Vehicle claim settlement loss assessment multi-scene risk analysis method, device, equipment and medium
By filtering, segmenting and detecting vehicle damage images, and combining pre-trained models to generate maintenance solutions, the problem of high maintenance costs in the existing technology is solved and more efficient claims risk detection is achieved.
Patent Information
- Application Number
- CN202411531261.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-30
- Publication Date
- 2025-02-07
AI Technical Summary
The existing vehicle claims and damage calculation multi-scenario risk analysis methods have high maintenance costs, especially in business scenarios with high traffic demand.
By obtaining the set of vehicle damage images uploaded by the user, non-vehicle image filtering, component segmentation and damage detection are carried out, and the maintenance plan is generated using the pre-trained reinforcement line detection model and the depression depth model, and the current scene is judged based on the scene parameters to determine the final or comprehensive Repair plan.
It effectively reduces the consumption of machine resources, reduces maintenance costs, and improves the accuracy of claims risk detection.
Smart Images
Figure CN119398940A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the fields of machine learning technology and financial technology, and in particular to a multi-scenario risk analysis method, device, equipment and medium for vehicle claims assessment. Background Art
[0002] With the rapid development of big data, artificial intelligence, and big model technology, intelligent vehicle damage assessment services have gradually matured, and their application scenarios are becoming more and more extensive. For example, for car owners or adjusters, being able to accurately identify the damaged parts and provide corresponding repair plans is an ideal effect. For insurance companies, how to strictly control the leakage of the results reported by claims adjusters and surveyors, and judge whether there are risks and water in the process such as excessive compensation, inflated parts, and repeated claims, is crucial to protecting the company's interests.
[0003] Building different services for different scenarios can effectively improve model accuracy, risk trigger rate and risk trigger accuracy. However, when the existing multi-scenario risk analysis method for vehicle claims assessment has a high traffic demand, the cost of building a complete damage assessment system is high, which increases the operating costs of the enterprise. Summary of the invention
[0004] The present invention provides a multi-scenario risk analysis method, device, equipment and medium for vehicle claim settlement and damage assessment, the main purpose of which is to solve the problem of high maintenance cost of the existing multi-scenario risk analysis method for vehicle claim settlement and damage assessment.
[0005] First, a multi-scenario risk analysis method for vehicle claims assessment is provided, including:
[0006] Acquire a vehicle damage image set uploaded by a user, filter the vehicle damage image set for non-vehicle images to obtain a filtered image set, and perform component segmentation and damage detection on the filtered image set to obtain a segmentation result and a damage detection result;
[0007] If the segmentation result does not meet the preset conditions or the damage detection result is empty, the user is prompted that it is impossible to locate the component and detect the damage according to the filtered image set; or if the segmentation result meets the preset conditions and the damage detection result is not empty, a pre-trained rib line detection model and a dent depth model are used to convert the maintenance plan according to the segmentation result and the damage detection result to obtain a first maintenance plan;
[0008] Using a pre-trained component maintenance solution detection model to perform maintenance solution detection according to the screened image set, obtain component information and a second maintenance solution, and determine whether the current scene is a self-service claim scene or an assisted claim recording scene according to pre-acquired scene parameters;
[0009] If the current scenario is a self-service claims scenario or an assisted recording scenario, the union of the first maintenance plan and the second maintenance plan is obtained to obtain the final maintenance plan; or if the current scenario is not a self-service claims scenario or an assisted recording scenario, the union and intersection of the first maintenance plan and the second maintenance plan are obtained to obtain a comprehensive maintenance plan, and claims risk detection is performed based on the comprehensive maintenance plan.
[0010] In the second aspect, a multi-scenario risk analysis device for vehicle claim settlement and damage assessment is provided, comprising:
[0011] An image detection module is used to obtain a vehicle damage image set uploaded by a user, filter the vehicle damage image set for non-vehicle images to obtain a screening image set, and perform component segmentation and damage detection on the screening image set to obtain a segmentation result and a damage detection result;
[0012] A first solution generation module, if the segmentation result does not meet the preset conditions or the damage detection result is empty, prompting the user that it is impossible to locate the component and detect the damage according to the screening image set, or, if the segmentation result meets the preset conditions and the damage detection result is not empty, using a pre-trained rib line detection model and a dent depth model to convert the maintenance solution according to the segmentation result and the damage detection result to obtain a first maintenance solution;
[0013] A scene judgment module, using a pre-trained component repair solution detection model to perform repair solution detection according to the screening image set, obtain component information and a second repair solution, and judge whether the current scene is a self-service claim scene or an assisted claim recording scene according to pre-acquired scene parameters;
[0014] The risk detection module obtains the union of the first maintenance plan and the second maintenance plan to obtain the final maintenance plan if the current scenario is a self-service claims scenario or an assisted recording scenario, or obtains the union and intersection of the first maintenance plan and the second maintenance plan to obtain a comprehensive maintenance plan if the current scenario is not a self-service claims scenario or an assisted recording scenario, and performs claims risk detection based on the comprehensive maintenance plan.
