Intelligent supervision method and system

Abstract

The application discloses a kind of wisdom supervision method and system, belong to wisdom supervision technical field, method includes: push supervision inspection task;According to supervision inspection task, determine inspection route;According to inspection route, each merchant is checked in turn;Get the image of the door head of merchant, extract door head information from door head image through image recognition technology, and compare with the historical door head data of background, to confirm whether to reach the merchant to be checked;Get the image of the business license of merchant, extract business license information from business license image through image recognition technology, and compare with the basic data of merchant in background, to confirm whether the business license of merchant is in compliance.In the present application, based on image recognition technology, the training data required for wisdom supervision is only business license image, it is difficult to obtain valuable data, the data acquisition range is extensive, the generalization ability of the model can be improved, and the supervision effect in the actual application process can be improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent supervision technology, specifically relating to an intelligent supervision method and system. Background Technology

[0002] In the modern business environment, market regulation plays a crucial role in maintaining fair competition, protecting consumer rights, and promoting market order. However, traditional market regulation methods are often limited by human resources, data analysis capabilities, and regulatory efficiency, making it difficult to comprehensively cover a wide range of businesses and prone to regulatory loopholes and information delays.

[0003] To address these challenges, intelligent regulatory approaches are gaining increasing attention. These approaches leverage advanced technologies such as cloud computing, big data analytics, artificial intelligence, and image recognition to improve the timeliness, accuracy, and coverage of regulation. Among existing technologies, early warning models based on backpropagation (BP) neural networks are widely used due to their high predictive accuracy. For example, a BP neural network-based early warning model for retailers selling counterfeit cigarettes can provide early warnings of such activities.

[0004] However, early warning models based on backpropagation (BP) neural networks typically require a large amount of high-quality, valuable data for training and optimization. Acquiring such data is costly, involving labor, equipment, and time. In some cases, the behaviors or events to be predicted may be sensitive, and merchants may be unwilling to disclose relevant data, making it difficult to obtain data on specific behaviors and thus limiting the training and application of the model. Furthermore, even if data is available, the scope of data collection may be limited, possibly covering only specific regions, specific types of merchants, or information within a specific time period, thus limiting the model's generalization ability and resulting in poor regulatory effectiveness in practical applications. Summary of the Invention

[0005] To address the technical problems of current intelligent supervision methods based on BP neural networks, such as difficulty in acquiring valuable data, limited data collection scope, restricted model generalization ability, and poor supervision effect in practical applications, this invention provides an intelligent supervision method and system.

[0006] First aspect

[0007] S101: Push regulatory inspection tasks;

[0008] S102: Determine the inspection route based on the aforementioned regulatory inspection task;

[0009] S103: Inspect each merchant sequentially according to the inspection route;

[0010] S104: Obtain the storefront image of the merchant, extract the storefront information from the storefront image using image recognition technology, and compare it with the historical storefront data in the background to confirm whether the merchant to be inspected has been reached.

[0011] S105: Obtain the image of the merchant's business license, extract the business license information from the image using image recognition technology, and compare it with the basic merchant data in the background to confirm whether the merchant's documents are compliant.

[0012] Second aspect

[0013] This invention provides an intelligent monitoring system for executing the intelligent monitoring method described in the first aspect.

[0014] Compared with the prior art, the present invention has at least the following beneficial technical effects:

[0015] (1) In this invention, intelligent supervision is based on image recognition technology. The business license image of the merchant is obtained, and the business license information is extracted from the image using image recognition technology. This information is then compared with the basic merchant data in the background to confirm whether the merchant's documents are compliant. The required training data is only the business license image, making it easy to obtain valuable data. The data collection scope is wide, which can improve the generalization ability of the model and enhance the supervision effect in practical applications.

