CN110570317A - Method and device for vehicle nuclear damage - Google Patents
Method and device for vehicle nuclear damage Download PDFInfo
- Publication number
- CN110570317A CN110570317A CN201811014319.4A CN201811014319A CN110570317A CN 110570317 A CN110570317 A CN 110570317A CN 201811014319 A CN201811014319 A CN 201811014319A CN 110570317 A CN110570317 A CN 110570317A
- Authority
- CN
- China
- Prior art keywords
- damage
- correlation
- injuries
- belong
- features
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/08—Insurance
Abstract
according to one implementation mode of the method, damage detection results of damaged vehicles are obtained, accident information at least comprising damage features and damage association features is extracted from the damage detection results, so that damage of each place is classified according to whether the damage belongs to the same accident or not, and a core damage result is determined according to the classification result. Since more abundant damage information is extracted to complete the core damage, the accuracy of the core damage can be improved.
Description
Technical Field
One or more embodiments of the present disclosure relate to the field of computer technology, and more particularly, to a method and apparatus for vehicle core loss via a computer.
Background
The vehicle is a generic term for various vehicles, which may include, for example, bicycles, cars, trucks, vans, trains, and the like. In the event of vehicle damage, the vehicle is often damaged by accident or human error, such as vehicle scraping, collision, etc. Subsequent processing of damaged vehicles often involves the identification of vehicle damage to provide basis for vehicle repair, insurance claims, and the like. Taking insurance claims as an example, a damage assessment process for a damaged vehicle may include: the insurance company sends out professional personnel for surveying and damage-fixing, gives out a maintenance scheme and an amount of compensation after surveying and damage-fixing, and uploads a damage-fixing picture to the background for the background checker to check the damage and check the price. The damage checking is generally to check the damage and check whether the determined damage is the damage of the accident, so as to avoid the situation that a user reports a plurality of accidents to an insurance company for cheating.
in the conventional technology, the nuclear damage process is often judged manually according to the damage assessment picture. This approach is labor intensive. In some implementations, it is also possible to determine whether the damage assessment result is reasonable by making simple rules, such as that the doors on both sides cannot be damaged simultaneously, so as to perform preliminary damage assessment. However, such rule making is too simple or absolute, often not covering all situations. For example, some injuries that meet the rules may not belong to the same incident. Therefore, it is necessary to provide a more reasonable way for checking the damaged vehicle by using more damage information, so as to improve the accuracy of checking the damaged vehicle.
disclosure of Invention
one or more embodiments of the present disclosure describe a method and an apparatus for vehicle core loss, which may extract damage features and damage associated features from a damage detection result of a damaged vehicle, and classify damages at various places based on the damage features and the damage associated features, so that core loss can be performed through richer damage information, and accuracy of core loss is improved.
according to a first aspect, there is provided a method for vehicle core loss, the method comprising: obtaining damage detection results of damaged vehicles, wherein the damage detection results comprise damage results of multiple damages; determining damage features and damage associated features of the damaged vehicle according to the damage detection results, wherein the damage features are extracted through the damage results of a single damage, and the damage associated features are used for describing the correlation of the damage results between at least two damages; and classifying the damage according to whether the damage belongs to the same accident or not at least based on the damage characteristics and the damage correlation characteristics so as to determine a core damage result based on the classification result.
In one embodiment, the impairment characterization comprises at least: a component on which the damage is located, a location of the damage on the damaged vehicle, a type of damage, a degree of damage.
In one embodiment, the damage-related features include one or more of a relative location feature of the damage, a distribution feature of the damage on the vehicle, a correlation feature between components where the damage is located, a damage-type correlation feature, and a damage-level correlation feature.
According to one embodiment, classifying the injuries according to whether the injuries belong to the same accident or not based on at least the injury characteristics and the injury correlation characteristics comprises: clustering the lesions based on at least the lesion features and the lesion correlation features; and determining the damage belonging to the same cluster in the clustering result as belonging to the same accident.
in one possible design, clustering based at least on the injury features and the injury-associated features includes: for any two injuries, inputting the respective injury characteristics of the two injuries and the injury correlation characteristics between the two injuries into a predetermined distance function to obtain the distance between the two injuries belonging to the same accident; and clustering the damage on the basis of the distance.
According to another embodiment, classifying the injuries according to whether the injuries belong to the same accident at least based on the injury characteristics and the injury correlation characteristics comprises: inputting the damage characteristics of the first damage, the damage characteristics of the second damage and the damage association characteristics between the first damage and the second damage into a pre-trained judgment model aiming at the first damage and the second damage in all the damages; and determining whether the first damage and the second damage belong to the same accident or not according to the output result of the judgment model.
In an alternative embodiment, the output comprises a probability that the first and second injuries belong to the same incident; and determining whether the first damage and the second damage belong to the same accident according to the output result of the judgment model comprises: determining that the first and second injuries belong to the same incident if the output is greater than a predetermined probability threshold.
In an alternative embodiment, the judgment model is obtained by training the following method: adding a plurality of damage sample pairs marked with whether the damage sample pairs belong to the same accident into a training sample set; respectively extracting characteristics of the damage samples of two damages in each damage sample pair; extracting the correlation characteristics of the damage samples between the two damages in each damage sample; and taking the damage sample characteristics and the damage sample correlation characteristics as input, and adjusting model parameters according to the labeling result of whether the damage sample characteristics and the damage sample correlation characteristics belong to the same accident so as to train the judgment model.
