Claim settlement fraud event identification method and related equipment

Abstract

The invention discloses a claim settlement fraud event identification method, and aims to solve the problem that in the prior art, as a claim settlement fraud event cannot be effectively identified, benefits of an insurance company are easily damaged. The method comprises the steps of obtaining an accident image of a target accident case; generating an accident description text of the target accident case according to the accident image; according to the accident description text and the report record of the target accident case, determining the semantic matching degree of the accident description text and the report record; and when the semantic matching degree is smaller than a preset matching degree threshold value, determining that the target accident case is a fraud event. The invention also discloses a claim settlement fraud event identification device, electronic equipment, a computer readable storage medium and a computer program product.

Description

Technical Field

[0001] The present application relates to the field of insurance business technology, and in particular to a method for identifying claims fraud incidents and related equipment. Background Art

[0002] In recent years, with the digitization of the accident reporting process, the efficiency of claim handling has been significantly improved. However, claims fraud has become increasingly frequent, posing a severe challenge to the insurance industry. This type of insurance fraud not only causes insurance companies to face huge economic losses, but also seriously disrupts the fair competition order of the market and damages the rights and interests of other legitimate customers.

[0003] Based on this, in practical applications, especially in specific scenarios involving accident scene images as a key source of evidence, how to effectively prevent and combat insurance fraud is a technical problem that urgently needs to be solved in this field. Summary of the invention

[0004] The embodiment of the present application provides a method for identifying claims fraud events, which is used to solve the problem in the prior art that claims fraud events cannot be effectively identified, thereby easily causing damage to the interests of insurance companies.

[0005] The embodiments of the present application also provide a claim fraud event identification device, an electronic device, a computer-readable storage medium, and a computer program product.

[0006] The present application embodiment adopts the following technical solutions: In a first aspect, the present application provides a method for identifying claims fraud events, comprising: Acquire accident images of target accident cases; Generate an accident description text of a target accident case according to the accident image; According to the accident description text and the report record of the target accident case, the semantic matching degree between the accident description text and the report record is determined; When the semantic matching degree is less than the preset matching degree threshold, the target accident case is determined to be a fraud incident.

[0007] Optionally, an accident description text of the target accident case is generated based on the accident image, including: Determine the vehicle subject, obstacle subject and background object of the target accident case based on the accident image; Generate an accident description text of the target accident case based on the vehicle body, obstacle body and background objects.

[0008] Optionally, the vehicle subject, obstacle subject, and background object of the target accident case are determined based on the accident image, including: Perform image segmentation processing on the accident image to obtain a sub-image of the image elements contained in the accident image; Determine the label and location information of the sub-graph; Based on the labels and location information, the vehicle subject, obstacle subject and background objects of the target accident case are determined.

[0009] Optionally, an accident description text of the target accident case is generated based on the vehicle subject, obstacle subject and background objects, including: Determine explicit information and implicit information of the target accident case based on the vehicle body and the obstacle body; where explicit information includes distance information between the vehicle body and the obstacle body; implicit information includes speed, collision angle and inertia of the vehicle body, as well as the movement trajectory and state change of the vehicle body before and after the accident; Introduce pseudo-entities for target accident cases based on vehicle subjects, obstacle subjects, and background objects; Determine the accident cause of the target accident case based on the pseudo-entity and the explicit information and implicit information of the target accident case; According to the accident causes, background objects, explicit information and implicit information, the accident description text of the target accident case is generated through the text generation model.

[0010] Optionally, obtain accident images of the target accident case, including: Determine the accident location of each accident case to be identified; Clustering the accident cases to be identified according to the accident locations to obtain multiple case groups of the accident cases to be identified; The case grouping in which the number of accident cases to be identified is greater than a preset number threshold among the multiple case groups is used as a target case grouping, and the accident cases to be identified in the target case grouping are determined as target accident cases; Obtain accident images of the target accident case.

[0011] Optionally, according to the accident description text and the report record of the target accident case, the semantic matching degree between the accident description text and the report record is determined, including: Determine the semantic overlap and keyword hit rate between the accident description text and the case report record according to the accident description text and the case report record; The semantic matching degree is determined based on the semantic overlap and keyword hit rate.

[0012] In a second aspect, the present application provides a claim fraud event identification device, including an image acquisition module, a description generation module, a matching degree calculation module and a fraud identification module, wherein: An image acquisition module, used to acquire accident images of a target accident case; A description generation module is used to generate an accident description text of a target accident case based on the accident image; A matching degree calculation module is used to determine the semantic matching degree between the accident description text and the case report record according to the accident description text and the case report record of the target accident case; The fraud identification module is used to determine that the target accident case is a fraud incident when the semantic matching degree is less than a preset matching degree threshold.

[0013] Optionally, describe the generation module, including: A subject determination unit, used to determine the vehicle subject, obstacle subject and background object of the target accident case according to the accident image; The description generation unit is used to generate an accident description text of a target accident case according to a vehicle body, an obstacle body and background objects.

[0014] Optionally, the subject determination unit is used to: Perform image segmentation processing on the accident image to obtain a sub-image of the image elements contained in the accident image; Determine the label and location information of the sub-graph; Based on the labels and location information, the vehicle subject, obstacle subject and background objects of the target accident case are determined.

