Vehicle accident loss assessment method and device, computer equipment and storage medium

Abstract

The embodiment of the application belongs to the field of artificial intelligence, and relates to a vehicle accident loss assessment method and device, a computer device and a storage medium. The method comprises the following steps: acquiring multi-modal description information of a vehicle accident; inputting report information, survey information and accident pictures in the multi-modal description information into an accident feature extraction model to obtain vehicle accident features; inputting the accident pictures and vehicle-mounted data into an object loss assessment model to obtain object loss features; inputting medical diagnosis information into a medical diagnosis assessment model to obtain human injury diagnosis features; fusing the vehicle accident features, the object loss features and the human injury diagnosis features to obtain case features; classifying the vehicle accident based on the case features, the object loss features and the human injury diagnosis features to obtain a vehicle accident classification result, and then inputting the vehicle accident classification result into a vehicle accident loss assessment model to obtain a loss assessment value. The application also relates to blockchain technology, and the multi-modal description information can be stored in the blockchain. The application improves the accuracy of vehicle accident loss assessment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a vehicle accident loss assessment method and device, a computer device and a storage medium. BACKGROUND

[0002] In the field of finance and insurance, when an insurance company receives a claim report, it needs to assess the loss of the case and obtain a loss assessment value. The insurance company needs to prepare corresponding resources based on the loss assessment value. Accurate loss assessment value is very important for the insurance company. If the resources are prepared too much, the operating cost of the insurance company will increase; if the resources are insufficient, it will cause disputes and operational risks. In vehicle accidents, the accident situation is complex, and damage to vehicles, vehicle-mounted objects and personnel casualties may occur, which further increases the difficulty of vehicle accident loss assessment.

[0003] Current vehicle accident loss assessment usually relies on loss assessors to conduct full-time investigation and loss assessment, which needs to assess the accident scene, accident process, damaged objects and personnel casualties. However, due to the diversity of accident types and the difficulty of accurately measuring the degree of accidents, especially in cases involving personnel casualties, the accuracy of vehicle loss assessment relying on manual experience is low. SUMMARY

[0004] The purpose of the embodiments of the present application is to provide a vehicle accident loss assessment method, device, computer device and storage medium to improve the accuracy of vehicle accident loss assessment.

[0005] To solve the above technical problems, the embodiments of the present application provide a vehicle accident loss assessment method, which adopts the following technical solutions:

[0006] Obtain multi-modal description information of a vehicle accident;

[0007] Input the claim information, investigation information and accident pictures in the multi-modal description information into an accident feature extraction model to obtain vehicle accident features;

[0008] Input the accident pictures and vehicle-mounted data in the multi-modal description information into an object loss assessment model to obtain object loss features;

[0009] Input the medical diagnosis information in the multi-modal description information into a medical diagnosis assessment model to obtain personal injury diagnosis features;

[0010] Fuse the vehicle accident features, the object loss features and the personal injury diagnosis features to obtain case features;

[0011] Classify the vehicle accident based on the case features, the object loss features and the personal injury diagnosis features to obtain a vehicle accident classification result;

[0012] inputting the vehicle accident classification result into a vehicle accident loss assessment model to obtain a loss assessment value.

[0013] To solve the above technical problems, the embodiment of the application further provides a vehicle accident loss assessment device, which adopts the technical scheme as follows:

[0014] a description information acquisition module, configured to acquire multi-modal description information of a vehicle accident;

[0015] an accident feature extraction module, configured to input report information, investigation information and accident pictures in the multi-modal description information into an accident feature extraction model to obtain vehicle accident features;

[0016] an article loss assessment module, configured to input accident pictures and vehicle-mounted data in the multi-modal description information into an article loss assessment model to obtain article loss features;

[0017] a medical diagnosis assessment module, configured to input medical diagnosis information in the multi-modal description information into a medical diagnosis assessment model to obtain human injury diagnosis features;

[0018] a case feature generation module, configured to fuse the vehicle accident features, the article loss features and the human injury diagnosis features to obtain case features;

[0019] a vehicle accident classification module, configured to classify the vehicle accident based on the case features, the article loss features and the human injury diagnosis features to obtain a vehicle accident classification result;

[0020] an accident loss assessment module, configured to input the vehicle accident classification result into a vehicle accident loss assessment model to obtain a loss assessment value.

[0021] To solve the above technical problems, the embodiment of the application further provides a computer device, which adopts the technical scheme as follows:

[0022] acquiring multi-modal description information of a vehicle accident;

[0023] inputting report information, investigation information and accident pictures in the multi-modal description information into an accident feature extraction model to obtain vehicle accident features;

[0024] inputting accident pictures and vehicle-mounted data in the multi-modal description information into an article loss assessment model to obtain article loss features;

[0025] inputting medical diagnosis information in the multi-modal description information into a medical diagnosis assessment model to obtain human injury diagnosis features;

[0026] Fusing the vehicle accident feature, the article loss feature and the person injury diagnosis feature to obtain a case feature;

[0027] Classifying the vehicle accident based on the case feature, the article loss feature and the person injury diagnosis feature to obtain a vehicle accident classification result;

[0028] Inputting the vehicle accident classification result into a vehicle accident loss assessment model to obtain a loss assessment value.

[0029] To solve the above technical problems, the embodiment of the present application further provides a computer readable storage medium, which adopts the technical scheme as follows:

[0030] Obtaining multi-modal description information of a vehicle accident;

[0031] Inputting report information, inspection information and accident pictures in the multi-modal description information into an accident feature extraction model to obtain a vehicle accident feature;

[0032] Inputting accident pictures and vehicle-mounted data in the multi-modal description information into an article loss assessment model to obtain an article loss feature;

[0033] Inputting medical diagnosis information in the multi-modal description information into a medical diagnosis assessment model to obtain a person injury diagnosis feature;

[0034] Fusing the vehicle accident feature, the article loss feature and the person injury diagnosis feature to obtain a case feature;

[0035] Classifying the vehicle accident based on the case feature, the article loss feature and the person injury diagnosis feature to obtain a vehicle accident classification result;

[0036] Inputting the vehicle accident classification result into a vehicle accident loss assessment model to obtain a loss assessment value.