According to a second aspect, there is provided an apparatus for vehicle core loss, the apparatus comprising: the system comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is configured to acquire damage detection results of damaged vehicles, and the damage detection results comprise damage results of multiple damages; a determination unit configured to determine a damage feature and a damage-related feature of the damaged vehicle according to the damage detection result, the damage feature being a feature extracted from a damage result of a single damage, the damage-related feature being used to describe a correlation of the damage results between at least two damages; and the classification unit is configured to classify all the injuries according to whether the injuries belong to the same accident or not at least based on the injury characteristics and the injury correlation characteristics so as to determine a core injury result based on a classification result.
According to a third aspect, there is provided a computer readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method of the first aspect.
According to a fourth aspect, there is provided a computing device comprising a memory and a processor, wherein the memory has stored therein executable code, and wherein the processor, when executing the executable code, implements the method of the first aspect.
According to the method and the device for vehicle core damage provided by the embodiment of the specification, damage detection results of damaged vehicles are obtained, accident information at least comprising damage features and damage association features is extracted from the damage detection results, so that damage of each place is classified according to whether the damage belongs to the same accident or not, and the core damage results are determined based on the classification results. Since more abundant damage information is extracted to complete the core damage, the accuracy of the core damage can be improved. Furthermore, the labor can be saved, and the core loss efficiency can be improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 illustrates a schematic diagram of an implementation scenario of an embodiment disclosed herein;
FIG. 2 illustrates a flow diagram of a method for vehicle loss of core, according to one embodiment;
FIG. 3 illustrates a specific example of clustering lesions;
FIG. 4 shows a specific example of classifying lesions by a decision model;
FIG. 5 shows a schematic block diagram of an apparatus for vehicle core loss according to one embodiment.
Detailed Description
The scheme provided by the specification is described below with reference to the accompanying drawings.
Fig. 1 is a schematic view of an implementation scenario of an embodiment disclosed in this specification. In the application scenario shown in fig. 1, it is a specific scenario of insurance claims. In the scene, when the vehicle has an accident, a user acquires the field information of the damaged vehicle through the field acquisition equipment, and the damaged vehicle can be detected through the field information. Then, the damage detection result is uploaded to the processing platform in fig. 1, and the processing platform completes the damage process and feeds back the damage process to the user. It should be noted that the damage detection process may be completed by the processing platform in fig. 1, or may be completed by other processing platforms, which is not limited in this specification. The method for vehicle loss checking in the embodiment of the specification is suitable for a loss checking process of a processing platform.
The processing platform may first obtain damage detection results for the damaged vehicle. It is understood that the lesion detection result may include one or more lesions. For the condition that only one damage is needed, the damage can be finished only by judging whether the damage meets the current accident or not. Under the condition that the damage detection result comprises damage results of multiple damages, the processing platform can further determine accident information of the damaged vehicle according to the damage detection result, classify all the damages according to whether the damages belong to the same accident or not according to the accident information, and determine a core damage result based on the classification result. The accident information may be information related to distinguishing whether the damage belongs to the same accident, and may include at least a damage characteristic and a damage-related characteristic. The damage features may be features extracted from the damage results of a single damage, such as one or more of the part on which the damage is located, the location of the damage on the damaged vehicle, the category of the damage, the degree of the damage, and so forth. The damage-associated feature may be a feature that describes a correlation of damage results between at least two damages, such as one or more of a positional relationship between two damages, a positional correlation between parts where the damages are located, a damage-type correlation, and so on.
It can be understood that the damage at each position is classified according to whether the damage belongs to the same accident or not according to the accident information, and each classification can correspond to one accident in the obtained classification result. Therefore, whether the current accident is met or not can be judged for each category, or the category meeting the current accident is selected from each category, and then the core loss result can be obtained.
therefore, the information of single damage can be fully utilized, and simultaneously, the relevance relation between the damages is utilized to classify all the damages according to whether the damages belong to the same accident or not, and the damages of the same category belong to the same accident. Therefore, more bases are provided for the core loss process, and the core loss result is more accurate.
In the field of insurance claim settlement, the above damage checking results can also be used for distinguishing whether each damage is damage caused by current accidents. The accuracy of the loss checking result is improved, and the method is helpful for more effectively preventing the actions of fraud, cheat insurance and the like of the user. It should be noted that fig. 1 is only an example of an application scenario, and does not limit the method for vehicle loss control according to the embodiment of the present disclosure in the application scenario. The method for vehicle core loss according to the embodiment of the present disclosure may also be applied to other scenarios, which are not illustrated here.
it will be appreciated that the processing platform of FIG. 1 may be a variety of devices, apparatuses, such as a desktop computer, a server, etc., having certain computing capabilities. It is understood that the processing platform may also be an equipment cluster formed by the above devices and equipment. The following describes a specific implementation of the method for vehicle core loss.
