WO2021027157A1

WO2021027157A1 - Vehicle insurance claim settlement identification method and apparatus based on picture identification, and computer device and storage medium

Info

Publication number: WO2021027157A1
Application number: PCT/CN2019/118323
Authority: WO
Inventors: 卢显锋
Original assignee: 平安科技（深圳）有限公司
Priority date: 2019-08-13
Filing date: 2019-11-14
Publication date: 2021-02-18
Also published as: CN110689440A

Abstract

Disclosed in embodiments of the present application are a vehicle insurance claim settlement identification method and apparatus based on picture identification, and a computer device and a storage medium. The present application is applied to the field of prediction models in intelligent decisions. The method comprises: collecting training pictures from a preset database and constructing a training sample according to the training pictures; based on the training sample, training a preset convolutional neural network model in a forward propagation and back propagation combined mode to obtain a trained convolutional neural network model; and if a claim settlement request uploaded by a user is received, inputting a claim settlement picture in the claim settlement request into the trained convolutional neural network model for prediction so as to output a claim settlement probability corresponding to the claim settlement picture, thereby determining whether claim settlement can be performed or not, generating a claim settlement estimated amount, and sending same to the user.

Description

Auto insurance claim settlement recognition method, device, computer equipment and storage medium based on picture recognition

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on August 13, 2019, the application number is 201910745383.8, and the application title is "Image recognition-based identification method, device, computer equipment and storage medium for auto insurance claims." The entire content is incorporated into this application by reference.

Technical field

This application relates to the field of artificial intelligence technology, and in particular to a method, device, computer equipment, and storage medium for auto insurance claims recognition based on image recognition.

Background technique

With the development of technology and economy, people's living standards are improving day by day, and more and more families own cars, which also drives the rapid development of auto insurance. Due to the huge inventory of the Chinese automobile market, various traffic accidents occur on the road every day, and the current auto insurance accident claims are mainly for insurance prospectors to identify whether to settle claims through on-site exploration. This method is not only time-consuming and labor-intensive, but also A lot of human resources are wasted, and some claims fraud cannot be effectively identified.

Summary of the invention

The embodiments of the application provide a method, device, computer equipment, and storage medium for auto insurance claim identification based on image recognition, aiming to solve the existing auto insurance claims that require insurance prospectors to identify whether to settle claims through on-site exploration, which is time-consuming, labor-intensive and wasteful Human resources are also unable to effectively identify some claims fraud problems.

In the first aspect, an embodiment of the present application provides a method for recognizing car insurance claims based on picture recognition, which includes: collecting training pictures from a preset database and constructing training samples based on the training pictures; Train the preset convolutional neural network model in a combination of forward propagation and back propagation to obtain the trained convolutional neural network model; if a claim request uploaded by the user is received, the claim settlement picture in the claim settlement request Input to the trained convolutional neural network model for prediction to output the claim settlement probability corresponding to the claim settlement picture; compare the claim settlement probability corresponding to the claim settlement picture with a preset threshold; if the claim settlement picture corresponds If the claim settlement probability is greater than the preset threshold, it is determined that the claim settlement picture can be settled; according to the claim settlement information in the claim settlement request uploaded by the user, an estimated claim settlement amount is generated according to preset rules and the estimated settlement amount is sent To the user.

In a second aspect, an embodiment of the present application also provides a car insurance claim recognition device based on picture recognition, which includes: a construction unit for collecting training pictures from a preset database and constructing training samples based on the training pictures; training unit , For training the preset convolutional neural network model by combining forward propagation and back propagation based on the training samples to obtain the trained convolutional neural network model; prediction unit, for receiving To the claim settlement request uploaded by the user, the claim settlement picture in the claim settlement request is input into the trained convolutional neural network model for prediction to output the claim settlement probability corresponding to the claim settlement picture; the comparison unit is used to compare the The claim settlement probability corresponding to the claim settlement picture is compared with a preset threshold; the determining unit is configured to determine that the claim settlement picture can be settled if the claim settlement probability corresponding to the claim settlement picture is greater than the preset threshold; the claim settlement unit uses According to the claim settlement information in the claim settlement request uploaded by the user, an estimated claim settlement amount is generated according to a preset rule and the estimated settlement amount is sent to the user.

In a third aspect, an embodiment of the present application also provides a computer device, which includes a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the A computer program to perform the following steps: collect training pictures from a preset database and construct training samples based on the training pictures; based on the training samples, use a combination of forward propagation and back propagation to convolve the preset The neural network model is trained to obtain the trained convolutional neural network model; if a claim settlement request uploaded by the user is received, the claim settlement picture in the claim settlement request is input into the trained convolutional neural network model for prediction. Output the claim settlement probability corresponding to the claim settlement picture; compare the claim settlement probability corresponding to the claim settlement picture with a preset threshold; if the claim settlement probability corresponding to the claims settlement picture is greater than the preset threshold, determine the claim settlement The picture can be settled; according to the claim settlement information in the claim settlement request uploaded by the user, an estimated claim settlement amount is generated according to preset rules and the estimated settlement amount is sent to the user.