[0015] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the steps of the multi-scenario risk analysis method for vehicle claims assessment are implemented.
[0016] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the multi-scenario risk analysis method for vehicle claims assessment are implemented.
[0017] In the scheme implemented by the above-mentioned vehicle claim damage assessment multi-scenario risk analysis method, device, computer equipment and storage medium, a set of vehicle damage images uploaded by a user can be obtained through a client, the vehicle damage image set is filtered for non-vehicle images to obtain a screened image set, and component segmentation and damage detection are performed on the screened image set to obtain a segmentation result and a damage detection result. If the segmentation result does not meet the preset conditions or the damage detection result is empty, the user is prompted that the component cannot be located and the damage detection cannot be performed based on the screened image set. If the segmentation result meets the preset conditions and the damage detection result is not empty, a pre-trained rib line detection model and a concave depth model are used to locate the component based on the segmentation result and the damage detection result. The damage detection result is converted into a maintenance plan to obtain a first maintenance plan, and a maintenance plan detection model for a pre-trained component maintenance plan is used to detect the maintenance plan according to the screening image set to obtain component information and a second maintenance plan. According to the pre-acquired scene parameters, it is determined whether the current scene is a self-service claim scene or an auxiliary recording scene. If the current scene is a self-service claim scene or an auxiliary recording scene, the union of the first maintenance plan and the second maintenance plan is obtained to obtain a final maintenance plan. If the current scene is not a self-service claim scene or an auxiliary recording scene, the union and intersection of the first maintenance plan and the second maintenance plan are obtained to obtain a comprehensive maintenance plan, and claims risk detection is performed according to the comprehensive maintenance plan. In the present invention, by adding a multi-label target detection model to the algorithm process, and adding scene-distinguishing pictures and case fusion logic to the query process, different requirements and different accuracies in different scenarios are met. The multi-label target detection model complements the original component segmentation model and the damage detection model, and can achieve effective complementary effects through different frameworks, training data, training methods, etc. Through this solution, when dealing with different high-traffic business scenarios, machine resources can be effectively saved and maintenance costs can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.
[0019] Figure 1 Schematic diagram of an application environment of a multi-scenario risk analysis method for vehicle claim assessment in one embodiment of the present invention;
[0020] Figure 2 It is a flow chart of a multi-scenario risk analysis method for vehicle claim settlement and damage assessment in one embodiment of the present invention;
[0021] Figure 3It is a structural schematic diagram of a multi-scenario risk analysis device for vehicle claim settlement and damage assessment in one embodiment of the present invention;
[0022] Figure 4 is a schematic diagram of a structure of a computer device in one embodiment of the present invention;
[0023] Figure 5 It is another structural schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION
[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0025] The multi-scenario risk analysis method for vehicle claim assessment provided by the embodiment of the present invention can be applied in the following aspects: Figure 1 In an application environment, the client obtains a vehicle damage image set uploaded by a user, filters the vehicle damage image set for non-vehicle images, obtains a screening image set, performs component segmentation and damage detection on the screening image set, obtains a segmentation result and a damage detection result, and if the segmentation result does not meet a preset condition or the damage detection result is empty, the user is prompted that the component cannot be located and the damage detection cannot be performed according to the screening image set; if the segmentation result meets the preset condition and the damage detection result is not empty, a pre-trained rib line detection model and a dent depth model are used to convert a maintenance plan according to the segmentation result and the damage detection result. , obtain the first maintenance plan, use the pre-trained component maintenance plan detection model to perform maintenance plan detection according to the screening image set, obtain component information and the second maintenance plan, and determine whether the current scene is a self-service claim scene or an auxiliary recording scene according to the pre-acquired scene parameters. If the current scene is a self-service claim scene or an auxiliary recording scene, obtain the union of the first maintenance plan and the second maintenance plan to obtain the final maintenance plan. If the current scene is not a self-service claim scene or an auxiliary recording scene, obtain the union and intersection of the first maintenance plan and the second maintenance plan to obtain a comprehensive maintenance plan, and perform claim risk detection according to the comprehensive maintenance plan. It can effectively improve the accuracy of claim risk detection. Among them, the client can be but not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices. The server can be implemented with an independent server or a server cluster composed of multiple servers. The present invention is described in detail below through specific embodiments.
[0026] See also Figure 2 As shown, Figure 2 A flowchart of a multi-scenario risk analysis method for vehicle claim assessment provided by an embodiment of the present invention includes the following steps:
[0027] S1. Obtain a vehicle damage image set uploaded by a user, filter the vehicle damage image set for non-vehicle images to obtain a filtered image set, and perform component segmentation and damage detection on the filtered image set to obtain a segmentation result and a damage detection result.