FIG. 2 illustrates a flow chart of a method for vehicle core loss according to one embodiment of the present disclosure. The execution subject of the method may be, for example, the processing platform of fig. 1. As shown in fig. 2, the method may include the steps of: step 21, obtaining damage detection results of the damaged vehicle, wherein the damage detection results comprise damage results of multiple damages; step 22, determining damage features and damage associated features of the damaged vehicle according to the damage detection result, wherein the damage features are extracted through the damage result of a single damage, and the damage associated features are used for describing the correlation of the damage results between two damages; and step 23, classifying all the injuries according to whether the injuries belong to the same accident or not at least based on the injury characteristics and the injury correlation characteristics so as to determine a core injury result based on a classification result. The specific implementation of the above steps is described below.
first, in step 21, a damage detection result of the damaged vehicle is acquired. It will be appreciated that prior to a nuclear damage, damage detection is required on the damaged vehicle. Nuclear damage, which is a check for detected damage. The damage detection result may include a damage result obtained by detecting all damages of the damaged vehicle, or may be a damage result obtained by detecting a part of the damages. Depending on the on-site information uploaded by the user. The scheme of the embodiment of the present specification is mainly directed to a case where the damage detection result includes damage results of a plurality of damages.
The damage detection results can be described in a variety of formats. For example: by pictures in the field, e.g. by rectanglesFrame or border lines to delineate different lesions; describing by a three-dimensional model, such as marking lesions everywhere in the three-dimensional model with a solid frame or an edge line; described in text or other formatted file, such as left front door { (x)1，y1，z1)，(x2，y2，z2) … …, etc. The examples in this specification do not limit this.
Then, in step 22, the damage characteristic and the damage-related characteristic of the damaged vehicle are determined from the damage detection result. It is understood that various accident information may be included in the damage detection result. The accident information may be information related to an accident, for example, the information related to a rear-end collision may include rear bumper deformation, headlamp chipping, windshield chipping, and the like. At least an injury feature and an injury-associated feature can be extracted from the incident information.
In one aspect, the incident information may include impairment characteristics. Lesion features are often features extracted from the lesion results of a single lesion. The impairment characteristics may for example include, but are not limited to, at least one of the following: the component on which the damage is located, the location of the damage on the damaged vehicle, the type of damage, the extent of the damage, and the like. For example, for the damage described herein as "rear bumper deformation," the damage characteristics may include: the part where the damage is located, the rear bumper, the damage category, is deformed. In the case where the damage detection result of the damaged vehicle is described by other means, for example, "left front door { (x)1，y1，z1)，(x2，y2，z2) … …, severe distortion ", the damage characteristics may include not only: the part where the damage is located-the left front door, the damage category-the deformation, and may also include the location of the damage on the damaged vehicle { (x)1，y1，z1)，(x2，y2，z2) … … }, degree of damage-severity, etc. In some implementations, the damage characteristic may also include a degree of freshness, such as whether the scratch rusts or not, etc. The rusty scratches and shiny scratches may be scratches belonging to different accidents.
in another aspect, the incident information may also include impairment correlation characteristics. The lesion association feature may be a feature describing a correlation of lesion outcomes between at least two lesions. The impairment association characteristics may include, for example, but are not limited to, at least one of: damage relative position relation characteristics, damage distribution characteristics on the vehicle, correlation characteristics between parts where damage exists, damage category correlation characteristics, damage degree correlation characteristics and the like.
In one embodiment, the lesion association features may include lesion relative position features that describe the relative positional relationship between the lesions. The description of the relative position features of the lesion is also different according to different description modes of the lesion detection result. For example, as described more generally: lesion 1 is in front of lesion 2. As described in more detail: the back border of lesion 1 is 50 cm horizontally from the front border of lesion 2, and so on.
in one embodiment, the impairment correlation characteristic may comprise a distribution characteristic of the impairment over the vehicle. The distribution characteristic of the damage on the vehicle can comprise the distribution characteristic of a plurality of damages, and can also comprise the distribution characteristic of two damages. For example: the damage may include left front bumper scratch, left front door scratch, left rear bumper scratch, which may be continued from left front bumper, left front door, left rear door to left rear bumper scratch, the damage distribution characteristic may be a continuous extension distribution, and so on.
In one embodiment, the damage-associated features may also include correlation features between the components where the damage is located. The correlation between the damaged parts may be used to indicate whether the parts where the damage is located are correlated in position or structure, the degree of correlation, and the like. For example, the front left light and the front left bumper are strongly correlated, the front left light and the rear right light are uncorrelated, and the like. The correlation characteristic between the parts where the damage is located can also be represented by quantized correlation, for example, the correlation is calculated according to whether two parts are damaged simultaneously in the accident, whether the damage degree is consistent, and the like. For example, if two components tend to be damaged at the same time, and the degree of damage is consistent, the correlation may be 100%.
In one embodiment, the lesion association features may also include lesion type correlation features. The correlation of the damage types can be used to indicate whether the damage types of different damages are correlated or not, the correlation degree, and the like. For example, the damage types "chipping" and "scratching" may be associated to a very small extent in two damages, a windshield chipping-off and a right rear door scratch, whereas the damage types "chipping" and "cracking" are associated to a relatively large extent in two damages, a left front vehicle light chipping and a windshield cracking. Similarly, the correlation characteristic of the damage type can be expressed by a quantified correlation, such as an absolute correlation of 1 (i.e., a correlation of 1 when two damage types are absolutely correlated), a "fragmentation" and "scratch" correlation of 0.2, a "fragmentation" and "fragmentation" correlation of 0.99, and so on.