[0016] (2) In this invention, the inspection route can be automatically determined according to the regulatory inspection task, and merchants with violation risks can be inspected first. The time and resources of regulatory personnel can be reasonably allocated, avoiding blind inspections and waste of resources. Limited resources can be concentrated on merchants with violation risks, and regulatory tasks can be implemented more efficiently. Attached Figure Description

[0017] The preferred embodiments will now be described in a clear and easy-to-understand manner, in conjunction with the accompanying drawings, to further explain the above-mentioned characteristics, technical features, advantages, and implementation methods of the present invention.

[0018] Figure 1 This is a flowchart illustrating a smart supervision method provided by the present invention. Detailed Implementation

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without any creative effort.

[0020] To keep the drawings concise, each figure only schematically shows the parts relevant to the invention, and these do not represent the actual structure of the product. Furthermore, to facilitate understanding, in some figures, only one of components with the same structure or function is schematically depicted, or only one is labeled. In this document, "one" not only means "only one," but can also mean "more than one."

[0021] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0022] In this document, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections. They can refer to mechanical connections or electrical connections. They can refer to direct connections or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0023] Furthermore, in the description of this invention, the terms "first," "second," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0024] Example 1

[0025] In one embodiment, refer to the appendix to the specification. Figure 1 The diagram shows a flowchart of a smart supervision method provided by the present invention.

[0026] The present invention provides a smart monitoring method, comprising:

[0027] S101: Push regulatory inspection tasks;

[0028] In one possible implementation, S101 specifically includes sub-steps S1011 and S1012:

[0029] S1011: Obtain market supervision data and management regulation data;

[0030] Market supervision data includes information about merchants and market operations, used to monitor and assess market compliance and order. Market supervision data includes, but is not limited to: merchant information, license status, violation records, business practices, and market data.

[0031] Among them, regulatory compliance data refers to the various rules, regulations, policies, and standards that merchants must follow in their market operations. These regulations typically originate from industry associations, government departments, and market regulatory agencies, aiming to ensure market order, fair competition, and consumer rights. Regulatory compliance data includes, but is not limited to: laws and regulations, industry standards, licenses and certificates, codes of conduct for business operations, and consumer rights protection.

[0032] S1012: When there are expired certificates, illegal reports, or violations in the market supervision data and management standard data, a regulatory inspection task is generated and pushed.

[0033] This invention can automatically detect anomalies in market supervision data and management regulations, such as expired licenses, illegal reports, or violations. Once these issues are detected, the system automatically generates regulatory inspection tasks and pushes them to regulatory personnel, enabling timely responses to potential problems. Simultaneously, automated detection and push can quickly identify merchants potentially at risk of violations, reducing the time cost of manual screening and judgment. This helps improve regulatory efficiency and ensures that resources are concentrated on inspecting high-risk merchants.

[0034] S102: Determine the inspection route based on the aforementioned regulatory inspection task;

[0035] In one possible implementation, S102 specifically includes sub-steps S1021 to S1025:

[0036] S1021: Obtain historical case data, time-series business indicator data, and merchant behavior data related to the regulatory inspection task, and build a data lake;

[0037] A data lake is a method of storing data in a system or repository in its natural format, which helps to configure data in various schemas and structural forms, typically object blocks or files. The main idea of ​​a data lake is to provide a unified storage for all data within an enterprise, transforming raw data (an exact copy of the data from the source system) into target data for various tasks such as reporting, visualization, analysis, and machine learning.

[0038] Time-series business metrics data refers to data that records relevant indicator values ​​of a merchant's or enterprise's business activities over a period of time. These metrics are typically organized and recorded according to time series to reflect the merchant's or enterprise's operating status, performance, and behavior. These metrics can include sales revenue, inventory levels, customer traffic, order volume, etc., recorded in a time series format.

[0039] S1022: A dynamic convolutional kernel model based on time-series multi-business indicators extracts the local intrinsic correlations between time-series business indicators;

[0040] Among them, the dynamic convolutional kernel model based on time-series multi-business indicators is a complex neural network model used to process time-series business indicator data of merchants or enterprises. This model is designed to dynamically generate adaptive convolutional kernel weights based on the correlation characteristics of business indicators in different time periods, thereby more accurately capturing the inherent relationships between time-series data and providing more valuable information for regulatory decisions.