[0015] Optionally, describes the generation unit, used to: Determine explicit information and implicit information of the target accident case based on the vehicle body and the obstacle body; where explicit information includes distance information between the vehicle body and the obstacle body; implicit information includes speed, collision angle and inertia of the vehicle body, as well as the movement trajectory and state change of the vehicle body before and after the accident; Introduce pseudo-entities for target accident cases based on vehicle subjects, obstacle subjects, and background objects; Determine the accident cause of the target accident case based on the pseudo-entity and the explicit information and implicit information of the target accident case; According to the accident causes, background objects, explicit information and implicit information, the accident description text of the target accident case is generated through the text generation model.

[0016] Optional, image acquisition module, for: Determine the accident location of each accident case to be identified; Clustering the accident cases to be identified according to the accident locations to obtain multiple case groups of the accident cases to be identified; The case grouping in which the number of accident cases to be identified is greater than a preset number threshold among the multiple case groups is used as a target case grouping, and the accident cases to be identified in the target case grouping are determined as target accident cases; Obtain accident images of the target accident case.

[0017] Optionally, a matching degree calculation module is used to: Determine the semantic overlap and keyword hit rate between the accident description text and the case report record according to the accident description text and the case report record; The semantic matching degree is determined based on the semantic overlap and keyword hit rate.

[0018] In a third aspect, the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method for identifying claims fraud events as described above.

[0019] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method for identifying claims fraud events as described above are implemented.

[0020] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method for identifying claims fraud events as described above.

[0021] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: By using the method provided in the embodiment of the present application, the accident description text of the target accident case can be generated based on the accident image of the target accident case, and then the target accident case can be identified as a fraudulent event based on the semantic matching degree between the accident description text and the case report record of the target accident case. In this way, the effective identification of claim fraud events can be achieved, avoiding the problem of insurance company interests being damaged. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1a A schematic diagram of the implementation process of a method for identifying claims fraud events provided in an embodiment of the present application; Figure 1b A schematic diagram of the implementation flow of a method for obtaining accident images of a target accident case provided in an embodiment of the present application; Figure 1c A schematic diagram of the implementation flow of a method for generating an accident description text of a target accident case based on an accident image provided in an embodiment of the present application; Figure 2 A specific structural diagram of a claim fraud event identification device is provided for an embodiment of the present application; Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0024] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0025] Example 1 In order to solve the problem in the prior art that claims fraud incidents cannot be effectively identified, which easily leads to damage to the interests of insurance companies, an embodiment of the present application provides a method for identifying claims fraud incidents.

[0026] The execution subject of the method may be various types of computing devices, or an application or application (Application, APP) installed on the computing device. The computing device may be, for example, a user terminal such as a mobile phone, a tablet computer, a smart wearable device, or a server.

[0027] For ease of description, the present application embodiment takes the execution subject of the method as an example of a server to introduce the method. Those skilled in the art can understand that the present application embodiment takes the server as an example to introduce the method, which is only an exemplary description and does not limit the scope of protection of the claims corresponding to the present solution.

[0028] Specifically, the implementation process of the method provided in the embodiment of the present application is as follows: Figure 1a As shown, the following steps are included: Step 11, obtaining the accident image of the target accident case.

[0029] Among them, the target accident cases include accident cases to be identified whether fraudulent behavior exists.

[0030] In the embodiments of the present application, accident images of the target accident case can be obtained through the Internet traffic safety comprehensive service management platform (for example, the traffic management 12123 platform), trip recorders, etc.

[0031] In an optional implementation, considering that claims fraud incidents usually mainly include repeated claims at the same accident address, when the number of accident cases to be identified is large, in order to reduce the resource overhead caused by full identification and improve identification efficiency, Figure 1b As shown, the accident image of the target accident case can also be obtained in the following manner: Step 111, determining the accident location of each accident case to be identified.

[0032] The occurrence address includes the latitude and longitude information of the vehicle's location when the accident occurred.

[0033] In the embodiment of the present application, the accident address of each accident case to be identified can be determined through accident reports, GPS positioning data, EXIF ​​data of accident scene photos, accident alarm records, records of road monitoring systems or Internet of Vehicles data of the Internet of Vehicles platform.

[0034] It should be noted that the above-mentioned method of determining the accident location of the accident case to be identified is only an exemplary description of the embodiment of the present application and does not impose any limitation on the embodiment of the present application.

[0035] Step 112, clustering the accident cases to be identified according to the accident locations to obtain a plurality of case groups of the accident cases to be identified.

[0036] In the embodiment of the present application, each accident case to be identified is clustered according to the accident address, that is, the similarity of the accident cases to be identified is measured by the proximity of the longitude and latitude of the accident address.

[0037] Specifically, in the embodiment of the present application, after obtaining the latitude and longitude information of the accident location of each accident case to be identified, each accident case to be identified can be clustered based on the latitude and longitude information through a clustering method such as K-Means or DBSCAN. After clustering, each accident case to be identified is divided into multiple case groups, and the accident cases to be identified in each case group are relatively close in geographical location.

[0038] For example, suppose there are 6 accident cases to be identified, and the longitude and latitude of the accident locations are: Case 1: (39.9042, 116.4074); Case 2: (39.9043, 116.4075); Case 3: (39.9062, 116.5912); Case 4: (39.9834, 116.3155); Case 5: (39.9837, 116.4551); Case 6: (39.9040, 116.4075) After clustering, the above 6 accident cases to be identified can be divided into 3 case groups: Group A: Case 1, Case 2 and Case 6.

[0039] Group B: Case 3.