In a fourth aspect, the embodiments of the present application also provide a computer-readable storage medium, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the following steps: The training pictures are collected in the preset database and training samples are constructed according to the training pictures; based on the training samples, the preset convolutional neural network model is trained by a combination of forward propagation and back propagation to obtain post-training A convolutional neural network model; if a claim settlement request uploaded by a user is received, the claim settlement picture in the claim settlement request is input to the trained convolutional neural network model for prediction to output the claim settlement probability corresponding to the claim settlement picture Comparing the claim settlement probability corresponding to the claim settlement picture with a preset threshold; if the claim settlement probability corresponding to the claims settlement picture is greater than the preset threshold, it is determined that the claim settlement picture can be settled; upload according to the user The claim settlement information in the claim settlement request generates an estimated claim settlement amount according to preset rules and sends the estimated settlement amount to the user.

Description of the drawings

In order to explain the technical solutions of the embodiments of the present application more clearly, the following will briefly introduce the drawings needed in the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present application. Ordinary technicians can obtain other drawings based on these drawings without creative work.

FIG. 1 is a schematic diagram of an application scenario of a method for recognizing auto insurance claims based on image recognition according to an embodiment of the application;

FIG. 2 is a schematic flowchart of a method for auto insurance claim settlement based on image recognition provided by an embodiment of the application;

3 is a schematic diagram of a sub-process of a method for recognizing auto insurance claims based on image recognition according to an embodiment of the application;

4 is a schematic diagram of a sub-process of a method for auto insurance claims settlement based on image recognition provided by an embodiment of the application;

FIG. 5 is a schematic diagram of a sub-process of a method for recognizing auto insurance claims based on image recognition according to an embodiment of the application;

FIG. 6 is a schematic flowchart of a method for recognizing auto insurance claims based on image recognition according to another embodiment of the application;

FIG. 7 is a schematic block diagram of a device for recognizing auto insurance claims based on image recognition according to an embodiment of the application;

FIG. 8 is a schematic block diagram of specific units of the device for recognizing auto insurance claims based on picture recognition according to an embodiment of the application; and

FIG. 9 is a schematic block diagram of a computer device provided by an embodiment of the application.

detailed description

The technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

It should be understood that when used in this specification and the appended claims, the terms "including" and "including" indicate the existence of the described features, wholes, steps, operations, elements and/or components, but do not exclude one or The existence or addition of multiple other features, wholes, steps, operations, elements, components, and/or collections thereof.

It should also be understood that the terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in the specification of this application and the appended claims, unless the context clearly indicates other circumstances, the singular forms "a", "an" and "the" are intended to include plural forms.

It should be further understood that the term "and/or" used in the specification and appended claims of this application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations .

Please refer to FIG. 1 and FIG. 2. FIG. 1 is a schematic diagram of an application scenario of a method for auto insurance claims settlement based on image recognition provided by an embodiment of the application. Fig. 2 is a schematic flow chart of a method for recognizing auto insurance claims based on image recognition according to an embodiment of the application. The method for recognizing auto insurance claims based on image recognition is applied to the server 20 and is implemented through the interaction between the terminal 10 and the server 20.

Fig. 2 is a schematic flowchart of a method for auto insurance claim settlement based on image recognition provided by an embodiment of the present application. As shown in the figure, the method includes the following steps S110-S160.

S110. Collect training pictures from a preset database and construct training samples according to the training pictures.

In one embodiment, the preset database refers to a database that stores auto insurance claim information. There are multiple auto insurance claim cases with completed claims in the preset database, such as all auto insurance claim cases of a certain auto insurance company. The case records the vehicle's claim settlement picture and claim settlement information. Among them, the claim picture is a picture taken by the insured person on the damaged part of the vehicle in a vehicle accident, for example, a picture of a rear-end vehicle, a picture of a vehicle rollover, and a picture of a vehicle scratching. Claim information refers to the insured uploading information related to the claim to the insurance company after a vehicle accident, such as the model of the car, the reason for the damage, and the policy number. Therefore, in order to make the model achieve a good training effect, the claims pictures in the completed claims cases are obtained from the preset database as the training pictures. After collecting the training images, it is also necessary to construct training samples for model input. The training images are labeled, and the training samples are constructed according to the training images and the corresponding labels of the training images, for example, (image, 1).

In an embodiment, as shown in FIG. 3, the step S110 may include steps: S111-S113.

S111. Collect training pictures from a preset database.

Specifically, each claim case is assigned a uniquely identified case number, and the claim cases are retrieved in batches from the preset database through the case number and the claims picture in each claim case is obtained.

