Intelligent accident vehicle identification method based on vehicle component information

By preprocessing data and training of vehicle component information and deep learning models, an accident vehicle intelligent identification model is built, which solves the problems of vehicle evaluation accuracy and low efficiency, and achieves more efficient and accurate vehicle evaluation, reducing evaluation costs.

CN118916783APending Publication Date: 2024-11-08BEIJING KUCHE YIMEI NETWORK TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411014613.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The accuracy and efficiency of vehicle evaluation in the prior art lead to high evaluation costs and the inability to effectively improve the accuracy and efficiency of vehicle evaluation.

Method used

By preprocessing the vehicle component information, encoding and training using deep learning models, an accident vehicle intelligent identification model is constructed, and the vehicle's accident vehicle probability is output through the model for identification.

Benefits of technology

It improves the accuracy and efficiency of vehicle evaluation, reduces the cost of vehicle evaluation, and provides more objective and reliable evaluation results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916783A_ABST
    Figure CN118916783A_ABST
Patent Text Reader

Abstract

The invention relates to the field of data processing, in particular to an accident vehicle intelligent identification method based on vehicle component information, and the method comprises the steps: S1, carrying out the data preprocessing, S2, inputting the preprocessed vehicle component information into an encoder layer through an input layer of a deep learning model, the method comprises the following steps: S1, pre-processing vehicle component information, S3, coding the pre-processed vehicle component information, S3, taking a data set of an input vector after coding as model construction data, S4, training a deep learning model, S5, testing the trained deep learning model, and taking the tested deep learning model as an accident vehicle intelligent identification model, and S6, carrying out intelligent identification on the accident vehicle. The real-time vehicle part information is input into the accident vehicle intelligent identification model; and S7, the accident vehicle is identified. Whether the vehicle is the accident vehicle model or not is predicted through the accident vehicle intelligent identification model, the accuracy and efficiency of vehicle evaluation are improved, and the cost of vehicle evaluation is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an intelligent identification method for accident vehicles based on vehicle component information. Background Art

[0002] The Accident Vehicle Intelligent Identification Model is an advanced technical tool used to accurately predict whether a vehicle is an accident victim. Application of this model can significantly improve the accuracy and efficiency of vehicle assessments. Traditional vehicle assessment methods typically rely on manual inspection and empirical judgment, which are subject to subjectivity, large errors, and time-consuming and labor-intensive processes. The introduction of the Accident Vehicle Intelligent Identification Model effectively addresses these issues. Leveraging big data analysis and machine learning algorithms, the model can rapidly process massive amounts of vehicle information and automatically determine whether a vehicle has been involved in an accident, providing objective and reliable assessment results. This not only reduces the cost and time of manual assessments, but also mitigates the risk of misjudgment caused by human error, improving the quality and efficiency of the entire vehicle assessment process. Furthermore, the use of the Accident Vehicle Intelligent Identification Model can help establish a more transparent and fair trading environment in the used car market, enhance trust between buyers and sellers, and further promote the healthy development of the used car market.

[0003] Chinese Patent Publication No. CN112651493A discloses a method and device for identifying accident vehicles based on a joint training model. This method relates to the technical field of accident vehicle identification. The joint training model comprises a linear model and a deep learning model. The method involves obtaining pre-processed insurance repair records for the current vehicle; obtaining vehicle structured data based on the pre-processed insurance repair records; inputting the vehicle structured data into the linear model and the pre-processed insurance repair records as unstructured vehicle data into the deep learning model to determine whether the current vehicle is an accident vehicle; and if so, identifying the cause of the accident based on the vehicle structured data. This method utilizes joint training of structured and unstructured data. However, this solution only identifies accident vehicles based on pre-processed insurance repair records and fails to improve the accuracy and efficiency of vehicle assessments. Summary of the Invention

[0004] To this end, the present invention provides an intelligent accident vehicle identification method based on vehicle component information to overcome the problem of high cost of vehicle evaluation caused by low accuracy and efficiency in vehicle evaluation in the prior art.

[0005] To achieve the above objectives, the present invention provides an intelligent accident vehicle identification method based on vehicle component information, comprising:

[0006] Step S1, performing data preprocessing on vehicle component information to obtain preprocessed vehicle component information;

[0007] Step S2: input the preprocessed vehicle component information into the encoder layer through the input layer of the deep learning model, encode the preprocessed vehicle component information, and obtain an encoded input vector;

[0008] Step S3, using the data set of the encoded input vector as model building data, and dividing the model building data into a training set and a test set;

[0009] Step S4: construct a deep learning model, and train the deep learning model using the training set to obtain a trained deep learning model;

[0010] Step S5: Testing the trained deep learning model based on the test set, and using the deep learning model that passes the test as the accident vehicle intelligent recognition model;

[0011] Step S6: input the real-time vehicle component information into the accident vehicle intelligent recognition model, and the accident vehicle intelligent recognition model outputs the accident vehicle probability of the vehicle;

[0012] Step S7: Identify the accident vehicle according to the accident vehicle probability.

[0013] Furthermore, in step S1, the data preprocessing method includes:

[0014] Step S10, collecting vehicle component information;

[0015] Step S11, performing data cleaning on the vehicle component information to obtain cleaned vehicle component information;

[0016] Step S12, performing format correction on the vehicle component information after data cleaning to obtain pre-processed vehicle component information.

[0017] Furthermore, in step S11, a text recognition processing tool is used to perform data cleaning on the vehicle component information.

[0018] Furthermore, in the step S12, the cleaned vehicle component information is processed into a digital format, and the text information is converted into digital information.

[0019] Furthermore, in step S2, the preprocessed vehicle component information is input into the encoder layer through the input layer of the deep learning model, and the preprocessed vehicle component information is encoded according to the data features of the preprocessed vehicle component information to obtain an encoded input vector.

[0020] Furthermore, in step S3, the data set of the encoded input vectors is used as model building data, and the model building data is divided into a 70% training set and a 30% test set.

[0021] Furthermore, in step S4, the method for constructing a deep learning model includes:

[0022] Step S40, setting the self-attention mechanism of the deep learning model;

[0023] Step S41, setting the first linear transformation of the deep learning model;

[0024] Step S42, setting the activation function of the deep learning model;

[0025] Step S43, setting the second linear change of the deep learning model;

[0026] Step S44, setting the output layer of the deep learning model.

[0027] Furthermore, in step S5, the trained deep learning model is tested according to the test set, and the pass status of the trained deep learning model is judged according to the test results, and the deep learning model that passes the test is output as the accident vehicle intelligent recognition model.

[0028] Furthermore, in step S6, the trained deep learning model is tested according to the test set, and the test accuracy N is output. The test accuracy N is compared with the preset test accuracy N0. The pass / fail status of the trained deep learning model is judged according to the comparison result, and the deep learning model that passes the test is output as the accident vehicle intelligent recognition model, wherein:

[0029] When N≥N0, it is determined that the trained deep learning model test has passed, and the deep learning model is output as the accident vehicle intelligent recognition model;

[0030] When N<N0, it is determined that the trained deep learning model test fails, the second vehicle component information is obtained, and steps S1 to S5 are repeated until the deep training model test passes.

[0031] Furthermore, in step S7, the accident vehicle probability value K obtained in step S6 is compared with a preset accident vehicle probability value K0, and the accident vehicle is identified based on the comparison result, wherein:

[0032] When K≥K0, the vehicle corresponding to the real-time vehicle component information is determined to be an accident vehicle;

[0033] When K<K0, the vehicle corresponding to the real-time vehicle component information is determined to be a non-accident vehicle.

[0034] Compared with the prior art, the beneficial effect of the present invention is that the method performs data preprocessing on the vehicle component information in step S1 to obtain the preprocessed vehicle component information, thereby improving the quality of the vehicle component information; the method inputs the preprocessed vehicle component information into the encoder layer through the input layer of the deep learning model in step S2, encodes the preprocessed vehicle component information, and obtains the encoded input vector to facilitate the input of subsequent model data; the method uses the data set of the encoded input vector as model construction data in step S3, and divides the model construction data into a training set and a test set to facilitate the model Training and testing: the method trains the deep learning model according to the training set in step S4 to obtain the trained deep learning model; the method verifies the trained deep learning model according to the test set in step S5, and uses the verified deep learning model as the accident vehicle intelligent recognition model; the method inputs real-time vehicle component information into the accident vehicle intelligent recognition model in step S6, and the accident vehicle intelligent recognition model outputs the accident vehicle probability of the vehicle; the method identifies the accident vehicle according to the accident vehicle probability of the vehicle in step S7, thereby improving the accuracy and efficiency of vehicle evaluation and reducing the cost of vehicle evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 Schematic diagram of the flow of the method for intelligently identifying accident vehicles based on vehicle component information according to this embodiment;

[0036] Figure 2 Schematic diagram of the data preprocessing method of this embodiment;

[0037] Figure 3 A flowchart of the method for constructing a deep learning model in this embodiment. DETAILED DESCRIPTION

[0038] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0039] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0040] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0041] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0042] See also Figure 1 FIG. 1 is a flow chart of an intelligent accident vehicle identification method based on vehicle component information according to an embodiment of the present invention. The method includes:

[0043] Step S1, performing data preprocessing on vehicle component information to obtain preprocessed vehicle component information;

[0044] Step S2: input the preprocessed vehicle component information into the encoder layer through the input layer of the deep learning model, encode the preprocessed vehicle component information, and obtain an encoded input vector;

[0045] Step S3, using the data set of the encoded input vector as model building data, and dividing the model building data into a training set and a test set;

[0046] Step S4: construct a deep learning model, and train the deep learning model using the training set to obtain a trained deep learning model;

[0047] Step S5: Testing the trained deep learning model based on the test set, and using the deep learning model that passes the test as the accident vehicle intelligent recognition model;

[0048] Step S6: input the real-time vehicle component information into the accident vehicle intelligent recognition model, and the accident vehicle intelligent recognition model outputs the accident vehicle probability of the vehicle;

[0049] Step S7: Identify the accident vehicle according to the accident vehicle probability.

[0050] Specifically, the method is set in an accident vehicle identification platform, and predicts whether a vehicle is an accident vehicle through a deep learning model, and obtains a more accurate prediction method through data information processing and continuous optimization of model results. In which, the method performs data preprocessing on the vehicle component information through step S1 to obtain preprocessed vehicle component information, thereby improving the quality of the vehicle component information. The method inputs the preprocessed vehicle component information into the encoder layer through the input layer of the deep learning model through step S2, encodes the preprocessed vehicle component information, and obtains an encoded input vector to facilitate the input of subsequent model data. The method uses the data set of the encoded input vector as the model construction data through step S3. According to the method, the model construction data is divided into a training set and a test set to facilitate training and testing of the model. The method trains the deep learning model according to the training set through step S4 to obtain the trained deep learning model. The method verifies the trained deep learning model according to the test set through step S5, and uses the verified deep learning model as the accident vehicle intelligent recognition model. The method inputs real-time vehicle component information into the accident vehicle intelligent recognition model through step S6, and the accident vehicle intelligent recognition model outputs the accident vehicle probability of the vehicle. The method identifies the accident vehicle according to the accident vehicle probability of the vehicle through step S7, thereby improving the accuracy and efficiency of vehicle evaluation and reducing the cost of vehicle evaluation.

[0051] Specifically, in step S1 , data preprocessing is performed on the vehicle component information to obtain preprocessed vehicle component information.

[0052] See also Figure 2 As shown in FIG. 1 , which is a flow chart of the data preprocessing method according to this embodiment, in step S1, the data preprocessing method includes:

[0053] Step S10, collecting vehicle component information;

[0054] Step S11, performing data cleaning on the vehicle component information to obtain cleaned vehicle component information;

[0055] Step S12, performing format correction on the vehicle component information after data cleaning to obtain pre-processed vehicle component information.

[0056] Specifically, in step S10, the vehicle component information refers to the detailed historical information of the vehicle components, including component type, maintenance record and service life. This embodiment does not limit the method of obtaining vehicle component information. Relevant technical personnel in this field can freely set it according to actual needs, such as obtaining it from a relevant vehicle accident inquiry website.

[0057] Specifically, in step S11, a text recognition processing tool is used to clean the vehicle component information, remove invalid and erroneous data records, and fill in missing values.

[0058] Specifically, data cleaning is performed through step S11 to improve the accuracy and completeness of the vehicle component information after data cleaning.

[0059] Specifically, this embodiment does not limit the text recognition and processing tool. Relevant personnel in this field can set it at will as long as it meets the text recognition and processing requirements. For example, the OCR.Space tool can be used. The OCR.Space tool is a powerful online character recognition tool with functions such as data format and content verification, removal of duplicate records, correction of structural errors, and filling of missing values.

[0060] Specifically, in step S12, the format correction refers to processing the vehicle component information after data cleaning into a digital format, and converting the text information into digital information.

[0061] Specifically, this embodiment does not limit the digital format processing method of vehicle component information after data cleaning. Relevant personnel in this field can set it at will, as long as it meets the digital format processing requirements. For example, component information can be classified and numbered according to type, the component start date can be converted into the number of days of use, and maintenance information can be converted into the number of maintenance times.

[0062] Specifically, in step S2, the preprocessed vehicle component information is input into the encoder layer through the input layer of the deep learning model, and the preprocessed vehicle component information is encoded according to the data features of the preprocessed vehicle component information to obtain an encoded input vector, wherein:

[0063] When the data features of the pre-processed vehicle component information are numerical features, step S2 normalizes the service life and maintenance times in the pre-processed vehicle component information to eliminate dimensionality effects and accelerate the convergence of the model;

[0064] When the data features of the pre-processed vehicle component information are categorical features, step S2 converts the pre-processed vehicle component information into a numerical vector using a one-hot encoding pair.

[0065] Specifically, step S2 is used to normalize the service life and maintenance times in the pre-processed vehicle component information to eliminate dimensionality effects and accelerate the convergence of the model.

[0066] Specifically, the one-hot encoding refers to an encoding method used to represent classified data. When the preprocessed vehicle component information is converted into a numerical vector using one-hot encoding in step S2, if there are four types of preprocessed vehicle component information, each type of preprocessed vehicle component information will be encoded into a four-element vector, in which one position is 1 and the rest are 0, and the preprocessed vehicle component information corresponding to the position 1 represents the component type of the four-element vector.

[0067] Specifically, in step S3, the data set of the encoded input vectors is used as model building data, and the model building data is divided into a 70% training set and a 30% test set.

[0068] Specifically, the training set refers to a data set used to train a deep learning model, and the test set refers to a data set used to evaluate the performance of a deep learning model.

[0069] See also Figure 3 As shown, it is a flow chart of the method for constructing a deep learning model in this embodiment. In step S4, the method for constructing a deep learning model includes:

[0070] Step S40, setting the self-attention mechanism of the deep learning model;

[0071] Step S41, setting the first linear transformation of the deep learning model;

[0072] Step S42, setting the activation function of the deep learning model;

[0073] Step S43, setting the second linear change of the deep learning model;

[0074] Step S44, setting the output layer of the deep learning model.

[0075] Specifically, in step S40, the self-attention mechanism enables the deep learning model to take into account the information of other components when processing the information of one component. The training set data passes through the self-attention mechanism to obtain the data of the self-attention mechanism.

[0076] Specifically, in step S41, when performing the first linear change, the data from the self-attention mechanism is linearly transformed through the linear layer, and the formula of the linear transformation is Y1=X1W1+b1, Y2=X2W_1+b_1, Y3=X3W+b1, where X1, X2 and X3 are input data, W1, W_1 and W1 are weight matrices, b1, b_1 and b1 are bias terms, and Y1, Y2 and Y3 are the first linear change data obtained after the first linear change.

[0077] Specifically, in step S42, when setting the activation function of the deep learning model, the first linear change data is activated according to the activation function A, and A=max[y,Y)] is set, y is the first linear change data, and Y is the preset activation value to obtain the activated data.

[0078] Specifically, the activation function A=max[y,Y)] means that when the data after the first linear change is greater than the preset activation value Y, the activated data A=y can be output. This embodiment does not limit the preset activation value Y. Relevant technicians in this field can freely set it according to actual needs. For example, the preset activation value can be set to 0, and when the first linear change data is greater than 0, it can be output through the activation function.

[0079] Specifically, in step S43, the activated data is subjected to a second linear transformation, and the second linear transformation formula is Z1=A1W2+b2, Z2=A2W_2+b_2, Z3=A3W2+b2, where A1, A2, A3 are the activated data, W2, W_2, W2 and b2, b_2, b2 are the weights and biases of the second linear layer respectively, and Z1, Z2 and Z3 are the second linear change data obtained after the second linear change.

[0080] Specifically, in step S44, when setting the output layer of the deep learning model, the output layer is set to output the probability K that the vehicle is an accident vehicle.

[0081] Specifically, in step S4, the training set is also input into the deep learning model to train the deep learning model, and the trained deep learning model is output.

[0082] Specifically, in step S5, the trained deep learning model is tested according to the test set, and the test accuracy N is output. The test accuracy N is compared with the preset test accuracy N0. The pass / fail status of the trained deep learning model is judged based on the comparison result, and the deep learning model that passes the test is output as the accident vehicle intelligent recognition model, wherein:

[0083] When N≥N0, it is determined that the trained deep learning model test has passed, and the deep learning model is output as the accident vehicle intelligent recognition model;

[0084] When N<N0, it is determined that the trained deep learning model test fails, the second vehicle component information is obtained, and steps S1 to S5 are repeated until the deep training model test passes.

[0085] Specifically, in step S6, the real-time vehicle component information is input into the accident vehicle intelligent identification model, and the accident vehicle intelligent identification model outputs the accident vehicle probability value K of the vehicle.

[0086] Specifically, in step S7, the accident vehicle probability value K obtained in step S6 is compared with a preset accident vehicle probability value K0, and the accident vehicle is identified based on the comparison result, wherein:

[0087] When K≥K0, the vehicle corresponding to the real-time vehicle component information is determined to be an accident vehicle;

[0088] When K<K0, the vehicle corresponding to the real-time vehicle component information is determined to be a non-accident vehicle.

[0089] Specifically, the preset accident vehicle probability value refers to a preset probability value reflecting that the vehicle is an accident vehicle. For example, the preset accident vehicle probability value K0 may be set to 0.8.

[0090] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. An intelligent identification method for accident vehicles based on vehicle component information, characterized in that: The method comprises: Step S1, performing data preprocessing on vehicle component information to obtain preprocessed vehicle component information; Step S2, inputting the preprocessed vehicle component information into the encoder layer through the input layer of the deep learning model, encoding the preprocessed vehicle component information to obtain an encoded input vector; Step S3, using the data set of the encoded input vector as model building data, and dividing the model building data into a training set and a test set; Step S4, constructing a deep learning model, and training the deep learning model with a training set to obtain a trained deep learning model; Step S5, testing the trained deep learning model according to the test set, and using the deep learning model that passes the test as the accident vehicle intelligent recognition model; Step S6, inputting the real-time vehicle component information into the accident vehicle intelligent identification model, and the accident vehicle intelligent identification model outputs the accident vehicle probability of the vehicle; Step S7, identifying the accident vehicle according to the accident vehicle probability.

2. The method for intelligently identifying accident vehicles based on vehicle component information according to claim 1, characterized in that: In step S1, the data preprocessing method includes: Step S10, collecting vehicle component information; Step S11, performing data cleaning on the vehicle component information to obtain vehicle component information after data cleaning; Step S12, performing format correction on the vehicle component information after data cleaning to obtain pre-processed vehicle component information.

3. The method for intelligently identifying accident vehicles based on vehicle component information according to claim 2 is characterized in that: In the step S11, a text recognition processing tool is used to perform data cleaning on the vehicle component information.

4. The method for intelligently identifying accident vehicles based on vehicle component information according to claim 3 is characterized in that: In the step S12, the cleaned vehicle component information is processed into a digital format, and the text information is converted into digital information.

5. The method for intelligently identifying accident vehicles based on vehicle component information according to claim 1, characterized in that: In step S2, the preprocessed vehicle component information is input into the encoder layer through the input layer of the deep learning model, and the preprocessed vehicle component information is encoded according to the data features of the preprocessed vehicle component information to obtain an encoded input vector.

6. The method for intelligently identifying accident vehicles based on vehicle component information according to claim 1, characterized in that: In step S3, the data set of the encoded input vectors is used as model building data, and the model building data is divided into a 70% training set and a 30% test set.

7. The method for intelligently identifying accident vehicles based on vehicle component information according to claim 1, characterized in that: In step S4, the method for constructing a deep learning model includes: Step S40, setting the self-attention mechanism of the deep learning model; Step S41, setting the first linear transformation of the deep learning model; Step S42, setting the activation function of the deep learning model; Step S43, setting the second linear change of the deep learning model; Step S44, setting the output layer of the deep learning model.

8. The method for intelligently identifying accident vehicles based on vehicle component information according to claim 1, characterized in that: In step S5, the trained deep learning model is tested according to the test set, and the pass status of the trained deep learning model is judged according to the test result, and the deep learning model that passes the test is output as the accident vehicle intelligent recognition model.

9. The method for intelligently identifying accident vehicles based on vehicle component information according to claim 1, characterized in that: In step S6, the trained deep learning model is tested according to the test set, and the test accuracy N is output. The test accuracy N is compared with the preset test accuracy N0. The pass of the trained deep learning model test is judged according to the comparison result, and the deep learning model that passes the test is output as the accident vehicle intelligent recognition model, where: When N≥N0, it is determined that the trained deep learning model test has passed, and the deep learning model is output as an intelligent recognition model for accident vehicles; When N<N0, it is determined that the trained deep learning model test fails, the second vehicle component information is obtained, and steps S1 to S5 are repeated until the deep training model test passes.

10. The method for intelligently identifying accident vehicles based on vehicle component information according to claim 1, characterized in that: In step S7, the accident vehicle probability value K obtained in step S6 is compared with the preset accident vehicle probability value K0, and the accident vehicle is identified according to the comparison result, wherein: When K≥K0, the vehicle corresponding to the real-time vehicle component information is determined to be an accident vehicle; When K<K0, the vehicle corresponding to the real-time vehicle component information is determined to be a non-accident vehicle.

Citation Information

Cited By

  • Inference model and real-time auxiliary system based on accident vehicle data training

    CN122222024A