[0040] Group C: Case 4 and Case 5.

[0041] Step 113, taking a case group in which the number of accident cases to be identified is greater than a preset number threshold among the multiple case groups as a target case group, and determining the accident cases to be identified in the target case group as the target accident cases.

[0042] In the embodiment of the present application, a target case group whose number of accident cases to be identified is greater than the preset number threshold can be selected from multiple case groups according to a preset number threshold. Then, the accident cases to be identified in the selected target case group are defined as target accident cases.

[0043] For example, using the above example, assuming that the six accident cases to be identified can be divided into Group A, Group B and Group C, and the preset number threshold is 2, then it can be determined that the target case group in which the number of accident cases to be identified in the group is greater than the preset number threshold (2) is Group A. Therefore, the accident cases 1, 2 and 3 to be identified in Group A are the target accident cases.

[0044] Step 114, acquiring the accident image of the target accident case.

[0045] In an embodiment of the present application, when obtaining the accident image of the target accident case, the accident image of the target accident case can be obtained through the Internet traffic safety comprehensive service management platform (for example, the traffic management 12123 platform), a trip recorder, etc.

[0046] Step 12: Generate an accident description text of the target accident case based on the accident image.

[0047] In the embodiment of the present application, after obtaining the accident image, the image processing and feature extraction can be performed on the accident image through image processing technology (such as YOLO, FasterR-CNN, etc.) to obtain the target object information (such as vehicles, pedestrians), road environment information (such as traffic lights and road signs), and vehicle damage characteristics contained in the accident image; then, based on the obtained target object information, road environment information, vehicle damage characteristics, etc., the specific cause and process of the accident can be inferred through semantic understanding and scene reasoning. Finally, based on the inferred specific cause and process of the accident, a predefined template or text generation model is used to generate an accident description text.

[0048] Among them, the template style of the predefined template can be as follows: At {time}, at {location}, vehicles {vehicle A information} and vehicles {vehicle B information} had {accident type} due to {reason}, resulting in {damage condition}.

[0049] For example, taking the template style of the above-mentioned predefined template as an example, the accident description text generated by using the predefined template based on the inferred specific cause and process of the accident can be as follows: "On the afternoon of November 18, 2024, at the XXXX intersection in Dongcheng District, Beijing, a white sedan with license plate number Beijing A12XXX and a black SUV with license plate number Beijing B56XXX rear-ended each other. The traffic light was red when the accident occurred. The black SUV braked suddenly, causing the white sedan to rear-end, and both vehicles suffered minor damage." It should be noted that the above-mentioned method of generating an accident description text based on an accident image is merely an exemplary description of an embodiment of the present application and does not impose any limitation on the embodiment of the present application.

[0050] In an optional embodiment, if Figure 1c As shown, the accident description text of the target accident case can be generated by the following steps 121 to 122: Step 121, determining the vehicle subject, obstacle subject and background object of the target accident case according to the accident image.

[0051] Among them, the vehicle entities include motor vehicles directly involved in or affected by the accident, including small cars, trucks, buses, motorcycles, etc.

[0052] The obstacle body refers to the object that causes the accident or has a direct impact on the vehicle in the accident, including dynamic obstacles and static obstacles.

[0053] Among them, dynamic obstacles: such as pedestrians, bicycles, electric vehicles, animals, etc.

[0054] Static obstacles: such as roadside guardrails, traffic signs, trees, potholes or construction equipment.

[0055] It should be noted that the obstacle bodies cited above are merely exemplary descriptions of the embodiments of the present application and do not impose any limitation on the embodiments of the present application.

[0056] Background objects refer to non-directly involved elements in the accident scene that provide environmental or condition information for accident description, including traffic environment, geographical location, weather conditions and time information.

[0057] In an optional implementation, the following steps (1) to (3) may be used to determine the vehicle subject, obstacle subject, and background object of the target accident case: (1) Perform image segmentation processing on the accident image to obtain a sub-image of the image elements contained in the accident image.

[0058] Among them, a sub-image refers to an image part extracted or segmented from an accident image that contains specific content or elements. In addition, each sub-image contains only one category of pixels (such as vehicles, pedestrians, roads); the position of elements in the sub-image is consistent with the original image, ensuring that the segmented elements can be correctly aligned in the original accident image.

[0059] In an embodiment of the present application, a deep learning model (such as DeepLab, Unet, etc.) can be used to classify and annotate the pixel-level content in the accident image so as to distinguish different elements or objects in the accident image and form independent areas or sub-images.

[0060] (2) Determine the label and location information of the subgraph.

[0061] The label of the sub-image is a semantic description of the pixel area category in the accident image, which is used to distinguish objects of different categories (for example, vehicles, pedestrians, roads, etc.).

[0062] The location information of the sub-image, including the location range of the target object in the sub-image, is usually expressed as a bounding box (BoundingBox) or a pixel outline.

[0063] In the embodiment of the present application, the category labels can be pre-defined according to the task requirements, such as "vehicle", "pedestrian", "road", "tree", etc. Then, the accident image is processed using a semantic segmentation or instance segmentation model (such as DeepLab, Mask R-CNN) to output a category mask.

[0064] Among them, each pixel value of the category mask image represents the category to which the pixel belongs, for example: Pixels with a value of 1 represent "vehicles".

[0065] Pixels with a value of 2 represent "pedestrians".

[0066] Pixels with a value of 3 represent "road".