[0041] Local intrinsic correlation refers to the correlation between parts of data within a certain range or local area. In time series data analysis, this means that within a specific time period or moment, some indicators or variables may increase or decrease simultaneously. This correlation may be more obvious within a local area, but may not necessarily exist in the entire dataset. Local intrinsic correlation emphasizes the inherent connection of data within a short period of time, which may arise due to certain external factors or business patterns.

[0042] S1023: A spatiotemporal cross-attention model based on a query sparse evaluation mechanism, which extracts the global intrinsic correlation between time-series business indicators;

[0043] Query Sparsity Measurement (QSM) is a technique used in machine learning and data processing to improve computational efficiency while maintaining model accuracy. QSM aims to evaluate a problem or query to determine the sparsity of the data that needs to be considered when solving it. By selectively processing data, QSM can reduce computation without sacrificing model quality, thereby accelerating model training and inference.

[0044] Among them, the spatiotemporal cross-attention model based on query sparsity evaluation mechanism is a complex neural network model for processing time series data. Its design aims to capture the correlation between different time points, while improving computational efficiency through query sparsity evaluation mechanism.

[0045] Global intrinsic correlation refers to the relationships that exist between different parts of the entire dataset or time series. These relationships may manifest over a longer time span, possibly due to trends, periodic changes, or other factors over a longer period. Global intrinsic correlation considers the overall trends and evolutionary patterns of the data, rather than just its performance within a specific local time period.

[0046] S1024: Sort the merchants according to the local intrinsic association and the global intrinsic association, so that merchants with violation risks are ranked at the top of the sorting results;

[0047] Specifically, the joint intrinsic correlation coefficient can be calculated based on the local intrinsic correlation and the global intrinsic correlation:

[0048] c = μ·c1 + (1-μ)·c2

[0049] Where c represents the joint intrinsic correlation coefficient, c1 represents the local intrinsic correlation, μ represents the weight of the local intrinsic correlation, c2 represents the global intrinsic correlation, and 1-μ represents the weight of the global intrinsic correlation.

[0050] Those skilled in the art can set the weight μ of the local intrinsic correlation according to the actual situation, and the present invention does not limit it.

[0051] Then, the merchants are sorted in order of increasing internal correlation coefficient.

[0052] It's important to note that local intrinsic correlations can help identify unusual trends in merchants within a specific timeframe. If a merchant's business metrics show significant inconsistencies with those of other merchants, it may indicate potential violations. Global intrinsic correlations, on the other hand, can help uncover cross-time-series relationships between merchants. If there are unusual correlations between a merchant's business metrics and those of other merchants, it may suggest potential collusion or violation.

[0053] S1025: Generate an inspection route based on the sorting results of each merchant, the geographical relationships of each merchant, and the configuration of market supervision personnel.

[0054] Specifically, the optimal inspection route can be determined using greedy algorithms, genetic algorithms, simulated annealing algorithms, etc.

[0055] In this invention, inspection routes can be automatically determined based on regulatory inspection tasks, prioritizing inspections of merchants with potential violations. This allows for the rational allocation of time and resources for regulatory personnel, avoiding blind inspections and resource waste. Limited resources can be concentrated on merchants with potential violations, enabling more efficient implementation of regulatory tasks.

[0056] S103: Inspect each merchant sequentially according to the inspection route;

[0057] S104: Obtain the storefront image of the merchant, extract the storefront information from the storefront image using image recognition technology, and compare it with the historical storefront data in the background to confirm whether the merchant to be inspected has been reached.

[0058] In this invention, image recognition technology allows the system to automatically extract storefront information, such as merchant names and trademarks, from storefront images. Comparing the extracted information with historical storefront data automatically verifies the merchant's identity, eliminating the tedious manual verification process. Furthermore, comparing storefront images with historical data allows for more accurate recording of merchants' historical information and changes. This helps establish merchant profiles and a historical database, providing a more accurate reference for future supervision.

