Underwriting method, underwriting device, electronic device, and storage medium

Abstract

The embodiment of the application provides a kind of underwriting method, underwriting device, electronic equipment and storage medium, belong to artificial intelligence technical field, it is applicable to the field of financial technology.The method comprises the following steps: in response to the insurance request of the insured object, the insurance type and the object attribute feature of the insured object are obtained from the insurance request, the face image of the insured object is obtained according to the insurance type, and face recognition is carried out on the face image to obtain face feature, health risk assessment is carried out according to the face feature and the object attribute feature, the health risk score of the insured object is obtained, the health proof material of the insured object is obtained according to the health risk score, the health risk score is updated according to the health proof material, the target risk score is obtained, and the insurance request is audited according to the target risk score, which can improve the underwriting efficiency.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and is applied to the field of financial technology, particularly to an underwriting method, underwriting device, electronic device, and storage medium. Background Technology

[0002] After an insured submits an insurance application to an insurance institution, the institution reviews the application to reduce the institution's risk of claims. However, the traditional underwriting process relies on manual review, where underwriters use their experience to determine if the insured has health risks. This process is time-consuming and cannot be conducted in real time, resulting in low underwriting efficiency. Summary of the Invention

[0003] The main objective of this application is to provide an underwriting method, underwriting device, electronic device, and storage medium, which aim to improve underwriting efficiency.

[0004] To achieve the above objectives, a first aspect of this application provides an underwriting method, the method comprising: In response to an insurance application request from an insured party, the insurance type and the object attribute characteristics of the insured party are obtained from the insurance application request; The facial image of the insured is obtained according to the type of insurance, and facial recognition is performed on the facial image to obtain facial features; A health risk assessment is performed based on the facial features and the object attribute features to obtain the health risk score of the insured object. Obtain the health certificate materials of the insured subject based on the health risk score; The health risk score is updated based on the health certification materials to obtain the target risk score; The insurance application is reviewed based on the target risk score.

[0005] In some embodiments, performing face recognition on the face image to obtain face features includes: Perform face detection on the face image to obtain the original face region; The original face region is aligned to obtain the aligned face region. Feature extraction is performed on the aligned face region to obtain the face features.

[0006] In some embodiments, performing face alignment on the original face region to obtain an aligned face region includes: Key point detection is performed on the original face region to obtain facial key points; Calculate the facial pose angle based on the aforementioned facial key points; Based on the facial key points and the facial pose angle, determine the alignment marker parameters; The original face region is aligned according to the alignment flag parameters and the face pose angle to obtain the aligned face region.

[0007] In some embodiments, the step of aligning the original face region according to the alignment flag parameters and the face pose angle to obtain the aligned face region includes: The illumination features of the original face region are extracted based on the alignment flag parameters; Based on the lighting characteristics and the face pose angle, a face alignment category is determined; wherein, the face alignment category includes a model alignment category or a key point alignment category; The face generation model is called according to the model alignment category, and the original face region is aligned using the called face generation model to obtain the aligned face region. The original face region is aligned according to the key point alignment category and the face key points to obtain the aligned face region.

[0008] In some embodiments, the step of extracting features from the aligned face region to obtain the face features includes: Key point detection is performed on the aligned face region to obtain the alignment key points; Feature extraction is performed on the alignment key points to obtain eye state features and facial expression features; Feature extraction is performed on the aligned face region to obtain skin color features and skin texture features; The facial features are determined based on the eye state features, the expression features, the skin color features, and the skin texture features.

[0009] In some embodiments, the step of performing a health risk assessment based on the facial features and the object attribute features to obtain a health risk score for the insured object includes: The facial features and the object attribute features are fused to obtain the target fused features; A health risk prediction model is used to assess the health risks of the target fusion feature, the facial feature, and the object attribute feature, respectively, to obtain a first risk score for the target fusion feature, a second risk score for the facial feature, and a third risk score for the object attribute feature. The health risk score is calculated based on the first risk score, the second risk score, and the third risk score.

[0010] In some embodiments, obtaining the health certification materials of the insured subject based on the health risk score includes: The risk category is determined based on the aforementioned health risk score; Calculate the trigger probability based on the health risk score and the risk category; The health certificate materials are obtained based on the trigger probability of the matching.

