CN117197498A

CN117197498A - Vehicle flaw similarity recognition method and system

Info

Publication number: CN117197498A
Application number: CN202311117765.9A
Authority: CN
Inventors: 谭建斌; 杨亚刚; 刘璐; 张钦格; 苑海川; 徐宁
Original assignee: Peoples Insurance Company of China
Current assignee: Peoples Insurance Company of China
Priority date: 2023-08-31
Filing date: 2023-08-31
Publication date: 2023-12-08

Abstract

The application discloses a vehicle flaw similarity recognition method and system, and belongs to the technical field of computer vision. The vehicle flaw similarity recognition method comprises the following steps: determining the number of respective vehicle flaws in the first image and the second image to be identified, and the type and position of each flaw; extracting sub-images in the first image and the second image according to the type and the position of the flaw, and classifying to obtain the image category of the sub-images; marking the characteristic points of each sub-image according to the image category and a preset characteristic recognition model, wherein the preset characteristic recognition model is used for determining coordinate vectors of the characteristic points of the sub-images; and comparing the coordinate vector of the characteristic point of each sub-image in the first image with the coordinate vector of the characteristic point of each sub-image in the second image, and determining the similarity between the vehicle flaw in the first image and the vehicle flaw in the second image. The method solves the problems of low efficiency of manual spot check and low accuracy of picture identification in the related art.

Description

Vehicle flaw similarity recognition method and system

Technical Field

The application belongs to the technical field of computer vision, and particularly relates to a vehicle flaw similarity recognition method and system.

Background

In the vehicle insurance claim service investigation and damage assessment links, photographs are taken for damaged parts of the vehicle, and the conditions that the vehicle damage investigation degree is inconsistent with the damage assessment time can be found sometimes through comparison of the investigation photographs and the damage assessment photographs, so that the occurrence of the conditions is avoided, and in the related art, review is usually carried out through manual sampling and picture similarity recognition.

Because of the large service scale and many photos, manual spot checks can only extract smaller proportion, and the case coverage rate is insufficient, so that the occurrence of the situations can not be effectively avoided. The picture similarity recognition is to align the pictures subjected to investigation and damage assessment through processing steps such as stretching, rotation, scaling, deformation and the like, and compare the similarity of the two aligned pictures.

Disclosure of Invention

The embodiment of the application provides a vehicle flaw similarity recognition method and system, which can solve the problems of low efficiency of manual spot check and low accuracy of picture recognition in the related technology.

In a first aspect, an embodiment of the present application provides a method for identifying a vehicle scar similarity, including:

determining the number of respective vehicle flaws in the first image and the second image to be identified, and the type and position of each flaw;

extracting sub-images in the first image and the second image according to the type and the position of the flaw, and classifying to obtain the image type of the sub-image, wherein the sub-image is an image of the position of the flaw in the first image and the second image;

marking the characteristic points of each sub-image according to the image category and a preset characteristic recognition model, wherein the preset characteristic recognition model is used for determining coordinate vectors of the characteristic points of the sub-images;

and comparing the coordinate vector of the characteristic point of each sub-image in the first image with the coordinate vector of the characteristic point of each sub-image in the second image, and determining the similarity between the vehicle flaw in the first image and the vehicle flaw in the second image.

In a second aspect, an embodiment of the present application provides a vehicle scar similarity recognition system, including:

the first determining module is used for determining the number of the vehicle flaws in the first image and the second image to be identified and the type and the position of each flaw;

the extraction module is used for extracting sub-images in the first image and the second image according to the type and the position of the flaw, and classifying the sub-images to obtain the image type of the sub-images, wherein the sub-images are images of the positions of the flaws in the first image and the second image;

the marking module is used for marking the characteristic points of each sub-image according to the image category and a preset characteristic recognition model, and the preset characteristic recognition model is used for determining coordinate vectors of the characteristic points of the sub-images;

and the second determining module is used for comparing the coordinate vector of the characteristic point of each sub-image in the first image with the coordinate vector of the characteristic point of each sub-image in the second image, and determining the similarity between the vehicle flaw in the first image and the vehicle flaw in the second image.