[0067] Finally, the category pixels in the sub-image are counted to determine the label of the sub-image.

[0068] In the embodiment of the present application, when determining the position information of the sub-image, the pixel coordinates of the target category can be extracted from the category mask image. For example, for the category "vehicle", all pixel coordinates with a pixel value of 1 in the category mask image are extracted. Then, a rectangular bounding box is determined based on the minimum and maximum coordinates of the target category pixels; and then the position information of the sub-image is determined based on the rectangular bounding box.

[0069] Alternatively, when determining the position information of the sub - figure, an image - processing tool (such as findContours in OpenCV) can also be used to extract the pixel contour of the target category from the category mask image. Then, the position information of the sub - figure is determined based on this pixel contour.

[0070] (3) Determine the vehicle body, obstacle body, and background objects of the target accident case according to the label and position information.

[0071] In the embodiment of the present application, when determining the vehicle body of the target accident case, the sub - figure with the label "vehicle" in the image segmentation result can be searched first. Then, according to the position information of this sub - figure, the vehicle closest to the collision point or the vehicle located in the road area is screened out as the vehicle body of the target accident case.

[0072] Similarly, when determining the obstacle body of the target accident case, the sub - figure with the label "obstacle" in the image segmentation result can be searched first, including roadblocks, guardrails, pedestrians, etc. Then, according to the position information, the obstacle that contacts the vehicle body and is close to the collision point is selected as the obstacle body.

[0073] Similarly, when determining the background objects of the target accident case, the sub - figure with the label "background" can be searched, including roads, traffic lights, buildings, etc. Then, the background information related to the accident scene is determined as the background objects.

[0074] Step 122: Generate the accident description text of the target accident case according to the vehicle body, obstacle body, and background objects.

[0075] In the embodiment of the present application, according to the vehicle body, obstacle body, and background objects, a predefined template or a text generation model can be used to generate the accident description text of the target accident case.

[0076] Among them, the template style of the predefined template can be as follows: [Time], [Location], vehicle [Vehicle A information] due to [Obstacle information] had [Behavior], resulting in [Accident type]. At that time, [Background object information].

[0077] For example, according to the vehicle body, obstacle body, and background objects, using a predefined template or a text generation model, the accident description text of the target accident case in the following form can be generated: "On November 18, 2024, at the crossroads in Dongcheng District, Beijing, vehicle A (Beijing A12XX) deviated due to avoiding the damaged guardrail in the middle of the road, resulting in a collision with vehicle B (Beijing B56XXX). The traffic light was red at the time of the accident, the road surface was wet and slippery, and the light was dim (night).

[0078] Alternatively, taking the text generation model as an example, when generating the accident description text of the target accident case based on the vehicle body, obstacle body and background objects, the accident description text of the target accident case can be generated based on the text generation model (for example, GPT, BERT and other models) and combined with the extracted information such as the vehicle body, obstacle body and background objects.

[0079] In an optional implementation, the accident description text of the target accident case may be generated according to the vehicle body, the obstacle body and the background object in the following manner: (1) Determine the explicit and implicit information of the target accident case based on the vehicle subject and the obstacle subject.

[0080] Among them, explicit information includes the distance information between the vehicle body and the obstacle body; implicit information includes the vehicle body's speed, collision angle, inertia, and the vehicle body's movement trajectory and state changes before and after the accident.

[0081] (2) Introduce pseudo-entities for the target accident case based on the vehicle subject, obstacle subject, and background objects.

[0082] Among them, the pseudo-entities include virtual vehicle pseudo-entities, obstacle pseudo-entities and background object pseudo-entities.

[0083] In an embodiment of the present application, taking into account that in actual applications, the target accident case may include some influencing factors that exist but are not directly observed, therefore, in order to ensure the accuracy of subsequent accident cause speculation and enhance the depth of accident reasoning, pseudo-entities can be introduced for the target accident case. These pseudo-entities may include virtual vehicles, virtual obstacles or virtual environmental changes to enhance the depth of accident reasoning.

[0084] It should be noted that when introducing pseudo-entities for target accident cases, any one or more of virtual vehicle pseudo-entities, virtual obstacle pseudo-entities and virtual environment pseudo-entities may be introduced according to actual needs.

[0085] When introducing a virtual vehicle pseudo-entity, we can first infer the possible vehicles in the nearby lanes based on the relative positions of the vehicle body and the obstacle body; or infer whether the vehicle body is disturbed by other vehicles based on the movement trajectory and state changes of the vehicle body before and after the accident. If it is determined that the vehicle body is disturbed by other vehicles, or it is inferred that there are possible vehicles in the nearby lanes, a virtual vehicle pseudo-entity is simulated for the vehicle body based on the deflection direction and acceleration change of the vehicle body, that is, the virtual vehicle pseudo-entity is introduced.

[0086] For example, based on the deflection direction and acceleration changes of the vehicle body, it is simulated that a "vehicle C" appears in the right lane of the vehicle body and causes the vehicle body to avoid it.

[0087] Similarly, when introducing a virtual obstacle pseudo-entity, the obstacle interference that the vehicle body may be subject to can be inferred based on the movement trajectory and state changes of the vehicle body before and after the accident. Then, a virtual obstacle pseudo-entity is generated based on the obstacle interference.

[0088] For example, a pseudo-entity "obstacle X" is inserted into the driving path of the vehicle body to represent possible fallen cargo, etc.