[0011] To achieve the above objectives, a second aspect of this application provides an underwriting device, the device comprising: The request acquisition module is used to respond to the insurance application request of the insured object and obtain the insurance type and the object attribute characteristics of the insured object from the insurance application request; The face recognition module is used to obtain the face image of the insured object according to the type of insurance, and to perform face recognition on the face image to obtain face features; The assessment module is used to conduct a health risk assessment based on the facial features and the object attribute features to obtain a health risk score for the insured object; The material acquisition module is used to acquire the health certificate materials of the insured object based on the health risk score; The update module is used to update the health risk score based on the health certificate materials to obtain the target risk score; The underwriting module is used to review the insurance application based on the target risk score.

[0012] To achieve the above objectives, a third aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0013] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0014] The underwriting method, underwriting device, electronic device, and computer-readable storage medium proposed in this application, in response to an insured's application, obtain the insured type and the insured's attribute characteristics from the application, and then review the application based on these characteristics. To automate the underwriting process, a facial image of the insured is obtained based on the insured type, and the insured's health status is assessed using this image. Facial recognition is performed on the facial image to capture facial features related to health status, resulting in facial features. A health risk assessment is performed based on the facial features and the insured's attribute characteristics, automatically identifying whether the insured has health risks and obtaining a health risk score. By comprehensively considering multimodal features such as facial features and insured attribute characteristics, the accuracy of the health risk assessment is ensured. To further improve the accuracy of the health risk assessment, health certificates are obtained based on the health risk score, and the health risk score is updated based on these certificates to obtain a target risk score. The application is then reviewed based on the target risk score, enabling automated application review and improving underwriting efficiency. Attached Figure Description

[0015] Figure 1 This is a flowchart of the underwriting method provided in the embodiments of this application; Figure 2 yes Figure 1 The flowchart of step S120 in the middle; Figure 3 yes Figure 2 The flowchart of step S220 in the text; Figure 4 yes Figure 3 The flowchart of step S340 in the text; Figure 5 yes Figure 2 The flowchart of step S230 in the middle; Figure 6 yes Figure 1 The flowchart of step S130 in the process; Figure 7 yes Figure 1 The flowchart of step S140 in the middle; Figure 8 This is a schematic diagram of the underwriting device provided in the embodiments of this application; Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0017] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0019] After an insured submits an insurance application to an insurance institution, the institution reviews the application to reduce the institution's risk of claims. However, the traditional underwriting process relies on manual review, where underwriters use their experience to determine if the insured has health risks. This process is time-consuming and cannot be conducted in real time, resulting in low underwriting efficiency.

[0020] Based on this, embodiments of this application provide an underwriting method, underwriting device, electronic device, and computer-readable storage medium, aiming to improve underwriting efficiency.

[0021] The underwriting method, underwriting device, electronic device, and computer-readable storage medium provided in this application are specifically described through the following embodiments. First, the underwriting method in this application embodiment is described.

[0022] The underwriting method provided in this application relates to the field of artificial intelligence technology. The underwriting method provided in this application can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the underwriting method, but is not limited to the above forms.

[0023] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0024] Figure 1 This is an optional flowchart of the underwriting method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S110 to S160.

[0025] Step S110: In response to the insurance application request of the insured, obtain the insurance type and the object attribute characteristics of the insured from the insurance application request; Step S120: Obtain the face image of the insured according to the type of insurance, and perform face recognition on the face image to obtain face features; Step S130: Conduct a health risk assessment based on facial features and object attribute features to obtain the health risk score of the insured object; Step S140: Obtain the health certificate materials of the insured based on the health risk score; Step S150: Update the health risk score based on the health certificate materials to obtain the target risk score; Step S160: Review the insurance application based on the target risk score.

[0026] Steps S110 to S160 as shown in the embodiments of this application assess the health status of the insured by comprehensively considering multimodal features such as facial features and object attribute features. This can automatically identify whether the insured has health risks without manual review, thus improving underwriting efficiency.

[0027] In step S110 of some embodiments, when an insured person requests to purchase an insurance product from an insurance institution, they submit an insurance application request through the business system provided by the insurance institution. After receiving the application request, the business system extracts the type of insurance and the insured person's attribute characteristics from it. The type of insurance is the type of insurance product selected by the insured person, which can be health insurance, life insurance, or accident insurance, etc. The attribute characteristics can be features related to the insured person's health status, such as gender, age, occupation, and lifestyle habits.