In a third aspect, embodiments of the present application provide a readable storage medium having stored thereon a program or instructions which when executed by a processor perform the steps of the method according to the first aspect.

In the embodiment of the application, firstly, the number of the respective vehicle flaws in a first image and a second image to be identified and the type and position of each flaw are determined, then, according to the type and position of the flaws, the sub-images in the first image and the second image are extracted and classified to obtain the image types of the sub-images, the sub-images are images of the positions of the flaws in the first image and the second image, the feature points of the sub-images are marked according to the image types and a preset feature recognition model, the preset feature recognition model is used for determining the coordinate vector of the feature point of the sub-image, and finally, the coordinate vector of the feature point of each sub-image in the first image is compared with the coordinate vector of the feature point of each sub-image in the second image, so that the similarity between the vehicle flaws in the first image and the vehicle flaws in the second image is determined. According to the embodiment of the application, the characteristic points of each flaw in the two images to be identified are marked, the coordinate vectors of a plurality of characteristic points of each flaw are determined, and then the similarity of the two images is determined by comparison, so that whether the loss degree of the vehicle flaw in the first image (the investigation image) is consistent with the loss degree of the vehicle flaw in the second image (the assessment image) can be automatically identified, and the accuracy and the efficiency of identification are improved.

Drawings

FIG. 1 is a flow chart of a method for identifying vehicle scar similarity provided by an embodiment of the present application;

fig. 2 is a block diagram of a vehicle flaw similarity recognition system according to an embodiment of the present application.

Detailed Description

The technical solutions of the embodiments of the present application will be clearly described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which are obtained by a person skilled in the art based on the embodiments of the present application, fall within the scope of protection of the present application.

The terms first, second and the like in the description and in the claims, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that embodiments of the application may be practiced otherwise than as specifically illustrated or described herein. Furthermore, in the description and claims, "and/or" means at least one of the connected objects, and the character "/", generally means that the associated object is an "or" relationship.

The method and the system for identifying the vehicle scratch similarity provided by the embodiment of the application are described in detail below through specific embodiments and application scenes thereof with reference to the accompanying drawings.

As shown in fig. 1, the embodiment of the application further provides a vehicle flaw similarity identifying method, and as shown in fig. 1, the vehicle flaw similarity identifying method may include: content shown in S101 to S104.

In S101, the number of respective vehicle flaws in the first image and the second image to be identified, and the type and position of each flaw are determined.

The first image may be an image of a damaged portion of the vehicle acquired in the exploration link, the second image may be an image of a damaged portion of the vehicle acquired in the assessment link, or may be an image acquired under other conditions, and the embodiment is not specifically limited based on practical application.

It should be noted that when the vehicle is damaged, there may be multiple damaged portions, and there may be multiple flaws on each portion.

In S102, according to the type and position of the flaw, sub-images in the first image and the second image are extracted and classified, and the image category of the sub-image is obtained.

The sub-images are images of positions of flaws in the first image and the second image. That is, the sub-images are partial images in the acquired whole image, so that the scar detail of the part can be displayed more clearly, and the subsequent comparison is more beneficial. For a clear presentation of the flaws, only one flaw may be included in each sub-image.

The image category refers to the type of scratch in the image, such as scratch, deformation, and fracture, and the embodiment is not limited to the specific application.

In S103, feature points of each sub-image are labeled according to the image category and the preset feature recognition model.

The preset feature recognition model is used for determining coordinate vectors of feature points of the sub-images.

That is, images may be labeled with different feature points according to the types of images. For example, the scratch type image can be marked by adopting characteristic points such as a scratch starting point, a scratch finishing point, a scratch center point, a positive diameter point of a scratch center, a point at the widest part of the scratch and the like; the deformation image can be marked by adopting characteristic points such as deformation edge points, deformation center points and the like; the fracture image can be marked by adopting characteristic points such as a fracture starting point, a fracture branching end point, a fracture center point, a fracture branching point and the like. Wherein each type of annotation may comprise a plurality, for example, the scratch start point may comprise scratch start point 1, scratch start point 2, scratch start point 3, and so on.