[0089] Correspondingly, when introducing virtual environment pseudo-entities, we can first determine the background changes that may exist but are not recorded in the accident scene of the target accident case, such as sudden weather changes (such as strong winds, strong light), traffic light failures, etc. Then, based on these background changes, generate virtual environment disturbance factors, that is, virtual environment pseudo-entities.

[0090] For example, a pseudo-entity “Red light signal delayed for 3 seconds” is introduced to indicate the potential confusion during the signal switching process.

[0091] In the embodiment of the present application, by introducing pseudo-entities, complex situations that may occur in the target accident case but are difficult to directly observe can be effectively simulated, thereby improving the comprehensiveness and rationality of the subsequent accident description text of the target accident case.

[0092] (3) Determine the cause of the target accident case based on the pseudo-entity and the explicit and implicit information of the target accident case.

[0093] In an embodiment of the present application, when determining the accident cause of a target accident case, multiple causal chains of the target accident case can be generated based on pseudo-entities, explicit information and implicit information of the target accident case; then, the accident cause probability under each causal chain is calculated respectively; finally, the causal chain corresponding to the largest accident cause probability is determined as the accident cause of the target accident case.

[0094] Next, with reference to examples, a specific implementation method for determining the cause of a target accident case according to a pseudo-entity and explicit information and implicit information of the target accident case in an embodiment of the present application is described: First, assume that the explicit information and implicit information are as follows: (1) Explicit information: Distance: The initial distance between the vehicle body and the obstacle body is 15 meters.

[0095] Collision point: The main body of the vehicle hits the right side of the guardrail.

[0096] (2) Implicit information: Speed: The vehicle speed of the main vehicle at the moment of collision was 48km / h.

[0097] Collision angle: The body of the vehicle hits the guardrail at an angle of 15°.

[0098] Movement trajectory: The vehicle body deviated to the right of the lane 3 seconds before the collision.

[0099] Inertia: The braking distance of the vehicle body is extended to 12 meters due to the slippery road surface.

[0100] Secondly, the specific information of the introduced pseudo-entity is as follows: (3) Virtual vehicle pseudo-entity: A "vehicle C" is generated in the right lane of the main vehicle, and it is inferred that the lane change behavior of "vehicle C" caused the main vehicle to avoid it.

[0101] The logic of introducing the virtual vehicle pseudo-entity is that the motion trajectory deviation and deceleration characteristics of the vehicle body are consistent with the avoidance behavior.

[0102] (4) Virtual obstacle pseudo-entity: A "dynamic obstacle X" is generated on the collision path of the vehicle body, presumably due to the fallen cargo causing the vehicle to deflect further.

[0103] The logic of introducing virtual obstacle pseudo-entities is because the vehicle body has an abnormal braking trajectory near the guardrail impact point.

[0104] (5) Virtual environment pseudo-entity: Simulate the delay in switching the state of a traffic light. Assume that a 3-second delay in the red light state causes the vehicle to make an incorrect judgment.

[0105] Logic: The status of the traffic light is related to the trajectory change of the vehicle.

[0106] Then, the following causal chain can be generated based on the explicit and implicit information of the pseudo-entities and target accident cases: Causal chain 1: Vehicle C changes lanes → the main vehicle avoids → braking is delayed.

[0107] Causal chain 2: Slippery road surface → increased braking distance.

[0108] Causal chain 3: Obstacle X impacts → Vehicle body hits guardrail.

[0109] Causal chain 4: Traffic light delay → vehicle misjudgment.

[0110] After that, the accident cause probability of each causal chain is calculated. For example, suppose the accident cause probabilities of causal chains 1 to 4 are 0.2, 0.1, 0.3, and 0.4, respectively.

[0111] Finally, the causal chain corresponding to the largest accident cause probability is determined as the accident cause of the target accident case. For example, using the above example, causal chain 4 is determined as the accident cause of the target accident case. In other words, the cause of the target accident case is that the vehicle subject misjudged the target accident case due to the delay of the traffic light.

[0112] (4) Generate the accident description text of the target accident case through the text generation model based on the accident cause, background objects, explicit information and implicit information.

[0113] In an embodiment of the present application, the cause of the accident, background objects, explicit information, and implicit information may be input into a trained text generation model to generate an accident description text for the target accident case.

[0114] Optionally, in the embodiment of the present application, when training the text generation model, the following method can be used: First, historical accident cases are collected as sample sets; Secondly, the accident causes, background objects, explicit information and implicit information of historical accident cases are determined as training samples for the text generation model; Afterwards, the accident causes, background objects, explicit information, and implicit information of the historical accident cases are input into network models with text generation functions, such as the Transformers model, the PLANET (Dynamic Content Planningin Autoregressive Transformers for Long-form Text Generation, PLANET) model, the recurrent neural network, the generative adversarial network, and the variational autoencoder, for training to obtain the accident description text of the historical accident case.

[0115] Calculate the semantic matching degree between the accident description text of the historical accident case and the report record of the historical accident case, and use the inverse of the semantic matching degree as the model loss of the network model with text generation function; Finally, the network model with text generation function is iteratively updated and trained according to the model loss until the model loss of the model meets the preset model loss requirements. The iterative update is stopped and the final model is used as the trained text generation model.

[0116] Step 13, according to the accident description text and the report record of the target accident case, determine the semantic matching degree between the accident description text and the report record.

[0117] In an embodiment of the present application, the semantic matching degree between the accident description text and the report record can be determined by comparing the matching degree between the scenes / events described in the accident description text and the report record of the target accident case.

[0118] Optionally, before determining the semantic matching degree between the accident description text and the report record, the embodiment of the present application also includes extracting the accident report voice record, and obtaining the accident process described by the reporter, that is, the report record, by performing tone separation and voice transcription on the accident report voice record.

[0119] In an optional implementation, the semantic matching degree between the accident description text and the case report record may be determined according to the accident description text and the case report record of the target accident case in the following manner: (1) Determine the semantic overlap and keyword hit rate between the accident description text and the case report record based on the accident description text and the case report record.

[0120] (2) Determine the semantic matching degree based on the semantic overlap and keyword hit rate.

[0121] Step 14: When the semantic matching degree is less than a preset matching degree threshold, the target accident case is determined to be a fraud incident.

[0122] By using the method provided in the embodiment of the present application, the accident description text of the target accident case can be generated based on the accident image of the target accident case, and then the target accident case can be identified as a fraudulent event based on the semantic matching degree between the accident description text and the case report record of the target accident case. In this way, the effective identification of claim fraud events can be achieved, avoiding the problem of insurance company interests being damaged.

[0123] Example 2 In order to solve the problem that the prior art cannot effectively identify claims fraud events, which easily leads to damage to the interests of insurance companies, the present application embodiment provides a claims fraud event identification device, the specific structural diagram of the device is as follows: Figure 2 As shown, it includes an image acquisition module 21, a description generation module 22, a matching degree calculation module 23 and a fraud identification module 24. The functions of each module are as follows: An image acquisition module 21 is used to acquire accident images of a target accident case; A description generation module 22, used to generate an accident description text of a target accident case based on the accident image; A matching degree calculation module 23 is used to determine the semantic matching degree between the accident description text and the case report record according to the accident description text and the case report record of the target accident case; The fraud identification module 24 is used to determine that the target accident case is a fraud event when the semantic matching degree is less than a preset matching degree threshold.

[0124] Optionally, describe the generation module, including: A subject determination unit, used to determine the vehicle subject, obstacle subject and background object of the target accident case according to the accident image; The description generation unit is used to generate an accident description text of a target accident case according to a vehicle body, an obstacle body and background objects.

[0125] Optionally, the subject determination unit is used to: Perform image segmentation processing on the accident image to obtain a sub-image of the image elements contained in the accident image; Determine the label and location information of the sub-graph; Based on the labels and location information, the vehicle subject, obstacle subject and background objects of the target accident case are determined.

[0126] Optionally, describes the generation unit, used to: Determine explicit information and implicit information of the target accident case based on the vehicle body and the obstacle body; where explicit information includes distance information between the vehicle body and the obstacle body; implicit information includes speed, collision angle and inertia of the vehicle body, as well as the movement trajectory and state change of the vehicle body before and after the accident; Introduce pseudo-entities for target accident cases based on vehicle subjects, obstacle subjects, and background objects; Determine the accident cause of the target accident case based on the pseudo-entity and the explicit information and implicit information of the target accident case; According to the accident causes, background objects, explicit information and implicit information, the accident description text of the target accident case is generated through the text generation model.

[0127] Optional, image acquisition module, for: Determine the accident location of each accident case to be identified; Clustering the accident cases to be identified according to the accident locations to obtain multiple case groups of the accident cases to be identified; The case grouping in which the number of accident cases to be identified is greater than a preset number threshold among the multiple case groups is used as a target case grouping, and the accident cases to be identified in the target case grouping are determined as target accident cases; Obtain accident images of the target accident case.

[0128] Optionally, a matching degree calculation module is used to: Determine the semantic overlap and keyword hit rate between the accident description text and the case report record according to the accident description text and the case report record; The semantic matching degree is determined based on the semantic overlap and keyword hit rate.

[0129] By using the device provided in the embodiment of the present application, the accident description text of the target accident case can be generated based on the accident image of the target accident case, and then the target accident case can be identified as a fraudulent event based on the semantic matching degree between the accident description text and the case report record of the target accident case. In this way, the effective identification of claim fraud events can be achieved, avoiding the problem of damage to the interests of insurance companies.

[0130] Example 3 Figure 3 A hardware structure diagram of an electronic device for implementing each embodiment of the present application, the electronic device 300 includes but is not limited to: a radio frequency unit 301, a network module 302, an audio output unit 303, an input unit 304, a sensor 305, a display unit 306, a user input unit 307, an interface unit 308, a memory 309, a processor 310, and a power supply 311. Those skilled in the art can understand that Figure 3 The electronic device structure shown in the figure does not constitute a limitation on the electronic device, and the electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. In the embodiments of the present application, the electronic device includes but is not limited to a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted terminal, a wearable device, and a pedometer.

[0131] Among them, the processor 310 is used to obtain an accident image of the target accident case; generate an accident description text of the target accident case based on the accident image; determine the semantic matching degree between the accident description text and the report record based on the accident description text and the report record of the target accident case; when the semantic matching degree is less than a preset matching degree threshold, determine that the target accident case is a fraud incident.

[0132] Optionally, an accident description text of the target accident case is generated based on the accident image, including: Determine the vehicle subject, obstacle subject and background object of the target accident case based on the accident image; Generate an accident description text of the target accident case based on the vehicle body, obstacle body and background objects.