[0037] Compared with the prior art, the embodiments of the present application have the following beneficial effects: multi-modal description information depicting the vehicle accident from multiple angles is obtained, the report information, the investigation information and the accident pictures in the multi-modal description information are input into an accident feature extraction model to obtain vehicle accident features representing the vehicle accident; the accident pictures and the vehicle data are input into an article loss evaluation model to obtain article loss features representing article loss; the medical diagnosis information is input into a medical diagnosis evaluation model to obtain injury diagnosis features representing medical treatment of the personnel; the vehicle accident features, the article loss features and the injury diagnosis features are fused to obtain case features comprehensively representing the vehicle accident; based on the case features, the article loss features and the injury diagnosis features, the vehicle accident can be accurately classified to obtain a vehicle accident classification result, and the vehicle accident classification result is input into a vehicle accident loss evaluation model to obtain a loss estimate representing vehicle loss. The present application performs all-round and automatic evaluation on the vehicle accident through the model, and improves the speed and accuracy of vehicle accident loss evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the schemes in the present application, the drawings needed in the description of the embodiments of the present application will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0039] Figure 1 is an exemplary system architecture diagram to which the present application can be applied;

[0040] Figure 2 is a flowchart of one embodiment of a vehicle accident loss evaluation method according to the present application;

[0041] Figure 3 is a structural schematic diagram of one embodiment of a vehicle accident loss evaluation device according to the present application;

[0042] Figure 4 is a structural schematic diagram of one embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0043] 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 the present application belongs; the terms used in the specification of the application are only for the purpose of describing specific embodiments and are not intended to limit the present application; the specification, claims and above-described drawing of the present application and the terms "include" and "have" and any variations thereof in the specification of the application are intended to cover non-exclusive inclusion. The specification, claims and above-described drawing of the present application or the terms "first", "second" and the like are used to distinguish different objects, not to describe a specific order.

[0044] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. As will be apparent to those of ordinary skill in the art, embodiments described herein can be combined with other embodiments.

[0045] In order to make the person skilled in the art better understand the scheme of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings.

[0046] As shown in Figure 1 The system architecture 100 can include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used to provide a communication link medium between the terminal devices 101, 102, 103 and the server 105. The network 104 can include various connection types, such as wired, wireless communication links, or optical fiber cables, and the like.

[0047] A user can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, and the like. Various communication client applications can be installed on the terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, and the like.

[0048] The terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, and the like.

[0049] The server 105 can be a server that provides various services, such as a background server that provides support for pages displayed on the terminal devices 101, 102, 103.

[0050] It should be noted that the vehicle accident loss assessment method provided by the embodiments of the present application is generally executed by a server, and accordingly, the vehicle accident loss assessment device is generally arranged in the server.

[0051] It should be understood that,Figure 1 The number of terminal devices, networks and servers in the system is only illustrative. According to the implementation needs, there can be any number of terminal devices, networks and servers.

[0052] With reference to the accompanying drawings, the embodiments of the present application will be described in detail. Figure 2 Fig. 1 shows a flow chart of one embodiment of the vehicle accident loss assessment method according to the present application. The vehicle accident loss assessment method comprises the following steps:

[0053] Step S201, acquiring multi-modal description information of the vehicle accident.

[0054] In this embodiment, the electronic device (for example, the server shown in Fig. 1) on which the vehicle accident loss assessment method runs can communicate with the terminal through wired connection or wireless connection. It should be pointed out that the above-mentioned wireless connection can include but not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other now known or future developed wireless connection. Figure 1

[0055] Specifically, when performing vehicle accident loss assessment, multi-modal description information related to the vehicle accident is acquired. The multi-modal description information includes information of multiple dimensions, which is used to comprehensively depict, describe and record the vehicle accident.

[0056] The multi-modal description information can include report information, survey information, accident pictures, vehicle data and medical diagnosis information. The report information can be the information reported by the user, which can include the time, place, cause of the vehicle accident, loss of objects (vehicles, goods, etc.) at the scene, and casualties, etc., denoted as In one embodiment, the report information can be in the form of voice.

[0057] The survey information can be the information given by the survey damage personnel after investigating the vehicle accident scene, which includes the cause of the accident, the responsibility of the accident, the handling method of the case, the risk of the case, etc., denoted as In one embodiment, the survey information can be in the form of text.

[0058] The accident pictures include all-around and multi-angle pictures of the damaged vehicle, pictures of the specific damaged parts of the vehicle, pictures of the traces at the scene and pictures of the environment at the scene, denoted as

[0059] ​​The vehicle data can be vehicle OBD (On-Board Diagnostics) data, and when the vehicle is damaged, the body, chassis, engine, electrical equipment and other parts will send information to the ECU (vehicle computer). Reading fault and damage data from the vehicle computer can be used for fault detection, damage detection and positioning, denoted as .

[0060] The medical diagnosis information can include medical invoice pictures, diagnosis conclusions and treatment plans generated after the accident victim goes to the hospital, and contains personnel casualties, treatment costs and the like, denoted as .

[0061] In step S202, the report information, the survey information and the accident pictures in the multi-modal description information are input into an accident feature extraction model to obtain vehicle accident features.

[0062] Specifically, the report information, the survey information and the accident pictures are input into the accident feature extraction model. The accident picture extraction model can be built based on a neural network, and can extract vehicle accident features representing vehicle accidents from the report information, the survey information and the accident pictures .

[0063] In step S203, the accident pictures and the vehicle data in the multi-modal description information are input into an object loss evaluation model to obtain object loss features.

[0064] Specifically, the accident pictures and the vehicle data in the multi-modal description information are input into the object loss evaluation model. The object loss evaluation model can be built based on a neural network. The object loss evaluation model processes the accident pictures and the vehicle data, thereby extracting vehicle damaged component information and damaged type information of damaged objects involved in the vehicle accident, to obtain object loss features .

[0065] In step S204, the medical diagnosis information in the multi-modal description information is input into a medical diagnosis evaluation model to obtain personal injury diagnosis features.

