management departments. Provide timely and effective decision support.

[0236] The embodiment of the present invention also realizes data visualization and interactive design, that is, the city status and governance effect are visualized in the form of data, charts, maps, etc., and the user operation interface and process are simple, intuitive and in line with the digital government scene, and user-friendly design concepts are adopted, such as clear navigation structure, intuitive operation buttons and real-time feedback mechanism, etc. It is convenient for users to intuitively understand the governance situation, solves the user learning burden caused by complex system interface and operation process, reduces the user's learning cost, and improves the system's usability and user satisfaction.

[0237] It should be noted that, for the method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that this implementation disclosure is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the present invention.

[0238] See also Figure 6 , is a block diagram of a training device for a government business scenario model provided by an embodiment of the present invention. The device comprises: a first acquisition module 601, a first training module 602, a second training module 603 and a knowledge distillation module 604, wherein:

[0239] The first acquisition module 601 is used to acquire a data set, wherein the data set includes: an unlabeled data set and a labeled small sample data set related to a target governance business scenario;

[0240] The first training module 602 is used to perform unsupervised training on the computer vision CV large model using the unlabeled data set, and iteratively optimize and adjust the parameters and algorithms of the CV large model according to the training results and the actual needs of the target governance business scenario, and obtain an intermediate CV large model when the optimization and adjustment result reaches the minimum reconstruction error;

[0241] The second training module 603 is used to optimize and train the intermediate CV large model using the labeled small sample data set to obtain a trained CV large model;

[0242] The knowledge distillation module 604 is used to perform knowledge distillation on the trained CV large model to obtain a small model corresponding to the target governance scenario.

[0243] Optionally, in another embodiment, based on the above embodiment, the device further includes:

[0244] An integration module, used to integrate the small model of the target governance scenario obtained by the knowledge distillation module into a preset scenario supermarket;

[0245] The encapsulation module is used to encapsulate the small models of the target governance scenario after integration by the integration module, and to set a matching user interface for each small model so that the user can select the small model corresponding to the target governance scenario through the user interface.

[0246] Optionally, in another embodiment, based on the above embodiment, the device further includes:

[0247] A preprocessing module, used for preprocessing the acquired unlabeled data set to obtain a processed unlabeled data set;

[0248] The first training module is specifically used to perform unsupervised training on the computer vision CV large model using the processed unlabeled data set, and iteratively optimize and adjust the parameters and algorithms of the CV large model according to the training results and the actual needs of the target governance business scenario. When the optimization and adjustment results achieve the minimum reconstruction error, an intermediate CV large model is obtained.

[0249] Optionally, in another embodiment, based on the above embodiment, the device further includes:

[0250] A second acquisition module is used to acquire city governance requirements before the first acquisition module acquires the data set. The city governance requirements include: data requirements, governance objectives, and user role requirements under different government target governance business scenarios;

[0251] The point selection determination module is used to determine the corresponding surveillance camera point selection and preset urban video perception layout planning suggestions according to the data requirements of the government in different target governance business scenarios, wherein the layout planning suggestions include: camera selection, installation location and coverage;

[0252] The video perception module is used to determine the application mode of each target governance business scenario and the algorithm that matches the corresponding application mode based on the processing speed and resource allocation of the computer vision CV large model;

[0253] The scenario-specific module is used to set corresponding scenario-specific functions based on each target governance business scenario after determining the algorithm that matches the application mode of each target governance business scenario, so as to quickly deploy and adapt to different urban governance needs. The scenario-specific functions include: scenario-specific algorithms, scene camera screening, decision analysis report generation and problem rectification record tracking.