[0059] S105: Obtain the image of the merchant's business license, extract the business license information from the image using image recognition technology, and compare it with the basic merchant data in the background to confirm whether the merchant's documents are compliant.

[0060] In this invention, intelligent supervision is achieved based on image recognition technology. The system acquires images of merchants' business licenses, extracts business license information from these images using image recognition, and compares this information with basic merchant data in the background to confirm the compliance of the merchant's documents. The required training data consists solely of business license images, making it easy to obtain valuable data. The wide data collection scope enhances the model's generalization ability and improves the supervisory effectiveness in practical applications.

[0061] In one possible implementation, after S105, the smart supervision method further includes:

[0062] S106: Obtain an image of the merchant's cigarette shelf, extract the cigarette category information from the image of the cigarette shelf using image recognition technology, and compare it with the order delivery data in the background to confirm whether the merchant is selling counterfeit or smuggled cigarettes.

[0063] In this invention, image recognition technology is used to automatically identify tobacco products on cigarette shelves. By comparing the identified product category information with order delivery data, it is possible to automatically detect whether there are unauthorized brands or illegal tobacco products, thereby helping regulatory authorities to promptly detect the sale of counterfeit cigarettes. Illegal cigarette sales refer to merchants selling tobacco products without authorization from legal channels. By comparing with order delivery data, it is possible to accurately identify whether merchants are selling illegal cigarettes, thus enabling targeted supervision and prevention of illegal cigarette sales.

[0064] In one possible implementation, S106 specifically includes sub-steps S1061 to S1065:

[0065] S1061: Extract the image semantic features of the cigarette shelf image through data-enhanced contrastive learning technology and cross-attention mechanism technology;

[0066] Data augmentation contrastive learning is a technique used to train models, particularly suitable for few-shot learning and small datasets. It generates augmented data samples by applying different transformations to the same sample, and then uses these augmented data samples for model training.

[0067] Cross-attention is a technique used to establish connections between different data sources. In this invention, cross-attention is used to extract semantic features from images of cigarette shelves, possibly to better capture the location, association, and features of different products on the shelf.

[0068] In this invention, by using data-enhanced contrastive learning and cross-attention mechanisms, the model can better understand the semantic features in images of cigarette shelves and accurately extract information on the types of cigarettes sold, which helps in subsequent analysis and judgment.

[0069] S1062: Input the image semantic features of the cigarette shelf image into the image recognition model;

[0070] S1063: Construct the loss function of the image recognition model based on the intersection-union ratio, generalized intersection-union ratio, and aspect ratio difference between the predicted bounding box and the ground truth bounding box;

[0071] In one possible implementation, the loss function of the image recognition model can be expressed as:

[0072] L=λ1IoU+λ2gIoU+λ3AR

[0073] Where L represents the loss function, IoU represents the cross-union ratio between the predicted box and the ground truth box, λ1 represents the weight of the cross-union ratio, gIoU represents the generalized cross-union ratio between the predicted box and the ground truth box, λ2 represents the weight of the generalized cross-union ratio, AR represents the aspect ratio difference between the predicted box and the ground truth box, and λ3 represents the weight of the aspect ratio difference.

[0074] Those skilled in the art can set the weights λ1 for the crossover-union ratio, λ2 for the generalized crossover-union ratio, and λ3 for the aspect ratio difference according to the actual situation; this invention does not impose any limitations.

[0075] In one possible implementation, the generalized intersection-union ratio (gIoU) is calculated as follows:

[0076]

[0077] d 2 =ρ 2 (b,b gt )

[0078]

[0079]

[0080] Where gIoU represents the generalized intersection-union ratio (IoU) between the predicted and ground truth boxes, IoU represents the IoU ratio between the predicted and ground truth boxes, d represents the Euclidean distance between the center points of the predicted and ground truth boxes, c represents the diagonal length of the minimum closure between the ground truth and predicted boxes, α represents the tradeoff parameter, v represents the consistency parameter, ρ() represents the Euclidean distance calculation function, and b represents the predicted box. gt Represents the true bounding box, w gt h represents the width of the actual bounding box. gt represents the height of the ground truth bounding box, w represents the width of the predicted bounding box, and h represents the height of the predicted bounding box.