S112: Mark the training picture according to the preset mark of the training picture.

Specifically, because a supervised learning model is used, the training data needs to be annotated before model training. Supervised learning refers to training the network using examples of known correct answers, where the annotation is the expected the result of. Specifically, after collecting the training pictures, Labelllmg is used to label the training pictures one by one. Labelllmg is a picture labeling tool. Among all the training pictures, some of the training pictures belong to the claims, and the other part of the pictures are not in the claim. Standard, among which, training pictures that belong to claims are marked, while training pictures that are not part of claims are unmarked. For training pictures that belong to claims, they are marked as 1, and training pictures that are not part of claims are marked as 0. , Until all the training pictures are marked. After the annotation is completed, the size of the image needs to be converted, and all training images are uniformly converted into 256*256 pixel images to maintain the stability of the features.

S113. Construct training samples according to the training pictures and the annotations corresponding to the training pictures.

Specifically, the training sample refers to the sample that can be input to the model. It is composed of training pictures and annotations. Each training image has a corresponding annotation. Each training image and the corresponding annotation form a training sample, for example, , (Image1, 1), (image2, 0). After constructing the training samples, the training samples are divided into training set and test set according to the preset ratio, and the training samples marked as 0 and 1 are equally divided in the training set and the test set, where the preset ratio can be The training set is 70% and the test set is 30%. Of course, it is understandable that other division ratios are also possible. Among them, the training set is used to train the model, and the test set is used to test the accuracy of the model.

S120. Based on the training samples, train a preset convolutional neural network model in a combination of forward propagation and back propagation to obtain a trained convolutional neural network model.

In one embodiment, the convolutional neural network, English called Convolutional Neural Networks, or CNN for short, is a type of feedforward neural network (Feedforward Neural Networks) that includes convolution or related calculations and has a deep structure. Learning) is one of the representative algorithms. The convolutional neural network model includes a convolutional layer, a pooling layer, and a fully connected layer. The convolutional layer and the pooling layer have multiple layers. The training process of the convolutional neural network model is divided into two stages. The first stage is data The stage of propagation from the low-level to the high-level is the forward propagation stage; the other stage is the stage where the error is propagated from the high level to the bottom level when the result of the forward propagation does not match the expectations, that is, the reverse To the propagation stage.

In an embodiment, as shown in FIG. 4, the step S120 may include the steps:

S121: Perform forward propagation of the training set in the training sample through the convolutional layer, the pooling layer, and the fully connected layer to obtain output values.

S122. Compare the output value with the target value to obtain a total error.

S123. Backpropagating the total error through the fully connected layer, the pooling layer, and the convolutional layer in order to update the weights in the network.

S124. Input the test set in the training sample into a convolutional neural network model for testing to obtain a trained convolutional neural network model.

Specifically, first initialize the ownership value in the network, that is, set it to a small random value close to 0, and then input the training image into the network to go through the convolutional layer, the pooling layer, and the fully connected layer in turn, where the volume There are multiple layers of accumulation and pooling, and the output value is finally obtained through forward propagation. Then the output value is compared with the target value. The target value refers to the label corresponding to the training picture in the training sample, and the total error between the output value and the target value is calculated. Since the training image passes through the convolutional layer, the pooling layer, and the fully connected layer from the input layer to the output layer, the data is inevitably lost in the process of transferring data between layers, which leads to errors. The error value caused by each layer is different, so when the total error of the network is obtained, the error needs to be propagated back to the network to find out how much the total error of each layer should bear. So in the end, gradient descent is performed, the error is returned layer by layer, the error of each layer is calculated, and then the weight is updated, that is, the total error is backpropagated, and the error of the fully connected layer is calculated according to the total error. The error of the pooling layer is obtained according to the error of the fully connected layer, the error of the convolution layer is obtained according to the error of the pooling layer, and the weight of each layer is updated according to the error of each layer. All training samples in the training set are input into the model for training to obtain a trained model, and then the training samples of the test set are input into the trained model to test the accuracy of the model. If the accuracy of the model reaches The preset requirements indicate that the model has been trained. If the accuracy of the model does not meet the preset requirements, the model needs to increase training samples to continue training.

S130: If a claim settlement request uploaded by the user is received, input the claim settlement picture in the claim settlement request into the trained convolutional neural network model for prediction to output a claim settlement probability corresponding to the claim settlement picture.

In one embodiment, the claim settlement request uploaded by the user includes a claim settlement picture and claim settlement information, where the claim settlement picture is a picture taken by the user on the damaged part of the vehicle during a vehicle accident, for example, a picture of a rear-end vehicle. Turned pictures and pictures of vehicles scratching. Claim information refers to the information related to the claim that the user uploads to the insurance company after a vehicle accident, such as the model of the car, the reason for the damage, and the policy number. After a vehicle accident occurs, the user fills in basic vehicle information and the cause of the accident through the mobile terminal 10, takes pictures of the damaged vehicle on the spot, and uploads the claim picture and claim information to the insurance company's server 20 to generate a claim request. After receiving the claim settlement request uploaded by the user, call the trained convolutional neural network model, input the claim settlement picture into the convolutional neural network model for prediction, and obtain the claim settlement probability of the claim settlement picture, that is, whether the claim settlement picture is It belongs to the probability of claim settlement.