[0028] In the embodiment of the present invention, the vehicle damage image set is an image set obtained by a user photographing damaged parts of a vehicle.
[0029] In the embodiment of the present invention, the vehicle damage image set is filtered for non-vehicle images to obtain a screened image set, which is to use a lightweight classification model to roughly analyze the vehicle damage image set, filter out non-vehicle images in the vehicle damage image set, and improve data quality.
[0030] In detail, common lightweight classification models include ImageNet18, RepViT, MobileNet, and ShuffleNetV2.
[0031] In the embodiment of the present invention, filtering the vehicle damage image set for non-vehicle images to obtain a filtered image set includes:
[0032] identifying a vehicle component contour of each image in the vehicle damage image set;
[0033] Matching the vehicle component contour of each image in the vehicle damage image set with pre-acquired standard component contour data to obtain a matching result;
[0034] According to the matching result, the images that have not been successfully matched in the vehicle damage image set are removed to obtain a filtered image set.
[0035] In an embodiment of the present invention, a vehicle damage image set uploaded by a user is obtained, and non-vehicle images are filtered on the vehicle damage image set to obtain a screening image set, thereby improving the efficiency of subsequent component segmentation and damage detection.
[0036] In the embodiment of the present invention, the component segmentation and damage detection for the screened image set is performed by using a pre-trained component segmentation model and a damage detection model to perform component segmentation and damage detection for the screened image set.
[0037] In detail, the component segmentation model and the damage detection model are obtained by acquiring business data, open source data and relevant network data through crawlers, and the acquired data are annotated according to business needs and used as training data to train the component segmentation model and the damage detection model. The component segmentation model and the damage detection model are obtained after training and tuning based on the open source deeplapV3 (semantic segmentation model) and yoloV7 (target detection model).
[0038] In the embodiment of the present invention, performing component segmentation and damage detection on the screening image set to obtain a segmentation result and a damage detection result includes:
[0039] Using a pre-trained segmentation model to perform component segmentation on the screening image set to obtain a component segmentation result;
[0040] A pre-trained damage detection model is used to perform damage detection based on the screened image set to obtain a damage detection result.
[0041] In detail, the component segmentation model can segment out different vehicle components in the filtered image set, such as the front cover, front bumper, front windshield, door, wheel, etc.
[0042] In detail, the damage detection model can detect the damage forms of different components, such as scratches, abrasions, dents, cracks, etc., based on the filtered image set.
[0043] In the embodiment of the present invention, by performing component segmentation and damage detection on the screened image set, segmentation results and damage detection results are obtained, thereby improving the efficiency of subsequent maintenance plan conversion.
[0044] S2. Determine whether the segmentation result meets the preset conditions and whether the damage detection result is not empty.
[0045] If the segmentation result does not meet the preset conditions or the damage detection result is empty, execute S3, prompt the user that it is impossible to locate the component and detect damage based on the filtered image set, or, if the segmentation result meets the preset conditions and the damage detection result is not empty, execute S4, use the pre-trained rib line detection model and the depression depth model to convert the maintenance plan according to the segmentation result and the damage detection result to obtain the first maintenance plan.
[0046] In the embodiment of the present invention, when the segmentation result does not meet the preset conditions or the damage detection result is empty, it means that the damage image uploaded by the user may not be clear enough, so component segmentation and damage detection cannot be performed.
[0047] In the embodiment of the present invention, before prompting the user that it is impossible to locate the component and detect damage according to the filtered image set, the method further includes:
[0048] By counting pixels of the segmentation results, the main component with the largest area is identified;
[0049] Calculate the area ratio of the main component in the image;
[0050] If the area ratio is less than a preset first ratio threshold, it is determined that the segmentation result does not meet the preset condition;
[0051] If the area ratio is greater than or equal to the first ratio threshold, calculating the ratio of the outermost contour of the segmented component in the image according to the image segmentation result to obtain the contour ratio;
[0052] If the contour ratio is greater than a preset second ratio threshold, it is determined that the segmentation result does not meet the preset condition;
[0053] If the contour proportion is less than or equal to the second proportion threshold, it is determined that the segmentation result meets the preset condition.
[0054] In detail, a contour detection algorithm may be used to calculate the proportion of the outermost contour of the segmented component in the image.
[0055] In an embodiment of the present invention, when the segmentation result does not meet the preset conditions or the damage detection result is empty, the efficiency of the vehicle claim assessment process is improved by prompting the user that it is impossible to locate the component and detect the damage based on the filtered image set.
[0056] In the embodiment of the present invention, when the segmentation result meets the preset conditions and the damage detection result is not empty, it means that the clarity of the vehicle damage image uploaded by the user meets the requirements. The image in the filtered image set can be accurately segmented into components.