[0133] Optionally, the vehicle subject, obstacle subject, and background object of the target accident case are determined based on the accident image, including: Perform image segmentation processing on the accident image to obtain a sub-image of the image elements contained in the accident image; Determine the label and location information of the sub-graph; Based on the labels and location information, the vehicle subject, obstacle subject and background objects of the target accident case are determined.

[0134] Optionally, an accident description text of the target accident case is generated based on the vehicle subject, obstacle subject and background objects, including: Determine explicit information and implicit information of the target accident case based on the vehicle body and the obstacle body; where explicit information includes distance information between the vehicle body and the obstacle body; implicit information includes speed, collision angle and inertia of the vehicle body, as well as the movement trajectory and state change of the vehicle body before and after the accident; Introduce pseudo-entities for target accident cases based on vehicle subjects, obstacle subjects, and background objects; Determine the accident cause of the target accident case based on the pseudo-entity and the explicit information and implicit information of the target accident case; According to the accident causes, background objects, explicit information and implicit information, the accident description text of the target accident case is generated through the text generation model.

[0135] Optionally, obtain accident images of the target accident case, including: Determine the accident location of each accident case to be identified; Clustering the accident cases to be identified according to the accident locations to obtain multiple case groups of the accident cases to be identified; The case grouping in which the number of accident cases to be identified is greater than a preset number threshold among the multiple case groups is used as a target case grouping, and the accident cases to be identified in the target case grouping are determined as target accident cases; Obtain accident images of the target accident case.

[0136] Optionally, according to the accident description text and the report record of the target accident case, the semantic matching degree between the accident description text and the report record is determined, including: Determine the semantic overlap and keyword hit rate between the accident description text and the case report record according to the accident description text and the case report record; The semantic matching degree is determined based on the semantic overlap and keyword hit rate.

[0137] The memory 309 is used to store a computer program that can be run on the processor 310. When the computer program is executed by the processor 310, the above functions implemented by the processor 310 are realized.

[0138] It should be understood that in the embodiment of the present application, the radio frequency unit 301 can be used for receiving and sending signals during information transmission or calls. Specifically, after receiving downlink data from the base station, it is sent to the processor 310 for processing; in addition, uplink data is sent to the base station. Generally, the radio frequency unit 301 includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, etc. In addition, the radio frequency unit 301 can also communicate with the network and other devices through a wireless communication system.

[0139] The electronic device provides users with wireless broadband Internet access through the network module 302, such as helping users to send and receive emails, browse web pages, and access streaming media.

[0140] The audio output unit 303 can convert the audio data received by the RF unit 301 or the network module 302 or stored in the memory 309 into an audio signal and output it as sound. Moreover, the audio output unit 303 can also provide audio output related to a specific function performed by the electronic device 300 (for example, a call signal reception sound, a message reception sound, etc.). The audio output unit 303 includes a speaker, a buzzer, a receiver, etc.

[0141] The input unit 304 is used to receive audio or video signals. The input unit 304 may include a graphics processor (GPU) 3041 and a microphone 3042. The graphics processor 3041 processes the image data of a static picture or video obtained by an image capture device (such as a camera) in a video capture mode or an image capture mode. The processed image frame can be displayed on the display unit 306. The image frame processed by the graphics processor 3031 can be stored in the memory 309 (or other storage medium) or sent via the radio frequency unit 301 or the network module 302. The microphone 3042 can receive sound and can process such sound into audio data. The processed audio data can be converted into a format output that can be sent to a mobile communication base station via the radio frequency unit 301 in the case of a telephone call mode.

[0142] The electronic device 300 also includes at least one sensor 305, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor includes an ambient light sensor and a proximity sensor, wherein the ambient light sensor can adjust the brightness of the display panel 3061 according to the brightness of the ambient light, and the proximity sensor can turn off the display panel 3061 and / or the backlight when the electronic device 300 is moved to the ear. As a type of motion sensor, the accelerometer sensor can detect the magnitude of acceleration in each direction (generally three axes), and can detect the magnitude and direction of gravity when stationary, which can be used to identify the posture of the electronic device (such as horizontal and vertical screen switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tapping), etc.; the sensor 305 can also include a fingerprint sensor, a pressure sensor, an iris sensor, a molecular sensor, a gyroscope, a barometer, a hygrometer, a thermometer, an infrared sensor, etc., which will not be repeated here.

[0143] The display unit 306 is used to display information input by the user or information provided to the user. The display unit 306 may include a display panel 3061, which may be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), or the like.

[0144] The user input unit 307 can be used to receive input digital or character information, and generate key signal input related to user settings and function control of the electronic device. Specifically, the user input unit 307 includes a touch panel 3071 and other input devices 3072. The touch panel 3071, also known as a touch screen, can collect user touch operations on or near it (such as operations performed by users using fingers, styluses, or any other suitable objects or accessories on or near the touch panel 3071). The touch panel 3071 may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch orientation, detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into contact point coordinates, and then sends it to the processor 310, receives the command sent by the processor 310 and executes it. In addition, the touch panel 3071 can be implemented in various types such as resistive, capacitive, infrared, and surface acoustic waves. In addition to the touch panel 3071, the user input unit 307 may also include other input devices 3072. Specifically, other input devices 3072 may include but are not limited to a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, and a joystick, which are not described in detail here.

[0145] Furthermore, the touch panel 3071 may be overlaid on the display panel 3061. When the touch panel 3071 detects a touch operation on or near it, it is transmitted to the processor 310 to determine the type of touch event. Then, the processor 310 provides corresponding visual output on the display panel 3061 according to the type of touch event. Figure 3 In the figure, the touch panel 3071 and the display panel 3061 are used as two independent components to realize the input and output functions of the electronic device. However, in some embodiments, the touch panel 3071 and the display panel 3061 can be integrated to realize the input and output functions of the electronic device, which is not limited here.

[0146] The interface unit 308 is an interface for connecting an external device to the electronic device 300. For example, the external device may include a wired or wireless headset port, an external power supply (or battery charger) port, a wired or wireless data port, a memory card port, a port for connecting a device with an identification module, an audio input / output (I / O) port, a video I / O port, a headphone port, etc. The interface unit 308 may be used to receive input (e.g., data information, power, etc.) from an external device and transmit the received input to one or more elements within the electronic device 300 or may be used to transmit data between the electronic device 300 and an external device.

[0147] The memory 309 can be used to store software programs and various data. The memory 309 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 309 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0148] The processor 310 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire electronic device. By running or executing software programs and / or modules stored in the memory 309 and calling data stored in the memory 309, it performs various functions of the electronic device and processes data, thereby monitoring the electronic device as a whole. The processor 310 may include one or more processing units; preferably, the processor 310 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 310.

[0149] The electronic device 300 may also include a power source 311 (such as a battery) for supplying power to various components. Preferably, the power source 311 may be logically connected to the processor 310 through a power management system, thereby implementing functions such as charging, discharging, and power consumption management through the power management system.

[0150] In addition, the electronic device 300 includes some functional modules not shown, which will not be described in detail here.

[0151] Preferably, an embodiment of the present application also provides an electronic device, including a processor 310, a memory 309, and a computer program stored in the memory 309 and executable on the processor 310. When the computer program is executed by the processor 310, each process of the above-mentioned claims fraud event identification method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0152] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, each process of the above-mentioned claim fraud event identification method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0153] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0154] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate a device for implementing the functions specified in one process or multiple processes in the flowchart and / or one box or multiple boxes in the block diagram.

[0155] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0156] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0157] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0158] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0159] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0160] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0161] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for identifying claims fraud events, characterized in that: include: Acquire accident images of the target accident case; generating an accident description text of the target accident case according to the accident image; Determining the semantic matching degree between the accident description text and the case report record according to the accident description text and the case report record of the target accident case; When the semantic matching degree is less than a preset matching degree threshold, the target accident case is determined to be a fraud event.

2. The method according to claim 1, characterized in that Generating the accident description text of the target accident case according to the accident image includes: Determine the vehicle subject, obstacle subject and background object of the target accident case according to the accident image; An accident description text of the target accident case is generated according to the vehicle body, the obstacle body and the background object.

3. The method according to claim 2, characterized in that Determining the vehicle subject, obstacle subject, and background object of the target accident case according to the accident image includes: Performing image segmentation processing on the accident image to obtain a sub-image of image elements contained in the accident image; Determining the label and position information of the sub-graph; According to the tag and the position information, the vehicle subject, obstacle subject and background object of the target accident case are determined.

4. The method according to claim 2, characterized in that Generating an accident description text of the target accident case according to the vehicle body, the obstacle body and the background object, including: Determine explicit information and implicit information of the target accident case according to the vehicle body and the obstacle body; wherein the explicit information includes distance information between the vehicle body and the obstacle body; the implicit information includes speed, collision angle and inertia of the vehicle body, and movement trajectory and state change of the vehicle body before and after the accident; Introducing a pseudo entity for the target accident case according to the vehicle subject, the obstacle subject, and the background object; Determine the cause of the target accident case according to the pseudo-entity, the explicit information and the implicit information of the target accident case; According to the accident cause, the background object, the explicit information and the implicit information, an accident description text of the target accident case is generated through a text generation model.

5. The method according to claim 1, characterized in that Obtain accident images of the target accident case, including: Determine the accident location of each accident case to be identified; Clustering the accident cases to be identified according to the accident addresses to obtain a plurality of case groups of the accident cases to be identified; The case group in which the number of the accident cases to be identified is greater than a preset number threshold value among the multiple case groups is used as a target case group, and the accident cases to be identified in the target case group are determined as the target accident cases; Acquire the accident image of the target accident case.

6. The method according to claim 1, characterized in that The determining, based on the accident description text and the report record of the target accident case, the semantic matching degree between the accident description text and the report record includes: Determining the semantic overlap and keyword hit rate between the accident description text and the case report record according to the accident description text and the case report record; The semantic matching degree is determined according to the semantic coincidence degree and the keyword hit rate.

7. A claim fraud event identification device, characterized in that: It includes an image acquisition module, a description generation module, a matching degree calculation module and a fraud identification module, wherein: An image acquisition module, used to acquire accident images of a target accident case; A description generation module, used to generate an accident description text of the target accident case according to the accident image; A matching degree calculation module, used to determine the semantic matching degree between the accident description text and the case report record according to the accident description text and the case report record of the target accident case; The fraud identification module is used to determine that the target accident case is a fraud incident when the semantic matching degree is less than a preset matching degree threshold.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method for identifying claims fraud events as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for identifying claims fraud events as described in any one of claims 1 to 6.

10. A computer program product, characterized in that It includes a computer program, which, when executed by a processor, implements the method for identifying claims fraud events as described in any one of claims 1 to 6.

Images