Image-based risk early warning method and device, equipment and storage medium

Abstract

The invention discloses an image-based risk early warning method and device, equipment and a storage medium, and the method comprises the steps: inputting an image related to a case into an image classification model, and obtaining an image type corresponding to the image; based on the image category, extracting structured key information from the image; based on a preset verification condition, the structured key information is verified, risk feature information is generated, and the risk feature information is used for representing an abnormal relation between the structured key information and the verification condition; and outputting a corresponding risk early warning result based on the risk feature information. In this way, the risk early warning result is output based on the risk feature information, dependence on manual auditing can be reduced, and the efficiency and timeliness of risk identification in the vehicle insurance claim settlement process are improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an image-based risk warning method, apparatus, device, and storage medium. Background Technology

[0002] With the continuous expansion of the insurance business, the number of insurance claims has continued to grow. Claims efficiency and risk control capabilities have become key factors in the operation and management of the insurance industry. In particular, the auto insurance claims process usually involves a large amount of image data related to the case. Effective analysis of image data and timely detection of potential risks are of great significance for preventing insurance fraud and reducing claims losses.

[0003] In the current auto insurance claims process, the identification and early warning of case risks mainly rely on manual review. Claims personnel need to manually review and judge the uploaded image data to determine whether the images meet the case requirements and whether there are any anomalies. However, this method heavily relies on human experience, resulting in a large workload. Moreover, when there are a large number of claims, it is prone to low review efficiency and difficulty in timely detection of potential risks. Summary of the Invention

[0004] This application provides an image-based risk warning method, apparatus, device, and storage medium, which can reduce reliance on manual review and improve the efficiency and timeliness of risk identification in the auto insurance claims process.

[0005] In a first aspect, embodiments of this application provide an image-based risk warning method, the method comprising:

[0006] Images related to the case are input into an image classification model to obtain the image category corresponding to the image;

[0007] Based on the image category, structured key information is extracted from the image;

[0008] Based on preset verification conditions, the structured key information is verified to generate risk feature information, which is used to characterize the abnormal relationship between the structured key information and the verification conditions.

[0009] Based on the risk characteristic information, the corresponding risk warning result is output.

[0010] One feasible implementation, wherein extracting structured key information from the image based on the image category, includes:

[0011] Based on the image category, determine the information extraction method corresponding to the image category;

[0012] Based on the information extraction method, information is extracted from the predefined structured fields in the image to obtain the structured key information.

[0013] One feasible implementation method, wherein the structured key information is verified based on preset verification conditions to generate risk feature information, includes:

[0014] Based on the preset verification conditions, the structured key information is subjected to consistency verification, integrity verification, or frequency verification to generate risk feature information.

[0015] One feasible implementation includes determining the information extraction method corresponding to the image category based on the image category, comprising:

[0016] The image categories are queried based on a pre-established extraction method table to obtain the query results;

[0017] Based on the query results, the information extraction method corresponding to the image category is determined.

[0018] One feasible implementation is that the consistency check is used to determine whether different structured key information conforms to a preset consistency relationship;

[0019] The integrity check is used to determine whether the structured key information meets the preset information integrity requirements;

[0020] The frequency verification is based on a historical case database to determine whether the frequency of the structured key information appearing within a preset time range exceeds a preset threshold.

[0021] One feasible implementation method, wherein the step of outputting a corresponding risk warning result based on the risk characteristic information, includes:

[0022] Based on the severity of the risk characteristic information, the risk warning results are output in a graded manner, and the risk warning results include at least a level one risk warning and a level two risk warning.

[0023] One possible implementation is that the image category includes a first image category and a second image category, wherein the second image category is a subcategory of the first image category.

[0024] Secondly, embodiments of this application provide an image-based risk warning device, including:

[0025] Thirdly, embodiments of this application provide an electronic device, the device including: a processor, a memory, and a system bus;

[0026] The processor and the memory are connected via the system bus;

[0027] The memory is used to store a program, the program including instructions that, when executed by the processor, cause the processor to perform any of the implementation steps of the image-based risk warning method described above.