In one embodiment, the impairment correlation characteristics may further comprise impairment degree correlation characteristics. The correlation of the degree of damage can be used to indicate whether the degree of damage is correlated, the degree of correlation, and the like. It is understood that in the damage detection result, a damage degree feature can also be extracted, for example, the damage degree of slight scratching is slight, etc. The degree of damage correlation may be relatively small for both "mild" and "severe" damage degrees. The damage-level-correlation characteristic may also be represented by a quantified correlation-level value, e.g., the damage-level-correlation characteristics of "mild" and "severe" are 0.3.
In other embodiments, the incident information may include other information for distinguishing whether the same incident is present, such as the type of incident entered by the user, a scene picture, a text description, etc., in addition to the damage features and damage-associated features. Wherein, the information can assist in judging whether the damages at each place are the same accident. For example, in the event of a three-vehicle rear-end collision, the front end damage and the rear end damage may be considered as two accidents for the intermediate vehicle only by the damage characteristic and the damage correlation characteristic. At this time, information such as the type of the accident (e.g., three-car rear-end collision), the scene picture of the three-car rear-end collision, the textual description "three-car rear-end collision" input by the user, and the like, which is input by the user, is combined, so that the accident classification can be performed more accurately.
and step 23, classifying all the injuries according to whether the injuries belong to the same accident or not at least based on the injury characteristics and the injury correlation characteristics so as to determine a core injury result based on a classification result. It is understood that the damage in each place can be classified according to the accident information, and the damage in each place in each category may be damage belonging to the same accident.
According to an aspect of an embodiment, the lesions may be clustered according to lesion features. The clustering method may or may not pre-specify how many categories to divide into (e.g., k-means algorithm specifies number of categories k). In the embodiment of the present specification, since how many accidents each damage actually belongs to is not generally determinable, a Clustering method that does not specify in advance how many categories to classify into, such as mean-shift Clustering, DBSCAN (Density-based Clustering method with Noise), or the like, is applicable.
It will be appreciated that clustering tends to be performed by the distance between objects. In the present specification embodiment, the target of the clustering is an injury. In some implementations, the distance between two injuries can be represented by the distance that two injuries belong to the same incident. The distance at which two lesions belong to the same incident can be determined by a distance function. For the distance function, the damage characteristics and the damage association characteristics of any two damages are input, and the distance of the two damages belonging to the same accident is output. The parameters of the distance function can be determined by a logistic regression or linear regression method using samples labeled as if they were the same accident, such that a smaller distance indicates a greater probability that two lesions are caused by the same accident. By using the distance function, the distance between any given plurality of injuries can be calculated by using the injury characteristics and the injury association characteristics of the injuries so as to perform clustering. In one implementation, the lesions may be mapped to points in space or plane according to the obtained distances, and the lesions may be clustered by the mapped points. As shown in fig. 3, each small black dot may correspond to 1 lesion. The distance between a point and a point may be the distance that the corresponding two injuries belong to the same accident. In another implementation, each damage and a standard point (standard damage point) may be mapped to a space or a plane, and the damage may be clustered by the mapped point. At this time, the distance between the point in the clustering process and the point is determined by the aforementioned distance function.
Referring to fig. 3, taking the DBSCAN method as an example, during the clustering process, the scan radius 33(r value) and the minimum inclusion point minPoints may be determined first. One non-visited point 32 is selected and all nearby points within a distance of (including r) the scan radius 33 are found. If the number of nearby points is greater than or equal to minPoints. If the number of points near point 32 in FIG. 3 is greater than minPoints, then point 32 forms a cluster with the points near it and the point 32 is marked as visited (visited). Then recursively, all points within the cluster that are not marked as accessed (visited), such as point 33, are processed in the same way, thereby expanding the cluster. If the cluster is sufficiently expanded, i.e., all points within the cluster are marked as visited, then the same algorithm is used to process the points that are not visited. Until all points have been visited.
In this way, a plurality of clusters are obtained by clustering, each cluster can correspond to a damage of an accident, thereby classifying the damage of each place of the damaged vehicle.
according to another embodiment, as shown in fig. 4, the damage features (in fig. 4, the first damage features and the second damage features, respectively) of any two damages (such as the first damage and the second damage, where "the first" and "the second" are only used for distinguishing different damages) in each damage, and the damage-related features between the two damages may be input into the pre-trained judgment model, so as to obtain the output result of the judgment model. Then, whether the two injuries belong to the same accident or not can be determined according to the output result of the judgment model. The judgment model can be obtained by training through the following method: adding a plurality of damage sample pairs marked with whether the damage sample pairs belong to the same accident into a training sample set; respectively extracting characteristics of the damage samples of two damages in each damage sample pair; extracting the correlation characteristics of each damage sample to the damage samples between two damages; and taking the damage sample characteristics and the damage sample correlation characteristics as input, and adjusting model parameters to train a judgment model according to the marking result of whether the corresponding damage sample pair belongs to the same accident. Each damage sample pair includes two damages, and the damage sample characteristics and the damage sample association characteristics are similar to the damage characteristics and the damage association characteristics, which are not described herein again.