[0028] If the insurance product requires identification of the insured's health risks, such as health insurance, life insurance, or critical illness insurance, then facial images of the insured are captured using imaging equipment. Facial images can reflect the insured's health condition; for example, hyperthyroidism can manifest as bulging eyes and eyelid edema, while anemia and heart failure can cause paleness. Facial recognition is then performed on the facial images to capture features related to health conditions, resulting in facial features that are used to determine whether the insured faces any health risks.

[0029] Please see Figure 2 In some embodiments, step S120 may include, but is not limited to, steps S210 to S230: Step S210: Perform face detection on the face image to obtain the original face region; Step S220: Perform face alignment on the original face region to obtain the aligned face region; Step S230: Extract features from the aligned face region to obtain face features.

[0030] In step S210 of some embodiments, background information in the face image may obscure key facial features, interfering with the extraction of facial features. To improve the accuracy and efficiency of face recognition, face detection is performed on the face image to locate the face region from the face image, exclude irrelevant background, and reduce computational load to obtain the original face region.

[0031] Specifically, an initial feature representation of the face image is extracted using a feature extraction network, such as ResNet50 or MobileNet. A Feature Pyramid Network (FPN) is then used to extract features from this initial representation at multiple scales, resulting in feature maps of different scales. This improves the detection capability for targets of different sizes in the face image. Anchor boxes are generated on each feature map. A classification branch classifies each feature map, obtaining the confidence score of the anchor box, which represents the probability that the anchor box contains a face. A regression branch regresses each feature map, obtaining the offset of the anchor box, which is the precise coordinate of the anchor box output by the regression calculation. Anchor boxes with a confidence score greater than a preset threshold are selected as candidate face regions. Candidate offsets for these candidate face regions are then selected from the offsets output by the regression branch. The original face region is located from the face image based on these candidate offsets.

[0032] In step S220 of some embodiments, the face in the face image may be deformed due to factors such as changes in expression, posture, lighting, and device. In order to improve the accuracy of face recognition, face alignment is performed on the original face region to adjust the original face region to a standard posture and position, thereby obtaining an aligned face region.

[0033] In step S230 of some embodiments, facial features of the aligned face region are extracted to identify the health status of the insured based on the facial features.

[0034] Steps S210 to S230 above, by performing face detection and face alignment operations on the face image, can eliminate the influence of factors such as pose and lighting on feature extraction, thereby improving the accuracy and efficiency of face feature extraction.

[0035] Please see Figure 3 In some embodiments, step S220 may include, but is not limited to, steps S310 to S340: Step S310: Perform key point detection on the original face region to obtain face key points; Step S320: Calculate the facial pose angle based on facial key points; Step S330: Determine the alignment marker parameters based on the facial key points and facial pose angle; Step S340: Align the original face region with the alignment marker parameters and face pose angle to obtain the aligned face region.

[0036] In step S310 of some embodiments, the original face region is input to a pre-trained keypoint detector for keypoint detection to obtain facial keypoints. Facial keypoints are coordinate points used to describe facial features, contours, or specific regions, such as eyebrows, eyes, nose, mouth, and jawline contours.

[0037] In step S320 of some embodiments, the facial pose angles include pitch angle, yaw angle, and roll angle. Pitch angle is the vertical tilt angle of the face, yaw angle is the horizontal rotation angle of the face, and roll angle is the forward and backward rotation angle of the face. The angle between the line connecting the tip of the nose and the chin and a vertical line is calculated based on facial key points. The yaw angle is obtained by calculating the angle between the line connecting the tip of the nose and the center points of both eyes and a vertical line based on facial key points. The roll angle is obtained by calculating the angle between the line connecting the center points of both eyes and a horizontal line based on facial key points.

[0038] In step S330 of some embodiments, the alignment flag parameter is a parameter used to indicate whether to perform face alignment on the original face region. The alignment flag parameter includes a first parameter or a second parameter. The first parameter indicates that face alignment is performed, and the second parameter indicates that face alignment is not performed. The system detects whether the positions of facial key points such as the eyes, nose, and mouth conform to the standard pose, and whether the face pose angle is within the allowable range. If the positions of the facial key points do not conform to the standard pose or the face pose angle is not within the allowable range, then the alignment flag parameter is determined to be the first parameter. If the positions of the facial key points conform to the standard pose and the face pose angle is within the allowable range, then the alignment flag parameter is determined to be the second parameter.