[0066] Specifically, the medical diagnosis information in the multi-modal description information is input into the medical diagnosis evaluation model. The medical diagnosis evaluation model can be built based on a neural network, and can extract personnel casualty information and medical information in the vehicle accident to obtain personal injury diagnosis features .

[0067] In step S205, the vehicle accident features, the object loss features and the personal injury diagnosis features are fused to obtain case features.

[0068] Specifically, the vehicle accident features, the object loss features and the personal injury diagnosis features are fused to obtain case features The case feature integrates the damage of the vehicle, the article, and the person, the disability coefficient, and the information related to the accident, and can comprehensively represent the accident case. In an embodiment, the vehicle accident feature, the article loss feature, and the person injury diagnosis feature are directly spliced to obtain the case feature.

[0069] In step S206, the vehicle accident is classified based on the case feature, the article loss feature, and the person injury diagnosis feature, to obtain a vehicle accident classification result.

[0070] Specifically, the case feature, the article loss feature, and the person injury diagnosis feature are obtained. The case feature describes the vehicle accident as a whole, the article loss feature describes the vehicle accident from the perspective of the vehicle and the article, and the person injury diagnosis feature describes the vehicle accident from the perspective of the casualties. According to the case feature, the article loss feature, and the person injury diagnosis feature, the vehicle accident can be classified to obtain a vehicle accident classification result.

[0071] In an embodiment, the case feature, the article loss feature, and the person injury diagnosis feature are input into an accident classification model, and the vehicle accident classification result is output by the accident classification model.

[0072] In step S207, the vehicle accident classification result is input into a vehicle accident loss assessment model to obtain a loss assessment value.

[0073] Specifically, the vehicle accident classification result is input into the vehicle accident loss assessment model. The vehicle accident loss assessment model can be built based on a neural network and can be a regression model. For each type of vehicle accident, the vehicle accident loss assessment model can output a corresponding loss assessment value.

[0074] The loss assessment value can be a loss assessment value of the vehicle accident. In an embodiment, the loss assessment value includes an unsettled reserve, which is a reserve of an insurance company for a claim case to deal with possible claims.

[0075] In an embodiment, the case feature, the article loss feature, the person injury diagnosis feature, and the vehicle accident classification result are input into the vehicle accident loss assessment model together to obtain the loss assessment value.

[0076] In this embodiment, multi-modal description information describing the vehicle accident from multiple angles is obtained, the report information, the inspection information and the accident pictures in the multi-modal description information are input into an accident feature extraction model to obtain vehicle accident features representing the vehicle accident; the accident pictures and the vehicle data are input into an article loss assessment model to obtain article loss features representing article loss; the medical diagnosis information is input into a medical diagnosis assessment model to obtain injury diagnosis features representing medical treatment of the injured; the vehicle accident features, the article loss features and the injury diagnosis features are fused to obtain case features comprehensively representing the vehicle accident; the vehicle accident can be accurately classified based on the case features, the article loss features and the injury diagnosis features to obtain a vehicle accident classification result, and the vehicle accident classification result is input into a vehicle accident loss assessment model to obtain a loss estimate representing vehicle loss. The vehicle accident is comprehensively and automatically evaluated by the model, and the speed and accuracy of vehicle accident loss assessment are improved.

[0077] Further, the step S202 can include: inputting the report information, the inspection information and the accident pictures in the multi-modal description information into an accident feature extraction model; the accident feature extraction model includes a speech recognition sub-model, a text processing sub-model and a picture recognition sub-model; the report information is subjected to feature extraction by the speech recognition sub-model to obtain report information features; the inspection information is subjected to feature extraction by the text processing sub-model to obtain inspection information features; the accident pictures are subjected to feature extraction by the picture recognition sub-model to obtain accident picture features; and the vehicle accident features are generated according to the report information features, the inspection information features and the accident picture features.

[0078] Specifically, the report information, the inspection information and the accident pictures in the multi-modal description information are input into an accident feature extraction model; the accident feature extraction model can be a composite model composed of multiple sub-models, including a speech recognition sub-model, a text processing sub-model and a picture recognition sub-model.

[0079] The report information in the form of speech is input into the speech recognition sub-model, and the speech recognition sub-model can perform ASR (Automatic Speech Recognition) processing to extract the report information features.

[0080] The inspection information is input into the text processing sub-model, and the text processing sub-model can perform NLP (Natural Language Processing) processing on the text-form inspection information to extract the inspection information features.

[0081] The accident picture is input into the picture recognition sub-model to perform feature extraction on the accident picture to obtain an accident picture feature. In an embodiment, the picture recognition sub-model can be constructed based on a CNN network (Convolutional Neural Networks).

[0082] The obtained report information feature, the survey information feature, and the accident picture feature can be spliced to generate a vehicle accident feature.

[0083] In an embodiment, the vehicle accident feature can be represented as follows:

[0084] (1)

[0085] wherein, represents the vehicle accident feature, represents the report information feature, represents the survey information feature, represents the accident picture feature, is a set of model parameters.

[0086] In this embodiment, the report information is subjected to feature extraction by the voice recognition sub-model, the survey information is subjected to feature extraction by the text processing sub-model, and the accident picture is subjected to feature extraction by the picture recognition sub-model, so as to generate a vehicle accident feature representing a vehicle accident according to the extracted features.

[0087] Further, the step S203 can include: inputting the accident picture and the vehicle data in the multi-modal description information into an article loss assessment model; the article loss assessment model includes a first assessment sub-model and a second assessment sub-model; processing the accident picture by the first assessment sub-model to obtain an external loss assessment result; processing the vehicle data by the second assessment sub-model to obtain an inside loss assessment result; and generating an article loss feature according to the external loss assessment result and the inside loss assessment result.

[0088] Specifically, the accident picture and the vehicle data in the multi-modal description information are input into an article loss assessment model; the article loss assessment model can include a first assessment sub-model and a second assessment sub-model.