In S104, the coordinate vector of the feature point of each sub-image in the first image is compared with the coordinate vector of the feature point of each sub-image in the second image, and the similarity between the vehicle flaw in the first image and the vehicle flaw in the second image is determined.

In one possible embodiment of the present application, determining the number of respective vehicle flaws in the first image and the second image to be identified, and the type and location of each flaw, may include: determining a first number of vehicle flaws in a first image to be identified, and the type and position of each flaw based on a preset detection model; based on a preset detection model, a second number of vehicle flaws in the second image to be identified, and the type and position of each flaw are determined.

In this embodiment, the number of vehicle scratches in the image, and the type of each scratch, for example, the type of scratch, deformation, fracture, etc., and the position, i.e., coordinates, of each scratch are determined by presetting a detection model. By classifying and positioning the vehicle flaws, each flaw can be acquired more clearly, so that comparison can be performed later, and the recognition accuracy is improved.

It should be noted that if the degree of the vehicle damage, i.e. the loss degree of the vehicle in the first image is the same as the degree of the vehicle damage, i.e. the loss degree of the vehicle in the second image, the first number is the same as the second number, but the first number may be different from the second number due to different factors such as shooting time, angle, environment, etc. of the first image and the second image.

In one possible embodiment of the present application, the training method of the preset detection model may include the contents shown in the steps one to three.

Step one, acquiring images of different types of vehicle flaws.

In this embodiment, different types of vehicle scar images may be extracted from the database. The vehicle scratch images of different types can include images of different shooting angles such as front face, side face, nodding, upward shooting, far shooting, close-up and the like, images of different light conditions such as natural light, night light, workshop light and the like, images of different vehicle types, images of different damage types such as scratches, deformation and cracking and the like, and images of other types of vehicle scratch images, and the embodiment is not limited according to practical application.

Marking each flaw in each vehicle flaw image to obtain flaw type and position coordinates of each flaw.

That is, each flaw of the obtained images of the different types of flaws of the vehicle is marked, and each flaw can be selected in the image sample by a picture marking tool for marking.

The marking may include marking a scratch type (damage-class) of each scratch, for example, a scratch, a deformation, and a fracture, and a position coordinate (x, y, width, height) of each scratch, for example, x, y is a coordinate of an upper left corner of the scratch in the image, width is a scratch width, height is a scratch height, or other manners may be used to represent the position coordinate of the scratch, which is based on practical application, and the embodiment is not specifically limited.

Inputting the vehicle flaw image and the corresponding flaw type and position coordinates into an initial detection model for training to obtain a preset detection model.

In this embodiment, the acquired vehicle scratch image and the corresponding label, that is, the scratch type and the position coordinate, are input into the initial detection model for training, and the trained preset detection model is obtained.

The initial detection model may be any model suitable for the field of machine learning and deep learning, and the embodiment of the application is not particularly limited.

In this embodiment, a preset detection model can be obtained through model training, and the type and position coordinates of the flaws in the vehicle flaw image can be automatically detected through the preset detection model, so that subsequent image comparison is facilitated, and the image comparison efficiency is improved.

In one possible embodiment of the present application, extracting and classifying the sub-images in the first image and the second image according to the type and the position of the flaw, to obtain the image class of the sub-image may include: extracting sub-images in the first image and the second image according to the positions of the flaws; and classifying each extracted sub-image according to the type of the scratch to obtain the image type of the sub-image.

Wherein there may be 0 to n scratches of unequal numbers in the image.