In one embodiment, the output of the judgment model may be a numerical value corresponding to the same accident and not belonging to the same accident, such as the output is 0 or 1, 0 indicates belonging to the same accident, 1 indicates not belonging to the same accident, and so on.
in another embodiment, the output result of the judgment model is the probability that the two damages corresponding to the input damage characteristic and the damage-related characteristic belong to the same accident. The greater the probability, the greater the likelihood that two lesions belong to the same incident. At this time, the output result may be further compared with a predetermined probability threshold, and in the case where the output result is greater than the predetermined probability threshold, it is determined that the two injuries belong to the same accident.
Therefore, whether every two damages belong to the same accident or not can be determined, so that the damages belonging to the same accident can be classified into one class, and the classification of the damages at each position on the damaged vehicle is completed.
according to the classification result, the core loss can be performed on each category to generate a core loss result, so that the manpower requirement and the manpower cost are greatly reduced.
For example, in some possible designs, after the classification results are obtained, the classification results may be presented to a core-loss worker. The nuclear casualty personnel only need to select one injury which is identical with the current accident. After receiving the selection of the damage by the nuclear damage personnel, the damage selected by the nuclear damage personnel and other damages belonging to the same category with the damage selected by the nuclear damage personnel can be automatically generated together to generate a nuclear damage result. In some implementations, incident information, such as the type of incident, may also be obtained for the loss core personnel to associate with the classification results for damage assessment. The incident type is, for example, a user declared case type. Optionally, applications for other categories of damage claims may also be rejected to the user.
the method for vehicle damage checking according to the embodiments of the present specification may be applied to anti-fraud processing in an insurance claim settlement process of an accident vehicle. For example, in the application of one insurance claim, a user combines the damages of a plurality of accidents into the current accident and reports the current accident to the insurance company. Although the damages are present according to the damage detection result, the damages may not be caused by the current accident, and fraud is caused for the insurance company. Therefore, the method for vehicle loss checking of the embodiment of the specification can assist an insurance company to carry out loss checking, so that the purpose of fraud prevention is achieved.
Reviewing the above process, the damage detection result of the damaged vehicle is obtained, accident information at least comprising damage features and damage correlation features is extracted from the damage detection result, so that the damage of each place is classified according to whether the damage belongs to the same accident, and the core damage result is determined based on the classification result. And because richer damage information is extracted for core damage, the accuracy of the core damage can be improved. Furthermore, the labor can be saved, and the core loss efficiency can be improved.
According to an embodiment of another aspect, there is also provided an apparatus for vehicle core loss. Fig. 5 shows a schematic block diagram of a vehicle impairment device 500 according to an embodiment. As shown in fig. 5, the apparatus 500 includes: an obtaining unit 51 configured to obtain a damage detection result of the damaged vehicle, wherein the damage detection result includes damage results of a plurality of damages; a determining unit 52 configured to determine a damage feature and a damage associated feature of the damaged vehicle according to the damage detection result, the damage feature being a feature extracted by a damage result of a single damage, the damage associated feature being used to describe a correlation of the damage results between at least two damages; and the classification unit 53 is configured to classify all the injuries according to whether the injuries belong to the same accident or not at least based on the injury characteristics and the injury correlation characteristics so as to determine a core injury result based on the classification result.
It will be appreciated that prior to a nuclear damage, damage detection is required on the damaged vehicle. The acquisition unit 51 may first acquire a damage detection result of damage detection of the damaged vehicle. The description form of the damage detection result may be a field picture, a three-dimensional model, a text, or a file with other format, which is not limited in the embodiments of the present specification.
the determination unit 52 may then determine accident information of the damaged vehicle from the damage detection result. Wherein the incident information includes at least an injury characteristic and an injury-associated characteristic. The lesion feature is a feature extracted by a lesion result of a single lesion, and may include, for example, but is not limited to, at least one of: the component on which the damage is located, the location of the damage on the damaged vehicle, the type of damage, the extent of the damage, and the like. The lesion association features may be used to describe the correlation of lesion outcomes between two lesions, which may include, but are not limited to, at least one of the following, for example: damage relative position relation characteristics, damage distribution characteristics on the vehicle, damage correlation characteristics between parts, damage degree correlation characteristics and the like. Optionally, the damage characteristics may also include how old and new the damage is, etc. In some embodiments, in addition to the injury feature and the injury related feature, the obtaining unit 51 may obtain other accident information for distinguishing whether the same accident is the same, such as the accident type input by the user, a scene picture, a text description, and the like.
next, the classification unit 53 may classify each of the injuries according to whether the injuries belong to the same accident or not based on at least the injury feature and the injury-related feature, so as to determine a core-damage result based on the classification result.
According to an aspect of the embodiment, the classification unit 53 may classify the damage by clustering the damage features. At this time, the classification unit 53 may calculate a distance between any two damages belonging to the same accident to perform clustering. The distance between the lesions may be calculated by a predetermined distance function. Alternatively, the classifying unit 53 may also map each damage to a point on a space or a plane according to the obtained distance, and cluster each damage according to the point obtained by mapping. Or, each damage and the standard point (standard damage point) may be mapped to a space or a plane, and the distance between the point and the point in the clustering process is determined by the distance function.