[0089] The accident picture is input into the first assessment sub-model for processing, and the first assessment sub-model can identify the vehicle type, the vehicle external damaged part, and other information in the accident picture to obtain an external loss assessment result. In an embodiment, the first assessment sub-model can be constructed based on a CNN network (Convolutional Neural Networks).

[0090] The vehicle-mounted data is input into the second evaluation sub-model for processing. The second evaluation sub-model can classify according to the vehicle-mounted data, obtain the damage of the engine and electrical equipment existing in the vehicle, and obtain the interior loss evaluation result. In an embodiment, the second evaluation sub-model can be constructed based on a tree model, which can be an LGB model (LightGBM), or a random forest, XGBOOST, GBDT, etc.

[0091] The exterior loss evaluation result and the interior loss evaluation result can be combined to form the object loss feature, which provides the damage of the vehicle from the outside and the inside of the vehicle.

[0092] In an embodiment, the object loss feature can be represented as:

[0093] (2)

[0094] wherein, represents the object loss feature, represents the exterior loss evaluation result, represents the interior loss evaluation result.

[0095] In this embodiment, the first evaluation sub-model is used to process the accident pictures to obtain the damage of the vehicle outside, and obtain the exterior loss evaluation result. The second evaluation sub-model is used to process the vehicle-mounted data to obtain the damage of the vehicle inside, and obtain the interior loss evaluation result. The object loss feature is generated according to the exterior loss evaluation result and the interior loss evaluation result, which can accurately measure the object loss.

[0096] Further, the above step S204 can include: obtaining medical diagnosis information in the multi-modal description information; performing character recognition or text processing on the medical diagnosis information through a medical diagnosis evaluation model to obtain a personal injury diagnosis feature.

[0097] Specifically, the medical diagnosis information in the multi-modal description information is input into the medical diagnosis evaluation model, which can include multiple sub-models. For picture data in the medical diagnosis information, such as single certificate pictures, character recognition (i.e., OCR, Optical Character Recognition) can be performed through a sub-model; for text data therein, text processing (NLP, Natural Language Processing) can be performed through a sub-model, so as to extract the injured part, injury type, surgery type, and medical expenses of the personnel, and obtain the personal injury diagnosis feature.

[0098] In an embodiment, the personal injury diagnosis feature can be represented as:

[0099] (3)

[0100] wherein, represents the human injury diagnosis feature, represents that the picture data in the medical diagnosis information can be subjected to OCR processing to extract picture text, and then the NLP is used to extract the injured part of the person, the injury type, the operation type, and the medical expenses and other features, to obtain the human injury diagnosis feature.

[0101] In this embodiment, after the medical diagnosis information is input into the medical diagnosis evaluation model, character recognition or text processing is performed to obtain the human injury diagnosis feature related to the injury of the person.

[0102] Further, the above step S205 can include: splicing the vehicle accident feature, the article loss feature, and the human injury diagnosis feature to obtain a spliced feature; inputting the spliced feature into a multilayer perception model; performing dimension reduction processing on the spliced feature by the multilayer perception model; and processing the dimension-reduced spliced feature by a full connection layer in the multilayer perception model to obtain a case feature.

[0103] Specifically, the vehicle accident feature, the article loss feature, and the human injury diagnosis feature are spliced to obtain a spliced feature, and then the spliced feature is input into a multilayer perception model. The multilayer perception model is built based on a multilayer perception mechanism. The multilayer perception mechanism (MLP, Multilayer Perceptron) is also called an artificial neural network (ANN, Artificial Neural Network). The bottom layer is an input layer, the middle layer is a hidden layer, and the last layer is an output layer. The layers of the multilayer perception mechanism can be fully connected.

[0104] The multilayer perception model in this application internally contains multiple embedding matrices. In this application, the types of objects (such as vehicles) involved are numerous, so the vector feature dimension output by each model is very high and relatively sparse. Therefore, the obtained spliced feature can be a tens of thousands of dimension sparse category type feature coding matrix. Multiplying the sparse coding matrix representing the spliced feature by the embedding matrix can obtain a low-dimensional dense vector, realizing dimension reduction. Dimension reduction processing is beneficial to subsequent storage and calculation.

[0105] Then, the dimension-reduced spliced feature is input into the full connection layer in the multilayer perception model for fusion to obtain a case feature.

[0106] In one embodiment, the case feature can be represented as follows:

[0107] (4)

[0108] wherein, concatenated features after dimension reduction processing of embedding matrix output, full-connect, case features.

[0109] In this embodiment, the vehicle accident features, the object loss features and the injury diagnosis features are concatenated to obtain the concatenated features. The concatenated features are first subjected to dimension reduction processing by the multi-layer perception model, and then input into the full-connect layer in the multi-layer perception model, so as to obtain the case features representing the vehicle accident.

[0110] Further, the step S206 can include: generating first features according to the case features and the object loss features; inputting the first features into the first accident classification model to obtain first classification results; generating second features according to the case features and the injury diagnosis features; inputting the second features into the second accident classification model to obtain second classification results; and generating the vehicle accident classification results according to the first classification results and the second classification results.

[0111] Specifically, the case features are concatenated with the object loss features to obtain the first features, and the case features are concatenated with the injury diagnosis features to obtain the second features.

[0112] The accident classification model can include the first accident classification model and the second accident classification model. The first features can be input into the first accident classification model, and the first accident classification model classifies according to the case features and the object loss features to obtain the first classification results. The second features can be input into the second accident classification model, and the second accident classification model classifies according to the case features and the injury diagnosis features to obtain the second classification results. The union of the first classification results and the second classification results can be used as the vehicle accident classification results.

[0113] In practice, a vehicle accident can only have object loss, and some vehicle accidents can have object loss and injury loss. The classification method in this application meets the actual situation.

[0114] In one embodiment, the case features, the object loss features and the injury diagnosis features are concatenated together and then input into the accident classification model for classification to obtain the vehicle accident classification results.

[0115] In one embodiment, the accident classification model can be constructed based on a tree model, for example, an LGB (LightGBM) model. The LGB model can classify and aggregate various vehicle accidents according to the input feature vectors to obtain a plurality of sub-clusters, and each sub-cluster is a subset of similar vehicle accidents.