In an embodiment, as shown in FIG. 5, the step S130 may include steps: S131-S134.

S131. Input the claim settlement picture to the convolutional layer in the convolutional neural network model for feature extraction to obtain first feature data.

Specifically, the recognition process of a claim picture is the same as the process of forward propagation. It should be noted that the picture is a bunch of numbers arranged in order in the computer. The common picture expression method is the RGB color model, namely red (Red) and green. (Green) and Blue (Blue) are the three primary colors, so the specific form of the claims picture is a three-dimensional tensor, called a feature map, which can be directly input to the model. The convolutional layer is used to extract features from the input data. It contains multiple convolution kernels. The parameters of the convolutional layer include the size of the convolution kernel, step size and padding. The parameters of the convolutional layer have been determined after the model is trained. , The input feature map is convolved through the convolution kernel to obtain the convolution result. The convolution process can be understood as using a filter (convolution kernel) to filter each small area of the image, so as to obtain the feature value of these small areas. Convolution results, and then normalize the convolution results to get the normalized results. Normalization is mainly to solve the problem of changes in the data distribution of the middle layer to prevent the gradient from disappearing or exploding, speed up the training speed, and finally pass activation The function activates the normalized result and outputs the first feature data, in which the activation function adopts the Relu function (Rectified linear unit stands for modified linear unit), and the function of the Relu function is to increase the nonlinear relationship between the layers of the neural network.

S132. Input the first feature data to the pooling layer in the convolutional neural network model to perform dimensionality reduction processing to obtain second feature data.

Specifically, the pooling layer is the next layer of network connected to the convolutional layer. The pooling layer is used to compress the amount of data and parameters to reduce overfitting, that is, to remove redundant information, and to remove the most important features. Extracting and performing feature dimensionality reduction can be simply understood as a compressed image. There are usually two methods of pooling, one is the maximum pooling method, and the other is the average pooling method. This embodiment adopts the maximum pooling method. Specifically, an area is selected in the first feature data, for example, a 2*2 area, and then the maximum value in the 2*2 area is selected as the value of a pixel. Select the maximum value of the next area as the value of the next pixel until all the pixels of the first feature data are extracted to obtain the second feature data, which can be simply understood as extracting the maximum value for each sub-matrix of the first feature data The second feature data is obtained by compressing in the way of value.

S133. Input the second feature data to the fully connected layer in the convolutional neural network model for mapping to obtain a one-dimensional feature vector.

Specifically, each node of the fully connected layer is connected to all nodes of the previous layer, and its function is to integrate the features extracted from the front, and map the features learned by the network to the label space of the sample. Therefore, the fully connected layer mapping is actually a convolution operation where the size of the convolution kernel is the size of the upper layer feature. The result of the convolution is a node, which corresponds to a point in the fully connected layer. The node of the fully connected layer is a one-dimensional feature. vector.

S134. Input the one-dimensional feature vector into a classifier for classification to obtain a claim settlement probability of the claim settlement picture.

Specifically, a classifier SVM (Support Vector Machine) is used for classification, which is a two-classification model for two-class classification of claims pictures, one is classified as belonging to claims, and the other is classified as not belonging to claims. When the one-dimensional feature vector is obtained, the one-dimensional feature vector is input into the classifier, and the one-dimensional feature vector is mapped to a value in the range of 0 to 1, and the value is the probability of claim settlement.

S140. Compare the claim settlement probability corresponding to the claim settlement picture with a preset threshold.

S150. If the claim settlement probability corresponding to the claim settlement picture is greater than the preset threshold value, it is determined that the claim settlement picture can be settled.

In one embodiment, after obtaining the claim probability of the claim settlement picture, the claim settlement probability is compared with a preset threshold value. The preset threshold value is 0.7. Of course, it can be understood that it can also be any other value. When the claim settlement probability corresponding to the claim settlement picture is greater than the preset threshold, it means that the settlement settlement picture is very close to the settlement settlement picture of the vehicle that has already been settled, and it is determined that the settlement settlement picture is claimable. When the claim settlement probability corresponding to the claim settlement picture is less than the preset threshold, it indicates that the claim settlement picture has the risk of insurance fraud or fraud, and it is determined that the claim settlement picture is not claimable. For example, when the claim settlement probability is 0.8, then the claim settlement probability is greater than the preset threshold, and it is determined that the claim settlement picture can be settled.