According to another embodiment, the classification unit 53 may also classify the damage by the judgment model. In this case, the classification unit 53 may input the damage features of two arbitrary damages in each of the damages and the damage-related features between the two damages to a pre-trained determination model, and obtain an output result of the determination model. Then, whether the two injuries belong to the same accident or not can be determined according to the output result of the judgment model. The output result of the judgment model may be a numerical value corresponding to the same accident or a numerical value not belonging to the same accident, for example, the output result is 0 or 1, or the probability that two damages belong to the same accident, which is not limited in the embodiments of the present specification. In case that the output result of the model is the probability that the two damages belong to the same accident, the classification unit 53 may further compare the output result with a predetermined probability threshold, and determine that the two damages belong to the same accident if the output result is greater than the predetermined probability threshold.
In one embodiment, the apparatus 500 may further include a training unit (not shown) configured to train a judgment model used by the classification unit 53 by: adding a plurality of damage sample pairs marked with whether the damage sample pairs belong to the same accident into a training sample set; respectively extracting characteristics of the damage samples of two damages in each damage sample pair; extracting damage sample correlation characteristics between two damages in each damage sample pair; and taking the damage sample characteristic and the damage sample correlation characteristic as input, and adjusting model parameters according to the labeling result of whether the same accident belongs to or not so as to train a judgment model.
Through above device, can extract abundanter damage information and accomplish the nuclear loss to improve the degree of accuracy of nuclear loss. Furthermore, the labor can be saved, and the core loss efficiency can be improved.
According to an embodiment of another aspect, there is also provided a computer-readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method described in connection with fig. 2.
according to an embodiment of yet another aspect, there is also provided a computing device comprising a memory and a processor, the memory having stored therein executable code, the processor, when executing the executable code, implementing the method described in connection with fig. 2.
Those skilled in the art will recognize that, in one or more of the examples described above, the functions described in this invention may be implemented in hardware, software, firmware, or any combination thereof. When implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium.
The above-mentioned embodiments, objects, technical solutions and advantages of the present invention are further described in detail, it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made on the basis of the technical solutions of the present invention should be included in the scope of the present invention.
Claims (18)
1. a method for vehicle core loss, the method comprising:
Obtaining damage detection results of damaged vehicles, wherein the damage detection results comprise damage results of multiple damages;
determining damage features and damage associated features of the damaged vehicle according to the damage detection results, wherein the damage features are extracted through the damage results of a single damage, and the damage associated features are used for describing the correlation of the damage results between at least two damages;
And classifying the damage according to whether the damage belongs to the same accident or not at least based on the damage characteristics and the damage correlation characteristics so as to determine a core damage result based on the classification result.
2. The method of claim 1, wherein the injury signature comprises at least: a component on which the damage is located, a location of the damage on the damaged vehicle, a type of damage, a degree of damage.
3. The method of claim 1, wherein the damage-associated features include one or more of damage-relative-location features, damage-distribution features on a vehicle, correlation features between parts where damage is located, damage-type correlation features, and damage-level correlation features.
4. The method of claim 1, wherein classifying the injuries according to whether the injuries belong to the same accident based on at least the injury signature and the injury correlation signature comprises:
clustering the lesions based on at least the lesion features and the lesion correlation features;
And determining the damage belonging to the same cluster in the clustering result as belonging to the same accident.
5. The method of claim 4, wherein clustering based at least on the impairment characterization and the impairment correlation characterization comprises:
For any two injuries, inputting the respective injury characteristics of the two injuries and the injury correlation characteristics between the two injuries into a predetermined distance function to obtain the distance between the two injuries belonging to the same accident;
And clustering the damage on the basis of the distance.
6. The method of claim 1, wherein classifying the injuries according to whether the injuries belong to the same accident based on at least the injury signature and the injury correlation signature comprises:
Inputting the damage characteristics of the first damage, the damage characteristics of the second damage and the damage association characteristics between the first damage and the second damage into a pre-trained judgment model aiming at the first damage and the second damage in all the damages;
And determining whether the first damage and the second damage belong to the same accident or not according to the output result of the judgment model.
7. The method of claim 6, wherein the output comprises a probability that the first and second injuries belong to the same incident; and
Determining whether the first damage and the second damage belong to the same accident according to the output result of the judgment model comprises:
determining that the first and second injuries belong to the same incident if the output is greater than a predetermined probability threshold.
8. The method of claim 6, wherein the decision model is trained by:
adding a plurality of damage sample pairs marked with whether the damage sample pairs belong to the same accident into a training sample set;
Respectively extracting characteristics of the damage samples of two damages in each damage sample pair;
extracting damage sample correlation characteristics between two damages in each damage sample pair;
and taking the damage sample characteristics and the damage sample correlation characteristics as input, and adjusting model parameters according to the labeling result of whether the damage sample characteristics and the damage sample correlation characteristics belong to the same accident so as to train the judgment model.