[0028] Fourthly, embodiments of this application provide a computer-readable storage medium for storing a computer program, which, when executed by a terminal device, implements any of the implementation steps of the image-based risk warning method described above.

[0029] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0030] In this embodiment, firstly, images related to the case are used as input to an image classification model to obtain image categories corresponding to the images. Then, based on the image categories, structured key information is extracted from the images. Next, the structured key information is validated based on preset validation conditions to generate risk feature information. This risk feature information characterizes the abnormal relationship between the structured key information and the validation conditions. Finally, based on the risk feature information, the corresponding risk warning result is output.

[0031] As can be seen, this solution classifies images related to the case using an image classification model and obtains structured key information from the images based on the obtained image categories, thus providing an effective data foundation for subsequent risk analysis. Furthermore, the structured key information is validated based on preset verification conditions, and risk feature information is generated to characterize abnormal relationships, achieving automated identification of case risks. Risk warning results are output based on this risk feature information, thereby reducing reliance on manual review and improving the efficiency and timeliness of risk identification in the auto insurance claims process. Attached Figure Description

[0032] Figure 1 A flowchart illustrating an image-based risk warning method provided in this application embodiment;

[0033] Figure 2 A flowchart illustrating a risk warning provided in an embodiment of this application;

[0034] Figure 3 This is a schematic diagram of the structure of an image-based risk warning device provided in an embodiment of this application. Detailed Implementation

[0035] As mentioned earlier, with the continuous expansion of the insurance business, the number of insurance claims has continued to grow. Claims efficiency and risk control capabilities have become key factors in the operation and management of the insurance industry. In particular, the auto insurance claims process usually involves a large amount of image data related to the case. Effective analysis of image data and timely detection of potential risks are of great significance for preventing insurance fraud and reducing claims losses.

[0036] In the existing auto insurance claims process, the identification and early warning of case risks mainly rely on manual review. Claims personnel need to manually review and judge the image data uploaded by the surveyor to determine whether the image meets the case requirements and whether there are any abnormalities.

[0037] However, in practice, due to some surveyors' non-standard operating procedures or lack of business experience, there have been instances where image data has been uploaded to the wrong directories. For example, images of ID cards have been uploaded to the vehicle information directory, or partial photos of damaged parts of the vehicle have been uploaded to the directory for overall accident scene photos. These erroneous uploads lead to chaotic classification of image data, increasing the difficulty for subsequent claims personnel to retrieve and verify the image data, thereby affecting the efficiency of claims processing and even causing claims disputes.

[0038] It is evident that the risk identification methods in the current auto insurance claims process rely heavily on human experience, resulting in a large overall workload. When there are a large number of claims, problems such as low review efficiency and delayed risk identification can easily occur, making it difficult to discover potential risks in a timely and effective manner and failing to meet the current requirements of auto insurance claims business for efficiency and risk control capabilities.

[0039] To address the aforementioned problems, this application provides an image-based risk warning method, apparatus, device, and storage medium. First, images related to the case are used as input to an image classification model to obtain image categories. Then, based on the image categories, structured key information is extracted from the images. Next, the structured key information is verified based on preset verification conditions to generate risk feature information. This risk feature information characterizes the abnormal relationship between the structured key information and the verification conditions. Finally, based on the risk feature information, the corresponding risk warning result is output.

[0040] As can be seen, this solution classifies images related to the case using an image classification model and obtains structured key information from the images based on the obtained image categories, thus providing an effective data foundation for subsequent risk analysis. Furthermore, the structured key information is validated based on preset verification conditions, and risk feature information is generated to characterize abnormal relationships, achieving automated identification of case risks. Risk warning results are output based on this risk feature information, thereby reducing reliance on manual review and improving the efficiency and timeliness of risk identification in the auto insurance claims process.