[0081] S1064: Train the image recognition model with the goal of minimizing the loss function;

[0082] S1065: Extract the type of cigarettes sold from the image of the cigarette shelf using the trained image recognition model.

[0083] In practical applications, similar methods can be used to train the storefront image recognition model and the business license image recognition model. To avoid repetition, this invention will not elaborate further.

[0084] In this invention, information on the types of cigarettes sold can be automatically extracted from images of cigarette shelves and compared with order data, providing an automated and efficient method for market supervision, improving the accuracy and efficiency of supervision, and protecting the rights and interests of legitimate brands and consumers.

[0085] In one possible implementation, after S106, the smart supervision method further includes:

[0086] S107: Back up and save the storefront image, the business license image, and the cigarette shelf image to the background.

[0087] In this invention, these images are backed up and stored in the background, ensuring data security and reliability. In the event of accidental data loss, corruption, or system failure, data can be restored using the backup data, preventing the loss of important information. Simultaneously, the backed-up image data can serve as evidence in necessary legal proceedings. In the event of violations or disputes, the backed-up images can support the actions of regulatory authorities. Furthermore, the backed-up image data can serve as a reference for ongoing monitoring. Regulatory authorities can track and compare historical data to ensure that merchants' behavior remains consistent across different time periods.

[0088] In one possible implementation, S107 specifically includes sub-steps S1071 and S1072:

[0089] S1071: Perform daily incremental backups of data that has been added or changed since the last full backup.

[0090] S1072: Back up all data weekly using a full backup method.

[0091] In this invention, the backup process is divided into daily incremental backups and weekly full backups, which can improve backup efficiency, save resources, shorten data recovery time, and increase data reliability and availability while ensuring data security.

[0092] In one possible implementation, after S107, the smart supervision method further includes:

[0093] S108: Desensitize sensitive information in the storefront image, the business license image, and the cigarette shelf image.

[0094] Specifically, desensitization can involve deleting or replacing symbols with asterisks (*).

[0095] In this invention, sensitive information in the storefront image, business license image, and cigarette shelf image is anonymized, which protects privacy and reduces risks, helps improve compliance and data security, and creates more possibilities for data sharing and use.

[0096] Compared with the prior art, the present invention has at least the following beneficial technical effects:

[0097] (1) In this invention, intelligent supervision is based on image recognition technology. The business license image of the merchant is obtained, and the business license information is extracted from the image using image recognition technology. This information is then compared with the basic merchant data in the background to confirm whether the merchant's documents are compliant. The required training data is only the business license image, making it easy to obtain valuable data. The data collection scope is wide, which can improve the generalization ability of the model and enhance the supervision effect in practical applications.

[0098] (2) In this invention, the inspection route can be automatically determined according to the regulatory inspection task, and merchants with violation risks can be inspected first. The time and resources of regulatory personnel can be reasonably allocated, avoiding blind inspections and waste of resources. Limited resources can be concentrated on merchants with violation risks, and regulatory tasks can be implemented more efficiently.

[0099] Example 2

[0100] In one embodiment, the present invention provides an intelligent monitoring system for executing the intelligent monitoring method in Embodiment 1.

[0101] The intelligent supervision system provided by this invention can realize the steps and effects of the intelligent supervision method in Embodiment 1 above. To avoid repetition, this invention will not repeat them.

[0102] Compared with the prior art, the present invention has at least the following beneficial technical effects:

[0103] (1) In this invention, intelligent supervision is based on image recognition technology. The business license image of the merchant is obtained, and the business license information is extracted from the image using image recognition technology. This information is then compared with the basic merchant data in the background to confirm whether the merchant's documents are compliant. The required training data is only the business license image, making it easy to obtain valuable data. The data collection scope is wide, which can improve the generalization ability of the model and enhance the supervision effect in practical applications.

[0104] (2) In this invention, the inspection route can be automatically determined according to the regulatory inspection task, and merchants with violation risks can be inspected first. The time and resources of regulatory personnel can be reasonably allocated, avoiding blind inspections and waste of resources. Limited resources can be concentrated on merchants with violation risks, and regulatory tasks can be implemented more efficiently.

[0105] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0106] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A smart supervision method, characterized in that, Applications in cloud data management systems include: S101: Push regulatory inspection tasks; S102: Determine the inspection route based on the aforementioned regulatory inspection task; S103: Inspect each merchant sequentially according to the inspection route; S104: Obtain the storefront image of the merchant, extract the storefront information from the storefront image using image recognition technology, and compare it with the historical storefront data in the background to confirm whether the merchant to be inspected has been reached. S105: Obtain the image of the merchant's business license, extract the business license information from the image using image recognition technology, and compare it with the basic merchant data in the background to confirm whether the merchant's documents are compliant. Specifically, S102 includes: S1021: Obtain historical case data, time-series business indicator data, and merchant behavior data related to the regulatory inspection task, and build a data lake; S1022: A dynamic convolutional kernel model based on time-series multi-business indicators extracts the local intrinsic correlations between time-series business indicators; S1023: A spatiotemporal cross-attention model based on a query sparse evaluation mechanism, which extracts the global intrinsic correlation between time-series business indicators; S1024: Sort the merchants according to the local intrinsic association and the global intrinsic association, so that merchants with violation risks are ranked at the top of the sorting results; S1025: Generate an inspection route based on the sorting results of each merchant, the geographical relationships of each merchant, and the configuration of market supervision personnel; The intelligent supervision method further includes, following S105: S106: Obtain an image of the merchant's cigarette shelf, extract the cigarette category information from the image of the cigarette shelf using image recognition technology, and compare it with the order delivery data in the background to confirm whether the merchant is selling counterfeit or smuggled cigarettes; Specifically, S106 includes: S1061: Extract the image semantic features of the cigarette shelf image through data-enhanced contrastive learning technology and cross-attention mechanism technology; S1062: Input the image semantic features of the cigarette shelf image into the image recognition model; S1063: Construct the loss function of the image recognition model based on the intersection-union ratio, generalized intersection-union ratio, and aspect ratio difference between the predicted bounding box and the ground truth bounding box; S1064: Train the image recognition model with the goal of minimizing the loss function; S1065: Extract the type of cigarettes sold from the image of the cigarette shelf using the trained image recognition model.

2. The intelligent supervision method according to claim 1, characterized in that, S101 specifically includes: S1011: Obtain market supervision data and management regulation data; S1012: When there are reports of illegal activities or violations in the market supervision data and the management standard data, a regulatory inspection task is generated and pushed out.

3. The intelligent supervision method according to claim 1, characterized in that, The loss function of the image recognition model is expressed as: in, L Represents the loss function. IoU This represents the intersection-over-union ratio (IoU) between the predicted bounding box and the ground truth bounding box. λ 1 represents the weight of the intersection-union ratio. gIoU This represents the generalized intersection-union ratio (GUC) between the predicted bounding box and the ground truth bounding box. λ 2 represents the weight of the generalized intersection-union ratio. AR This represents the aspect ratio difference between the predicted bounding box and the ground truth bounding box. λ 3 indicates the weight of the aspect ratio difference.

4. The intelligent supervision method according to claim 1, characterized in that, Following S106, the following is also included: S107: Back up and save the storefront image, the business license image, and the cigarette shelf image to the background.

5. The intelligent supervision method according to claim 4, characterized in that, S107 specifically includes: S1071: Perform daily incremental backups of data that has been added or changed since the last full backup. S1072: Back up all data weekly using a full backup method.

6. The intelligent supervision method according to claim 4, characterized in that, Following S107, the following is also included: S108: Desensitize sensitive information in the storefront image, the business license image, and the cigarette shelf image.

7. A smart monitoring system, characterized in that, Used to perform the smart supervision method according to any one of claims 1 to 6.

Images