S160. Generate an estimated claim settlement amount according to preset rules according to the claim settlement information in the claim settlement request uploaded by the user and send the estimated claim settlement amount to the user.

In an embodiment, as shown in FIG. 6, the step S160 includes steps: S161-S162.

S161. Filter the claim information of the already settled claims from a preset database and match the filtered claim information with the claim information in the claim request uploaded by the user according to a preset rule.

S162. Obtain a claim settlement amount from a successfully matched claim settlement case as an estimated claim settlement amount and send the estimated settlement amount to the user.

In one embodiment, firstly, all the claim cases are screened from the preset database, the claim cases that have been settled are selected, and the claim information of the selected claim cases is obtained. Among them, the claim cases that have been settled are marked after the settlement is completed , Therefore, you can filter the claims that have been settled by selecting the marked claim cases; then, the claim information of the filtered claim cases and the claim information uploaded by the user are matched according to preset rules, where the preset rules are, for example, Match according to vehicle model, vehicle age, and damaged part. For example, if the vehicle model uploaded by the user is Type A, the vehicle age is 2 years, and the damaged part is the rear of the vehicle, then look up one by one from the claim information accordingly. Match the claim information, and use the amount of the claim in the matched claim information as the estimated amount of the claim, and send the estimated amount of the claim to the user.

The embodiment of the present application shows a method for recognizing auto insurance claims based on picture recognition, by collecting training pictures from a preset database and constructing training samples based on the training pictures; based on the training samples, forward propagation and back propagation are used The preset convolutional neural network model is trained in a combined manner to obtain the trained convolutional neural network model; if a claim settlement request uploaded by the user is received, the claim settlement picture in the claim settlement request is input to the post-training Prediction is performed in the convolutional neural network model to output the claim settlement probability corresponding to the claim settlement picture; the claim settlement probability corresponding to the claims settlement picture is compared with a preset threshold; if the claim settlement probability corresponding to the claims settlement picture is greater than The preset threshold value determines that the claim settlement picture can be settled; according to the claim settlement information in the claim settlement request uploaded by the user, an estimated claim settlement amount is generated according to preset rules and the estimated settlement amount is sent to the user. Improve claims efficiency, save human resources, and effectively identify insurance fraud.

FIG. 7 is a schematic block diagram of a device 200 for auto insurance claim settlement recognition based on image recognition provided by an embodiment of the present application. As shown in FIG. 7, corresponding to the above car insurance claim recognition method based on picture recognition, the present application also provides a car insurance claim recognition device 200 based on picture recognition. The device 200 for recognizing car insurance claims based on picture recognition includes a unit for executing the aforementioned method for recognizing car insurance claims based on picture recognition, and the device may be configured in a server. Specifically, referring to FIG. 7, the car insurance claim settlement recognition device 200 based on image recognition includes: a construction unit 210, a training unit 220, a prediction unit 230, a comparison unit 240, a determination unit 250 and a claim settlement unit 260.

The construction unit 210 is configured to collect training pictures from a preset database and construct training samples according to the training pictures.

In one embodiment, as shown in FIG. 8, the construction unit 210 includes: a collection unit 211, a labeling unit 212 and a construction subunit 213.

The collecting unit 211 is used to collect training pictures from a preset database.

The labeling unit 212 is configured to label the training picture according to the preset label of the training picture.

The construction subunit 213 is configured to construct training samples according to the training pictures and the annotations corresponding to the training pictures.

The training unit 220 is configured to train a preset convolutional neural network model in a combination of forward propagation and backward propagation based on the training samples to obtain a trained convolutional neural network model.

In an embodiment, as shown in FIG. 8, the training unit 220 includes: a forward propagation unit 221, an error unit 222, a back propagation unit 223 and a testing unit 224.

The forward propagation unit 221 is configured to forward the training set in the training sample through the convolutional layer, the pooling layer, and the fully connected layer to obtain output values.

The error unit 222 is used to compare the output value with the target value to obtain the total error.

The back-propagation unit 223 is configured to back-propagate the total error through the fully connected layer, the pooling layer, and the convolutional layer in order to update the weights in the network.

The testing unit 224 is configured to input the test set in the training sample into the convolutional neural network model for testing to obtain a trained convolutional neural network model.

The prediction unit 230 is configured to, if a claim settlement request uploaded by the user is received, input the claim settlement picture in the claim settlement request into the trained convolutional neural network model for prediction to output a claim settlement probability corresponding to the claim settlement picture.

In an embodiment, as shown in FIG. 8, the prediction unit 230 includes: a feature extraction unit 231, a dimensionality reduction unit 232, a mapping unit 233 and a classification unit 234.

The feature extraction unit 231 is configured to input the claim settlement picture into the convolutional layer in the convolutional neural network model to perform feature extraction to obtain first feature data.