[0041] It should be noted that the implementation subject of the image-based risk warning method is not limited in the embodiments of this application. For example, the image-based risk warning method of this application embodiment can be applied to information processing devices such as servers or terminal devices. The server can be a standalone server, a cluster server, or a cloud server. The terminal device can be an electronic device such as a smartphone, computer, personal digital assistant (PDA), or tablet computer.

[0042] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0043] Figure 1 A flowchart illustrating an image-based risk warning method provided in an embodiment of this application. (Combined with...) Figure 1 As shown, it may include steps S101-S104.

[0044] S101: Input the images related to the case into the image classification model to obtain the image category corresponding to the image.

[0045] In this embodiment, images related to the case are input into a preset image classification model to obtain the image category corresponding to the image. It should be noted that this image classification model is trained based on image samples from a large number of auto insurance claims cases. These image samples are labeled according to pre-defined image categories so that the image classification model can output the corresponding image category.

[0046] The image categories in this application embodiment include a first image category and a second image category, with the second image category being a subcategory of the first image category. Specifically, the first image category includes categories such as personnel, identities, vehicles, accident scenes, materials, and others.

[0047] It should be noted that the second image category for personnel can include 10 categories, such as surveyors, insured persons, and third-party personnel; the second image category for identity can include 16 categories, such as driver's licenses, vehicle registration certificates, and ID cards; the second image category for vehicles includes 4 categories, such as license plate numbers, whole vehicle photos, and photos of damaged parts of the vehicle; the second image category for accident scenes includes 15 categories, such as overall photos of the accident scene, partial photos of the accident scene, and photos of the road environment; the second image category for materials includes 160 categories, such as repair lists, traffic accident liability certificates, and invoices; other categories are used to characterize image categories that do not belong to the above-mentioned personnel, identity, vehicle, accident scene, or material categories, such as auxiliary images that are less relevant to the case or are not currently involved in the key information extraction process.

[0048] Furthermore, the image classification model trained in this embodiment possesses lightweight deployment characteristics. It employs the MobileViT model (a lightweight deep learning model for image classification tasks) as its basic algorithm framework, integrating the advantages of convolutional neural networks and visual Transformer models to construct a lightweight image classification model suitable for auto insurance claims scenarios. Compared to existing lightweight convolutional neural networks, this image classification model maintains lower computational complexity while more fully representing the global semantic information of images, thereby improving the classification performance for complex claims scenarios. Testing has verified that the image classification model constructed in this embodiment achieves a high classification accuracy in image classification tasks, reaching a maximum of 98.75%.

[0049] S102: Extract structured key information from images based on image categories.

[0050] In this embodiment, based on the obtained image category, an information extraction method corresponding to that image category is determined. Based on this extraction method, information is extracted from predefined structured fields in the image to obtain structured key information. The predefined structured fields are a set of data fields pre-defined for different image categories to represent key information in the image. Structured fields are represented by field names and values, and the content of the structured fields differs for different image categories. For example, for identity images, predefined structured fields may include name, ID number, date of birth, and ID validity period; for vehicle images, the predefined structured fields may include license plate number, vehicle model, and vehicle color.

[0051] In this embodiment, an image category is queried based on a pre-established extraction method table to obtain query results, and the information extraction method corresponding to the image category is determined based on the query results. The extraction method table establishes a one-to-one correspondence between image categories and information extraction methods, indicating the information extraction method to be used for different image categories. The information extraction method may include at least one of text recognition, object detection, and key information extraction. The specific information extraction method used can be selected or combined according to the image category.

[0052] S103: Based on preset verification conditions, the structured key information is verified to generate risk feature information. The risk feature information is used to characterize the abnormal relationship between the structured key information and the verification conditions.

[0053] In this embodiment, based on preset verification conditions, structured key information is subjected to consistency verification, integrity verification, or frequency verification to generate risk characteristic information. Specifically, consistency verification determines whether different pieces of structured key information conform to a preset consistency relationship; integrity verification determines whether the structured key information meets preset information integrity requirements; and frequency verification, based on a historical case database, determines whether the frequency of occurrence of structured key information within a preset time range exceeds a preset threshold.