9. An apparatus for vehicle core loss, the apparatus comprising:
The system comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is configured to acquire damage detection results of damaged vehicles, and the damage detection results comprise damage results of multiple damages;
A determination unit configured to determine a damage feature and a damage-related feature of the damaged vehicle according to the damage detection result, the damage feature being a feature extracted from a damage result of a single damage, the damage-related feature being used to describe a correlation of the damage results between at least two damages;
and the classification unit is configured to classify all the injuries according to whether the injuries belong to the same accident or not at least based on the injury characteristics and the injury correlation characteristics so as to determine a core injury result based on a classification result.
10. the apparatus of claim 9, wherein the injury signature comprises at least: a component on which the damage is located, a location of the damage on the damaged vehicle, a type of damage, a degree of damage.
11. The apparatus of claim 9, wherein the damage-associated features include one or more of a relative location of damage feature, a distribution of damage on a vehicle feature, a correlation between parts where damage is located feature, a type of damage correlation feature, a degree of damage correlation feature.
12. The apparatus of claim 9, wherein the classification unit is further configured to:
Clustering the lesions based on at least the lesion features and the lesion correlation features;
and determining the damage belonging to the same cluster in the clustering result as belonging to the same accident.
13. The apparatus of claim 12, wherein the classification unit is further configured to:
For any two injuries, inputting the respective injury characteristics of the two injuries and the injury correlation characteristics between the two injuries into a predetermined distance function to obtain the distance between the two injuries belonging to the same accident;
And clustering the damage on the basis of the distance.
14. The apparatus of claim 9, wherein the classification unit is further configured to:
inputting the damage characteristics of the first damage, the damage characteristics of the second damage and the damage association characteristics between the first damage and the second damage into a pre-trained judgment model aiming at the first damage and the second damage in all the damages;
And determining whether the first damage and the second damage belong to the same accident or not according to the output result of the judgment model.
15. The apparatus of claim 14, wherein the output comprises a probability that the first and second injuries belong to the same incident; and
The classification unit is further configured to include:
Determining that the first and second injuries belong to the same incident if the output is greater than a predetermined probability threshold.
16. The apparatus of claim 14, wherein the apparatus further comprises a training unit configured to train the decision model by:
adding a plurality of damage sample pairs marked with whether the damage sample pairs belong to the same accident into a training sample set;
respectively extracting characteristics of the damage samples of two damages in each damage sample pair;
extracting damage sample correlation characteristics between two damages in each damage sample pair;
And taking the damage sample characteristics and the damage sample correlation characteristics as input, and adjusting model parameters according to the labeling result of whether the damage sample characteristics and the damage sample correlation characteristics belong to the same accident so as to train the judgment model.
17. A computer-readable storage medium, on which a computer program is stored which, when executed in a computer, causes the computer to carry out the method of any one of claims 1-8.
18. A computing device comprising a memory and a processor, wherein the memory has stored therein executable code that, when executed by the processor, performs the method of any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811014319.4A CN110570317B (en) | 2018-08-31 | 2018-08-31 | Method and device for vehicle nuclear damage |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811014319.4A CN110570317B (en) | 2018-08-31 | 2018-08-31 | Method and device for vehicle nuclear damage |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110570317A true CN110570317A (en) | 2019-12-13 |
| CN110570317B CN110570317B (en) | 2023-06-20 |
Family
ID=68772388
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811014319.4A Active CN110570317B (en) | 2018-08-31 | 2018-08-31 | Method and device for vehicle nuclear damage |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110570317B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113361424A (en) * | 2021-06-11 | 2021-09-07 | 爱保科技有限公司 | Intelligent loss assessment image acquisition method, device, medium and electronic equipment for vehicle |
| CN114596439A (en) * | 2022-03-20 | 2022-06-07 | 邦邦汽车销售服务(北京)有限公司 | Image-based intelligent damage assessment and check method and device for car insurance and electronic equipment |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5130116A (en) * | 1988-10-12 | 1992-07-14 | Centocor, Inc. | Radiotherapeutic immunoconjugates labeled with iodine-125 |
| JPH1166147A (en) * | 1997-08-21 | 1999-03-09 | Tsubasa Syst Kk | Estimation system for repair cost for vehicle involved in accident and recording medium thereof |
| CN201571156U (en) * | 2009-12-31 | 2010-09-01 | 合肥优视嵌入式技术有限责任公司 | Remote loss-confirming device for vehicle |
| US8799034B1 (en) * | 2013-03-08 | 2014-08-05 | Allstate University Company | Automated accident detection, fault attribution, and claims processing |
| US20150087279A1 (en) * | 2013-09-20 | 2015-03-26 | Better Mousetrap, LLC | Mobile accident processing system and method |