The dimensionality reduction unit 232 is configured to input the first feature data into the pooling layer in the convolutional neural network model to perform dimensionality reduction processing to obtain second feature data.

The mapping unit 233 is configured to input the second feature data to the fully connected layer in the convolutional neural network model for mapping to obtain a one-dimensional feature vector.

The classification unit 234 is configured to input the one-dimensional feature vector into a classifier for classification to obtain the claim settlement probability of the claim settlement picture.

The comparison unit 240 is configured to compare the claim settlement probability corresponding to the claim settlement picture with a preset threshold.

The determining unit 250 is configured to determine that the claim settlement picture can be settled if the claim settlement probability corresponding to the claim settlement picture is greater than the preset threshold.

The claim settlement unit 260 is configured to generate an estimated claim settlement amount according to preset rules according to the claim settlement information in the claim settlement request uploaded by the user and send the estimated settlement amount to the user.

In an embodiment, as shown in FIG. 8, the claims settlement unit 260 includes: a matching unit 261 and a sending unit 262.

The matching unit 261 is configured to filter the claim information of the settled claims from a preset database and match the filtered claim information with the claim information in the claim request uploaded by the user according to the preset rules.

The sending unit 262 is configured to obtain a claim settlement amount as an estimated claim settlement amount from a successfully matched claim settlement case and send the estimated settlement amount to the user.

It should be noted that those skilled in the art can clearly understand that the above-mentioned car insurance claims recognition device 200 based on image recognition and the specific implementation process of each unit can refer to the corresponding description in the foregoing method embodiment, for the convenience of description and It's concise, so I won't repeat it here.

The above-mentioned car insurance claim recognition device based on picture recognition can be implemented in the form of a computer program, and the computer program can be run on a computer device as shown in FIG. 9.

Please refer to FIG. 9, which is a schematic block diagram of a computer device according to an embodiment of the present application. The computer device 500 may be a server, the server may be an independent server, or a server cluster composed of multiple servers.

Referring to FIG. 9, the computer device 500 includes a processor 502, a memory, and a network interface 505 connected through a system bus 501, where the memory may include a non-volatile storage medium 503 and an internal memory 504.

The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions. When the program instructions are executed, the processor 502 can execute an auto insurance claim recognition method based on image recognition.

The processor 502 is used to provide calculation and control capabilities to support the operation of the entire computer device 500.

The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can make the processor 502 execute an auto insurance claim recognition method based on image recognition.

The network interface 505 is used for network communication with other devices. Those skilled in the art can understand that the structure shown in FIG. 9 is only a block diagram of part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device 500 to which the solution of the present application is applied. The specific computer device 500 may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

Wherein, the processor 502 is configured to run a computer program 5032 stored in a memory to implement the method for recognizing a car insurance claim based on image recognition in the embodiment of the present application.

It should be understood that, in this embodiment of the application, the processor 502 may be a central processing unit (Central Processing Unit, CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (Digital Signal Processors, DSPs), Application Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the foregoing embodiments can be implemented by computer programs instructing relevant hardware. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the foregoing method embodiments.

Therefore, the embodiment of the present application also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for recognizing auto insurance claims based on image recognition described in the above embodiments.

The storage medium may be a U disk, a mobile hard disk, a read-only memory (Read-Only Memory, ROM), a magnetic disk or an optical disk, and other computer-readable storage media that can store computer programs.

The above are only specific implementations of this application, but the protection scope of this application is not limited to this. Anyone familiar with the technical field can easily think of various equivalents within the technical scope disclosed in this application. Modifications or replacements, these modifications or replacements shall be covered within the protection scope of this application. Therefore, the protection scope of this application shall be subject to the protection scope of the claims.

Claims

An auto insurance claim recognition method based on picture recognition, including:

Collecting training pictures from a preset database and constructing training samples according to the training pictures;

Based on the training samples, training a preset convolutional neural network model in a combination of forward propagation and back propagation to obtain a trained convolutional neural network model;

If a claim settlement request uploaded by the user is received, input the claim settlement picture in the claim settlement request into the trained convolutional neural network model for prediction to output a claim settlement probability corresponding to the claim settlement picture;

Comparing the claim settlement probability corresponding to the claim settlement picture with a preset threshold;

If the claim settlement probability corresponding to the claim settlement picture is greater than the preset threshold, determining that the claim settlement picture can be settled;

According to the claim settlement information in the claim settlement request uploaded by the user, an estimated claim settlement amount is generated according to a preset rule and the estimated claim settlement amount is sent to the user.
The method for recognizing auto insurance claims based on picture recognition according to claim 1, wherein the collecting training pictures from a preset database and constructing training samples according to the training pictures comprises:

Collect training pictures from the preset database;

Label the training picture according to the preset mark of the training picture;

Constructing training samples according to the training pictures and the annotations corresponding to the training pictures.
The method for recognizing auto insurance claims based on image recognition according to claim 1, wherein said training samples are based on the training samples, using a combination of forward propagation and back propagation to train a preset convolutional neural network model to Get the trained convolutional neural network model, including:

Forward propagation of the training set in the training sample through the convolutional layer, the pooling layer, and the fully connected layer to obtain output values;

Comparing the output value with the target value to obtain the total error;

Backpropagating the total error through the fully connected layer, the pooling layer, and the convolutional layer in order to update the weights in the network;

The test set in the training sample is input into the convolutional neural network model for testing to obtain a trained convolutional neural network model.
The method for recognizing auto insurance claims based on picture recognition according to claim 1, wherein said inputting said claim picture into said convolutional neural network model for prediction to output a claim probability corresponding to said claim picture comprises:

Input the claim settlement picture to the convolutional layer in the convolutional neural network model for feature extraction to obtain first feature data;

Inputting the first feature data to the pooling layer in the convolutional neural network model to perform dimensionality reduction processing to obtain second feature data;

Inputting the second feature data to the fully connected layer in the convolutional neural network model for mapping to obtain a one-dimensional feature vector;

The one-dimensional feature vector is input into a classifier for classification to obtain the claim settlement probability of the claim settlement picture.
The auto insurance claim identification method based on image recognition according to claim 1, wherein the estimated claim settlement amount is generated according to preset rules according to the claim settlement information in the claim settlement request uploaded by the user and the estimated settlement amount is sent To the user, including:

Filtering the claim information of the already settled claims from a preset database and matching the filtered claim information with the claim information in the claim request uploaded by the user according to the preset rules;

Obtain the claim settlement amount from the matched claim settlement case as the estimated settlement amount and send the estimated settlement amount to the user.
The method for recognizing auto insurance claims based on picture recognition according to claim 2, wherein after the marking the training picture according to the preset mark of the training picture, the method further comprises:

Convert the annotated training picture into a 256*256 pixel picture.
The method for recognizing auto insurance claims based on image recognition according to claim 4, wherein said inputting said first feature data to a pooling layer in said convolutional neural network model performs dimensionality reduction processing to obtain second feature data ,include:

The maximum value pooling method is used to perform dimensionality reduction processing on the first feature to obtain second feature data.
A car insurance claim settlement recognition device based on picture recognition, including:

The construction unit is used for collecting training pictures from a preset database and constructing training samples according to the training pictures;

A training unit, configured to train a preset convolutional neural network model based on the training samples in a combination of forward propagation and back propagation to obtain a trained convolutional neural network model;

A prediction unit, configured to, if a claim settlement request uploaded by a user is received, input the claim settlement picture in the claim settlement request into the trained convolutional neural network model for prediction to output a claim settlement probability corresponding to the claim settlement picture;

A comparison unit, configured to compare the claim settlement probability corresponding to the claim settlement picture with a preset threshold;

A determining unit, configured to determine that the claim settlement picture can be settled if the claim settlement probability corresponding to the claim settlement picture is greater than the preset threshold;

The claim settlement unit is configured to generate an estimated claim settlement amount according to preset rules according to the claim settlement information in the claim settlement request uploaded by the user and send the estimated settlement amount to the user.
The car insurance claim settlement recognition device based on picture recognition according to claim 8, wherein the training unit comprises:

The forward propagation unit is configured to forward the training set in the training sample through the convolutional layer, the pooling layer and the fully connected layer to obtain the output value;

An error unit for comparing the output value with the target value to obtain the total error;

A backpropagation unit, configured to backpropagate the total error through a fully connected layer, a pooling layer, and a convolutional layer in order to update the weights in the network;

The test unit is used to input the test set in the training sample into the convolutional neural network model for testing to obtain the trained convolutional neural network model.
The device for auto insurance claims settlement recognition based on picture recognition according to claim 8, wherein the prediction unit comprises:

A feature extraction unit, configured to input the claims picture into the convolutional layer in the convolutional neural network model to perform feature extraction to obtain first feature data;

A dimensionality reduction unit, configured to input the first feature data into the pooling layer in the convolutional neural network model to perform dimensionality reduction processing to obtain second feature data;

A mapping unit, configured to input the second feature data to the fully connected layer in the convolutional neural network model for mapping to obtain a one-dimensional feature vector;

The classification unit is configured to input the one-dimensional feature vector into a classifier for classification to obtain the claim settlement probability of the claim settlement picture.
A computer device includes a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the computer program stored in the memory to perform the following steps:

Collecting training pictures from a preset database and constructing training samples according to the training pictures;

Based on the training samples, training the preset convolutional neural network model in a combination of forward propagation and back propagation to obtain a trained convolutional neural network model;

If a claim settlement request uploaded by the user is received, input the claim settlement picture in the claim settlement request into the trained convolutional neural network model for prediction to output a claim settlement probability corresponding to the claim settlement picture;

Comparing the claim settlement probability corresponding to the claim settlement picture with a preset threshold;

If the claim settlement probability corresponding to the claim settlement picture is greater than the preset threshold, determining that the claim settlement picture can be settled;

According to the claim settlement information in the claim settlement request uploaded by the user, an estimated claim settlement amount is generated according to a preset rule and the estimated claim settlement amount is sent to the user.
The computer device according to claim 11, wherein said collecting training pictures from a preset database and constructing training samples according to the training pictures comprises:

Collect training pictures from the preset database;

Label the training picture according to the preset mark of the training picture;

Constructing training samples according to the training pictures and the annotations corresponding to the training pictures.
The computer device according to claim 11, wherein, based on the training samples, the preset convolutional neural network model is trained by a combination of forward propagation and back propagation to obtain the trained convolution Neural network models, including:

Forward propagation of the training set in the training sample through the convolutional layer, the pooling layer, and the fully connected layer to obtain output values;

Comparing the output value with the target value to obtain the total error;

Backpropagating the total error through the fully connected layer, the pooling layer, and the convolutional layer in order to update the weights in the network;

The test set in the training sample is input into the convolutional neural network model for testing to obtain a trained convolutional neural network model.
11. The computer device according to claim 11, wherein the inputting the claim settlement picture into the convolutional neural network model for prediction to output a claim settlement probability corresponding to the claim settlement picture comprises:

Input the claim settlement picture to the convolutional layer in the convolutional neural network model for feature extraction to obtain first feature data;

Inputting the first feature data to the pooling layer in the convolutional neural network model to perform dimensionality reduction processing to obtain second feature data;

Input the second feature data to the fully connected layer in the convolutional neural network model for mapping to obtain a one-dimensional feature vector;

The one-dimensional feature vector is input into a classifier for classification to obtain the claim settlement probability of the claim settlement picture.
The computer device according to claim 11, wherein said generating an estimated amount of claim settlement according to preset rules according to the claim settlement information in the claim settlement request uploaded by the user and sending the estimated amount of settlement to the user comprises :

Filtering the claim information of the already settled claims from a preset database and matching the filtered claim information with the claim information in the claim request uploaded by the user according to the preset rules;

Obtain the claim settlement amount from the matched claim settlement case as the estimated settlement amount and send the estimated settlement amount to the user.
The computer device according to claim 12, wherein, after the marking the training picture according to the preset mark of the training picture, the method further comprises:

Convert the annotated training picture into a 256*256 pixel picture.
The computer device according to claim 14, wherein the inputting the first feature data to the pooling layer in the convolutional neural network model to perform dimensionality reduction processing to obtain the second feature data comprises:

The maximum value pooling method is used to perform dimensionality reduction processing on the first feature to obtain second feature data.
A computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the processor executes the following steps:

Collecting training pictures from a preset database and constructing training samples according to the training pictures;

Based on the training samples, training a preset convolutional neural network model in a combination of forward propagation and back propagation to obtain a trained convolutional neural network model;

If a claim settlement request uploaded by the user is received, input the claim settlement picture in the claim settlement request into the trained convolutional neural network model for prediction to output a claim settlement probability corresponding to the claim settlement picture;

Comparing the claim settlement probability corresponding to the claim settlement picture with a preset threshold;

If the claim settlement probability corresponding to the claim settlement picture is greater than the preset threshold, determining that the claim settlement picture can be settled;

According to the claim settlement information in the claim settlement request uploaded by the user, an estimated claim settlement amount is generated according to a preset rule and the estimated claim settlement amount is sent to the user.
18. The computer-readable storage medium according to claim 18, wherein, based on the training samples, the preset convolutional neural network model is trained by a combination of forward propagation and backward propagation to obtain post-training The steps of the convolutional neural network model include:

Forward propagation of the training set in the training sample through the convolutional layer, the pooling layer, and the fully connected layer to obtain output values;

Comparing the output value with the target value to obtain the total error;

Backpropagating the total error through the fully connected layer, the pooling layer, and the convolutional layer in order to update the weights in the network;

The test set in the training sample is input into the convolutional neural network model for testing to obtain a trained convolutional neural network model.
18. The computer-readable storage medium of claim 18, wherein the step of inputting the claim settlement picture into the convolutional neural network model for prediction to output a claim settlement probability corresponding to the claim settlement picture comprises:

Input the claim settlement picture to the convolutional layer in the convolutional neural network model for feature extraction to obtain first feature data;

Inputting the first feature data to the pooling layer in the convolutional neural network model to perform dimensionality reduction processing to obtain second feature data;

Inputting the second feature data to the fully connected layer in the convolutional neural network model for mapping to obtain a one-dimensional feature vector;

The one-dimensional feature vector is input into a classifier for classification to obtain the claim settlement probability of the claim settlement picture.