[0054] It should be noted that risk characteristic information is used to characterize the abnormal relationship between structured key information and verification conditions, and the verification conditions can be pre-set according to insurance claims business rules. Compared with the existing risk control methods in the auto insurance claims process, which are mostly concentrated in the post-event review stage and lack a mechanism for pre-identification and early warning of risks, the embodiments of this application, by verifying the structured key information extracted from case-related images, can identify potential abnormal situations in the early stage of the claims process and generate corresponding risk characteristic information, thereby providing a basis for subsequent risk warning and realizing pre-identification and proactive prevention of risks.

[0055] Meanwhile, in actual claims processing, claims personnel often can only review individual cases independently, making it difficult to promptly detect abnormal behaviors across cases or accumulated over a long period. For example, when some repair shops use old parts to impersonate new parts for claims, this application embodiment uses a frequency verification method. Based on a historical case database, it statistically analyzes the frequency of occurrence of relevant structured key information within a preset time range. When the frequency exceeds a preset threshold, corresponding risk characteristic information is generated, thereby achieving early identification and risk warning of such abnormal claims behavior.

[0056] S104: Based on risk characteristic information, output the corresponding risk warning results.

[0057] In this embodiment of the application, the risk warning results are output in a graded manner based on the severity of the risk characteristic information, and the risk warning results include at least a first-level risk warning and a second-level risk warning.

[0058] Taking risk warnings for vehicle headlight tags as an example, during the claims process, the relevant images are first input into an image classification model for classification. When the image classification model determines that the image belongs to the headlight tag category within the materials category, the corresponding headlight tag detection and recognition model is invoked to further process the image based on this category. Specifically, the headlight tag detection and recognition model extracts key information such as the serial number and production time from the headlight tag image, thereby obtaining the corresponding structured key information.

[0059] After obtaining the structured key information of the headlight tag, frequency verification is performed on the structured key information based on the historical case database. When the same or highly similar headlight tag serial numbers are detected in different claims cases within a preset time range, high-risk feature information is generated; or when the frequency of headlight tags associated with the same repair entity exceeds a preset threshold, medium-risk feature information is generated.

[0060] Furthermore, based on the severity of the risk characteristic information, corresponding risk warning results are output. Specifically, for high-risk characteristic information, a Level 1 risk warning is output to prompt claims personnel to pay close attention to and thoroughly investigate the corresponding cases; for medium-risk characteristic information, a Level 2 risk warning is output to alert claims personnel to potential abnormal claims behavior and to selectively conduct further investigations.

[0061] Through the above methods, the embodiments of this application can achieve automatic identification and graded early warning of vehicle parts claims risks, effectively making up for the shortcomings of the existing claims process that relies solely on manual review of individual cases and is difficult to detect abnormal behavior across cases in a timely manner.

[0062] It should be noted that the headlight label detection and recognition model used in this application embodiment is based on YOLOv8-OBB (a rotational target detection model that supports directed bounding boxes) to achieve rotational target detection. The training data for this headlight label detection and recognition model comes from a large number of headlight label images involved in real insurance claims cases, and the coordinates of the four vertices of the bounding rectangle of the headlight label and the image category are labeled. Furthermore, due to the large number of similar headlight labels in the training data, which could easily lead to false detections, an additional 107 headlight label images with negative samples were collected and combined with real labels to generate 2970 enhanced training data images. The headlight label detection and recognition model was validated on a dataset containing 38050 claim images. Only 29 of the claim images were headlight label images, and the model successfully detected 28 of them. The one image that was not detected was a severely blurred headlight label image. Its accuracy and recall both exceeded 99%, fully verifying the high-precision localization capability of the headlight label detection and recognition model.