| US20160358081A1 (en) * | 2015-06-08 | 2016-12-08 | International Business Machines Corporation | Automated vehicular accident detection |
| US9773281B1 (en) * | 2014-09-16 | 2017-09-26 | Allstate Insurance Company | Accident detection and recovery |
| CN107392218A (en) * | 2017-04-11 | 2017-11-24 | 阿里巴巴集团控股有限公司 | A kind of car damage identification method based on image, device and electronic equipment |
| CN107403424A (en) * | 2017-04-11 | 2017-11-28 | 阿里巴巴集团控股有限公司 | A kind of car damage identification method based on image, device and electronic equipment |
| CN108364253A (en) * | 2018-03-15 | 2018-08-03 | 北京威远图易数字科技有限公司 | Car damage identification method, system and electronic equipment |
| CN108446618A (en) * | 2018-03-09 | 2018-08-24 | 平安科技（深圳）有限公司 | Car damage identification method, device, electronic equipment and storage medium |
-
2018
- 2018-08-31 CN CN201811014319.4A patent/CN110570317B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5130116A (en) * | 1988-10-12 | 1992-07-14 | Centocor, Inc. | Radiotherapeutic immunoconjugates labeled with iodine-125 |
| JPH1166147A (en) * | 1997-08-21 | 1999-03-09 | Tsubasa Syst Kk | Estimation system for repair cost for vehicle involved in accident and recording medium thereof |
| CN201571156U (en) * | 2009-12-31 | 2010-09-01 | 合肥优视嵌入式技术有限责任公司 | Remote loss-confirming device for vehicle |
| US8799034B1 (en) * | 2013-03-08 | 2014-08-05 | Allstate University Company | Automated accident detection, fault attribution, and claims processing |
| US20150087279A1 (en) * | 2013-09-20 | 2015-03-26 | Better Mousetrap, LLC | Mobile accident processing system and method |
| US9773281B1 (en) * | 2014-09-16 | 2017-09-26 | Allstate Insurance Company | Accident detection and recovery |
| US20160358081A1 (en) * | 2015-06-08 | 2016-12-08 | International Business Machines Corporation | Automated vehicular accident detection |
| CN107392218A (en) * | 2017-04-11 | 2017-11-24 | 阿里巴巴集团控股有限公司 | A kind of car damage identification method based on image, device and electronic equipment |
| CN107403424A (en) * | 2017-04-11 | 2017-11-28 | 阿里巴巴集团控股有限公司 | A kind of car damage identification method based on image, device and electronic equipment |
| CN108446618A (en) * | 2018-03-09 | 2018-08-24 | 平安科技（深圳）有限公司 | Car damage identification method, device, electronic equipment and storage medium |
| CN108364253A (en) * | 2018-03-15 | 2018-08-03 | 北京威远图易数字科技有限公司 | Car damage identification method, system and electronic equipment |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113361424A (en) * | 2021-06-11 | 2021-09-07 | 爱保科技有限公司 | Intelligent loss assessment image acquisition method, device, medium and electronic equipment for vehicle |
| CN114596439A (en) * | 2022-03-20 | 2022-06-07 | 邦邦汽车销售服务(北京)有限公司 | Image-based intelligent damage assessment and check method and device for car insurance and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110570317B (en) | 2023-06-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111488875B (en) | Vehicle insurance claim settlement loss checking method and device based on image recognition and electronic equipment | |
| US20240087102A1 (en) | Automatic Image Based Object Damage Assessment | |
| US11887064B2 (en) | Deep learning-based system and method for automatically determining degree of damage to each area of vehicle | |
| CN110569701B (en) | Computer-implemented vehicle damage assessment method and device | |
| CN109784170B (en) | Vehicle risk assessment method, device, equipment and storage medium based on image recognition | |
| US11127200B1 (en) | Photo deformation techniques for vehicle repair analysis | |
| US9014432B2 (en) | License plate character segmentation using likelihood maximization | |
| CN115605889A (en) | Method for determining damage to vehicle parts | |
| US11113582B2 (en) | Method and system for facilitating detection and identification of vehicle parts | |
| US11574395B2 (en) | Damage detection using machine learning | |
| CN111311540A (en) | Vehicle damage assessment method and device, computer equipment and storage medium | |
| US11410287B2 (en) | System and method for artificial intelligence based determination of damage to physical structures | |
| CN115810134B (en) | Image acquisition quality inspection method, system and device for vehicle insurance anti-fraud | |
| CN111222409A (en) | Vehicle brand labeling method, device and system | |
| CN109215119B (en) | Method and device for establishing three-dimensional model of damaged vehicle | |
| CN110570317A (en) | Method and device for vehicle nuclear damage | |
| CN112861567A (en) | Vehicle type classification method and device | |
| CN113408364B (en) | Temporary license plate recognition method, system, device and storage medium | |
| CN113673304B (en) | Vehicle-mounted expected functional safety hazard analysis and evaluation method based on scene semantic driving | |
| CN116091496B (en) | Defect detection method and device based on improved Faster-RCNN | |
| CN112241127B (en) | Automatic driving safety scoring method, automatic driving safety scoring device, computer equipment and storage medium | |
| CN110728582B (en) | Information processing method, device, storage medium and processor | |
| US20230342937A1 (en) | Vehicle image analysis | |
| CN110728276A (en) | License plate recognition method and device | |
| CN114842198A (en) | Intelligent loss assessment method, device and equipment for vehicle and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40018748 Country of ref document: HK |
|
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200924 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Innovative advanced technology Co.,Ltd. Address before: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant before: Advanced innovation technology Co.,Ltd. Effective date of registration: 20200924 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Advanced innovation technology Co.,Ltd. Address before: A four-storey 847 mailbox in Grand Cayman Capital Building, British Cayman Islands Applicant before: Alibaba Group Holding Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |