CN111461266B

CN111461266B - Vehicle damage assessment abnormity identification method, device, server and storage medium

Info

Publication number: CN111461266B
Application number: CN202010557295.8A
Authority: CN
Inventors: 胡俊; 王思涵; 曹莹; 刘成秀; 赵艳民
Original assignee: Aibao Technology Co ltd
Current assignee: Aibao Technology Co ltd
Priority date: 2020-06-18
Filing date: 2020-06-18
Publication date: 2020-12-01
Anticipated expiration: 2040-06-18
Also published as: CN111461266A

Abstract

The invention discloses a vehicle damage assessment abnormity identification method, a vehicle damage assessment abnormity identification device, a server and a readable storage medium. The method comprises the following steps: firstly, determining a first damage combination in current damage information, wherein the occurrence probability of the first damage combination meets the preset requirement in a historical case; removing the part of the first damage combination from the current damage information to form a second damage combination, wherein the second damage combination comprises second damage information, and determining the matching probability of the first damage combination and each piece of second damage information; and if second damage information with the matching probability within the first preset numerical range exists, judging that the current vehicle damage assessment list is abnormal in damage assessment. The invention divides case information into a high-probability damage component combination and a suspicious damage component combination, calculates the matching degree of the high-probability damage component and the suspicious damage component according to historical case data, determines whether the damage assessment of the current case is abnormal or not, and further improves the core damage efficiency and the core damage accuracy on the basis of reducing the core damage difficulty degree.

Description

Vehicle damage assessment abnormity identification method, device, server and storage medium

Technical Field

The invention relates to the technical field of vehicle accident loss checking, in particular to a vehicle loss assessment abnormity identification method, a device, a server and a storage medium.

Background

With the continuous increase of the total amount of automobiles in China, the occurrence frequency of automobile accidents is gradually increased, and the damage assessment of the automobile accidents becomes a key link for automobile insurance claim settlement. After the staff carries out vehicle accident damage assessment, further verification needs to be carried out according to damage assessment information, namely a damage verification link. In the process of nuclear damage, workers need to judge whether the case is a false case or not from a large amount of data, the workload is too large, and mistakes and omissions are easy to occur. Secondly, the case has the problem that the vehicle is damaged newly or old, which is difficult to be confirmed by naked eyes, and the case has the situations of over repair and over replacement, so that the problem positioning by a nuclear damage worker from a large list is time-consuming and labor-consuming. Therefore, a method for performing efficient core loss by using machine learning is urgently needed in the core loss process.

In the prior art, a method and a system for determining damage of a vehicle accident have been proposed, in the prior art, the probability of damage of a key vehicle component is calculated based on the similarity between loss data included in historical data in a database and loss data of a case to be processed, and the combination mode of vehicle components in a new case is determined according to the probability of the historical data.

However, the following problems exist in the prior art: in the prior art, only the combination probability of historical data is based, the insurance scene and the vehicle type are not considered, the vehicle composition principle, the internal and external structures and the mutual relation of the vehicle composition principle and the internal and external structures are not considered, and whether the loss combination probability in the historical case is reasonable under different scenes is not considered, so that the prior art can only be limited to the judgment of the occurrence frequency of the loss combination and cannot accurately infer the rationality of the case corresponding to the loss assessment information.

Disclosure of Invention

In order to solve the technical problems, embodiments of the present invention provide a vehicle damage assessment abnormality identification method, which checks damage assessment information in combination with historical data, a risk scene, and a vehicle type, in consideration of various factors such as a vehicle composition principle, internal and external structures, and a mutual relationship thereof, in the technical field of vehicle damage assessment, so as to further improve damage assessment efficiency and damage assessment accuracy, and reduce the difficulty level of damage assessment. The technical scheme is as follows:

in a first aspect, a vehicle damage assessment abnormality identification method includes:

acquiring current case information, wherein the current case information comprises a current vehicle damage assessment list and current accident information, the current vehicle damage assessment list comprises at least one piece of current damage information, and the current damage information comprises current damaged part information and corresponding repair information;

determining a first damage combination in the current damage information according to the current accident information, wherein the first damage combination comprises first damage information, the first damage information comprises first damage component information and corresponding repair information, and the occurrence probability of the first damage combination meets preset requirements in historical cases;

sequentially acquiring a part except the first damage combination in the current damage information to form a second damage combination, wherein the second damage combination comprises at least one piece of second damage information, and the second damage information comprises second damage part information and corresponding repair information;

determining a matching probability of the first damage combination and each piece of the second damage information;

and if the second damage information with the matching probability within a first preset numerical range exists, judging that the damage assessment of the current vehicle damage assessment list is abnormal.

According to the vehicle damage assessment abnormity identification method provided by the embodiment, a first damage combination in current damage information is determined according to current accident information, the occurrence probability of the first damage combination meets the preset requirement in a historical case, namely the first damage combination is a high-probability damage possible component combination in the historical case; removing the part of the first damage combination from the current damage information to form a second damage combination, and determining the matching probability of the first damage combination and each piece of second damage information; and the matching probability represents the occurrence probability of the damage combination in the historical case, and if second damage information with the matching probability within a first preset numerical range exists, namely the occurrence probability of the second damage combination does not accord with the statistical result of the historical case data, the current vehicle damage assessment list is judged to be abnormal in damage assessment.

Further optionally, before the determining the current damage information according to the current accident information, the method further includes:

calculating the similarity between each piece of current damage information and an abnormal data record in a preset abnormal data group;

judging whether the preset abnormal data group contains abnormal data records with the similarity larger than a second preset numerical range or not;

if yes, judging that the current vehicle loss assessment list is abnormal in loss assessment;

and if not, determining a first damage combination in the current damage information according to the current accident information.

Further optionally, the method further includes:

when the current vehicle damage assessment list is determined to be abnormal in damage assessment, combining the first damage combination and the second damage information with the matching probability within a first preset numerical range into an abnormal data record; and adding the synthesized abnormal data record into the preset abnormal data group.

Further optionally, before the obtaining of the current case information, the method further includes:

and constructing the preset abnormal data set based on the vehicle space structures of vehicles of different vehicle types.

Further optionally, the determining a first damage combination in the current damage information according to the current accident information includes:

acquiring accident topics matched with the current case from a first preset database according to the current accident information, wherein each accident topic comprises a plurality of part damage combinations, each part damage combination comprises necessary damage component information and the occurrence probability of the necessary damage component information in the historical case, and the necessary damage component information comprises the necessary damage component combination and a corresponding replacement mode;

and determining the part damage combination with the occurrence probability meeting the preset conditions in the accident topic as the first damage combination.

Further optionally, the determining the matching probability of the first impairment combination and each piece of the second impairment information is: and sequentially acquiring the matching probability of the first damage combination and each piece of second damage information from a second preset database.

Further optionally, the first preset database and the second preset database include a plurality of accident topics, and before the acquiring of the current case information, the method further includes:

acquiring historical case information, wherein the historical case information comprises historical accident information and a historical vehicle damage assessment list, and the historical accident information comprises one or more of collision scene information, vehicle type information and vehicle accessory information;

and constructing the first preset database and the second preset database by utilizing machine learning according to the historical case information, wherein the part damage combination of the second preset database comprises suspicious damaged part combination information and the necessary damaged part combination, and the suspicious damaged part combination information comprises suspicious damaged part combination and corresponding repair information.

Further optionally, the constructing the first preset database by using machine learning includes:

extracting the historical accident information and the historical vehicle damage assessment lists, and forming a preset accident topic set through a clustering algorithm, wherein the preset accident topic set comprises preset accident topics, damage component information extracted from the historical vehicle damage assessment lists and corresponding repair information;

under each preset accident topic, analyzing the damaged part information and the corresponding repair information through a correlation analysis algorithm to form a plurality of first damaged information and corresponding first preset values, and setting the first damaged information and the corresponding first preset values as a first preset database; and/or the presence of a gas in the gas,

the building the second pre-set database using machine learning comprises:

sequentially acquiring second damaged part information and corresponding repair information, wherein the second damaged part information and the corresponding repair information are parts of the historical vehicle damage assessment list except the first damaged part information and the corresponding repair information;

and combining the first damage information with each piece of second damage part information and corresponding repair information, calculating a second preset value of the first damage information, each piece of second damage part information and corresponding repair information, and setting the first damage information, the second damage part information, the corresponding repair information and the simultaneous occurrence probability as a second preset database.

Further optionally, the forming the preset accident topic set includes:

calculating the occurrence frequency of each part in each type of vehicle model, and selecting the part with the occurrence frequency larger than a second preset numerical range as a first candidate key part;

calculating the evaluation price of each part in each type of vehicle model, and selecting the part with the evaluation price larger than a third preset numerical value range as a second candidate key part;

determining the intersection of the first candidate key part and the second candidate key part as a current vehicle type candidate key part;

and constructing the preset accident topic set according to the current vehicle type key spare parts, the collision scene, the vehicle type and the vehicle accessory information.

Further optionally, the current case information is input into the historical case information, and the first preset database and the second preset database are updated.

In a second aspect, a vehicle damage assessment abnormality recognition apparatus includes

The system comprises a first determining module, a second determining module and a third determining module, wherein the first determining module is used for acquiring current case information, the current case information comprises a current vehicle damage assessment list and current accident information, the current vehicle damage assessment list comprises at least one piece of current damage information, and the current damage information comprises current damaged part information and corresponding repair information;

a second determining module, configured to determine a first damage combination in the current damage information according to the current accident information, where the first damage combination includes first damage information, the first damage information includes first damage component information and corresponding repair information, and an occurrence probability of the first damage combination satisfies a preset requirement in a historical case;

the identification module is used for sequentially acquiring a part, except the first damage combination, of the current damage information to form a second damage combination, wherein the second damage combination comprises at least one piece of second damage information, and the second damage information comprises second damage part information and corresponding repair information;

the calculation module is used for determining the matching probability of the first damage combination and each piece of second damage information;

and the output module is used for judging that the damage assessment of the current vehicle damage assessment list is abnormal if the second damage information with the matching probability within a first preset numerical range exists.

In a third aspect, a server comprises:

at least one processor; and

a memory coupled to the at least one processor; wherein,

the memory stores a computer program executable by the at least one processor to implement the vehicle damage assessment anomaly identification method described above.

In a fourth aspect, a computer-readable storage medium has stored therein a computer program which, when executed, is capable of implementing the above-described vehicle damage assessment abnormality identification method.

The embodiment of the invention identifies the vehicle damage assessment abnormity based on the judgment of the theme and key combination and the damage rationality calculation. Has the following beneficial effects:

1. by using the LDA model algorithm to determine the theme, the time spent for determining the theme is reduced, and the classification efficiency of the historical input information is greatly improved. Comprehensive classification coverage of historical input information can be achieved.

The determination of the theme by the LDA model algorithm not only comprises historical damage data, but also comprises various aspects such as a risk scene, a vehicle model, a vehicle composition principle, a vehicle internal and external structure, vehicle accessory information and the like. The accuracy of the identification method is greatly improved by the determination mode.

3. By using the association analysis algorithm to determine the key combination, association analysis is performed on the loss characteristics of the input information included in the theme under the condition that the theme is determined, and the accuracy of the identification method is further improved.

4. In the process of determining the key combination, the efficiency of the identification method is further improved by preferentially selecting the key combination with lower support degree. Meanwhile, the accuracy and efficiency of the identification method are further improved by a mode of calculating the reasonability of the loss through key combination and combination of other loss characteristics.

Drawings

Fig. 1 shows a flow chart of a vehicle damage assessment abnormality identification method according to an embodiment of the invention.

Fig. 2 shows a flow chart of another vehicle damage assessment abnormality identification method according to the embodiment of the invention.

Fig. 3 shows a flow chart of generating the first preset database in the vehicle damage assessment abnormality identification method according to the embodiment of the invention.

Fig. 4 shows a flow chart of generation of the second preset database in the vehicle damage assessment abnormality identification method according to the embodiment of the invention.

Fig. 5 is a schematic diagram showing a functional structure of a vehicle damage assessment abnormality recognition apparatus according to an embodiment of the present invention.

Detailed Description

The content of the invention will now be discussed with reference to a number of exemplary embodiments. It is to be understood that these examples are discussed only to enable those of ordinary skill in the art to better understand and thus implement the teachings of the present invention, and are not meant to imply any limitations on the scope of the invention.

As used herein, the term "include" and its variants are to be read as open-ended terms meaning "including, but not limited to. The term "based on" is to be read as "based, at least in part, on". The terms "one embodiment" and "an embodiment" are to be read as "at least one embodiment". The term "another embodiment" is to be read as "at least one other embodiment".

Example 1:

as shown in fig. 1, a flowchart of a vehicle damage assessment abnormality identification method according to an embodiment of the present invention is provided, where the vehicle damage assessment abnormality identification method includes:

101: and acquiring current case information.

The current case information comprises a current vehicle damage assessment list and current accident information, the current vehicle damage assessment list comprises at least one piece of current damage information, and the current damage information comprises current damaged part information and corresponding repair information; the current accident information includes: reporting description information to determine keywords and further determine a theme; accident reasons including actual accident causes such as collision, flooding, storm and the like, so that the theme of the preset accident theme set is accurately determined; and also collision scene, vehicle type and vehicle accessory information.

102: and determining a first damage combination in the current damage information according to the current accident information.

The first damage combination comprises first damage information, the first damage information comprises first damage part information and corresponding repair information, and the occurrence probability of the first damage combination meets the preset requirement in the historical case; determining the first damage combination according to the current accident information can fit to form a plurality of subjects, and the data source of the first damage combination comprises collision scenes, vehicle types, vehicle accessory information and other aspects. The accuracy of the identification method is greatly improved by the determination mode.

103: and sequentially acquiring the parts except the first damage combination in the current damage information to form a second damage combination.

The second damage combination comprises at least one piece of second damage information, and the second damage information comprises second damage component information and corresponding repair information;

104: and determining the matching probability of the first damage combination and each piece of second damage information.

The matching mode provided by the embodiment can further improve the matching speed, and the matching accuracy can be improved and the identification accuracy can be ensured through comparison of the matching probability.

105: and judging that second damage information with the matching probability within the first preset numerical range exists, and if the second damage information exists, executing the step 106.

And 106, judging that the damage assessment of the current vehicle damage assessment list is abnormal.

Example 2:

as an improvement of embodiment 1, another vehicle damage assessment abnormality identification method according to an embodiment of the present invention is provided, with reference to fig. 2, and includes:

and 201, acquiring historical case information.

The historical case information comprises historical accident information and a historical vehicle damage assessment list, wherein the historical accident information comprises one or more of collision scene, vehicle type and vehicle accessory information; the data structure of the historical accident information is embodied in the embodiment as follows: [ (rear-end collision), (bmad 740), (front bumper skin change, middle net change, left headlight change) ].

And 202, constructing a first preset database and a second preset database by utilizing machine learning according to the historical case information.

The part damage combination of the second preset database comprises suspicious damaged part combination information and the necessary damaged part combination, and the suspicious damaged part combination information comprises suspicious damaged part combination and corresponding repair information.

And 203, constructing a preset abnormal data set based on the vehicle accessory information of the vehicles of different vehicle types.

206, judging whether the preset abnormal data group contains abnormal data records with the similarity larger than a second preset numerical range, if so, executing step 211, and if not, executing step 207;

and 207, determining a first damage combination in the current damage information according to the current accident information.

The first damage combination comprises first damage information, the first damage information comprises first damage component information and corresponding repair information, and the occurrence probability of the first damage combination meets the preset requirement in the historical case;

sequentially acquiring parts except the first damage combination in the current damage information to form a second damage combination, wherein the second damage combination comprises at least one piece of second damage information, and the second damage information comprises second damage part information and corresponding repair information;

a probability of a match of the first impairment combination with each piece of second impairment information is determined 209.

210, judging whether second damage information with the matching probability within a first preset numerical range exists, and if so, executing a step 211;

if not, outputting the current vehicle damage assessment list that the damage assessment is normal (not shown in the figure).

And 211, judging that the damage assessment of the current vehicle damage assessment list is abnormal.

Combining the first damage combination and second damage information with the matching probability within a first preset value range into an abnormal data record 212;

the synthesized exception data record is added 213 to the preset exception data set.

According to the vehicle damage assessment abnormity identification method provided by the embodiment, key damaged parts and parts of different vehicle types under different scene themes and maintenance logic combinations thereof are defined to comprehensively judge the rationality of case vehicle loss. Not only can judge the rationality of case accessory combination and maintenance man-hour combination, but also can judge the rationality of the whole case.

Example 3

For the convenience of the reader to understand, the following describes the vehicle damage assessment abnormality identification method in detail with reference to specific examples, where the method includes generating a flowchart by using a first preset database and a second preset database, and the vehicle damage assessment abnormality identification method provided in this embodiment includes:

step 1: and acquiring historical case information.

The historical case information comprises historical accident information and a historical vehicle damage assessment list, wherein the historical accident information comprises one or more of collision scene, vehicle type and vehicle accessory information; the historical vehicle damage assessment list comprises damaged component information and corresponding repair information.

And generating a first preset database and a second preset database by using machine learning according to the historical case information. According to the method, the database is established by collecting various information such as collision scenes, vehicle types and vehicle accessory information corresponding to the collision scenes, the vehicle types and the vehicle accessory information in the historical case information, and the accuracy of the vehicle damage assessment abnormity identification method is further improved.

Generating the first preset database includes:

301: the preset accident topic set is constructed through a clustering algorithm according to the information of the vehicle type key spare parts, the collision scene, the vehicle type and the vehicle accessories, and the LDA model algorithm is adopted as the clustering algorithm to construct the preset accident topic set in the embodiment. The clustering accuracy can be further improved by using the LDA algorithm, so that the theme of the preset accident theme set is closer to the real collision condition. The generation mode of the key spare parts of the vehicle type is as follows: calculating the occurrence frequency of each part in each vehicle type, and selecting the part with the occurrence frequency larger than a fourth preset numerical range as a first candidate key part; calculating the evaluation price of each part in each vehicle type, and selecting the part with the evaluation price larger than a fifth preset numerical value range as a second candidate key part; and determining the intersection of the first candidate key part and the second candidate key part as the current vehicle type key candidate part. In this embodiment, the evaluation price is calculated in the following manner: and calculating the average value of the prices of each part of each vehicle type in the last N years, wherein the first 10 percent and the second 10 percent of the price interval of each part are required to be removed in the calculation process of the average value, and then the average value is calculated. Through the mode of selecting the key parts, the accuracy of the theme generated by the preset accident theme set can be further improved.

302: and storing the historical case information into corresponding subjects of the preset accident subject set in a classified manner, so that a plurality of pieces of historical case information are stored in each subject of the preset accident subject set.

303: and extracting damaged part information and corresponding repair information from the plurality of historical case information, and forming a historical first damage combination through a correlation analysis algorithm, wherein the historical first damage combination comprises historical first damage information, and the historical first damage information comprises the plurality of historical first damage part information and the corresponding repair information. In this embodiment, an FP-growth correlation analysis algorithm is preferably used to perform statistics on the occurrence frequencies of the damaged component information and the corresponding repair information, and the first preset database is determined according to the occurrence frequencies, where the first preset database includes not only the historical first damage combination but also the occurrence probability of the historical first damage combination, that is, the form of the multiple historical first damage information and the occurrence probability.

304: and judging the rationality of the first preset database according to the internal structure of the vehicle in the historical accident information, and screening combinations with low rationality in the first preset judgment rule to construct a preset abnormal data group. Because the historical first damage information in the first preset database comprises a plurality of pieces of historical first damage part information and corresponding repair information, the possibility and the reasonability of the simultaneous occurrence of the plurality of pieces of historical first damage part information and the corresponding repair information can be judged through the internal structure of the vehicle, and the historical first damage information with low possibility and reasonability is added into the preset abnormal data group as abnormal data. The generation method can further improve the precision and accuracy of vehicle damage assessment abnormity identification.

Generating the second preset database includes:

305: and sequentially acquiring second damage information of a second damage combination according to the historical case information, wherein the second damage combination is the damaged part information and the replacement information of the historical first damage combination in the historical vehicle damage assessment list.

306: and counting the probability of the simultaneous occurrence of each second damage information in the second damage combination and the first damage combination, and according to the matching probability, carrying out a second preset database. The second preset database is a form that a plurality of historical first damage combinations are provided with historical second damage part information, corresponding repair information and matching probability in parallel.

Step 2: acquiring current case information, wherein the current case information comprises a current vehicle damage assessment list and current accident information, and the current vehicle damage assessment list comprises current damage information; the current accident information includes information such as collision scene, vehicle type, and vehicle accessory information.

And step 3: comparing the current damage component information of the current damage information with the abnormal data records in the preset abnormal data group, and calculating the similarity between the current vehicle damage assessment list and the abnormal data records; judging whether the preset abnormal data group contains abnormal data records with the similarity larger than a second preset numerical range or not;

if yes, judging that the current vehicle damage assessment list is abnormal; and determining that the first damage combination in the current vehicle damage assessment list is abnormal data according to the current accident information and a first preset database. And when the current vehicle damage assessment list is determined to be abnormal in damage assessment, synthesizing the first damage combination and second damage information with the matching probability within a first preset numerical range into an abnormal data record and storing the abnormal data record into a preset abnormal database.

And if the preset abnormal data group does not contain the abnormal data record with the similarity larger than the third preset numerical range, determining a first damage combination in the current vehicle damage assessment list according to the current accident information and a first preset judgment rule. The abnormal data recording and judging method can further improve the identification precision.

And step 3: and determining a first damage combination in the current vehicle damage assessment list according to the current accident information and a first preset database. The method comprises the steps of extracting a plurality of pieces of current damaged part information and corresponding repair information of current damaged information in a current vehicle damage assessment list, combining the current damaged part information and the corresponding repair information, screening the combination according to a first preset database, extracting the combination with the lowest occurrence probability to judge, and determining the combination as a first damaged combination if historical first damaged part information and corresponding repair information in the combination can be matched with the current damaged part information and corresponding repair information in the current damaged information one by one. Otherwise, extracting the combination with the second lowest occurrence probability for the above judgment. And sequentially judging from low to high in occurrence probability, and selecting the current damaged part information with the largest covering quantity and the corresponding changed historical first damaged combination from the first preset database as the first damaged combination of the current accident information. The identification mode from low occurrence probability to high occurrence probability is more in line with the principle of vehicle damage accidents, and the identification speed of the vehicle damage assessment abnormity identification method can be improved.

And 4, step 4: and sequentially acquiring second damage information of a second damage combination according to the current case information, wherein the second damage combination is the damaged part information excluding the first damage combination and the corresponding repair information in the current vehicle damage assessment list. And sequentially acquiring second damage information in the second damage combination. And sequentially extracting the matching probability of each second damage component in the second damage combination, the repair information of the second damage component and the first damage combination according to the current accident information and a second preset database, and determining the second damage component with the matching probability within a first preset numerical range and the repair information of the second damage component as abnormal damage assessment information. In this embodiment, there are two ways to confirm the impairment anomaly information: firstly, after a first damage combination is determined, inputting the first damage combination into a second preset database to obtain a plurality of historical second damage combinations, combining each second damage component and repair information thereof in the current second damage combination with the first damage combination, extracting matching probability in the second preset database, comparing the matching probability with a first preset numerical range, and if the matching probability is in the first preset numerical range, combining the first damage combination with the second damage component and repair information thereof in the corresponding second damage combination and marking as fixed-damage abnormal information. Secondly, after a second damage component and the repair information thereof of a second damage combination are determined, the second damage component and the repair information thereof are input into a second preset database to obtain a corresponding historical first damage combination. And performing matching probability calculation on the obtained multiple historical first damage combinations and the first damage combinations in the current vehicle damage assessment list, and if the matching probability is within a first preset value range, combining the first damage combinations with the second damage components and the repair and replacement information in the corresponding second damage combinations and marking as abnormal damage assessment information.

And 5: and inputting the current vehicle damage assessment list and the current accident information in the current case information into historical case information, and updating the first preset database and the second preset database.

Example 4:

the embodiment specifically discloses specific steps of a vehicle damage assessment abnormality identification method, as shown in fig. 3 and 4, a first preset database and a second preset database need to be determined in advance:

step 1: and extracting historical case information to generate a preset accident topic set. And calling an LDA model algorithm, wherein the algorithm extracts and analyzes each piece of data of historical case information including historical accident information and a historical vehicle damage assessment list, and classifies the data according to word frequency. And classifying the classification mode according to data in multiple aspects of vehicle type key spare parts, collision scenes, vehicle types and vehicle accessory information included in the historical case information so as to construct a preset accident topic set. The collision scene is the accident reason, and the collision scene includes: accident type, collision classification and collision subdivision. In the present embodiment, the accident cause includes: single-party accidents, double-party accidents; the collision scene of the single accident comprises the following steps: the collision of stone pier-shaped objects, the collision of wall bodies, the collision of columnar objects, the collision of irregular objects and the support of the bottom. The collision scene of the accidents of the two parties comprises: meeting, turning, rear-end collision, sliding, backing and doubling. The collision subdivision is detailed in the following table:

TABLE 1

In this embodiment, the selection manner of the key spare parts of the vehicle type is as follows: firstly, calculating the occurrence frequency of each vehicle accessory, and in the embodiment, selecting the vehicle accessory with the occurrence frequency of more than 5 per thousand as a 'candidate key part 1'; second, the price of each vehicle component is calculated, and in the present embodiment, a vehicle component having an evaluation price of more than 100 dollars is selected as the "candidate key part 2". And thirdly, taking the intersection of the 'alternative key part 1' and the 'alternative key part 2' as a final alternative key part of the LDA model algorithm.

According to the set number X of the topics, the LDA model algorithm classifies the words according to the word frequency of various words appearing in the historical case information, and therefore X topics are formed. Each topic item comprises historical case information which is divided according to an LDA model algorithm, and the historical case information comprises historical damage component information and corresponding repair information. In the present embodiment, the data structure of the historical damaged component information and the corresponding repair information included in the historical case information is represented in the form of "[ (vehicle failure component + (repair or replacement), support) ]". The damaged part information and the corresponding repair information in one historical case information may include a plurality of single point data with different numbers, that is, the single damaged part information and the corresponding repair information. For example, in this embodiment, the damaged component information and the corresponding repair information in the history information include: "[ (front bumper beam, 0.87), (headlamp (right) beam, 0.77), (front fender beam, 0,56), (mid net beam, 0.41), (doubling, 0.40) ]", which the LDA model algorithm classifies into the subject designation "right front doubling". In the embodiment, the collision scene, the vehicle type and the vehicle accessory information in the current accident information and the vehicle type alternative key parts are extracted in the theme generation process for analysis, so that the accuracy of the theme is improved, and the theme is closer to the real situation of the accident.

Step 2: a first combination of lesions under the same topic is determined. On the basis of determining the theme, extracting all historical case information in the theme, searching relevance and causal structures of different damage component information and corresponding repair information included in the historical case information by using a relevance analysis algorithm according to damage features in the historical case information, wherein an FP-growth algorithm is used in the embodiment. The method comprises the steps of extracting occurrence frequencies of damaged part information and corresponding repair information combinations in historical case information, forming support degrees through calculation, selecting combinations higher than a certain support degree (supportRate) to mark as historical first damaged combinations, wherein the historical first damaged combinations comprise a plurality of historical first damaged part information and corresponding repair information, and then determining occurrence probability of the historical first damaged parts. The first damage combination and the corresponding occurrence probability are used as data values to form a first preset database.

The screened historical first damaged part information and the corresponding repair information in the historical first damaged combination may not completely cover all damaged part information in the historical case information, and part of damaged part information may not enter the historical first damaged combination. And marking the damaged part information and the corresponding repair information which do not enter the historical first damage combination as the historical second damage information and the corresponding repair information, and extracting the damaged part information and the corresponding repair information to form the historical second damage combination. And combining the historical second damage component information and the corresponding repair information in the historical first damage combination and the historical second damage combination, calculating the common occurrence probability of the historical first damage combination and the historical second damage combination, and further determining the matching probability of the historical first damage combination and the historical second damage combination. And forming a second preset database by combining the combination result of each piece of historical second damage component information and corresponding repair information in the historical first damage combination and the historical second damage combination.

And performing secondary screening of the historical first damage combination and the second damage combination in the first preset database according to the vehicle accessory information, marking the combination which does not accord with the vehicle accessory information in the historical first damage combination as abnormal, and storing the combination into a preset abnormal data group of a corresponding theme. In the embodiment, in the secondary screening process, the mutual combination relationship among the vehicle accessories is obtained through the information of the vehicle accessories, the rationality judgment of the simultaneous occurrence of the historical first damage combination and the historical second damage combination is carried out through the mutual combination relationship among the vehicle accessories, and the unreasonable historical first damage combination is marked as the abnormal data record.

In this embodiment, the above steps complete the determination of the preset accident topic set, the first damage combination, the occurrence probability, the first preset database and the second damage combination, the matching probability, and the second preset database. Based on this, a vehicle damage assessment abnormality identification method proposed by an embodiment of the present invention includes:

step 1: acquiring current case information, wherein the current case information comprises a current vehicle damage assessment list and current accident information, and the current vehicle damage assessment list comprises current damage information; the current accident information includes information such as collision scene, vehicle type, and vehicle accessory information.

Step 2: comparing the current damage information with abnormal data records in a preset abnormal data group, and calculating the similarity between a current vehicle damage assessment list and the abnormal data records; judging whether the preset abnormal data group contains abnormal data records with the similarity larger than a second preset numerical range or not; if yes, judging that the current vehicle damage assessment list is abnormal in damage assessment; when the current vehicle damage assessment list is determined to be abnormal in damage assessment, combining the first damage combination and second damage information with the matching probability within a first preset numerical range into an abnormal data record; and adding the synthesized abnormal data record into a preset abnormal data group.

And if the preset abnormal data group does not contain the abnormal data record with the similarity larger than the second preset numerical range, determining a first damage combination in the current vehicle damage assessment list according to the current accident information and the first preset database.

And step 3: and determining a first damage combination in the current vehicle damage assessment list according to the current accident information and a first preset database. The method comprises the steps of extracting current damage information in a current vehicle damage assessment list to combine, screening the combination according to a first preset database, and sequentially extracting historical first damage combinations corresponding to the occurrence probability from low to high to compare the historical first damage combinations with current damaged component information and corresponding repair information in the current damage information. And determining the combination which has the lowest occurrence probability and covers the current damaged part information and the corresponding repair information most in the corresponding historical first damaged combination as the first damaged combination. For example, in the present embodiment, the current vehicle damage assessment list with the current case information is: [ front bumper frame (repair), front bumper skin (exchange), condenser (exchange), middle net (exchange), glass kettle (exchange) ]. The first damage combination and the first preset database formed under the LDA model algorithm and the correlation analysis algorithm are as follows:

TABLE 2

Key combination	Key support
		Front bumper frame (repair)	0.098123749
Front bumper frame (repair), middle net (change)	0.068369521
		Front bumper frame (repair), front bumper skin (exchange)	0.094179559
Front bumper frame (repair), middle net (change), front bumper skin (change)	0.065843387
		Front bumper frame (repair), front bumper skin (exchange), condenser (exchange)	0.035127162
Front bumper frame (exchange), middle net (exchange), front bumper skin (exchange), condenser (exchange)	0.029034511

According to the vehicle collision principle, the more damaged part information included in the first damage combination means the more damaged vehicle accessories, which means the higher severity of the vehicle accident and the lower probability of occurrence, but at the same time, the more damaged part information in the current case information can be covered by the first damage combination, and the data similarity with the current case information is higher. Therefore, the key combination with lower support degree is preferentially selected to be matched with the damage characteristics in the input information. The matching method improves the identification speed to a certain extent, so that the overall identification method is improved.

In this embodiment, the lowest support [ front bumper frame (exchange), middle net (exchange), front bumper skin (exchange), condenser (exchange) ] is selected to match with the damaged component information and the corresponding repair information in the current vehicle damage assessment list. The matching fails because the repair patterns of the front bumper frame are different. And then selecting the (front bumper framework (repair), front bumper skin (replacement) and condenser (replacement)) with the second to last support degree to match with the current vehicle damage assessment list, wherein the matching is successful at the moment. And after the matching is successful, calling a second preset database for further judgment.

TABLE 3

Key combination	Characteristics of injury	Rationality
			Front bumper frame (repair), front bumper skin (exchange), condenser (exchange)	Middle net (trade)	87.3%
Front bumper frame (repair), front bumper skin (exchange), condenser (exchange)	Front bumper lining (changeable)	83.9%
			Front bumper frame (repair), front bumper skin (exchange), condenser (exchange)	Bass horn (trade)	86.2%
Front bumper frame (repair), front bumper skin (exchange), condenser (exchange)	Radiator frame (changeable)	82.8%
			Front bumper frame (repair), front bumper skin (exchange), condenser (exchange)	Radiator frame (repair)	77.4%
Front bumper frame (repair), front bumper skin (exchange), condenser (exchange)	Carafe (exchange)	0.1%
			Front bumper frame (repair), front bumper skin (exchange), condenser (exchange)	Radiator (changeable), radiator frame (changeable)	0.5%

And 4, step 4: and combining the successfully matched first damage combination with other damage part information which is not included in the first damage combination in the current case information, namely calculating the simultaneous occurrence probability of the second damage combination and the first damage combination, calculating the matching probability and a second preset numerical range, combining the second damage combination which is lower than the second preset numerical range with the first damage combination, marking the second damage combination as the abnormal damage assessment information, and outputting the abnormal damage assessment information. In this embodiment, the first lesion is combined: [ front bumper skeleton (repair), front bumper skin (trade), condenser (trade) ] and the second damage combination: after the combination of [ carafe (trade) ], the matching probability was 0.1%. And marking the data as an abnormal data record and outputting the data.

Example 5

A vehicle damage assessment abnormality recognition apparatus according to an embodiment of the present invention is, as shown in fig. 5, provided with:

the first determining module 51 is configured to obtain current case information, where the current case information includes a current vehicle damage assessment list and current accident information, the current vehicle damage assessment list includes at least one piece of current damage information, and the current damage information includes current damaged component information and corresponding repair information;

a second determining module 52, configured to determine a first damage combination in the current damage information according to the current accident information, where the first damage combination includes first damage information, the first damage information includes first damage component information and corresponding repair information, and an occurrence probability of the first damage combination satisfies a preset requirement in a historical case;

the identification module 53 is configured to sequentially obtain a portion of the current damage information excluding the first damage combination to form a second damage combination, where the second damage combination includes at least one piece of second damage information, and the second damage information includes second damage component information and corresponding repair information;

a calculating module 54, configured to determine a matching probability of the first impairment combination and each piece of the second impairment information;

and the output module 55 is configured to determine that the damage assessment of the current vehicle damage assessment list is abnormal if the second damage information with the matching probability within the first preset numerical range exists.

Example 6

In accordance with another aspect of the present invention, there is provided a server comprising:

at least one processor; and

a memory coupled to the at least one processor; wherein,

the memory stores a computer program executable by the at least one processor to implement the vehicle damage assessment abnormality identification method of the present invention.

Example 7

According to another aspect of the present invention, there is provided a computer-readable storage medium having stored thereon a computer program which, when executed, is capable of implementing the method of the present invention.

Those of ordinary skill in the art will appreciate that the various illustrative modules and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described apparatuses and devices may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is merely a logical division, and in actual implementation, there may be other divisions, for example, multiple modules or components may be combined or integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or modules, and may be in an electrical, mechanical or other form.

The modules described as separate parts may or may not be physically separate, and parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the embodiment of the present invention.

In addition, functional modules in the embodiments of the present invention may be integrated into one processing module, or each of the modules may exist alone physically, or two or more modules are integrated into one module.

The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method for transmitting/receiving the power saving signal according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a U disk, a removable hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

The above description is only a preferred embodiment of the application and is illustrative of the principles of the technology employed. It will be appreciated by a person skilled in the art that the scope of the invention as referred to in the present application is not limited to the embodiments with a specific combination of the above-mentioned features, but also covers other embodiments with any combination of the above-mentioned features or their equivalents without departing from the inventive concept. For example, the above features may be replaced with (but not limited to) features having similar functions disclosed in the present application.

It should be understood that the order of execution of the steps in the summary of the invention and the embodiments of the present invention does not absolutely imply any order of execution, and the order of execution of the steps should be determined by their functions and inherent logic, and should not be construed as limiting the process of the embodiments of the present invention.

Claims

1. A vehicle damage assessment abnormality identification method is characterized by comprising the following steps:

acquiring historical case information, wherein the historical case information comprises historical accident information and a historical vehicle damage assessment list; the historical accident information comprises one or more of collision scene, vehicle type and vehicle accessory information;

extracting the historical accident information and the historical vehicle damage assessment list, calculating the occurrence frequency of each part in each type of vehicle, and selecting the part with the occurrence frequency larger than a second preset value range as a first candidate key part; calculating the evaluation price of each part in each type of vehicle model, and selecting the part with the evaluation price larger than a third preset numerical value range as a second candidate key part; determining the intersection of the first candidate key part and the second candidate key part as a current vehicle type candidate key part;

forming a preset accident topic set through a clustering algorithm according to the current vehicle type alternative key parts, the collision scene, the vehicle type and the vehicle accessory information;

the preset accident theme set comprises a preset accident theme, damage component information extracted from a plurality of historical vehicle damage assessment lists and corresponding repair information; analyzing the damaged part information and the corresponding repair information through a correlation analysis algorithm on each preset accident topic in the preset accident topic set, and constructing a first preset database and a second preset database by using machine learning, wherein the construction of the first preset database by using machine learning is as follows: forming the first preset database according to an LDA model algorithm and an FP-growth correlation analysis algorithm;

determining a first damage combination in the current damage information according to the current accident information and a first preset database, wherein the first damage combination comprises first damage information, the first damage information comprises first damage component information and corresponding repair information, and the occurrence probability of the first damage combination meets preset requirements in historical cases;

determining the matching probability of the first damage combination and each piece of second damage information according to a second preset database;

2. The vehicle damage assessment abnormality identification method according to claim 1, further comprising, before said determining a first damage combination in said current damage information from said current accident information:

3. The vehicle damage assessment abnormality identification method according to claim 2, further comprising:

when the current vehicle damage assessment list is determined to be abnormal in damage assessment, combining the first damage combination and the second damage information with the matching probability within a first preset numerical range into an abnormal data record;

and adding the synthesized abnormal data record into the preset abnormal data group.

4. The vehicle damage assessment abnormality identification method according to claim 2, further comprising, before said acquiring current case information:

and constructing the preset abnormal data set based on the vehicle accessory information of vehicles of different vehicle types.

5. The vehicle damage assessment abnormality identification method according to claim 1, wherein said determining a first damage combination in said current damage information from said current accident information includes:

acquiring accident topics matched with the current case from a first preset database according to the current accident information, wherein each accident topic comprises a plurality of part damage combinations and accident reasons, each part damage combination comprises necessary damage component information and the occurrence probability of the necessary damage component information in the historical case, and the necessary damage component information comprises the necessary damage component combination and a corresponding replacement mode;

6. The vehicle damage assessment abnormality identification method according to claim 5, wherein said determining a matching probability of said first damage combination with each of said second damage information is: and sequentially acquiring the matching probability of the first damage combination and each piece of second damage information from a second preset database.

7. The vehicle damage assessment abnormality identification method according to claim 6, wherein said first preset database and said second preset database include a plurality of said accident topics, and before said acquiring current case information, said method further comprises:

8. The vehicle damage assessment abnormality identification method according to claim 7, wherein said constructing said first preset database by machine learning comprises:

the building the second pre-set database using machine learning comprises:

9. The vehicle damage assessment abnormality identification method according to claim 1, wherein said current case information is inputted into said historical case information, and said first preset database and said second preset database are updated.

10. A vehicle damage assessment abnormality recognition apparatus is characterized by comprising

The system comprises a first determination module, a second determination module and a third determination module, wherein the first determination module is used for acquiring historical case information, and the historical case information comprises historical accident information and a historical vehicle damage assessment list; the historical accident information comprises one or more of collision scene, vehicle type and vehicle accessory information; extracting the historical accident information and the historical vehicle damage assessment list, calculating the occurrence frequency of each part in each type of vehicle, and selecting the part with the occurrence frequency larger than a second preset value range as a first candidate key part; calculating the evaluation price of each part in each type of vehicle model, and selecting the part with the evaluation price larger than a third preset numerical value range as a second candidate key part; determining the intersection of the first candidate key part and the second candidate key part as a current vehicle type candidate key part; forming a preset accident topic set through a clustering algorithm according to the current vehicle type alternative key parts, the collision scene, the vehicle type and the vehicle accessory information; the preset accident theme set comprises a preset accident theme, damage component information extracted from a plurality of historical vehicle damage assessment lists and corresponding repair information; analyzing the damaged part information and the corresponding repair information through a correlation analysis algorithm on each preset accident topic in the preset accident topic set, and constructing a first preset database and a second preset database by using machine learning, wherein the construction of the first preset database by using machine learning is as follows: forming the first preset database according to an LDA model algorithm and an FP-growth correlation analysis algorithm; acquiring current case information, wherein the current case information comprises a current vehicle damage assessment list and current accident information, the current vehicle damage assessment list comprises at least one piece of current damage information, and the current damage information comprises current damaged part information and corresponding repair information;

a second determining module, configured to determine a first damage combination in the current damage information according to the current accident information and a first preset database, where the first damage combination includes first damage information, the first damage information includes first damage component information and corresponding repair information, and an occurrence probability of the first damage combination satisfies a preset requirement in a historical case;

the calculation module is used for determining the matching probability of the first damage combination and each piece of second damage information according to a second preset database;

11. A server, comprising:

at least one processor; and

a memory coupled to the at least one processor; wherein,

the memory stores a computer program executable by the at least one processor to implement the vehicle damage assessment abnormality identification method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that a computer program is stored in the computer-readable storage medium, which, when executed, is capable of implementing the vehicle damage assessment abnormality identification method according to any one of claims 1 to 9.