[0063] In practical applications, the trained headlight label detection and recognition model can accurately locate headlight label images and perform standard rectangular perspective transformation cropping on the detected headlight labels, retaining only the headlight label area to quickly extract key information such as serial number and production time from the headlight label images. Testing showed that the average accuracy of this headlight label detection and recognition model reached 99.2% among the key information extracted from 385 headlight label images.

[0064] Based on the relevant content of steps S101-S104 above, in this embodiment, firstly, images related to the case are used as input to an image classification model to obtain image categories corresponding to the images. Then, based on the image categories, structured key information is extracted from the images. Next, based on preset verification conditions, the structured key information is verified to generate risk feature information. This risk feature information is used to characterize the abnormal relationship between the structured key information and the verification conditions. Finally, based on the risk feature information, the corresponding risk warning result is output. It can be seen that this solution classifies images related to the case using an image classification model and obtains structured key information from the images based on the obtained image categories, thus providing an effective data foundation for subsequent risk analysis. Furthermore, verifying the structured key information based on preset verification conditions and generating risk feature information to characterize abnormal relationships achieves automated identification of case risks. Outputting risk warning results based on this risk feature information reduces reliance on manual review and improves the efficiency and timeliness of risk identification in the auto insurance claims process.

[0065] Furthermore, embodiments of this application can construct a system integrating image classification and risk warning. In specific applications, Figure 2 This application provides a flowchart for a risk warning system. When collecting evidence for claims, the surveyor only needs to click on the corresponding image category to upload the image to the correct location, or directly use the photo function to capture the image and automatically save it to the corresponding category to achieve image upload. Next, the system automatically categorizes the images and determines if there are any inconsistencies in the uploaded image paths. If the system detects that the uploaded image category is inconsistent with the selected image category, it will promptly prompt the surveyor to verify and re-upload, thereby effectively improving the accuracy of uploaded claims images.

[0066] Next, after confirming that the image category and upload path are consistent, the system further calls the risk database (i.e., the historical case database in the above embodiment) under the corresponding category (i.e. the image category in the above embodiment) for comparison. Once any possible abnormal or high-risk behavior is found, the system will immediately issue a risk warning to the surveyor, guiding him to take timely verification or avoidance measures, thereby effectively preventing the spread of risk and ensuring the safety and efficiency of the claims process.

[0067] It should be noted that the image information (including but not limited to identity information, material information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) related to the case involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0068] Furthermore, Figure 3 This is a schematic diagram of the structure of an image-based risk warning device provided in an embodiment of this application. (Combined with...) Figure 3 As shown, the image-based risk warning device 300 provided in this application embodiment may include:

[0069] The image input module 301 is used to input images related to the case into the image classification model to obtain the image category corresponding to the image.

[0070] The key information acquisition module 302 is used to extract structured key information from the image based on the image category;

[0071] The risk feature generation module 303 is used to verify the structured key information based on preset verification conditions and generate risk feature information, which is used to characterize the abnormal relationship between the structured key information and the verification conditions.

[0072] The risk warning output module 304 is used to output the corresponding risk warning result based on the risk characteristic information.

[0073] Optionally, the key information acquisition module 302 may include:

[0074] The extraction method determination module is used to determine the information extraction method corresponding to the image category based on the image category;

[0075] The field extraction module is used to extract information from predefined structured fields in the image based on the information extraction method, so as to obtain the structured key information.

[0076] Optionally, the risk feature generation module 303 may include:

[0077] An anomaly verification module is used to perform consistency verification, integrity verification, or frequency verification on the structured key information based on the preset verification conditions, and generate risk feature information.

[0078] Optionally, the extraction method determination module is specifically used for:

[0079] The image categories are queried based on a pre-established extraction method table to obtain the query results;

[0080] Based on the query results, the information extraction method corresponding to the image category is determined.

[0081] Optionally, the anomaly verification module is specifically used for:

[0082] The consistency check is used to determine whether different structured key information conforms to a preset consistency relationship;

[0083] The integrity check is used to determine whether the structured key information meets the preset information integrity requirements;

[0084] The frequency verification is based on a historical case database to determine whether the frequency of the structured key information appearing within a preset time range exceeds a preset threshold.

[0085] Optionally, the risk warning output module 304 is specifically used for:

[0086] Based on the severity of the risk characteristic information, the risk warning results are output in a graded manner, and the risk warning results include at least a level one risk warning and a level two risk warning.

[0087] Optionally, the image category includes a first image category and a second image category, wherein the second image category is a subcategory of the first image category.

[0088] Furthermore, embodiments of this application also provide an electronic device, including: a processor, a memory, and a system bus;

[0089] The processor and the memory are connected via the system bus;

[0090] The memory is used to store one or more programs, the one or more programs including instructions that, when executed by the processor, cause the processor to perform any of the implementation steps of the image-based risk warning method described above.

[0091] Furthermore, embodiments of this application also provide a computer-readable storage medium for storing a computer program, which, when executed by a terminal device, implements any of the implementation steps of the image-based risk warning method described above.

[0092] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that all or part of the steps in the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application. It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on describing the differences from other embodiments. The same or similar parts between the various embodiments can be referred to mutually.

[0093] The system disclosed in the embodiments is described in a relatively simple manner because it corresponds to the method disclosed in the embodiments. For relevant details, please refer to the method section.

[0094] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0095] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An image-based risk warning method, characterized in that, The method includes: Images related to the case are input into an image classification model to obtain the image category corresponding to the image; Based on the image category, structured key information is extracted from the image; Based on preset verification conditions, the structured key information is verified to generate risk feature information, which is used to characterize the abnormal relationship between the structured key information and the verification conditions. Based on the risk characteristic information, the corresponding risk warning result is output.

2. The method according to claim 1, characterized in that, The extraction of structured key information from the image based on the image category includes: Based on the image category, determine the information extraction method corresponding to the image category; Based on the information extraction method, information is extracted from the predefined structured fields in the image to obtain the structured key information.

3. The method according to claim 1, characterized in that, The structured key information is verified based on preset verification conditions to generate risk feature information, including: Based on the preset verification conditions, the structured key information is subjected to consistency verification, integrity verification, or frequency verification to generate risk feature information.

4. The method according to claim 2, characterized in that, The step of determining the information extraction method corresponding to the image category based on the image category includes: The image categories are queried based on a pre-established extraction method table to obtain the query results; Based on the query results, the information extraction method corresponding to the image category is determined.

5. The method according to claim 3, characterized in that, The consistency check is used to determine whether different structured key information conforms to a preset consistency relationship; The integrity check is used to determine whether the structured key information meets the preset information integrity requirements; The frequency verification is based on a historical case database to determine whether the frequency of the structured key information appearing within a preset time range exceeds a preset threshold.

6. The method according to claim 1, characterized in that, The step of outputting a corresponding risk warning result based on the risk characteristic information includes: Based on the severity of the risk characteristic information, the risk warning results are output in a graded manner, and the risk warning results include at least a level one risk warning and a level two risk warning.

7. The method according to claim 1, characterized in that, The image categories include a first image category and a second image category, wherein the second image category is a subcategory of the first image category.

8. An image-based risk warning device, characterized in that, include: An image input module is used to input images related to the case into an image classification model to obtain the image category corresponding to the image. The key information acquisition module is used to extract structured key information from the image based on the image category; The risk feature generation module is used to verify the structured key information based on preset verification conditions and generate risk feature information, which is used to characterize the abnormal relationship between the structured key information and the verification conditions. The risk warning output module is used to output the corresponding risk warning result based on the risk characteristic information.

9. An electronic device, characterized in that, The device includes: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, the program including instructions that, when executed by the processor, cause the processor to perform the steps of the image-based risk warning method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program that, when executed by a terminal device, implements the steps of the image-based risk warning method according to any one of claims 1 to 7.

Images