In this embodiment, each flaw is cut out from the original image according to the position coordinates to obtain a sub-image, and the sub-images are classified according to the types of the flaws to obtain a sample set of various image categories. For example, the types of scratches include scratches, deformations, breaks, and sub-image class sample sets classified according to these types include sample-scratches { sample-images _{Scratch mark} Sample-deformation { sample-images }, sample-deformation _{Deformation of} Sample-disruption { sample-images }, sample-disruption } _{Rupture of} Three image class sample sets.

The embodiment of the application is described by taking a training method of a preset feature recognition model of one image class as an example, and the model training methods of other image classes are the same.

In one possible embodiment of the present application, the training method of the preset feature recognition model includes the contents shown in the steps one to three.

Step one, a plurality of sub-images of a first image class are acquired.

And acquiring a plurality of sub-images of the first image category in the plurality of image category sample sets. For example, scratch sample set { sample-images }, is obtained _{Scratch mark} A plurality of sub-images in a }.

And secondly, marking the characteristic points of each sub-image respectively to obtain the coordinate vector of each characteristic point.

The characteristic points comprise a starting point, a central point, an end point, an orthogonal diameter point passing through the central point, a widest point, an edge point and a branching point.

It should be noted that the types of flaws in each image category are different, and therefore, the characteristic points of flaws in each image category are also marked differently.

For example, the characteristic points of the scratch may include a scratch start point 1, a scratch start point 2, a scratch start point 3, a scratch end point 1, a scratch end point 2, a scratch end point 3, a scratch center point 1, a scratch center point 2, a scratch center point 3, a scratch center point 4, a scratch center orthogonal diameter 1 point 2, a scratch center orthogonal diameter 1 point 3, a scratch center orthogonal diameter 1 point 4, a scratch center orthogonal diameter 1 point 5, a scratch center orthogonal diameter 2 point 1, a scratch center orthogonal diameter 2 point 2, a scratch center orthogonal diameter 2 point 3, a scratch center orthogonal diameter 2 point 4, a scratch center orthogonal diameter 2 point 5, a scratch widest point 1, a scratch widest point 2, a scratch widest point 3, 24 characteristic points may be marked, or the characteristic points may be appropriately reduced according to the size of the scratch, and the embodiment is not limited specifically based on practical application.

The deformed feature points may include: the deformation edge point 1, the deformation edge point 2, the deformation edge point 3, the deformation edge point 4, the deformation edge point 5, the deformation edge point 6, the deformation edge point 7, the deformation edge point 8, the deformation edge point 9, the deformation edge point 10, the deformation center point 1, the deformation center point 2, the deformation center point 3, the deformation center point 4 and the deformation center point 5 can be marked with 15 characteristic points, or the characteristic points can be properly reduced according to the deformation size, and the embodiment is not limited by practical application.

The characteristic points of the rupture may include: the starting point 1, the starting point 2, the starting point 3, the ending point 1, the ending point 2, the ending point 3, the ending point 2, the ending point 1, the ending point 2, the ending point 3, the ending point 1, the ending point 2, the ending point 3, the ending point 4, the ending point 5, the ending point 1, the ending point 2, the ending point 3, the ending point 4, and the ending point 5 of the crack can be marked with 19 feature points, or the feature points can be properly reduced according to the degree of the crack, which is not limited by the embodiment according to practical application.

Inputting the multiple sub-images of the first image category and the coordinate vectors of the multiple feature points corresponding to each sub-image into an initial feature recognition model for training to obtain a preset feature recognition model corresponding to the first image category.

In this embodiment, the classified sub-images and the labels corresponding to each sub-image, that is, the feature points in each sub-image, are input into the initial feature recognition model for training, so as to obtain a trained preset feature recognition model.

The initial feature recognition model may be a residual network deep learning algorithm, or may be any model suitable for the fields of machine learning and deep learning, which is not particularly limited in the embodiment of the present application.

In this embodiment, a preset feature recognition model can be obtained through model training, and the positions of flaws in each sub-image can be marked through the preset feature recognition model, so that subsequent image comparison can be facilitated, and the accuracy of image comparison can be improved.