[0254] Optionally, the video perception module is specifically used to determine the application mode of multiple target governance business scenarios in the city according to the urban governance needs; and to set the alarm event push method of video artificial intelligence AI perception based on the processing speed and resource allocation, data statistics and visual display of the computer vision CV large model; and to determine the algorithm that matches the application mode of each target governance business scenario based on the decision analysis of the AI ​​scene topic, wherein the alarm event push method includes: platform display, topic subscription overview, SMS notification, at least one way to push the matching algorithm detection and recognition results, and the scenario decision analysis includes: customized recommendation and decision suggestion text setting for the corresponding scenario data.

[0255] Optionally, in another embodiment, based on the above embodiment, the device further includes:

[0256] The testing module is used to perform system testing on the set video artificial intelligence AI perception and AI scene topics, as well as the corresponding application modes, to ensure the stable operation of the system in various target governance scenarios, wherein the system testing includes: functional testing, performance testing and user experience testing.

[0257] , see also Figure 7 , a block diagram of a control device for government business scenario governance is provided in an embodiment of the present invention, the device includes: a selection module 701, a data acquisition module 702, an intelligent analysis module 703, a generation module 704 and a display module 705, wherein,

[0258] The selection module 701 is used to select a small model of a target governance scenario and corresponding scenario parameters from the scenario supermarket through the user interface in response to the user's operation instruction, wherein the small model of the target governance scenario is a small model obtained by training a computer vision CV large model and performing knowledge distillation on the trained CV large model;

[0259] The data acquisition module 702 is used to acquire the video perception data of the city to be detected of the target governance scenario;

[0260] The intelligent analysis module 703 is used to perform intelligent analysis on the video perception data of the city to be detected based on the selected small model of the target governance scene and the corresponding scene parameters, and obtain a decision analysis report on the governance result of the corresponding target governance scene;

[0261] A generation module 704 is used to generate and push an alarm event for a corresponding scenario when the urban governance requirements of the corresponding scenario are not met in the decision analysis report;

[0262] The display module 705 is used to visually display the decision analysis report of the target governance scenario governance results and the corresponding alarm events.

[0263] Optionally, in another embodiment, based on the above embodiment, the device further includes:

[0264] An evaluation result acquisition module is used to regularly obtain the use evaluation results of the small model of the target governance scenario integrated in the scenario supermarket and the user feedback data when the decision analysis report meets the urban governance needs of the corresponding scenario;

[0265] An adjustment and optimization module is used to adjust and optimize the parameters of the small model of the target governance scenario integrated in the scenario supermarket by using the usage evaluation results and the user feedback data.

[0266] Optionally, in another embodiment, based on the above embodiment, the device further includes:

[0267] The maintenance module is used to regularly perform routine maintenance on the scene supermarket system, and the routine maintenance includes at least one of data backup, system update and troubleshooting.

[0268] Optionally, an embodiment of the present invention further provides an electronic device, including:

[0269] processor;

[0270] a memory for storing instructions executable by the processor;

[0271] In which, the processor is configured to execute the instructions to implement the training method of the digital government business scenario model as described above or the control method of government business scenario governance as described above.

[0272] Optionally, an embodiment of the present invention also provides a computer-readable storage medium, which, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to execute the training method of the digital government business scenario model as described above or the control method of government business scenario governance as described above.

[0273] Optionally, an embodiment of the present invention also provides a computer program product, including a computer program or instructions, which, when executed by a processor of an electronic device, implements the training method of the digital government business scenario model as described above or the control method of government business scenario governance as described above.

[0274] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0275] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0276] Figure 8800 is a block diagram of an electronic device 800 provided in an embodiment of the present invention. For example, the electronic device 800 may be a mobile terminal or a server, and the electronic device is a mobile terminal as an example for description in the embodiment of the present invention. For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a message transceiver device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0277] Reference Figure 8 , the electronic device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0278] The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0279] The memory 804 is configured to store various types of data to support operations on the device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0280] The power supply component 806 provides power to the various components of the electronic device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 800.

[0281] The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and the rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0282] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the electronic device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0283] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: home button, volume button, start button, and lock button.

[0284] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open / closed state of the device 800, the relative positioning of components, such as the display and keypad of the electronic device 800, and the sensor assembly 814 can also detect the position change of the electronic device 800 or a component of the electronic device 800, the presence or absence of contact between the user and the electronic device 800, the orientation or acceleration / deceleration of the electronic device 800, and the temperature change of the electronic device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0285] The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on a communication standard, such as WiFi, a carrier network (such as 2G, 3G, 4G or 5G), or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0286] In an embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to execute the training method of the digital government business scenario model or the control method of government business scenario governance shown above.

[0287] In an embodiment, a computer-readable storage medium is also provided, and when the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device 800 can execute the training method of the digital government business scenario model or the control method of government business scenario governance shown above. For example, the computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0288] In an embodiment, a computer program product is also provided, including a computer program or instructions. When the computer program or instructions are executed by the processor 820 of the electronic device 800, the electronic device 800 executes the above-mentioned training method of the digital government business scenario model or the control method of government business scenario governance.

[0289] Fig. 9 900 is a block diagram of a device 900 for training a digital government business scenario model or controlling government business scenario governance provided by an embodiment of the present invention. For example, the device 900 may be provided as a server. Fig. 9, the apparatus 900 includes a processing component 922, which further includes one or more processors, and a memory resource represented by a memory 932 for storing instructions, such as an application, that can be executed by the processing component 922. The application stored in the memory 932 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 922 is configured to execute the instructions to perform the above method.

[0290] The device 900 may also include a power supply component 926 configured to perform power management of the device 900, a wired or wireless network interface 950 configured to connect the device 900 to a network, and an input / output (I / O) interface 958. The device 900 may operate based on an operating system stored in the memory 932, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or the like.

[0291] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed by the present invention. The specification and examples are to be considered exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.

[0292] It should be understood that the present invention is not limited to the exact construction that has been described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A training method for a government business scenario model, characterized in that: include: Acquire a data set, wherein the data set includes: an unlabeled data set and a labeled small sample data set related to a target governance business scenario; The unlabeled data set is used to perform unsupervised training on the computer vision CV large model, and the parameters and algorithms of the CV large model are iteratively optimized and adjusted according to the training results and the actual needs of the target governance business scenario. When the optimization and adjustment results achieve the minimum reconstruction error, an intermediate CV large model is obtained; Utilizing the labeled small sample data set to optimize and train the intermediate CV large model to obtain a trained CV large model; The trained large CV model is subjected to knowledge distillation to obtain a small model corresponding to the target governance scenario.

2. The training method of the government business scenario model according to claim 1 is characterized in that: The method further comprises: Integrate the corresponding small model of the target governance scenario into the preset scenario supermarket; The integrated small models of the target governance scenario are encapsulated, and a matching user interface is set for each small model, so that the user can select the small model corresponding to the target governance scenario through the user interface.

3. The training method of the government business scenario model according to claim 1 or 2, characterized in that: The method further comprises: Preprocessing the acquired unlabeled data set to obtain a processed unlabeled data set; The unsupervised training of the computer vision CV large model using the unlabeled data set specifically includes: unsupervised training of the computer vision CV large model using the processed unlabeled data set, and iteratively optimizing and adjusting the parameters and algorithms of the CV large model according to the training results and the actual needs of the target governance business scenario, and obtaining an intermediate CV large model when the optimization and adjustment result achieves the goal of minimizing the reconstruction error.

4. The training method of the government business scenario model according to claim 1 or 2, characterized in that: Before acquiring the data set, the method further includes: Obtaining city governance requirements, including data requirements, governance objectives, and user role requirements under different government governance business scenarios; According to the data requirements of the government in different target governance business scenarios, determine the selection of corresponding surveillance camera points and preset urban video perception layout planning suggestions, where the layout planning suggestions include: camera selection, installation location and coverage; Based on the processing speed and resource allocation of the computer vision CV large model, determine the application mode of each target governance business scenario and the algorithm that matches the corresponding application mode; After determining the algorithm that matches the application mode of each target governance business scenario, corresponding scenario-specific functions are set based on each target governance business scenario to quickly deploy and adapt to different urban governance needs. The scenario-specific functions include: scenario-specific algorithms, scene camera screening, decision analysis report generation, and problem rectification record tracking.

5. The training method of the government business scenario model according to claim 4 is characterized in that: The processing speed and resource allocation of the computer vision CV large model are used to determine the application mode of each target governance business scenario and the algorithm matching the corresponding application mode, including: Determine the application mode of multiple target governance business scenarios in the city based on the urban governance needs; and based on the processing speed and resource allocation, data statistics and visualization display of the computer vision CV large model, set the video artificial intelligence AI-perceived alarm event push method, and determine the algorithm that matches the application mode of each target governance business scenario based on the decision analysis of the AI ​​scenario.

6. A control method for government business scenario governance, characterized in that: include: In response to the user's operation instruction, the small model of the target governance scene and the corresponding scene parameters are selected from the scene supermarket through the user interface. The small model of the target governance scene is a small model obtained by training the computer vision CV large model and performing knowledge distillation on the trained CV large model; Acquire the video perception data of the city to be detected for the target governance scenario; Based on the selected small model of the target governance scenario and the corresponding scenario parameters, intelligent analysis is performed on the video perception data of the city to be detected to obtain a decision analysis report on the governance results of the corresponding target governance scenario; When the decision analysis report does not meet the urban governance requirements of the corresponding scenario, an alarm event of the corresponding scenario is generated and pushed; The decision analysis report of the target governance scenario governance results and the corresponding alarm events are displayed visually.

7. The control method for government business scenario governance according to claim 6 is characterized in that: The method further comprises: When the decision analysis report meets the urban governance needs of the corresponding scenario, regularly obtain the use evaluation results of the small model of the target governance scenario integrated in the scenario supermarket, as well as user feedback data; The parameters of the small model of the target governance scenario integrated in the scenario supermarket are adjusted and optimized by using the usage evaluation results and the user feedback data.

8. A training device for a government business scenario model, characterized in that: include: A first acquisition module is used to acquire a data set, wherein the data set includes: an unlabeled data set and a labeled small sample data set related to a target governance business scenario; The first training module is used to perform unsupervised training on the computer vision CV large model using the unlabeled data set, and iteratively optimize and adjust the parameters and algorithms of the CV large model according to the training results and the actual needs of the target governance business scenario, and obtain the intermediate CV large model when the optimization and adjustment result achieves the minimum reconstruction error; The second training module is used to optimize and train the intermediate CV large model using the labeled small sample data set to obtain a trained CV large model; The knowledge distillation module is used to perform knowledge distillation on the trained CV large model to obtain a small model corresponding to the target governance scenario.

9. A control device for government business scenario management, characterized in that: include: A selection module is used to respond to the user's operation instruction and select the required small model of the target governance scene and the corresponding scene parameters from the scene supermarket through the user interface. The small model of the target governance scene is a small model obtained by training the computer vision CV large model and performing knowledge distillation on the trained CV large model; A data acquisition module, used to acquire the video perception data of the city to be detected in the target governance scenario; An intelligent analysis module, which is used to perform intelligent analysis on the video perception data of the city to be detected based on the selected small model of the target governance scenario and the corresponding scenario parameters, and obtain a decision analysis report on the governance results of the corresponding target governance scenario; A generation module, used to generate and push an alarm event for a corresponding scenario when the urban governance requirements of the corresponding scenario are not met in the decision analysis report; The display module is used to visualize the decision analysis report of the target governance scenario governance results and the corresponding alarm events.

10. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; In which, the processor is configured to execute the instructions to implement the training method of the digital government business scenario model as described in any one of claims 1 to 5 or the control method of government business scenario governance as described in claim 6 or 7.

Images