[0116] In the embodiment, the first accident classification model classifies according to the case features and the object loss features to obtain a first classification result; the second accident classification model classifies according to the case features and the human injury diagnosis features to obtain a second classification result, so as to classify the vehicle accident from different angles and obtain a vehicle accident classification result.

[0117] Further, the step S207 can include: inputting the first classification result in the vehicle accident classification result into a vehicle accident loss assessment model to obtain a first loss assessment value of a historical vehicle accident corresponding to the first classification result; inputting the second classification result in the vehicle accident classification result into the vehicle accident loss assessment model to obtain a second loss assessment value of a historical vehicle accident corresponding to the second classification result; and calculating a loss assessment value of the vehicle accident according to the first loss assessment value and the second loss assessment value.

[0118] Specifically, the first classification result in the vehicle accident classification result is input into the vehicle accident loss assessment model. Since the first classification result can represent a vehicle accident category obtained according to the first features, the category has a large number of similar historical vehicle accidents. The vehicle accident loss assessment model can obtain a first loss assessment value of the historical vehicle accidents of the category, which can be regarded as a loss assessment value obtained according to the first classification result.

[0119] Similarly, the second classification result in the vehicle accident classification result is input into the vehicle accident loss assessment model. The second classification result can represent a vehicle accident category obtained according to the second features, the category has a large number of similar historical vehicle accidents. The vehicle accident loss assessment model can obtain a second loss assessment value of the historical vehicle accidents of the category, which can be regarded as a loss assessment value obtained according to the second classification result.

[0120] The first loss assessment value can be an average value of the loss assessment values of the historical vehicle accidents corresponding to the first classification result; and the second loss assessment value can be an average value of the loss assessment values of the historical vehicle accidents corresponding to the second classification result.

[0121] In one embodiment, the vehicle accident loss assessment model input by the first classification result and the second classification result can be the same neural network model, or can be two parallel neural network sub-models in the vehicle accident loss assessment model.

[0122] The first loss estimate and the second loss estimate can be combined to obtain the loss estimate of the current vehicle accident. The first loss estimate and the second loss estimate can be directly added together, or the first loss estimate and the second loss estimate can be weighted. The historical vehicle accidents corresponding to the classification results have time information, such as the time of occurrence of the claim. The number of historical vehicle accidents in which the time information is within a preset time period (which is close to the current time point, for example, from a certain time point in the past to the current time point) can be obtained through the historical vehicle accidents corresponding to the first classification result, and the number of historical vehicle accidents corresponding to the second classification result can be obtained through the historical vehicle accidents corresponding to the second classification result. The ratio of the first number to the second number can be used as the ratio of the first loss estimate to the second loss estimate, so as to focus on the loss estimate of the vehicle accident close to the current time.

[0123] In one embodiment, the loss estimate can be represented as:

[0124] (5)

[0125] wherein rbns (report but not settled) can be the loss estimate of the current vehicle accident, for example, can be the unsettled reserve. represents the first classification result; represents the second classification result. represents the vehicle accident loss assessment model, which can be a regression model.

[0126] In one embodiment, the case features, the object loss features, and the injury diagnosis features are spliced together and then input into the accident classification model for classification to obtain the vehicle accident classification result. The accident classification result is then input into the vehicle accident loss assessment model. The vehicle accident loss model takes the average of the loss estimates of the historical vehicle accidents corresponding to the vehicle accident classification result as the loss estimate of the current vehicle accident.

[0127] In this embodiment, the corresponding historical vehicle accidents are queried according to the classification results, the first loss estimate and the second loss estimate are obtained according to the historical vehicle accidents, and the loss estimate of the current vehicle accident is calculated according to the first loss estimate and the second loss estimate, so as to obtain the case loss estimate according to the historical vehicle accidents.

[0128] It should be emphasized that, in order to further ensure the privacy and security of the above-mentioned multi-modal description information, the above-mentioned multi-modal description information can also be stored in a node of a block chain.

[0129] The blockchain referred to in the present application is a new application mode of distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm and other computer technologies. The blockchain is essentially a decentralized database, which is a series of data blocks associated using cryptographic methods, each data block containing information of a batch of network transactions, for verifying the validity of the information (anti-fake) and generating the next block. The blockchain can include a blockchain underlying platform, a platform product service layer, and an application service layer.

[0130] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. Among them, artificial intelligence (AI) is to use digital computers or machine controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0131] The basic technology of artificial intelligence generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. The software technology of artificial intelligence mainly includes computer vision technology, robot technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc.

[0132] The present application can be applied to intelligent transportation in the field of smart city, thereby promoting the construction of smart city.

[0133] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by computer readable instructions instructing related hardware, and the computer readable instructions can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments of each method. Among them, the storage medium can be a magnetic disc, an optical disc, a read-only memory (ROM) and other non-volatile storage media, or a random access memory (RAM) and the like.

[0134] It should be understood that, although each step in the flowchart of the accompanying drawings is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless explicitly stated herein, the execution of these steps is not strictly limited in sequence, and they can be executed in other sequences. Moreover, at least part of the steps in the flowchart of the accompanying drawings can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence is not necessarily sequential, but can be alternately executed with at least part of other steps or sub-steps or stages of other steps.

[0135] Further referring to Figure 3 , as an implementation of the method shown in the above Figure 2 , the present application provides an embodiment of a vehicle accident loss assessment device, which corresponds to the method embodiment shown in Figure 2 , and the device can be applied to various electronic devices.

[0136] As shown in Figure 3 , the vehicle accident loss assessment device 300 described in the embodiment includes a description information acquisition module 301, an accident feature extraction module 302, an article loss assessment module 303, a medical diagnosis assessment module 304, a case feature generation module 305, a vehicle accident classification module 306, and an accident loss assessment module 307, wherein:

[0137] The description information acquisition module 301 is configured to acquire multi-modal description information of a vehicle accident.

[0138] The accident feature extraction module 302 is configured to input the report information, the inspection information, and the accident pictures in the multi-modal description information into an accident feature extraction model to obtain vehicle accident features.