It should be noted that a plurality of models, such as a scratch feature recognition model M-recognize, can be obtained by training the models described above _{Scratch mark} Deformation identification model M-recognize _{Deformation of} Fracture identification model M-recognize _{Rupture of} And (3) model, namely marking the characteristic points of the corresponding flaw types, returning vectors containing coordinates of all the characteristic points in the image, and returning empty coordinate vectors with the length equal to the number of the characteristic points of the flaw types if the corresponding flaw types are not available in the image.

In one possible embodiment of the present application, comparing the coordinate vector of the feature point of each sub-image in the first image with the coordinate vector of the feature point of each sub-image in the second image, determining the similarity between the vehicle flaw in the first image and the vehicle flaw in the second image may include: converting the coordinate vector of the characteristic points of each sub-image of the first image category in the first image into a 128-dimensional value vector through dimension increasing operation to obtain a first vector; converting the coordinate vector of the characteristic points of each sub-image of the first image category in the second image into a 128-dimensional value vector through dimension increasing operation to obtain a second vector; the Euclidean distance of the first vector from the second vector is determined.

The Euclidean distance has a value range of [0,1], and the difference is more obvious when the value is closer to 1, and the flaw is more similar when the value is closer to 0.

In this embodiment, before comparing the coordinate vector in the first image with the coordinate vector in the second image, the coordinate vector may be subjected to an up-scaling operation to obtain more scar details, so that the comparison is more accurate.

In one possible embodiment of the present application, the vehicle flaw similarity identifying method may further include: judging that the vehicle flaw in the first image and the vehicle flaw in the second image are the same flaw when the Euclidean distance is smaller than the distance threshold; when the Euclidean distance is greater than the distance threshold, the vehicle flaw in the first image and the vehicle flaw in the second image are determined to be different flaws.

That is, a distance threshold may be set, and when the determined euclidean distance is less than the distance threshold, the flaw may be determined to be a uniform flaw, or else not be the same flaw.

It should be noted that, in the vehicle flaw similarity identifying method provided by the embodiment of the present application, the execution body may be a vehicle flaw similarity identifying system, or a control module for executing the vehicle flaw similarity identifying method in the vehicle flaw similarity identifying system. In the embodiment of the application, the vehicle flaw similarity recognition system provided by the embodiment of the application is described by taking the method for executing the vehicle flaw similarity recognition by the vehicle flaw similarity recognition system as an example.

As shown in fig. 2, the embodiment of the application further provides a vehicle flaw similarity recognition system. The vehicle scar similarity recognition system may include: a first determination module 201, an extraction module 202, a labeling module 203, and a second determination module 204.

Wherein, the first determining module 201 is configured to determine the number of respective vehicle flaws in the first image and the second image to be identified, and the type and the position of each flaw; the extracting module 202 is configured to extract sub-images in the first image and the second image according to the type and the position of the flaw, and classify the sub-images to obtain an image type of the sub-image, where the flaw in the first image and the second image is located; the labeling module 203 is configured to label feature points of each sub-image according to an image category and a preset feature recognition model, where the preset feature recognition model is used to determine coordinate vectors of feature points of the sub-image; the second determining module 204 is configured to compare the coordinate vector of the feature point of each sub-image in the first image with the coordinate vector of the feature point of each sub-image in the second image, and determine a similarity between the vehicle flaw in the first image and the vehicle flaw in the second image.

In the embodiment of the present application, first, the first determining module 201 determines the number of respective vehicle flaws in the first image and the second image to be identified, and the type and the position of each flaw, then the extracting module 202 extracts sub-images in the first image and the second image according to the type and the position of the flaw, classifies the sub-images to obtain an image type of the sub-images, the sub-images are images of positions of flaws in the first image and the second image, the labeling module 203 labels feature points of the sub-images according to the image type and a preset feature recognition model, the preset feature recognition model is used for determining coordinate vectors of feature points of the sub-images, and finally the second determining module 204 compares the coordinate vectors of the feature points of the sub-images in the first image with the coordinate vectors of the feature points of the sub-images in the second image, so as to determine similarity between the vehicle flaws in the first image and the vehicle flaws in the second image. According to the embodiment of the application, the characteristic points of each flaw in the two images to be identified are marked, the coordinate vectors of a plurality of characteristic points of each flaw are determined, and then the similarity of the two images is determined by comparison, so that whether the loss degree of the vehicle flaw in the first image (the investigation image) is consistent with the loss degree of the vehicle flaw in the second image (the assessment image) can be automatically identified, and the accuracy and the efficiency of identification are improved.

In one possible embodiment of the present application, the first determining module 201 may be configured to: determining a first number of vehicle flaws in a first image to be identified, and the type and position of each flaw based on a preset detection model; based on a preset detection model, a second number of vehicle flaws in the second image to be identified, and the type and position of each flaw are determined.

In one possible embodiment of the present application, the first determining module 201 may be configured to: acquiring different types of vehicle flaw images; marking each flaw in each vehicle flaw image to obtain flaw type and position coordinates of each flaw; and inputting the vehicle flaw image and the corresponding flaw type and position coordinates into an initial detection model for training to obtain a preset detection model.

In one possible embodiment of the present application, the extraction module 202 may be configured to: extracting sub-images in the first image and the second image according to the positions of the flaws; and classifying each extracted sub-image according to the type of the scratch to obtain the image type of the sub-image.

In one possible embodiment of the present application, the labeling module 203 may be configured to: acquiring a plurality of sub-images of a first image class; marking the characteristic points of each sub-image to obtain coordinate vectors of the characteristic points, wherein the characteristic points comprise a starting point, a central point, an end point, an orthogonal diameter point passing through the central point, a widest point, an edge point and a branching point; and inputting the multiple sub-images of the first image category and the coordinate vectors of the multiple feature points corresponding to each sub-image into an initial feature recognition model for training to obtain a preset feature recognition model corresponding to the first image category.

In one possible embodiment of the present application, the second determining module 204 may be configured to: converting the coordinate vector of the characteristic points of each sub-image of the first image category in the first image into a 128-dimensional value vector through dimension increasing operation to obtain a first vector; converting the coordinate vector of the characteristic points of each sub-image of the first image category in the second image into a 128-dimensional value vector through dimension increasing operation to obtain a second vector; the Euclidean distance of the first vector from the second vector is determined.

In one possible embodiment of the present application, the second determining module 204 may be configured to: judging that the vehicle flaw in the first image and the vehicle flaw in the second image are the same flaw when the Euclidean distance is smaller than the distance threshold; when the Euclidean distance is greater than the distance threshold, the vehicle flaw in the first image and the vehicle flaw in the second image are determined to be different flaws.

The vehicle flaw similarity recognition system in the embodiment of the application can be a device, and can also be a component, an integrated circuit or a chip in a terminal. The device may be a mobile electronic device or a non-mobile electronic device. By way of example, the mobile electronic device may be a cell phone, tablet computer, notebook computer, palm computer, vehicle mounted electronic device, wearable device, ultra-mobile personal computer (ultra-mobile personal computer, UMPC), netbook or personal digital assistant (personal digital assistant, PDA), etc., and the non-mobile electronic device may be a server, network attached storage (Network Attached Storage, NAS), personal computer (personal computer, PC), television (TV), teller machine or self-service machine, etc., and embodiments of the present application are not limited in particular.

The vehicle flaw similarity recognition system in the embodiment of the application can be a device with an operating system. The operating system may be an Android operating system, an ios operating system, or other possible operating systems, and the embodiment of the present application is not limited specifically.

The vehicle flaw similarity recognition system provided by the embodiment of the application can realize each process realized by the method embodiment of fig. 1, achieves the same technical effect, and is not repeated here for avoiding repetition.

The embodiment of the application also provides a readable storage medium, wherein the readable storage medium stores a program or an instruction, and the program or the instruction realizes each process of the vehicle flaw similarity identification method embodiment provided by any embodiment when being executed by a processor. And the same technical effects can be achieved, and in order to avoid repetition, the description is omitted here.

Among them, the readable storage medium includes a computer readable storage medium such as a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk or an optical disk, and the like.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element. Furthermore, it should be noted that the scope of the methods and apparatus in the embodiments of the present application is not limited to performing the functions in the order shown or discussed, but may also include performing the functions in a substantially simultaneous manner or in an opposite order depending on the functions involved, e.g., the described methods may be performed in an order different from that described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a computer software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) comprising instructions for causing a terminal (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method according to the embodiments of the present application.

The embodiments of the present application have been described above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those having ordinary skill in the art without departing from the spirit of the present application and the scope of the claims, which are to be protected by the present application.

Claims

1. A vehicle flaw similarity recognition method, characterized by comprising:

2. The method for identifying the similarity of vehicle flaws according to claim 1, wherein the determining the number of the respective vehicle flaws in the first image and the second image to be identified, and the type and the position of each flaw, comprises:

determining a first number of vehicle flaws in a first image to be identified, and the type and position of each flaw based on a preset detection model;

and determining a second number of vehicle flaws in a second image to be identified, and the type and position of each flaw based on the preset detection model.

3. The vehicle flaw similarity recognition method according to claim 2, wherein the training method of the preset detection model includes:

acquiring different types of vehicle flaw images;

marking each flaw in each vehicle flaw image to obtain flaw type and position coordinates of each flaw;

inputting the vehicle flaw image and the corresponding flaw type and position coordinates into an initial detection model for training to obtain the preset detection model.

4. The method for identifying the similarity of the flaws of the vehicle according to claim 1, wherein the steps of extracting the sub-images in the first image and the second image according to the type and the position of the flaws and classifying the sub-images to obtain the image types of the sub-images include:

extracting sub-images in the first image and the second image according to the positions of the flaws;

and classifying the extracted sub-images according to the types of the scratches to obtain the image types of the sub-images.

5. The vehicle flaw similarity recognition method according to claim 1, wherein the training method of the preset feature recognition model includes:

acquiring a plurality of sub-images of a first image class;

marking characteristic points of each sub-image respectively to obtain coordinate vectors of the characteristic points, wherein the characteristic points comprise a starting point, a central point, an end point, an orthogonal diameter point passing through the central point, a widest point, an edge point and a branching point;

and inputting the multiple sub-images of the first image category and the coordinate vectors of the multiple feature points corresponding to each sub-image into an initial feature recognition model for training to obtain a preset feature recognition model corresponding to the first image category.

6. The vehicle flaw similarity identifying method according to claim 1, wherein the comparing the coordinate vector of the feature point of each sub-image in the first image with the coordinate vector of the feature point of each sub-image in the second image, determines the similarity between the vehicle flaw in the first image and the vehicle flaw in the second image, includes:

converting the coordinate vector of the characteristic points of each sub-image of the first image category in the first image into a 128-dimensional value vector through dimension-lifting operation to obtain a first vector;

converting the coordinate vector of the characteristic points of each sub-image of the first image category in the second image into a 128-dimensional value vector through dimension-increasing operation to obtain a second vector;

and determining the Euclidean distance between the first vector and the second vector.

7. The vehicle flaw similarity recognition method according to claim 6, characterized in that the method further comprises:

when the Euclidean distance is smaller than a distance threshold value, judging that the vehicle flaw in the first image and the vehicle flaw in the second image are the same flaw;

and when the Euclidean distance is larger than the distance threshold value, judging that the vehicle flaw in the first image and the vehicle flaw in the second image are different flaws.

8. A vehicle scar similarity recognition system, comprising:

9. The vehicle scar similarity recognition system of claim 8, wherein the extraction module is configured to:

10. A readable storage medium, characterized in that the readable storage medium has stored thereon a program or instructions which, when executed by a processor, implement the steps of the vehicle flaw similarity recognition method according to any one of claims 1-7.