[0057] In an embodiment of the present invention, the pre-trained rib line detection model and the depression depth model are used to convert the maintenance plan according to the segmentation result and the damage detection result to obtain a first maintenance plan, the rib line detection model is used to detect the rib line position of each component in the segmentation result, and the depression depth model is used to detect the depression depth of each component, and a maintenance plan is given in combination with the rib line position and the depression depth of the component.
[0058] In the embodiment of the present invention, the use of the pre-trained rib line detection model and the dent depth model to convert the maintenance plan according to the segmentation result and the damage detection result to obtain the first maintenance plan includes:
[0059] Using the rib line detection model to detect the rib line position of each component according to the segmentation result;
[0060] Using the depression depth model to detect the depression position, depression area ratio and depression depth of each component according to the segmentation result and the damage detection result;
[0061] A targeted maintenance plan is generated according to the rib line position, depression position, depression area ratio and depression depth of each component to obtain a first maintenance plan.
[0062] In detail, generating a targeted maintenance plan based on the rib line position, the depression position, the depression area ratio and the depression depth of each component means judging whether the damage of the component is located on the rib line based on the rib line position, the depression position, the depression area ratio and the depression depth, and providing a maintenance plan in combination with the damage area ratio. For example, if the damage of the cotyledon is located on the rib line and the damage area ratio is greater than 40%, it needs to be replaced, otherwise it only needs to be repaired; if the damage of the door is not located on the rib line, it only needs to be sheet metal and painted; if the damage area of the hood is large and the depression depth is deep, it needs to be replaced, otherwise it only needs to be repaired.
[0063] In detail, the ribs are an important element in automobile design to enhance visual effects and functionality. The ribs can increase the rigidity of the cover and prevent tearing or deformation caused by wind when driving at high speeds. In the event of a collision, the rib design of the hood helps absorb and disperse the impact force, reducing damage to pedestrians. At the same time, the ribs also help control the deformation direction of the hood in a collision and protect the occupants in the car. Therefore, whether the damaged position of the vehicle component is located on the ribs is a key factor in providing a repair plan.
[0064] In an embodiment of the present invention, a pre-trained rib line detection model and a depression depth model are used to convert the maintenance plan according to the segmentation result and the damage detection result to obtain a first maintenance plan, thereby improving the efficiency of subsequently generating a final maintenance plan and a comprehensive maintenance plan.
[0065] S5. Use a pre-trained component maintenance plan detection model to perform maintenance plan detection based on the screened image set to obtain component information and a second maintenance plan, and execute S6. Determine whether the current scene is a self-service claim scene or an assisted recording scene based on pre-acquired scene parameters.
[0066] In the embodiment of the present invention, the component maintenance solution detection model is a multi-target detection model, which can directly generate component information and maintenance solution information based on the input screening image set.
[0067] In the embodiment of the present invention, before the maintenance solution detection is performed according to the screening image set using the pre-trained component maintenance solution detection model to obtain component information and the second maintenance solution, the method further includes:
[0068] Obtaining a preset number of vehicle damage images with component annotations and repair plan annotations to obtain training data;
[0069] Performing one-hot encoding on the training data to obtain encoded training data;
[0070] Replace the original loss function of the pre-built target detection model to obtain an optimized target detection model;
[0071] The optimized target detection model is trained according to the training data using a random sampling algorithm for a preset number of iterations to obtain a component maintenance solution detection model.
[0072] In detail, the one-hot encoding is a commonly used categorical data representation method that converts each category into a binary vector in which only one element is 1 and the rest are 0.
[0073] In detail, the one-hot encoding of the training data to obtain the encoded training data refers to converting the component labels and maintenance plan labels of the training data into binary labels with only one element being 1. For example, the car door can be converted into "10000", the wheel hub can be converted into "01000", and the boot can be converted into "00100"; further, the maintenance plan of "replace" can be converted into "000001", and the maintenance plan of "maintenance" can be converted into "000010".
[0074] In detail, the original loss function is usually a softmax function, which is suitable for single-label scenarios. When facing multi-label scenarios, the loss value cannot converge.
[0075] In detail, the replacement of the original loss function of the pre-built target detection model refers to replacing the original loss function with a concurrent-softmax function, which is a loss function used in multi-label classification problems that takes into account the co-occurrence probability between categories. This method is different from the traditional softmax function, which not only considers the probability that a sample belongs to a certain category, but also considers the probability of the relationship between categories. This method of considering co-occurrence probability can help the model better handle multi-label problems, especially when there is some dependency or association between categories.
[0076] In detail, the random sampling algorithm is a data preprocessing technique commonly used in machine learning and deep learning, which involves randomly selecting a portion of data from the entire dataset for training the model. This method helps improve the generalization ability of the model, prevent overfitting, and speed up training.
[0077] In the embodiment of the present invention, a pre-trained component maintenance plan detection model is used to perform maintenance plan detection according to the screening image set to obtain component information and a second maintenance plan, thereby improving the efficiency of subsequently generating a final maintenance plan and a comprehensive maintenance plan.
[0078] In the embodiment of the present invention, the pre-acquired scene parameters are acquired synchronously when acquiring the vehicle damage image uploaded by the user.
[0079] In detail, judging whether the current scene is a self-service claim scene or an auxiliary recording scene based on the pre-acquired scene parameters means that when the scene parameters are Self-service claim or Auxiliary entry, it represents that the current scene is an auxiliary claim scene or an auxiliary recording scene.
[0080] If the current scenario is a self-service claims scenario or an assisted recording scenario, execute S7, obtain the union of the first maintenance plan and the second maintenance plan, and obtain the final maintenance plan; or, if the current scenario is not a self-service claims scenario or an assisted recording scenario, execute S8, obtain the union and intersection of the first maintenance plan and the second maintenance plan, and obtain a comprehensive maintenance plan, and perform claims risk detection based on the comprehensive maintenance plan.
[0081] In the embodiment of the present invention, obtaining the union of the first maintenance plan and the second maintenance plan to obtain the final maintenance plan is to obtain all detected components and their maintenance plans, which can improve the risk detection probability and risk triggering accuracy when performing repeated claims detection.
[0082] In the embodiment of the present invention, when the current scene is not a self-service claims scene or an assisted order recording scene, it indicates that this is the original leakage control scene.
[0083] In the embodiment of the present invention, the step of obtaining the union and intersection of the first maintenance plan and the second maintenance plan to obtain a comprehensive maintenance plan includes:
[0084] Obtaining the intersection of the first maintenance plan and the second maintenance plan to obtain an accurate maintenance plan;
[0085] Obtaining a union of the first maintenance plan and the second maintenance plan to obtain a union maintenance plan;
[0086] Performing a fusion logic on the combined maintenance plan according to the identification result of each component included in the component information to obtain a fusion maintenance plan;
[0087] The precise maintenance plan, the combined maintenance plan and the fusion maintenance plan are summarized to obtain a comprehensive maintenance plan.
[0088] In detail, the union maintenance plan is logically integrated according to the recognition results of each component contained in the component information to obtain an integrated maintenance plan, which means confirming the model processing results of each individual image according to the component information, and then fusing the model processing results of each image (such as taking the most serious damage of all components) to obtain the final case-level maintenance plan return result.
[0089] In the embodiment of the present invention, the claim risk detection according to the comprehensive maintenance plan refers to determining whether there is duplicate claims based on the user's claim record and the comprehensive maintenance plan.
[0090] In an embodiment of the present invention, when the current scenario is a self-service claims scenario or an assisted recording scenario, a comprehensive repair plan is obtained by obtaining the union and intersection of the first repair plan and the second repair plan, thereby improving the efficiency of multi-scenario risk analysis of vehicle claims and damage assessment.
[0091] It can be seen that in the embodiment of the present invention, by adding a multi-label target detection model to the algorithm process and adding scene-distinguishing pictures and case fusion logic to the query process, different requirements and different levels of accuracy in different scenarios can be met. The multi-label target detection model complements the original component segmentation model and damage detection model, and can achieve effective complementary effects through different frameworks, training data, training methods, etc. Through this solution, when dealing with different high-traffic business scenarios, it can effectively save machine resources and reduce maintenance costs.
[0092] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.
[0093] In one embodiment, a multi-scenario risk analysis device for vehicle claim assessment is provided, and the multi-scenario risk analysis device for vehicle claim assessment corresponds one-to-one to the multi-scenario risk analysis method for vehicle claim assessment in the above embodiment. Figure 3 As shown, the vehicle claim settlement multi-scenario risk analysis device includes a data screening module 101, an image detection module 102, a first solution generation module 103, a scenario judgment module 104, and a final solution generation module 105. The functional modules are described in detail as follows:
[0094] The image detection module 101 is used to obtain a vehicle damage image set uploaded by a user, filter the vehicle damage image set for non-vehicle images to obtain a filtered image set, and perform component segmentation and damage detection on the filtered image set to obtain a segmentation result and a damage detection result;
[0095] The first solution generation module 102 prompts the user that it is impossible to locate the component and detect damage according to the screened image set if the segmentation result does not meet the preset conditions or the damage detection result is empty, or, if the segmentation result meets the preset conditions and the damage detection result is not empty, converts the maintenance solution according to the segmentation result and the damage detection result using a pre-trained rib line detection model and a dent depth model to obtain a first maintenance solution;
[0096] The scene judgment module 103 uses a pre-trained component maintenance solution detection model to perform maintenance solution detection according to the screening image set, obtains component information and a second maintenance solution, and judges whether the current scene is a self-service claim scene or an auxiliary order recording scene according to pre-acquired scene parameters;
[0097] The risk detection module 104, if the current scenario is a self-service claims scenario or an assisted recording scenario, obtains the union of the first maintenance plan and the second maintenance plan to obtain a final maintenance plan, or, if the current scenario is not a self-service claims scenario or an assisted recording scenario, obtains the union and intersection of the first maintenance plan and the second maintenance plan to obtain a comprehensive maintenance plan, and performs claims risk detection based on the comprehensive maintenance plan.
[0098] In one embodiment, the image detection module 101, when performing non-vehicle image filtering on the vehicle damage image set to obtain the filtered image set, is used to:
[0099] identifying a vehicle component contour of each image in the vehicle damage image set;
[0100] Matching the vehicle component contour of each image in the vehicle damage image set with pre-acquired standard component contour data to obtain a matching result;
[0101] According to the matching result, the images that have not been successfully matched in the vehicle damage image set are removed to obtain a filtered image set.
[0102] In one embodiment, the image detection module 101, when performing component segmentation and damage detection on the screening image set to obtain the segmentation result and the damage detection result, is used to:
[0103] Using a pre-trained segmentation model to perform component segmentation on the screening image set to obtain a component segmentation result;
[0104] A pre-trained damage detection model is used to perform damage detection based on the screened image set to obtain a damage detection result.
[0105] In one embodiment, the first solution generating module 102, before prompting the user that it is impossible to locate the component and detect damage according to the filtered image set, is used to:
[0106] By counting pixels of the segmentation results, the main component with the largest area is identified;
[0107] Calculate the area ratio of the main component in the image;
[0108] If the area ratio is less than a preset first ratio threshold, it is determined that the segmentation result does not meet the preset condition;
[0109] If the area ratio is greater than or equal to the first ratio threshold, calculating the ratio of the outermost contour of the segmented component in the image according to the image segmentation result to obtain the contour ratio;
[0110] If the contour ratio is greater than a preset second ratio threshold, it is determined that the segmentation result does not meet the preset condition;
[0111] If the contour proportion is less than or equal to the second proportion threshold, it is determined that the segmentation result meets the preset condition.
[0112] In one embodiment, the first solution generating module 102, when converting the maintenance solution according to the segmentation result and the damage detection result using the pre-trained rib line detection model and the dent depth model to obtain the first maintenance solution, is used to:
[0113] Detecting the position of the ribs of each component according to the segmentation result using the rib detection model;
[0114] Using the depression depth model to detect the depression position, depression area ratio and depression depth of each component according to the segmentation result and the damage detection result;
[0115] A targeted maintenance plan is generated according to the rib line position, depression position, depression area ratio and depression depth of each component to obtain a first maintenance plan.
[0116] In one embodiment, the scene determination module 103, before using the pre-trained component maintenance solution detection model to perform maintenance solution detection according to the screening image set to obtain component information and the second maintenance solution, is used to:
[0117] Obtaining a preset number of vehicle damage images with component annotations and repair plan annotations to obtain training data;
[0118] Performing one-hot encoding on the training data to obtain encoded training data;
[0119] Replace the original loss function of the pre-built target detection model to obtain an optimized target detection model;
[0120] The optimized target detection model is trained according to the training data using a random sampling algorithm for a preset number of iterations to obtain a component maintenance solution detection model.
[0121] In one embodiment, when the risk detection module 104 obtains the union and intersection of the first maintenance plan and the second maintenance plan to obtain a comprehensive maintenance plan, it is used to:
[0122] Obtaining the intersection of the first maintenance plan and the second maintenance plan to obtain an accurate maintenance plan;
[0123] Obtaining a union of the first maintenance plan and the second maintenance plan to obtain a union maintenance plan;
[0124] Performing a fusion logic on the combined maintenance plan according to the identification result of each component included in the component information to obtain a fusion maintenance plan;
[0125] The precise maintenance plan, the combined maintenance plan and the fusion maintenance plan are summarized to obtain a comprehensive maintenance plan.
[0126] The present invention provides a multi-scenario risk analysis device for vehicle claims and damage assessment. Based on the basic idea of a tree model, the parent node of the tree is used as the main operating unit, and its left and right child nodes are used as the basic branches for execution success / failure. The left and right child nodes are automatically selected according to the operation status of the parent node (success or failure). The left and right child nodes are respectively subtrees that meet the same design structure. Through the nesting of trees, the originally complex jump process is converted into the selection of tree nodes with a clear nested structure. Combined with the processing of iterators, complex logic that meets different jump conditions can be flexibly designed, because changes in nodes do not change the stability of the tree structure itself. Compared with the traditional if-else structure, the tree mode proposed in the embodiment of the present invention can effectively maintain the structural stability of the code, improve the maintainability of the code, and greatly improve the processing efficiency of complex processes.
[0127] For the specific limitations of the multi-scenario risk analysis device for vehicle claims and damage assessment, please refer to the limitations of the intelligent question and answer processing method above, which will not be repeated here. Each module in the above-mentioned multi-scenario risk analysis device for vehicle claims and damage assessment can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0128] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the service side of a multi-scenario risk analysis method for vehicle claims and damage assessment.
[0129] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps on the client side of a multi-scenario risk analysis method for vehicle claim assessment
[0130] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:
[0131] Acquire a vehicle damage image set uploaded by a user, filter the vehicle damage image set for non-vehicle images to obtain a filtered image set, and perform component segmentation and damage detection on the filtered image set to obtain a segmentation result and a damage detection result;
[0132] If the segmentation result does not meet the preset conditions or the damage detection result is empty, the user is prompted that it is impossible to locate the component and detect the damage according to the filtered image set; or if the segmentation result meets the preset conditions and the damage detection result is not empty, a pre-trained rib line detection model and a dent depth model are used to convert the maintenance plan according to the segmentation result and the damage detection result to obtain a first maintenance plan;
[0133] Using a pre-trained component maintenance solution detection model to perform maintenance solution detection according to the screened image set, obtain component information and a second maintenance solution, and determine whether the current scene is a self-service claim scene or an assisted claim recording scene according to pre-acquired scene parameters;
[0134] If the current scenario is a self-service claims scenario or an assisted recording scenario, the union of the first maintenance plan and the second maintenance plan is obtained to obtain the final maintenance plan; or if the current scenario is not a self-service claims scenario or an assisted recording scenario, the union and intersection of the first maintenance plan and the second maintenance plan are obtained to obtain a comprehensive maintenance plan, and claims risk detection is performed based on the comprehensive maintenance plan.
[0135] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
[0136] Acquire a vehicle damage image set uploaded by a user, filter the vehicle damage image set for non-vehicle images to obtain a filtered image set, and perform component segmentation and damage detection on the filtered image set to obtain a segmentation result and a damage detection result;
[0137] If the segmentation result does not meet the preset conditions or the damage detection result is empty, the user is prompted that it is impossible to locate the component and detect the damage according to the filtered image set; or if the segmentation result meets the preset conditions and the damage detection result is not empty, a pre-trained rib line detection model and a dent depth model are used to convert the maintenance plan according to the segmentation result and the damage detection result to obtain a first maintenance plan;
[0138] Using a pre-trained component maintenance solution detection model to perform maintenance solution detection according to the screened image set, obtain component information and a second maintenance solution, and determine whether the current scene is a self-service claim scene or an assisted claim recording scene according to pre-acquired scene parameters;
[0139] If the current scenario is a self-service claims scenario or an assisted recording scenario, the union of the first maintenance plan and the second maintenance plan is obtained to obtain the final maintenance plan; or if the current scenario is not a self-service claims scenario or an assisted recording scenario, the union and intersection of the first maintenance plan and the second maintenance plan are obtained to obtain a comprehensive maintenance plan, and claims risk detection is performed based on the comprehensive maintenance plan.
[0140] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.
[0141] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0142] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0143] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.
[0144] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
[0145] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.
[0146] Finally, it should be noted that if software tools or components other than those of the company appear in the application embodiments, they are only used for illustration and do not represent actual use. The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the above embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features can be replaced by equivalents; and these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.
Claims
1. A multi-scenario risk analysis method for vehicle claim settlement, characterized in that: include: Acquire a vehicle damage image set uploaded by a user, filter the vehicle damage image set for non-vehicle images to obtain a filtered image set, and perform component segmentation and damage detection on the filtered image set to obtain a segmentation result and a damage detection result; If the segmentation result does not meet the preset conditions or the damage detection result is empty, the user is prompted that it is impossible to locate the component and detect the damage according to the filtered image set; or, if the segmentation result meets the preset conditions and the damage detection result is not empty, a pre-trained rib line detection model and a dent depth model are used to convert the maintenance plan according to the segmentation result and the damage detection result to obtain a first maintenance plan; Using a pre-trained component maintenance solution detection model to perform maintenance solution detection according to the screened image set, obtain component information and a second maintenance solution, and determine whether the current scene is a self-service claim scene or an assisted claim recording scene according to pre-acquired scene parameters; If the current scenario is a self-service claims scenario or an assisted recording scenario, the union of the first maintenance plan and the second maintenance plan is obtained to obtain the final maintenance plan. Alternatively, if the current scenario is not a self-service claims scenario or an assisted recording scenario, the union and intersection of the first maintenance plan and the second maintenance plan are obtained to obtain a comprehensive maintenance plan, and claims risk detection is performed based on the comprehensive maintenance plan.
2. The multi-scenario risk analysis method for vehicle claim assessment as claimed in claim 1, characterized in that: The step of filtering the vehicle damage image set for non-vehicle images to obtain a filtered image set includes: identifying a vehicle component contour of each image in the vehicle damage image set; Matching the vehicle component contour of each image in the vehicle damage image set with pre-acquired standard component contour data to obtain a matching result; According to the matching result, the images that have not been successfully matched in the vehicle damage image set are removed to obtain a filtered image set.
3. The multi-scenario risk analysis method for vehicle claim settlement and damage assessment as claimed in claim 1, characterized in that: The performing component segmentation and damage detection on the screening image set to obtain a segmentation result and a damage detection result includes: Using a pre-trained segmentation model to perform component segmentation on the screening image set to obtain a component segmentation result; A pre-trained damage detection model is used to perform damage detection based on the screened image set to obtain a damage detection result.
4. The multi-scenario risk analysis method for vehicle claim settlement and damage assessment as claimed in claim 1, characterized in that: Before prompting the user that it is impossible to locate the component and detect damage according to the filtered image set, the method further includes: By counting pixels of the segmentation results, the main component with the largest area is identified; Calculate the area ratio of the main component in the image; If the area ratio is less than a preset first ratio threshold, it is determined that the segmentation result does not meet the preset condition; If the area ratio is greater than or equal to the first ratio threshold, calculating the ratio of the outermost contour of the segmented component in the image according to the segmentation result to obtain the contour ratio; If the contour ratio is greater than a preset second ratio threshold, it is determined that the segmentation result does not meet the preset condition; If the contour proportion is less than or equal to the second proportion threshold, it is determined that the segmentation result meets the preset condition.
5. The multi-scenario risk analysis method for vehicle claim settlement and damage assessment as claimed in claim 1, characterized in that: The method of converting a maintenance plan according to the segmentation result and the damage detection result using the pre-trained rib line detection model and the dent depth model to obtain a first maintenance plan includes: Detecting the position of the ribs of each component according to the segmentation result using the rib detection model; Using the depression depth model to detect the depression position, depression area ratio and depression depth of each component according to the segmentation result and the damage detection result; A targeted maintenance plan is generated according to the rib line position, depression position, depression area ratio and depression depth of each component to obtain a first maintenance plan.
6. The multi-scenario risk analysis method for vehicle claim assessment as claimed in claim 1, characterized in that: Before performing maintenance solution detection according to the screening image set using the pre-trained component maintenance solution detection model to obtain component information and a second maintenance solution, the method further includes: Obtaining a preset number of vehicle damage images with component annotations and repair plan annotations to obtain training data; Performing one-hot encoding on the training data to obtain encoded training data; Replace the original loss function of the pre-built target detection model to obtain an optimized target detection model; The optimized target detection model is trained according to the training data using a random sampling algorithm for a preset number of iterations to obtain a component maintenance solution detection model.
7. The multi-scenario risk analysis method for vehicle claim settlement and damage assessment as claimed in claim 1, characterized in that: The obtaining of the union and intersection of the first maintenance plan and the second maintenance plan to obtain a comprehensive maintenance plan includes: Obtaining the intersection of the first maintenance plan and the second maintenance plan to obtain an accurate maintenance plan; Obtaining a union of the first maintenance plan and the second maintenance plan to obtain a union maintenance plan; Performing a fusion logic on the combined maintenance plan according to the identification result of each component included in the component information to obtain a fusion maintenance plan; The precise maintenance plan, the combined maintenance plan and the fusion maintenance plan are summarized to obtain a comprehensive maintenance plan.
8. A multi-scenario risk analysis device for vehicle claim settlement and damage assessment, characterized in that: include: An image detection module is used to obtain a vehicle damage image set uploaded by a user, filter the vehicle damage image set for non-vehicle images to obtain a screening image set, and perform component segmentation and damage detection on the screening image set to obtain a segmentation result and a damage detection result; A first solution generation module, if the segmentation result does not meet the preset conditions or the damage detection result is empty, prompting the user that it is impossible to locate the component and detect the damage according to the screening image set, or, if the segmentation result meets the preset conditions and the damage detection result is not empty, using a pre-trained rib line detection model and a dent depth model to convert the maintenance solution according to the segmentation result and the damage detection result to obtain a first maintenance solution; A scene judgment module, using a pre-trained component repair solution detection model to perform repair solution detection according to the screening image set, obtain component information and a second repair solution, and judge whether the current scene is a self-service claim scene or an assisted claim recording scene according to pre-acquired scene parameters; The risk detection module obtains the union of the first maintenance plan and the second maintenance plan to obtain the final maintenance plan if the current scenario is a self-service claims scenario or an assisted recording scenario, or obtains the union and intersection of the first maintenance plan and the second maintenance plan to obtain a comprehensive maintenance plan if the current scenario is not a self-service claims scenario or an assisted recording scenario, and performs claims risk detection based on the comprehensive maintenance plan.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the multi-scenario risk analysis method for vehicle claim assessment are implemented as described in 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, the steps of the multi-scenario risk analysis method for vehicle claim assessment are implemented as described in any one of claims 1 to 7.