[0139] The article loss assessment module 303 is configured to input the accident pictures and the vehicle-mounted data in the multi-modal description information into an article loss assessment model to obtain article loss features.

[0140] The medical diagnosis assessment module 304 is configured to input the medical diagnosis information in the multi-modal description information into a medical diagnosis assessment model to obtain human injury diagnosis features.

[0141] The case feature generation module 305 is configured to fuse the vehicle accident features, the article loss features, and the human injury diagnosis features to obtain case features.

[0142] The vehicle accident classification module 306 is configured to classify the vehicle accident based on the case features, the article loss features, and the human injury diagnosis features to obtain a vehicle accident classification result.

[0143] The accident loss assessment module 307 is configured to input the vehicle accident classification result into a vehicle accident loss assessment model to obtain a loss assessment value.

[0144] In this embodiment, multi-modal description information describing the vehicle accident from multiple perspectives is obtained, the report information, the inspection information, and the accident pictures are input into an accident feature extraction model to obtain vehicle accident features representing the vehicle accident; the accident pictures and the vehicle data are input into an object loss assessment model to obtain object loss features representing the object loss; the medical diagnosis information is input into a medical diagnosis assessment model to obtain injury diagnosis features on the medical treatment level; the vehicle accident features, the object loss features, and the injury diagnosis features are fused to obtain case features comprehensively representing the vehicle accident; the vehicle accident can be accurately classified based on the case features, the object loss features, and the injury diagnosis features to obtain a vehicle accident classification result, and the vehicle accident classification result is input into a vehicle accident loss assessment model to obtain a loss assessment value representing the vehicle loss. The vehicle accident is comprehensively and automatically assessed by the model, and the speed and accuracy of the vehicle accident loss assessment are improved.

[0145] In some optional implementation manners of this embodiment, the accident feature extraction module 302 can include an input extraction sub-module, a report extraction sub-module, an inspection extraction sub-module, a picture extraction sub-module, and a feature generation sub-module, where:

[0146] The input extraction sub-module is configured to input the report information, the inspection information, and the accident pictures in the multi-modal description information into the accident feature extraction model; the accident feature extraction model includes a speech recognition sub-model, a text processing sub-model, and a picture recognition sub-model.

[0147] The report extraction sub-module is configured to perform feature extraction on the report information by using the speech recognition sub-model to obtain report information features.

[0148] The inspection extraction sub-module is configured to perform feature extraction on the inspection information by using the text processing sub-model to obtain inspection information features.

[0149] The picture extraction sub-module is configured to perform feature extraction on the accident pictures by using the picture recognition sub-model to obtain accident picture features.

[0150] The feature generation sub-module is configured to generate vehicle accident features according to the report information features, the inspection information features, and the accident picture features.

[0151] In this embodiment, the report information is subjected to feature extraction by using the speech recognition sub-model, the inspection information is subjected to feature extraction by using the text processing sub-model, and the accident pictures are subjected to feature extraction by using the picture recognition sub-model, so as to generate the vehicle accident features representing the vehicle accident according to the extracted features.

[0152] In some optional implementations of the embodiment, the article loss assessment module 303 can include an article assessment input submodule, an external assessment submodule, an interior assessment submodule, and an article loss generation submodule, wherein:

[0153] The article assessment input submodule is configured to input the accident pictures and the vehicle-mounted data in the multi-modal description information into the article loss assessment model; the article loss assessment model includes a first assessment submodule and a second assessment submodule.

[0154] The external assessment submodule is configured to process the accident pictures through the first assessment submodule to obtain an external loss assessment result.

[0155] The interior assessment submodule is configured to process the vehicle-mounted data through the second assessment submodule to obtain an interior loss assessment result.

[0156] The article loss generation submodule is configured to generate an article loss feature according to the external loss assessment result and the interior loss assessment result.

[0157] In the embodiment, the accident pictures are processed through the first assessment submodule to obtain the external damage of the vehicle, and the external loss assessment result is obtained; the vehicle-mounted data are processed through the second assessment submodule to obtain the internal damage of the vehicle, and the interior loss assessment result is obtained; and the article loss feature is generated according to the external loss assessment result and the interior loss assessment result, so that the article loss can be accurately measured.

[0158] In some optional implementations of the embodiment, the medical diagnosis assessment module 304 can include a diagnosis acquisition submodule and a personal injury diagnosis submodule, wherein:

[0159] The diagnosis acquisition submodule is configured to acquire the medical diagnosis information in the multi-modal description information.

[0160] The personal injury diagnosis submodule is configured to perform character recognition or text processing on the medical diagnosis information through the medical diagnosis assessment model to obtain a personal injury diagnosis feature.

[0161] In the embodiment, the medical diagnosis information is input into the medical diagnosis assessment model to perform character recognition or text processing, and the personal injury diagnosis feature related to the personal injury is obtained.

[0162] In some optional implementations of the embodiment, the case feature generation module 305 can include a feature splicing submodule, a perception input submodule, a feature dimension reduction submodule, and a fusion processing submodule, wherein:

[0163] The feature splicing submodule is configured to splice the vehicle accident feature, the article loss feature, and the personal injury diagnosis feature to obtain a spliced feature.

[0164] The perception input submodule is configured to input the spliced features into the multi-layer perception model.

[0165] The feature dimension reduction submodule is configured to perform dimension reduction processing on the spliced features by using the multi-layer perception model.

[0166] The fusion processing submodule is configured to perform processing on the dimension-reduced spliced features by using the full connection layer in the multi-layer perception model, to obtain the case features.

[0167] In this embodiment, the vehicle accident features, the object loss features, and the human injury diagnosis features are spliced to obtain the spliced features; the multi-layer perception model is used to perform dimension reduction processing on the spliced features, and then the full connection layer in the multi-layer perception model is used, so as to obtain the case features that comprehensively represent the vehicle accident.

[0168] In some optional implementation manners of this embodiment, the vehicle accident classification module 306 can include a first generation submodule, a first classification submodule, a second generation submodule, a second classification submodule, and a result generation submodule, where:

[0169] The first generation submodule is configured to generate first features according to the case features and the object loss features.

[0170] The first classification submodule is configured to input the first features into the first accident classification model, to obtain the first classification result.

[0171] The second generation submodule is configured to generate second features according to the case features and the human injury diagnosis features.

[0172] The second classification submodule is configured to input the second features into the second accident classification model, to obtain the second classification result.

[0173] The result generation submodule is configured to generate the vehicle accident classification result according to the first classification result and the second classification result.

[0174] In this embodiment, the first classification result is obtained by using the first accident classification model to classify the case features and the object loss features; the second classification result is obtained by using the second accident classification model to classify the case features and the human injury diagnosis features, so that the vehicle accident classification is performed from different angles, and the vehicle accident classification result is obtained.

[0175] In some optional implementation manners of this embodiment, the accident loss assessment module 307 can include a first loss assessment submodule, a second loss assessment submodule, and a loss assessment submodule, where:

[0176] The first loss assessment submodule is configured to input the first classification result in the vehicle accident classification result into the vehicle accident loss assessment model, to obtain a first loss assessment value of a historical vehicle accident corresponding to the first classification result.

[0177] The second damage assessment submodule is used to input the second classification result from the vehicle accident classification results into the vehicle accident loss assessment model to obtain the second damage assessment estimate of the historical vehicle accident corresponding to the second classification result.

[0178] The damage assessment submodule is used to calculate the damage assessment value of a vehicle accident based on the first damage assessment value and the second damage assessment value.

[0179] In this embodiment, the corresponding historical vehicle accidents are queried according to the classification results. The first loss assessment estimate and the second loss assessment estimate are obtained based on the statistics of historical vehicle accidents. The loss assessment estimate of the current vehicle accident is calculated based on the first loss assessment estimate and the second loss assessment estimate, thereby obtaining the case loss assessment based on the historical vehicle accidents.

[0180] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0181] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0182] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0183] The memory 41 includes at least one type of readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or a memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit and the external storage device of the computer device 4. In this embodiment, the memory 41 is generally used to store an operating system and various application software installed on the computer device 4, such as computer readable instructions of the vehicle accident loss assessment method, etc. In addition, the memory 41 can also be used to temporarily store various data that have been output or will be output.

[0184] The processor 42 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer readable instructions or process data stored in the memory 41, such as computer readable instructions of the vehicle accident loss assessment method.

[0185] The network interface 43 can include a wireless network interface or a wired network interface, and is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0186] The computer device provided in this embodiment can execute the vehicle accident loss assessment method described above. Here, the vehicle accident loss assessment method can be the vehicle accident loss assessment method of each of the embodiments described above.

[0187] In the embodiment, multi-modal description information describing the vehicle accident from multiple angles is obtained, the report information, the inspection information and the accident pictures are input into the accident feature extraction model to obtain vehicle accident features representing the vehicle accident; the accident pictures and the vehicle data are input into the object loss evaluation model to obtain object loss features representing the object loss; the medical diagnosis information is input into the medical diagnosis evaluation model to obtain injury diagnosis features on the medical treatment level; the vehicle accident features, the object loss features and the injury diagnosis features are fused to obtain case features comprehensively representing the vehicle accident; the vehicle accident can be accurately classified based on the case features, the object loss features and the injury diagnosis features to obtain a vehicle accident classification result, the vehicle accident classification result is input into the vehicle accident loss evaluation model to obtain a loss estimate value representing the vehicle loss, and the vehicle accident is comprehensively and automatically evaluated by the model, so that the speed and the accuracy of the vehicle accident loss evaluation are improved.

[0188] The application also provides another embodiment, that is, a computer readable storage medium storing computer readable instructions, the computer readable instructions being executable by at least one processor to make the at least one processor execute the steps of the vehicle accident loss evaluation method as described above.

[0189] In the embodiment, multi-modal description information describing the vehicle accident from multiple angles is obtained, the report information, the inspection information and the accident pictures are input into the accident feature extraction model to obtain vehicle accident features representing the vehicle accident; the accident pictures and the vehicle data are input into the object loss evaluation model to obtain object loss features representing the object loss; the medical diagnosis information is input into the medical diagnosis evaluation model to obtain injury diagnosis features on the medical treatment level; the vehicle accident features, the object loss features and the injury diagnosis features are fused to obtain case features comprehensively representing the vehicle accident; the vehicle accident can be accurately classified based on the case features, the object loss features and the injury diagnosis features to obtain a vehicle accident classification result, the vehicle accident classification result is input into the vehicle accident loss evaluation model to obtain a loss estimate value representing the vehicle loss, and the vehicle accident is comprehensively and automatically evaluated by the model, so that the speed and the accuracy of the vehicle accident loss evaluation are improved.

[0190] Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and the necessary general hardware platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the method described in each embodiment of the present application.

[0191] Obviously, the above-described embodiments are only some of the embodiments of the present application, not all the embodiments, and the drawings show the preferred embodiments of the present application, but do not limit the patent scope of the present application. The present application can be implemented in many different forms, and conversely, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing specific embodiments, or make equivalent replacements to some of the technical features. Any equivalent structure made by using the content of the specification and drawings, directly or indirectly applied to other related technical fields, is also within the scope of the patent protection of the present application.

Claims

1. A method for assessing vehicle accident losses, characterized in that, Includes the following steps: Obtain multimodal description information of vehicle accidents; The report information, investigation information, and accident images from the multimodal description information are input into the accident feature extraction model to obtain vehicle accident features; The accident images and vehicle data from the multimodal description information are input into the object loss assessment model to obtain the object loss characteristics; The medical diagnostic information in the multimodal description information is input into the medical diagnostic assessment model to obtain the human injury diagnostic features; The vehicle accident characteristics, the property loss characteristics, and the personal injury diagnosis characteristics are fused to obtain the case characteristics; Based on the case characteristics, the property loss characteristics, and the personal injury diagnosis characteristics, the vehicle accident is classified to obtain the vehicle accident classification result; The vehicle accident classification results are input into the vehicle accident loss assessment model to obtain the estimated damage value; The step of fusing the vehicle accident characteristics, the property loss characteristics, and the personal injury diagnosis characteristics to obtain the case characteristics includes: The vehicle accident features, the object loss features, and the personal injury diagnosis features are spliced ​​together to obtain spliced ​​features; The spliced ​​features are input into the multilayer perceptron model; The spliced ​​features are reduced in dimensionality using the multilayer perceptron model. The spliced ​​features after dimensionality reduction are processed by the fully connected layer in the multilayer perceptron model to obtain case features; The step of classifying the vehicle accident based on the case characteristics, the property loss characteristics, and the personal injury diagnosis characteristics to obtain the vehicle accident classification result includes: A first feature is generated based on the case characteristics and the object loss characteristics; The first feature is input into the first accident classification model to obtain the first classification result; A second feature is generated based on the case characteristics and the personal injury diagnosis characteristics; The second feature is input into the second accident classification model to obtain the second classification result; A vehicle accident classification result is generated based on the first classification result and the second classification result; Among them, the multilayer perception model is built on the multilayer perceptron. Its bottom layer is the input layer, the middle layer is the hidden layer, and the last layer is the output layer. The multilayer perception model contains multiple layers of embedding matrices.

2. The vehicle accident loss assessment method according to claim 1, characterized in that, The step of inputting the report information, investigation information, and accident images from the multimodal description information into the accident feature extraction model to obtain vehicle accident features includes: The report information, investigation information, and accident images from the multimodal description information are input into the accident feature extraction model; the accident feature extraction model includes a speech recognition sub-model, a text processing sub-model, and an image recognition sub-model; The report information features are obtained by extracting features from the speech recognition sub-model. The text processing sub-model is used to extract features from the survey information to obtain the survey information features; The accident image features are obtained by extracting features from the accident image using the image recognition sub-model. Vehicle accident features are generated based on the characteristics of the reported incident information, the characteristics of the investigation information, and the characteristics of the accident images.

3. The vehicle accident loss assessment method according to claim 1, characterized in that, The step of inputting the accident images and vehicle data from the multimodal description information into the object loss assessment model to obtain the object loss features includes: The accident images and vehicle data from the multimodal description information are input into the object loss assessment model; the object loss assessment model includes a first assessment sub-model and a second assessment sub-model. The accident images are processed using the first evaluation sub-model to obtain external loss assessment results; The vehicle data is processed by the second evaluation sub-model to obtain the internal loss evaluation result; The object loss characteristics are generated based on the external loss assessment results and the internal loss assessment results.

4. The vehicle accident loss assessment method according to claim 1, characterized in that, The step of inputting the medical diagnostic information from the multimodal description information into the medical diagnostic assessment model to obtain the human injury diagnostic features includes: Obtain medical diagnostic information from the multimodal description information; The medical diagnostic information is subjected to text recognition or text processing by a medical diagnostic assessment model to obtain human injury diagnostic features.

5. The vehicle accident loss assessment method according to claim 1, characterized in that, The step of inputting the vehicle accident classification results into the vehicle accident loss assessment model to obtain the damage assessment estimate includes: Input the first classification result from the vehicle accident classification results into the vehicle accident loss assessment model to obtain the first loss estimate of the historical vehicle accident corresponding to the first classification result. Input the second classification result from the vehicle accident classification results into the vehicle accident loss assessment model to obtain the second loss assessment estimate of the historical vehicle accident corresponding to the second classification result; The damage assessment estimate for the vehicle accident is calculated based on the first damage assessment estimate and the second damage assessment estimate.

6. A vehicle accident loss assessment device, characterized in that, include: The description information acquisition module is used to acquire multimodal description information of vehicle accidents; The accident feature extraction module is used to input the report information, investigation information and accident images from the multimodal description information into the accident feature extraction model to obtain vehicle accident features; The object loss assessment module is used to input the accident images and vehicle data from the multimodal description information into the object loss assessment model to obtain object loss features; The medical diagnosis and assessment module is used to input the medical diagnosis information from the multimodal description information into the medical diagnosis and assessment model to obtain human injury diagnosis features; The case feature generation module is used to fuse the vehicle accident features, the object loss features, and the personal injury diagnosis features to obtain case features; The vehicle accident classification module is used to classify the vehicle accident based on the case characteristics, the property loss characteristics, and the personal injury diagnosis characteristics, and obtain the vehicle accident classification result; The accident loss assessment module is used to input the vehicle accident classification results into the vehicle accident loss assessment model to obtain the estimated damage value. The case feature generation module includes: a feature splicing submodule, a perception input submodule, a feature dimensionality reduction submodule, and a fusion processing submodule. Specifically: the feature splicing submodule splices vehicle accident features, object loss features, and personal injury diagnosis features to obtain spliced ​​features; the perception input submodule inputs the spliced ​​features into a multilayer perception model; the feature dimensionality reduction submodule performs dimensionality reduction processing on the spliced ​​features through the multilayer perception model; and the fusion processing submodule processes the dimensionality-reduced spliced ​​features through a fully connected layer in the multilayer perception model to obtain the case features. The vehicle accident classification module includes: a first generation submodule, a first classification submodule, a second generation submodule, a second classification submodule, and a result generation submodule. Specifically: the first generation submodule generates a first feature based on case characteristics and property damage characteristics; the first classification submodule inputs the first feature into a first accident classification model to obtain a first classification result; the second generation submodule generates a second feature based on case characteristics and personal injury diagnosis characteristics; the second classification submodule inputs the second feature into a second accident classification model to obtain a second classification result; and the result generation submodule generates a vehicle accident classification result based on the first and second classification results. Among them, the multilayer perception model is built on the multilayer perceptron. Its bottom layer is the input layer, the middle layer is the hidden layer, and the last layer is the output layer. The multilayer perception model contains multiple layers of embedding matrices.

7. A computer device comprising a memory and a processor, the memory storing computer-readable instructions, wherein the processor, when executing the computer-readable instructions, implements the steps of the vehicle accident loss assessment method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the vehicle accident loss assessment method as described in any one of claims 1 to 5.

Images