[0039] Specifically, it is determined whether the center points of both eyes are on the same horizontal line. If the result is yes, the first offset distance of the nose tip from the vertical center line is calculated, and the alignment mark parameter is determined based on the first offset distance. If the result is no, the alignment mark parameter is determined to be the first parameter.

[0040] The process of determining the alignment flag parameter based on the first offset distance is as follows: If the first offset distance is greater than a preset distance threshold, then the alignment flag parameter is determined to be the first parameter. The preset distance threshold can be set according to the actual situation, such as 5% of the image width. If the first offset distance is less than or equal to the preset distance threshold, then the second offset distance of the mouth center point from the vertical center line is calculated, and the alignment flag parameter is determined based on the second offset distance.

[0041] The process of determining the alignment flag parameter based on the second offset distance is as follows: If the second offset distance is greater than a preset distance threshold, the alignment flag parameter is determined to be the first parameter. If the second offset distance is less than or equal to the preset distance threshold, it is determined whether the pitch angle, yaw angle, and roll angle are greater than or equal to the pitch threshold, yaw angle, and roll angle are respectively checked. If at least one of these checks is correct, the alignment flag parameter is determined to be the first parameter. If all three checks are incorrect, the alignment flag parameter is determined to be the second parameter.

[0042] In step S340 of some embodiments, if the alignment flag parameter is the first parameter, then the original face region is aligned according to the face pose angle to obtain the aligned face region. If the alignment flag parameter is the second parameter, then there is no need to align the original face region, and the face features of the original face region can be directly extracted.

[0043] Through the above steps S310 to S340, the face can be adjusted to a standard pose and position, thereby improving the accuracy of facial feature extraction.

[0044] Please see Figure 4 In some embodiments, step S340 may include, but is not limited to, steps S410 to S440: Step S410: Extract the illumination features of the original face region based on the alignment marker parameters; Step S420: Determine the face alignment category based on the lighting features and face pose angle; wherein, the face alignment category includes model alignment category or key point alignment category; Step S430: Call the face generation model according to the model alignment category, and perform face alignment on the original face region using the called face generation model to obtain the aligned face region; Step S440: Perform face alignment on the original face region according to the key point alignment category and face key points to obtain the aligned face region.

[0045] In step S410 of some embodiments, if the alignment flag parameter is the first parameter, then the illumination features of the original face region are extracted. The illumination features are feature representations that reflect the illumination conditions of the face image.

[0046] Specifically, the average value of all pixels in the original face region is calculated to obtain the average brightness. Average brightness is used to measure illumination intensity and is directly proportional to illumination intensity. The maximum and minimum pixel values ​​of the original face region are obtained. The difference between the maximum and minimum pixel values ​​is obtained, and the sum is obtained. The difference is used as the numerator, and the sum is used as the denominator. The ratio between the difference and the sum is calculated to obtain the contrast. Contrast is used to measure illumination uniformity; the lower the contrast, the more uneven the illumination. Edge detection or thresholding is performed on the original face region to obtain the shadow region, and the proportion of the shadow region to the original face region is calculated. The shadow region is the set of pixels with significantly lower illumination intensity than the surrounding area. The proportion of the shadow region is used to determine whether the illumination conditions are complex. A large proportion of the shadow region indicates complex illumination conditions. The original face region is converted to HSV space to obtain saturation and brightness. The coefficients of variation for saturation and brightness are calculated. The coefficient of variation is the ratio between the standard deviation and the mean. The coefficients of variation for saturation and brightness are calculated based on the standard deviation and mean of saturation and brightness, respectively. The coefficient of variation is used to measure the uniformity of color distribution. The larger the coefficient of variation, the less uniform the color distribution.

[0047] Average brightness, contrast, proportion of shadow areas, coefficient of variation of saturation, and coefficient of variation of brightness are used as lighting characteristics.

[0048] In step S420 of some embodiments, a first brightness threshold and a second brightness threshold are set. The first brightness threshold is a lower limit value, such as 50, and the second brightness threshold is an upper limit value, such as 220. The values ​​of the first brightness threshold and the second brightness threshold are within the range of [0, 255], and the first brightness threshold is less than the second brightness threshold. If the average brightness is less than the first brightness threshold, it indicates that the lighting in the original face area is dark; or if the average brightness is greater than the second brightness threshold, it indicates that the lighting in the original face area is overexposed. In this case, the lighting conditions are considered complex, and the brightness score is set to 1. If the average brightness is greater than or equal to the first brightness threshold and less than or equal to the second brightness threshold, the lighting conditions are considered simple, and the brightness score is set to 0.

[0049] Set a contrast threshold, with a value in the range [0,1], such as 0.4. If the contrast is greater than or equal to the contrast threshold, it indicates high contrast and a clear image, suggesting simple lighting conditions, and the contrast score is set to 0. If the contrast is less than the contrast threshold, it indicates low contrast and potential issues such as fog or backlighting, suggesting complex lighting conditions, and the contrast score is set to 1.

[0050] Set a shadow ratio threshold, with a value in the range [0,1], such as 0.3. If the proportion of the shadow area is greater than or equal to the shadow ratio threshold, it indicates a high shadow ratio, and the lighting conditions are considered complex, so the shadow score is set to 1. If the proportion of the shadow area is less than the shadow ratio threshold, it indicates a low shadow ratio, and the lighting conditions are considered simple, so the shadow score is set to 0.

[0051] We set a threshold for the coefficient of variation (COP) of saturation to obtain the first COP threshold, and set a threshold for the COP of brightness to obtain the second COP threshold. If the COP of saturation is less than the first COP threshold and the COP of brightness is less than the second COP threshold, it indicates that the color distribution is uniform, and the lighting conditions are considered simple, so the color score is set to 0. If the COP of saturation is greater than or equal to the first COP threshold, or the COP of brightness is greater than or equal to the second COP threshold, it indicates that the color distribution is uneven, and the lighting conditions are considered complex, so the color score is set to 1.

[0052] The luminance score, contrast score, shadow score, and color score are summed to obtain the illumination score. An illumination threshold is set, with a value within the range [1,4], such as 3. If the illumination score is greater than or equal to the illumination threshold, the lighting conditions are complex. If the illumination score is less than the illumination threshold, the lighting conditions are simple.

[0053] If the absolute value of the pitch angle is greater than or equal to the preset pitch threshold, the absolute value of the yaw angle is greater than or equal to the preset yaw threshold, or the absolute value of the roll angle is greater than or equal to the preset roll threshold, the deviation angle is considered large. The preset pitch threshold can be 20°, the preset yaw threshold can be 30°, and the preset roll threshold can be 15°. When the absolute value of the pitch angle is ≥20°, it indicates excessive head tilting or lowering, with the chin or forehead obscured. When the absolute value of the yaw angle is ≥30°, it indicates obvious left and right side profiles, with half of the face not visible. When the absolute value of the roll angle is ≥15°, it indicates in-face tilt, with the line connecting the eyes clearly not horizontal. If the absolute value of the pitch angle is less than the preset pitch threshold, the absolute value of the yaw angle is less than the preset yaw threshold, and the absolute value of the roll angle is less than the preset roll threshold, the deviation angle is considered small.

[0054] When lighting conditions are complex or the deviation angle is large, it indicates that the image environment of the original face region is complex. To improve the accuracy of face alignment, the face alignment category is determined to be the model alignment category. When lighting conditions are simple and the deviation angle is small, it indicates that the image environment of the original face region is simple. To improve the efficiency of face alignment, the face alignment category is determined to be the keypoint alignment category.

[0055] In step S430 of some embodiments, if the face alignment category is the model alignment category, then the face generation model is invoked to perform face alignment on the original face region, thereby generating a high-quality face-aligned image under complex face poses and lighting conditions, and obtaining the aligned face region. The face generation model can employ generative adversarial networks, diffusion models, convolutional neural networks, etc.

[0056] In step S440 of some embodiments, if the face alignment category is a keypoint alignment category, the angle between the line connecting the center points of the two eyes and the horizontal line is calculated based on the facial keypoints to obtain the rotation angle. The midpoint of the center points of the two eyes is calculated to obtain the rotation center. A two-dimensional affine transformation matrix is ​​constructed based on the rotation angle and the rotation center, and the original face region is subjected to affine transformation based on the two-dimensional affine transformation matrix to obtain the aligned face region.

[0057] Steps S410 to S440 above select different face alignment methods based on the complexity of the image environment, which can effectively balance the accuracy and efficiency of face alignment.

[0058] Please see Figure 5 In some embodiments, step S230 may include, but is not limited to, steps S510 to S540: Step S510: Perform key point detection on the aligned face region to obtain the alignment key points; Step S520: Extract features from the alignment key points to obtain eye state features and expression features; Step S530: Extract features from the aligned face region to obtain skin color features and skin texture features; Step S540: Determine facial features based on eye state features, expression features, skin color features, and skin texture features.

[0059] In step S510 of some embodiments, the aligned face region is input to a pre-trained keypoint detector for keypoint detection to obtain aligned keypoints.

[0060] In step S520 of some embodiments, the eye region is obtained from the aligned face region based on the coordinates of the alignment key points. The vertical distance between the upper and lower boundaries of the eyes is calculated based on the eye region, and the horizontal distance between the left and right boundaries of the eyes is calculated based on the eye region. The eye opening degree is calculated based on the vertical distance and the ratio distance. The formula for calculating the eye opening degree is expressed as: , Where A and B are the vertical distances between the upper and lower boundaries of the left and right eyes, respectively; and C is the horizontal distance between the left and right boundaries of the left or right eye.

[0061] The eye region is converted to HSV color space, and the pixel average value of the saturation channel (S channel) is calculated to obtain the eye redness level. Eye opening and closing degree and eye redness level are used as eye state features.

[0062] Feature extraction is performed on the aligned keypoints to obtain keypoint features. These keypoint features describe the shape, posture, and facial expression changes, such as the degree of eyebrow elevation and mouth corner upturn. Expression recognition is then performed based on the keypoint features to obtain expression categories. Expression categories reflect the insured's emotions; categories can include anger, sadness, fear, etc. These expression categories are then used as expression features.

[0063] In step S530 of some embodiments, the aligned face region is converted to HSV space to obtain an intermediate face image. Lower and upper limits for hue (H), saturation (S), and brightness are defined to obtain a skin tone range. Skin tone segmentation is performed on the intermediate face image based on the skin tone range to obtain a skin tone mask. The skin tone region of the aligned face region is extracted based on the skin tone mask. The mean and standard deviation of all pixel values ​​in the skin tone region are calculated to obtain skin tone features.

[0064] Skin texture features of the aligned face region are extracted using gray-level co-occurrence matrix, local binary mode, or Gabor filter. These skin texture features are the texture features of the facial skin, such as wrinkles, fine lines, and pores.

[0065] In step S540 of some embodiments, eye state features, expression features, skin color features and skin texture features are spliced ​​or fused to obtain facial features.

[0066] Through the above steps S510 to S540, facial features of the face image can be obtained, so as to predict the health risk of the insured based on the facial features.

[0067] Please see Figure 6 In some embodiments, step S130 may include, but is not limited to, steps S610 to S630: Step S610: Perform feature fusion on facial features and object attribute features to obtain target fused features; Step S620: The health risk assessment is performed on the target fusion feature, face feature and object attribute feature respectively by the health risk prediction model to obtain the first risk score of the target fusion feature, the second risk score of the face feature and the third risk score of the object attribute feature. Step S630: Calculate the health risk score based on the first risk score, the second risk score, and the third risk score.

[0068] In step S610 of some embodiments, facial features and object attribute features are mapped to the same dimension to obtain mapped facial features and mapped attribute features. In the health risk identification process, facial features are more important than object attribute features. Facial features provide dominant information for risk identification, while object attribute features provide auxiliary information. Therefore, facial features are selected to determine query features, and object attribute features are selected to determine key and value features. Specifically, query weights are multiplied by mapped attribute features to obtain query features. Key weights are multiplied by mapped attribute features to obtain key features. Value weights are multiplied by mapped attribute features to obtain value features. Attention weights are calculated based on query features and key features. The formula for calculating attention weights is as follows: , Where X represents the query feature; Y represents the key feature; T represents the transpose operation; and d represents the feature dimension of the key feature.

[0069] The attention features are obtained by weighted summation of the attention weights and value features. The target fusion features are then obtained by adding the mapped face features and attention features in their respective positions.

[0070] In step S620 of some embodiments, the health risk prediction model is used to predict the health risk score of the insured. The higher the health risk score, the greater the probability that the insured has a health risk. The health risk score is a non-negative number. Target fusion features can be input into the health risk prediction model to assess the health risk of the insured, obtain a first risk score, and use the first risk score as the health risk score.

[0071] To improve the accuracy of health risk prediction, a multi-feature index is used for health risk assessment. Specifically, facial features are input into the health risk prediction model for risk assessment, resulting in a second risk score. Object attribute features are then input into the health risk prediction model for risk assessment, resulting in a third risk score.

[0072] In step S630 of some embodiments, the average of the first risk score, the second risk score, and the third risk score is calculated to obtain a health risk score. If the first risk score, the second risk score, and the third risk score are represented as F1, F2, and F3, respectively, then the health risk score is represented as (F1+F2+F3) / 3, where / represents a division operation.

[0073] Through steps S610 to S630, the health risk score of the insured can be obtained, which can automatically identify the health status of the insured based on the health risk score, thereby improving underwriting efficiency and reducing the payout risk of the insurance institution.

[0074] Please see Figure 7In some embodiments, step S140 may include, but is not limited to, steps S710 to S730: Step S710: Determine the risk category based on the health risk score; Step S720: Calculate the trigger probability of the investigation based on the health risk score and risk category; Step S730: Obtain health certificate materials based on the trigger probability of the contract.

[0075] In step S710 of some embodiments, a first risk threshold and a second risk threshold are set, both of which are positive numbers, with the first risk threshold being less than the second risk threshold. If the health risk score is less than the first risk threshold, the risk category is determined to be low risk. If the health risk score is greater than or equal to the first risk threshold and less than the second risk threshold, the risk category is determined to be medium risk. If the health risk score is greater than or equal to the second risk threshold, the risk category is determined to be high risk.

[0076] In step S720 of some embodiments, if the risk category is low risk, there is no need to trigger the investigation process, and the review result of the insurance application is determined to be approved. The insured may provide inaccurate health information due to negligence or intentional concealment of their health condition. To ensure the accuracy of health information, if the risk category is medium or high risk, the investigation trigger probability is calculated based on the health risk score. The investigation process is then triggered based on this probability, requiring the insured to provide further health documentation. The formula for calculating the investigation trigger probability is as follows: , Where p represents the trigger probability; x is the health risk score; w is the weight parameter; b is the bias parameter; and T represents the transpose operation.

[0077] In step S730 of some embodiments, if the probability of triggering the investigation is greater than or equal to a preset probability threshold, the investigation process is triggered to obtain the health certificate materials of the insured. The health certificate materials are documents used to prove an individual's health status, such as medical examination reports, medical records, etc. If the probability of triggering the investigation is less than the preset probability threshold, the review result of the insurance application is determined to be approved. The preset probability threshold is located in the range [0,1], such as 0.4.

[0078] Steps S710 to S730 above identify high-risk individuals in advance and trigger the due diligence process to obtain health certificates, thereby ensuring the authenticity and reliability of the insured's health status and reducing the insurance company's payout risk.

[0079] In step S150 of some embodiments, the health certificate materials are vectorized to obtain health certificate features. The target fusion features and health certificate features are concatenated and input into a health risk prediction model to assess the health risk of the insured, obtaining a reference risk score. The reference risk score is used as the target risk score, and the health risk score is updated to the target risk score.

[0080] Alternatively, health certificate features can be input into a health risk prediction model to assess the health risk of the insured, resulting in a fourth risk score. The target risk score is then calculated by averaging the second, third, and fourth risk scores.

[0081] In step S160 of some embodiments, the insurance application is reviewed in real time based on the target risk score. Specifically, if the target risk score is greater than or equal to a preset scoring threshold, it indicates that the health risk of the insured is high, and the review result of the insurance application is that the application is rejected. If the target risk score is less than the preset scoring threshold, it indicates that the health risk of the insured is low, and the review result of the insurance application is that the application is approved.

[0082] Please see Figure 8 This application also provides an underwriting device that can implement the above-described underwriting method. The underwriting device includes: The request acquisition module 810 is used to respond to the insurance application request of the insured object and obtain the insurance type and object attribute characteristics of the insured object from the insurance application request; The face recognition module 820 is used to obtain the face image of the insured according to the type of insurance, and to perform face recognition on the face image to obtain face features; The assessment module 830 is used to conduct health risk assessment based on facial features and object attribute features to obtain the health risk score of the insured object. The material acquisition module 840 is used to obtain the health certificate materials of the insured based on the health risk score. The update module 850 is used to update the health risk score based on the health certificate materials to obtain the target risk score; The underwriting module 860 is used to review insurance applications based on the target risk score.

[0083] The specific implementation of this underwriting device is basically the same as the specific implementation of the underwriting method described above, and will not be repeated here.

[0084] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned underwriting method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0085] Please see Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 910 can be implemented using a general-purpose central processing unit (CPU), microprocessor, application specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 920 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 920 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 920 and is called and executed by the processor 910 using the underwriting method of the embodiments of this application. The input / output interface 930 is used to implement information input and output; The communication interface 940 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 950 transmits information between various components of the device (e.g., processor 910, memory 920, input / output interface 930, and communication interface 940); The processor 910, memory 920, input / output interface 930 and communication interface 940 are connected to each other within the device via bus 950.

[0086] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described underwriting method.

[0087] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0088] The underwriting method, underwriting device, electronic device, and computer storage medium provided in this application embodiment assess the health status of the insured by comprehensively considering multimodal features such as facial features and object attribute features. This enables automatic identification of whether the insured has health risks without the need for manual review, thus improving underwriting efficiency.

[0089] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0090] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0091] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0092] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0093] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0094] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0095] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0096] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0097] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0098] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0099] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. An underwriting method, characterized in that, The method includes: In response to an insurance application request from an insured party, the insurance type and the object attribute characteristics of the insured party are obtained from the insurance application request; The facial image of the insured is obtained according to the type of insurance, and facial recognition is performed on the facial image to obtain facial features; A health risk assessment is performed based on the facial features and the object attribute features to obtain the health risk score of the insured object. Obtain the health certificate materials of the insured subject based on the health risk score; The health risk score is updated based on the health certification materials to obtain the target risk score; The insurance application is reviewed based on the target risk score.

2. The method according to claim 1, characterized in that, The step of performing face recognition on the face image to obtain face features includes: Perform face detection on the face image to obtain the original face region; The original face region is aligned to obtain the aligned face region. Feature extraction is performed on the aligned face region to obtain the face features.

3. The method according to claim 2, characterized in that, The step of aligning the original face region to obtain the aligned face region includes: Key point detection is performed on the original face region to obtain facial key points; Calculate the facial pose angle based on the aforementioned facial key points; Based on the facial key points and the facial pose angle, determine the alignment marker parameters; The original face region is aligned according to the alignment flag parameters and the face pose angle to obtain the aligned face region.

4. The method according to claim 3, characterized in that, The step of aligning the original face region according to the alignment flag parameters and the face pose angle to obtain the aligned face region includes: The illumination features of the original face region are extracted based on the alignment flag parameters; Based on the lighting characteristics and the face pose angle, a face alignment category is determined; wherein, the face alignment category includes a model alignment category or a key point alignment category; The face generation model is called according to the model alignment category, and the original face region is aligned using the called face generation model to obtain the aligned face region. The original face region is aligned according to the key point alignment category and the face key points to obtain the aligned face region.

5. The method according to claim 2, characterized in that, The step of extracting features from the aligned face region to obtain the face features includes: Key point detection is performed on the aligned face region to obtain the alignment key points; Feature extraction is performed on the alignment key points to obtain eye state features and facial expression features; Feature extraction is performed on the aligned face region to obtain skin color features and skin texture features; The facial features are determined based on the eye state features, the expression features, the skin color features, and the skin texture features.

6. The method according to any one of claims 1 to 5, characterized in that, The step of conducting a health risk assessment based on the facial features and the object attribute features to obtain a health risk score for the insured object includes: The facial features and the object attribute features are fused to obtain the target fused features; A health risk prediction model is used to assess the health risks of the target fusion feature, the facial feature, and the object attribute feature, respectively, to obtain a first risk score for the target fusion feature, a second risk score for the facial feature, and a third risk score for the object attribute feature. The health risk score is calculated based on the first risk score, the second risk score, and the third risk score.

7. The method according to any one of claims 1 to 5, characterized in that, The process of obtaining health certification materials for the insured based on the health risk score includes: The risk category is determined based on the aforementioned health risk score; Calculate the trigger probability based on the health risk score and the risk category; The health certificate materials are obtained based on the trigger probability of the matching.

8. An underwriting device, characterized in that, The device includes: The request acquisition module is used to respond to the insurance application request of the insured object and obtain the insurance type and the object attribute characteristics of the insured object from the insurance application request; The face recognition module is used to obtain the face image of the insured object according to the type of insurance, and to perform face recognition on the face image to obtain face features; The assessment module is used to conduct a health risk assessment based on the facial features and the object attribute features to obtain a health risk score for the insured object; The material acquisition module is used to acquire the health certificate materials of the insured object based on the health risk score; The update module is used to update the health risk score based on the health certificate materials to obtain the target risk score; The underwriting module is used to review the insurance application based on the target risk score.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 7.

Images