CN111666995A - Vehicle damage assessment method, device, equipment and medium based on deep learning model - Google Patents
Vehicle damage assessment method, device, equipment and medium based on deep learning model Download PDFInfo
- Publication number
- CN111666995A CN111666995A CN202010478087.9A CN202010478087A CN111666995A CN 111666995 A CN111666995 A CN 111666995A CN 202010478087 A CN202010478087 A CN 202010478087A CN 111666995 A CN111666995 A CN 111666995A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- damage
- pictures
- damaged
- model
- 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
- 230000006378 damage Effects 0.000 title claims abstract description 366
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000013136 deep learning model Methods 0.000 title claims abstract description 35
- 238000001514 detection method Methods 0.000 claims abstract description 191
- 230000011218 segmentation Effects 0.000 claims abstract description 128
- 238000012549 training Methods 0.000 claims abstract description 68
- 238000003860 storage Methods 0.000 claims abstract description 10
- 230000000694 effects Effects 0.000 claims description 56
- 208000027418 Wounds and injury Diseases 0.000 claims description 23
- 208000014674 injury Diseases 0.000 claims description 22
- 238000005457 optimization Methods 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 13
- 238000007790 scraping Methods 0.000 claims description 7
- 230000006735 deficit Effects 0.000 claims description 5
- 238000002372 labelling Methods 0.000 claims description 5
- 230000002829 reductive effect Effects 0.000 abstract description 4
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 4
- 230000002238 attenuated effect Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 206010039203 Road traffic accident Diseases 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000001771 impaired effect Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 208000037974 severe injury Diseases 0.000 description 1
- 230000009528 severe injury Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Accounting & Taxation (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Finance (AREA)
- Evolutionary Biology (AREA)
- Development Economics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- General Business, Economics & Management (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of artificial intelligence, and discloses a vehicle damage assessment method, a device, equipment and a storage medium based on a deep learning model, wherein a damaged main part of a vehicle is determined by acquiring a plurality of damaged pictures to be recognized of the vehicle, segmenting a model according to the plurality of pictures to be recognized and a preset part obtained by training, determining a plurality of damage detection frames of the damaged main part according to the plurality of pictures to be recognized and the preset damage detection model obtained by training, and outputting a damage assessment result of the vehicle according to the damaged main part and the plurality of damage detection frames; according to the method, the damaged main part of the vehicle is determined by utilizing the preset part segmentation model, the plurality of damage detection frames of the damaged main part are obtained by utilizing the preset damage detection model, and then the damage assessment result with the heaviest damage of the vehicle is output according to the damaged main part and the plurality of damage detection frames of the damaged main part, so that the problem of inconsistency of the damage assessment result of the vehicle is solved, the accuracy of the damage assessment result is improved, the labor for damage assessment is reduced, and the damage assessment efficiency is improved.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a vehicle damage assessment method, device, equipment and medium based on a deep learning model.
Background
With the rapid increase of the automobile holding capacity, the traffic collision accident rapidly increases, and the vehicle damages more and more. Therefore, insurance companies continuously launch insurance services of vehicle insurance to guarantee the vehicle property safety of the masses, when the insured vehicles have traffic accidents, vehicle damage assessment is needed to be carried out on the damaged vehicles, and vehicle insurance claim amount is determined according to the heaviest condition of vehicle damage, and the vehicle insurance damage assessment is a key link of claim settlement.
At present, the judgement to the vehicle damage mainly relies on the manual work to estimate, the level of damage personnel carries out the scene survey judgement at the vehicle accident scene, can spend a large amount of times through the impaired mode of level of damage personnel manual classification, will invest a large amount of cost of labor, inefficiency is unfavorable for the quick realization of car insurance claim, and the level of damage personnel need classify the different images of gathering and distinguish the damage that the part received according to the image, receive the influence of various subjective factors easily and can't guarantee the impaired accuracy to the vehicle.
Therefore, in order to solve the above problems, the application of computer vision image recognition technology to the scene of vehicle damage is a new choice. The existing computer loss assessment process is as follows: the damaged condition of the vehicle is shot by adopting an image acquisition tool, and the damaged condition reflected in the picture is automatically identified by utilizing a computer, so that the damaged condition of the vehicle is judged. However, the method cannot accurately identify the heaviest condition of the damaged vehicle in the image, the damage assessment result may be inconsistent, and the damage assessment accuracy of the vehicle still needs to be further improved.
Disclosure of Invention
The invention provides a vehicle damage assessment method, a vehicle damage assessment device, vehicle damage assessment equipment and a vehicle damage assessment medium based on a deep learning model, and aims to solve the problem that the vehicle damage assessment accuracy is insufficient due to the fact that the heaviest condition of vehicle damage in an image cannot be accurately identified in the existing damage assessment process.
A vehicle damage assessment method based on a deep learning model comprises the following steps:
acquiring a plurality of pictures to be identified of a vehicle, wherein the pictures to be identified are pictures reflecting the damaged condition of the vehicle;
determining a damaged main component of the vehicle according to the plurality of pictures to be recognized and a preset component segmentation model obtained through training, wherein the preset component segmentation model is a model with the best vehicle component segmentation effect;
determining a plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and a preset damage detection model obtained through training, wherein the preset damage detection model is a model with the best vehicle damage judgment effect;
and outputting a damage assessment result of the vehicle according to the damaged main component and the plurality of damage detection frames, wherein the damage assessment result comprises the heaviest damaged component and the heaviest damaged category of the vehicle.
Further, the determining the damaged main component of the vehicle according to the plurality of pictures to be recognized and the trained preset component segmentation model includes:
inputting the multiple pictures to be recognized into the preset component segmentation model, and acquiring all damaged components contained in the multiple pictures to be recognized;
acquiring the area of each damaged part, and comparing the area of all the damaged parts;
and taking the damaged part with the largest area as a damaged main part of the vehicle.
Further, the determining the damaged main component of the vehicle according to the plurality of pictures to be recognized and the trained preset component segmentation model includes:
inputting the multiple pictures to be recognized into the preset component segmentation model to obtain a damage detection frame of each picture to be recognized;
determining a plurality of damage detection frames of the damaged main part according to the damaged main part and the damage detection frame of each picture to be identified.
Further, the determining a plurality of damage detection frames of the damaged main part according to the damaged main part and the damage detection frame of each picture to be identified comprises:
determining a central point of each injury detection frame, and determining the position of the central point;
judging whether the position of the central point is positioned in the damaged main part;
if the position of the central point is located in the damaged main part, reserving a damage detection frame corresponding to the central point, and taking the corresponding damage detection frame as a damage detection frame of the damaged main part;
and if the position of the central point is not located in the damaged main part, deleting the damage detection frame corresponding to the central point.
Further, the obtaining manner of the preset component segmentation model includes:
acquiring a plurality of historical vehicle loss pictures, wherein the historical vehicle loss pictures comprise previous vehicle loss pictures of all vehicle types on the market;
classifying and labeling the plurality of historical vehicle loss pictures according to components of the vehicle to obtain a component segmentation database, wherein the components of the vehicle comprise a fender, a vehicle door, a bumper and a hood;
taking the part segmentation database as a training set and a test set of vehicle part segmentation, and establishing a part segmentation model for analyzing the vehicle part segmentation effect;
and adjusting the learning rate, the training times and the optimization mode to obtain a part segmentation model with the best vehicle part segmentation effect as the preset part segmentation model.
Further, the obtaining method of the preset flaw detection model includes:
acquiring a plurality of historical vehicle loss pictures, wherein the historical vehicle loss pictures comprise previous vehicle loss pictures of all vehicle types on the market;
performing rectangular marking on the plurality of historical vehicle damage pictures according to the damage degree of the vehicle to obtain a damage detection database, wherein the damage degree of the vehicle comprises scraping, sinking and missing;
taking the damage detection database as a training set and a test set for vehicle damage judgment, and establishing a damage detection model for analyzing the vehicle damage judgment effect;
and performing data enhancement on the picture data in the injury detection database, adjusting the learning rate, the training times and the optimization mode, and obtaining an injury detection model with the best vehicle injury judgment effect as the preset injury detection model.
A vehicle damage assessment device based on a deep learning model comprises:
the system comprises an acquisition module, a recognition module and a recognition module, wherein the acquisition module is used for acquiring a plurality of pictures to be recognized of a vehicle, and the pictures to be recognized are pictures reflecting the damaged condition of the vehicle;
the first determination module is used for determining a damaged main component of the vehicle according to the plurality of pictures to be recognized and a preset component segmentation model obtained through training, wherein the preset component segmentation model is a model with the best segmentation effect on vehicle components;
the second determining module is used for determining a plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and a preset damage detection model obtained through training, wherein the preset damage detection model is a model with the best vehicle damage judgment effect;
and the output module is used for outputting a damage assessment result of the vehicle according to the damaged main component and the damage detection frames, wherein the damage assessment result comprises the heaviest damaged component and the heaviest damaged category of the vehicle.
Further, the first determining module is specifically configured to:
inputting the multiple pictures to be recognized into the preset component segmentation model, and acquiring all damaged components contained in the multiple pictures to be recognized;
acquiring the area of each damaged part, and comparing the area of all the damaged parts;
and taking the damaged part with the largest area as a damaged main part of the vehicle.
A computer device comprising a memory, a processor and a computer program stored in said memory and executable on said processor, said processor implementing the steps of the deep learning model based vehicle impairment method as described above when executing said computer program.
A readable storage medium storing a computer program which, when executed by a processor, performs the steps of the deep learning model-based vehicle damage assessment method as described above.
According to the vehicle damage assessment method, the vehicle damage assessment device, the computer equipment and the storage medium based on the deep learning model, a plurality of pictures to be recognized of a vehicle are obtained, the pictures to be recognized are pictures reflecting the damaged condition of the vehicle, the damaged main part of the vehicle is determined according to the pictures to be recognized and a preset part segmentation model obtained through training, the preset part segmentation model is a model with the best segmentation effect on the vehicle part, a plurality of damage detection frames of the damaged main part are determined according to the pictures to be recognized and the preset damage detection model obtained through training, the preset damage detection model is a model with the best judgment effect on the damage of the vehicle, and the damage assessment result of the vehicle is output according to the damaged main part and the plurality of damage detection frames, and comprises the most damaged part and the most damaged category of the vehicle; according to the vehicle damage assessment method, the damaged main part of the vehicle is determined by utilizing the preset part segmentation model, the plurality of damage detection frames of the damaged main part are obtained by utilizing the preset damage detection model, and then the damage assessment result with the heaviest damage of the vehicle is output according to the damaged main part and the plurality of damage detection frames of the damaged main part, so that the problem of inconsistency of the vehicle damage assessment result is solved, the accuracy of the damage assessment result is improved, manpower in the damage assessment process is reduced, the damage assessment efficiency is improved, the vehicle insurance claim settlement amount is determined according to the damage assessment result, the customer experience is optimized, and the labor cost is saved.
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 of the present invention will be 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 that other drawings can be obtained according to these drawings without inventive labor.
FIG. 1 is a schematic diagram of an application environment of a deep learning model-based vehicle damage assessment method according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of a vehicle damage assessment method based on a deep learning model according to an embodiment of the present invention;
FIG. 3 is a flowchart illustrating implementation of step S20 of the deep learning model-based vehicle damage assessment method according to an embodiment of the present invention;
FIG. 4 is a flowchart illustrating implementation of step S30 of the deep learning model-based vehicle damage assessment method according to an embodiment of the present invention;
FIG. 5 is a schematic structural diagram of a vehicle damage assessment device based on a deep learning model according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a computer device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The vehicle damage assessment method based on the deep learning model provided by the embodiment of the invention can be applied to an application environment shown in figure 1, wherein a client communicates with a server through a network, the server receives a plurality of pictures to be recognized of a vehicle, the pictures to be recognized are pictures reflecting the damaged condition of the vehicle, the damaged main part of the vehicle is determined according to the pictures to be recognized and a preset part segmentation model obtained through training, the preset part segmentation model is a model with the best vehicle part segmentation effect, a plurality of damage detection frames of the damaged main part are determined according to the pictures to be recognized and the preset damage detection model obtained through training, the preset damage detection model is a model with the best vehicle damage judgment effect, the damage assessment result of the vehicle is output to the client according to the damaged main part and the damage detection frames, and the damage assessment result comprises the most damaged part of the vehicle, and the damage assessment result is obtained through training, The most severe injury category.
Among other things, the client may be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of a plurality of servers.
In an embodiment, as shown in fig. 2, a vehicle damage assessment method based on a deep learning model is provided, which is described by taking the application of the method to the service end in fig. 1 as an example, and includes the following steps:
s10: the method comprises the steps of obtaining a plurality of pictures to be identified of the vehicle, wherein the pictures to be identified are pictures reflecting the damaged condition of the vehicle.
When a traffic accident occurs to an insured vehicle, vehicle damage assessment needs to be performed on the damaged insured vehicle, a plurality of pictures to be identified of the damaged insured vehicle need to be acquired, the heaviest damage condition of the vehicle is determined according to the damage conditions of the vehicle in the pictures to be identified, and therefore the vehicle damage settlement amount of the vehicle is determined.
The picture to be recognized is an accident picture of a vehicle in a vehicle accident scene, and the picture to be recognized can be provided by a vehicle owner or other people related to the vehicle through a client.
S20: and determining the damaged main component of the vehicle according to the plurality of pictures to be recognized and the trained preset component segmentation model, wherein the preset component segmentation model is the model with the best segmentation effect on the vehicle components.
The method comprises the steps of inputting a plurality of pictures to be recognized into a preset component segmentation model obtained through training, and determining damaged main components of a vehicle according to the pictures to be recognized and the preset component segmentation model, wherein the preset component segmentation model is a model with the best vehicle component segmentation effect, and the preset component segmentation model is obtained through deep learning training.
For example, based on a preset component segmentation model obtained by pre-training, a plurality of pictures to be recognized are input into the preset component segmentation model obtained by training, which components are contained in each picture to be recognized are counted, damaged components in the pictures are determined, and the damaged components with the largest area are determined as damaged main components according to the principle that the area of the components is the largest.
In this embodiment, determining the damaged component with the largest area as the damaged main component according to the principle of the largest component area is merely an exemplary illustration, and in other embodiments, the determination manner of the damaged main component is other, and details are not described here.
S30: and determining a plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and a preset damage detection model obtained by training, wherein the preset damage detection model is the model with the best vehicle damage judgment effect.
After the damaged main part of the vehicle is determined, a plurality of pictures to be recognized are input into a preset damage detection model obtained through training to obtain damage detection frames of all damaged parts in each picture to be recognized, and then the plurality of damage detection frames of the damaged main part are determined according to the damage detection frames of the damaged parts in all the pictures to be recognized, wherein the preset damage detection model is a model with the best vehicle damage judgment effect, and the preset damage detection model is obtained through deep learning training.
S40: and outputting a damage assessment result of the vehicle according to the damaged main component and the plurality of damage detection frames, wherein the damage assessment result comprises the most serious damage component and the most serious damage category of the vehicle.
After acquiring a damaged main part of the vehicle and a plurality of damage detection frames of the damaged main part, screening the heaviest damage of the main part in the plurality of damage detection frames of the damaged main part, and outputting the main part and the heaviest damage category as a damage assessment result of the vehicle, wherein the damage assessment result comprises the heaviest damage part, the heaviest damage category and the like of the vehicle.
For example, the damaged main part of the vehicle is a fender, the damaged category includes scratch, dent and loss, the fender is simultaneously scratched and dented according to reflection of a plurality of flaw detection frames of the fender, the dent is determined as the most serious flaw, the output damage assessment result of the vehicle is the fender dent, and a subsequent damage assessment worker can determine the vehicle insurance claim amount according to the fender dent of the vehicle.
In the embodiment, a plurality of pictures to be recognized of a vehicle are obtained, the pictures to be recognized are pictures reflecting the damage condition of the vehicle, the damaged main part of the vehicle is determined according to the pictures to be recognized and a preset part segmentation model obtained through training, the preset part segmentation model is a model with the best segmentation effect on the vehicle part, a plurality of damage detection frames of the damaged main part are determined according to the pictures to be recognized and the preset damage detection model obtained through training, the preset damage detection model is a model with the best judgment effect on the damage of the vehicle, and the damage assessment result of the vehicle is output according to the damaged main part and the plurality of damage detection frames, and comprises the heaviest damage part and the heaviest damage category of the vehicle; according to the vehicle damage assessment method, the damaged main part of the vehicle is determined by utilizing the preset part segmentation model, the plurality of damage detection frames of the damaged main part are obtained by utilizing the preset damage detection model, and then the damage assessment result with the heaviest damage of the vehicle is output according to the damaged main part and the plurality of damage detection frames of the damaged main part, so that the problem of inconsistency of the vehicle damage assessment result is solved, the accuracy of the damage assessment result is improved, the labor cost in the damage assessment process is reduced, the damage assessment efficiency is improved, the vehicle risk claim settlement amount can be determined according to the damage assessment result in the subsequent process, the customer experience is optimized, and the labor cost is saved.
In an embodiment, as shown in fig. 3, in step S20, determining the damaged main component of the vehicle according to the plurality of pictures to be recognized and the trained preset component segmentation model includes:
s21: and inputting a plurality of pictures to be recognized into the preset component segmentation model, and acquiring all damaged components contained in the plurality of pictures to be recognized.
After a plurality of pictures to be recognized of the vehicle are obtained, the pictures to be recognized are input into a preset component segmentation model obtained through training, and all damaged components contained in the pictures to be recognized are obtained.
For example, the multiple pictures to be recognized of the vehicle include three vehicle components, namely a fender, a vehicle door and a bumper, but the vehicle door in the picture is intact, and the fender and the bumper are damaged, all damaged components included in the obtained multiple pictures to be recognized include the fender and the bumper.
In this embodiment, the three vehicle components including the fender, the door, and the bumper in the multiple pictures to be recognized are only exemplary illustrations, and in other embodiments, the multiple pictures to be recognized also include other components, which are not described herein again.
S22: the area of each damaged part is obtained and compared with the area of all damaged parts.
After all damaged parts contained in a plurality of pictures to be identified are obtained, the actual area of each damaged part is calculated according to the area size occupied by each part in the pictures, so that the area sizes of all the damaged parts are compared according to the areas.
For example, the multiple pictures to be recognized of the vehicle comprise three vehicle components, namely a fender, a vehicle door and a bumper, wherein the fender and the bumper are damaged, pixel ratios of the fender and the bumper in all the pictures are respectively determined, areas of the fender and the bumper are further determined, the area is larger when the pixel ratio is larger, and the areas of the fender and the bumper in the multiple pictures to be recognized are compared.
In this embodiment, determining the area of the damaged component according to the pixel proportion of the component in all the pictures is only an exemplary illustration, and in other embodiments, the area size of the damaged component may also be obtained by other methods, which is not described herein again.
S23: and taking the damaged part with the largest area as a damaged main part of the vehicle.
For example, comparing the areas of the fender and the bumper in the multiple pictures to be recognized, determining that the area of the fender is larger than that of the bumper, and determining that the area of the fender is the largest, wherein the fender is a damaged main part of the vehicle.
In this embodiment, a plurality of pictures to be recognized are input into the preset component segmentation model, all damaged components included in the plurality of pictures to be recognized are obtained, the area of each damaged component is obtained, the area of all damaged components is compared, the damaged component with the largest area is used as the damaged main component of the vehicle, the step of determining the damaged main component of the vehicle according to the area of each damaged component of the vehicle is detailed, the judgment process of the main damaged condition of the vehicle is further defined, and the accuracy of vehicle damage judgment is improved.
In an embodiment, as shown in fig. 4, in step S30, determining a plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and the trained predetermined damage detection model includes:
s31: and inputting a plurality of pictures to be recognized into a preset component segmentation model to obtain a damage detection frame of each picture to be recognized.
After a plurality of pictures to be recognized of the vehicle are obtained, the pictures to be recognized are input into a preset part segmentation model obtained through training, and a damage detection frame of a damaged part in each picture to be recognized is obtained.
For example, a plurality of pictures to be recognized of the vehicle comprise three vehicle components, namely a fender, a vehicle door and a bumper, but the vehicle door in the picture is intact, and the fender and the bumper are damaged, a damage detection frame of the damaged fender and a damage detection frame of the damaged bumper in each picture to be recognized are obtained.
S32: and determining a plurality of damage detection frames of the damaged main part according to the damaged main part and the damage detection frame of each picture to be identified.
After all the damage detection frames contained in the multiple pictures to be identified are obtained, determining multiple damage detection frames of the damaged main part according to the damaged main part and the damage detection frames of each picture to be identified.
For example, if the damaged main part of the vehicle is a fender, the damage detection frame belonging to the fender is determined among all the damage detection frames included in the plurality of pictures to be recognized.
In the embodiment, the plurality of pictures to be recognized are input into the preset component segmentation model to obtain the damage detection frame of each picture to be recognized, and then the plurality of damage detection frames of the damaged main component are determined according to the damaged main component and the damage detection frame of each picture to be recognized, so that the step of determining the plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and the preset damage detection model obtained by training is refined, the acquisition process of the damage detection frame of the damaged main component of the vehicle is further clarified, and the accuracy of vehicle damage judgment is improved.
In an embodiment, in step S32, determining a plurality of damage detection frames of the damaged main part according to the damaged main part and the damage detection frame of each of the pictures to be recognized includes:
s321: and determining the center point of each wound detection frame, and determining the position of the center point.
After the damaged main part of the vehicle is determined, the calculation is circulated to determine the central point of each damage detection frame in the picture and determine the position of the central point of the damage detection frame.
S322: and judging whether the position of the central point is positioned in the damaged main part.
After the position of the center point of each damage detection frame is determined, whether the position of the center point is located on the damaged main part or not is detected.
S323: and if the position of the central point is positioned in the damaged main part, keeping the damage detection frame corresponding to the central point, and taking the corresponding damage detection frame as the damage detection frame of the damaged main part.
And if the central point of the damage detection frame is positioned in the damaged main part, keeping the damage detection frame corresponding to the central point, and taking the corresponding damage detection frame as the damage detection frame of the damaged main part.
S324: and if the position of the central point is not located in the damaged main part, deleting the damage detection frame corresponding to the central point.
And if the central point of the damage detection frame is not positioned in the damaged main part, deleting the damage detection frame corresponding to the central point.
For example, the damaged main part of the vehicle is determined to be a fender, and the picture to be recognized has two flaw detections, wherein the center point of the flaw detection frame 1 is located on the fender, and the center point of the flaw detection frame 2 is located on the bumper, the flaw detection frame 2 is deleted, and the flaw detection frame 1 is reserved.
In the embodiment, whether the central point position of each damage detection frame is in the damaged main component is judged, if yes, the damage detection frames are reserved, and if not, all the damage detection frames of the damaged main component are further determined, so that the influence of the damage detection frames of other components on the damage assessment result is reduced, the actual heaviest damage condition of the vehicle is determined according to the damage detection frames of the damaged main component in the subsequent process, and the accuracy of vehicle damage assessment is improved.
In an embodiment, before determining a damaged main component of a vehicle according to a plurality of pictures to be recognized and a trained preset component segmentation model, the preset component segmentation model needs to be trained in advance, and an obtaining manner of the preset component segmentation model includes:
SA 1: and acquiring a plurality of historical vehicle loss pictures, wherein the historical vehicle loss pictures comprise vehicle loss pictures of all vehicle types on the market.
For example, a plurality of historical loss pictures are acquired from a loss database of an insurance company, and the historical loss pictures comprise past loss pictures of all vehicle types on the market.
SA 2: and classifying and labeling the plurality of historical vehicle loss pictures according to components of the vehicle to obtain a component division database, wherein the components of the vehicle comprise a fender, a vehicle door, a bumper and a hood.
For example, after a plurality of historical damage pictures are acquired, the plurality of historical damage pictures are classified in a manual labeling mode, each pixel in each picture needs to be classified and labeled, and then a component division database is obtained according to a plurality of component structures such as a fender, a vehicle door, a bumper and a hood.
In this embodiment, the components of the vehicle including the fender, the door, the bumper, and the hood are merely exemplary, and in other embodiments, the vehicle further includes other components, which are not described herein again.
SA 3: and establishing a part segmentation model for analyzing the vehicle part segmentation effect by taking the part segmentation database as a training set and a test set of the vehicle part segmentation.
For example, after obtaining the component segmentation database, the component segmentation database is used as a training set and a test set, and a component segmentation model for analyzing the vehicle component segmentation effect of the depeplab-based DPC network is constructed based on the tensoflow framework. The number of the car damage pictures in the part segmentation database training set is twenty thousand, and the number of the car damage pictures in the part segmentation database training set is thirty thousand.
In this embodiment, based on the tensoflow frame, the component segmentation model for analyzing the vehicle component segmentation effect, which is constructed based on the depeplab DPC network, is only an exemplary description, and in other embodiments, the component segmentation model may also be constructed in other manners, which is not described herein again.
SA 4: and adjusting the learning rate, the training times and the optimization mode to obtain a part segmentation model with the best vehicle part segmentation effect as a preset part segmentation model.
After a part segmentation model for analyzing the vehicle part segmentation effect is established, the part segmentation model with the best vehicle part segmentation effect on a test set is finally selected as a preset part segmentation model through continuously adjusting the learning rate, the training times, the optimization mode and other hyper-parameters.
For example, in the process of training the preset component segmentation model, the initial learning rate is 0.005, the training times are 18 times, the 12 th epoch is attenuated to 0.0005, the 15 th epoch is attenuated to 0.00005, and the training is finished after 18 epochs are trained, wherein the optimization mode adopts SGD with momentum being 0.9.
The mIoU of the preset component segmentation model obtained in the embodiment can reach 0.98 on the test set, the accuracy of vehicle component segmentation is higher, and the effect is optimal.
In this embodiment, the initial learning rate is 0.005, the training times are 18 times, and the optimization manner is SGD, which are only exemplary illustrations, and in other embodiments, the initial learning rate, the training times, and the optimization manner may be other, and are not described herein again.
In the embodiment, a plurality of historical car loss pictures are obtained, the historical car loss pictures comprise the previous car loss pictures of all the car types on the market, the historical car loss pictures are classified and labeled according to the parts of the car to obtain a part segmentation database, the parts of the car comprise a fender, a car door, a bumper and a bonnet, the part segmentation database is used as a training set and a test set for vehicle part segmentation, a part segmentation model for analyzing the vehicle part segmentation effect is established, a part segmentation model with the best vehicle part segmentation effect is obtained by adjusting the learning rate, the training times and the optimization mode and is used as a preset part segmentation model, the obtaining mode of the preset part segmentation model is further refined, the historical car loss pictures of all the car types are adopted and labeled, the accuracy and the reliability of the preset part segmentation model are improved, and the accuracy of the vehicle part segmentation is higher, the effect is optimal.
In an embodiment, before determining the plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and the trained preset damage detection model, the preset damage detection model needs to be trained in advance, and the obtaining method of the set damage detection model includes:
SB 1: and acquiring a plurality of historical vehicle loss pictures, wherein the historical vehicle loss pictures comprise previous vehicle loss pictures of all vehicle types on the market.
For example, from an insurance company vehicle loss database, a plurality of historical vehicle loss pictures are obtained, wherein the historical vehicle loss pictures comprise vehicle loss pictures of all vehicle types on the market.
SB 2: and carrying out rectangular marking on the plurality of historical vehicle damage pictures according to the damage degree of the vehicle to obtain a damage detection database, wherein the damage degree of the vehicle comprises scraping, sinking and missing.
For example, after a plurality of historical car damage pictures are acquired, each damage in the pictures is subjected to rectangular marking in a manual marking mode, and then a damage detection database is obtained according to various damage structures such as scraping, sinking and missing.
In this embodiment, the damage degree of the vehicle including scraping, sinking, and missing is only an exemplary illustration, and in other embodiments, the damage degree of the vehicle may also include other damages, which is not described herein again.
SB 3: and (4) establishing a damage detection model for analyzing the vehicle damage judgment effect by taking the damage detection database as a training set and a test set for vehicle damage judgment.
For example, after obtaining the damage detection database, the damage detection database is used as a training set and a test set, and a damage detection model for analyzing the vehicle damage judgment effect based on the RetinaNet network is constructed based on the tensierflow framework. The number of the car damage pictures in the training set of the damage detection database is twenty thousand, and the number of the test set pictures is thirty thousand.
In this embodiment, based on the tensoflow frame, the establishment of the injury detection model for analyzing the vehicle injury judgment effect based on the RetinaNet network is only an exemplary description, and in other embodiments, the injury detection model may also be obtained in other manners, which is not described herein again.
SB 4: and performing data enhancement on the picture data in the injury detection database, adjusting the learning rate, the training times and the optimization mode, and obtaining an injury detection model with the best vehicle injury judgment effect as a preset injury detection model.
After a damage detection model for analyzing the vehicle damage judgment effect is established, data enhancement such as random cutting, random overturning and the like is carried out on a training set picture in the damage detection model, learning rate, training times and optimization mode hyper-parameters are continuously adjusted, and then the damage detection model with the best vehicle damage judgment effect on a test set is selected as a preset damage detection model.
For example, the training number is 24, 24 epochs are trained in total, the learning rate is initially 0.005, batch _ size is 8, momentum is 0.9, weight _ decay is 0.0001, the learning rate is attenuated to 0.0005 and 0.00005 at 16 epochs and 23 epochs, and the SGD is adopted as the optimization mode.
The mAP of the preset damage detection model obtained in the embodiment on the test set can reach 0.82, the accuracy of judging the damage of the vehicle is higher, and the effect is optimal.
In this embodiment, the initial learning rate is 0.005, the training times are 24 times, and the optimization manner is SGD, which are only exemplary illustrations, and in other embodiments, the initial learning rate, the training times, and the optimization manner may be other, and are not described herein again.
In the embodiment, a plurality of historical car damage pictures are obtained, the historical car damage pictures comprise the previous car damage pictures of all the car types on the market, the historical car damage pictures are subjected to rectangular marking according to the damage degree of the car to obtain a damage detection database, the damage degree of the car comprises scraping, sinking and missing, the damage detection database is used as a training set and a testing set for judging the damage of the car, a damage detection model for analyzing the damage judgment effect of the car is established, the picture data in the damage detection database is subjected to data enhancement, the learning rate, the training times and the optimization mode are adjusted to obtain a damage detection model with the best damage judgment effect of the car, the damage detection model is used as a preset damage detection model, the obtaining mode of the preset damage detection model is further refined, the historical car damage pictures of all the car types are adopted and the damage degree of the car is subjected to rectangular marking, and the accuracy and reliability of the preset damage detection model are improved, the accuracy of judging the damage of the vehicle is higher, and the effect is optimal.
In the embodiment, the vehicle damage assessment method based on component segmentation and damage detection is a brand-new vehicle damage assessment problem solving method, solves the problem that damage assessment results in vehicle damage assessment are inconsistent, is high in identification accuracy, short in consumed time (only 400ms), fast in output result, applicable to various vehicle types, wide in photographing angle range and various scenes. The method can enable insurance enterprises to quickly claim damages, promote the progress of the industry, optimize the customer experience, save the labor cost and enable the enterprises to be more competitive.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
In one embodiment, a vehicle damage assessment device based on a deep learning model is provided, and the vehicle damage assessment device based on the deep learning model corresponds to the vehicle damage assessment method based on the deep learning model in the above embodiment one to one. As shown in fig. 5, the deep learning model-based vehicle damage assessment apparatus includes an obtaining module 501, a first determining module 502, a second determining module 503, and an output module 504. The functional modules are explained in detail as follows:
the acquiring module 501 is configured to acquire a plurality of pictures to be identified of a vehicle, where the pictures to be identified are pictures that reflect a damaged condition of the vehicle;
a first determining module 502, configured to determine a damaged main component of the vehicle according to the multiple pictures to be recognized and a trained preset component segmentation model, where the preset component segmentation model is a model with an optimal segmentation effect on a vehicle component;
a second determining module 503, configured to determine a plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and a preset damage detection model obtained through training, where the preset damage detection model is a model with a best vehicle damage judgment effect;
an output module 504, configured to output a damage assessment result of the vehicle according to the damaged main component and the plurality of damage detection frames, where the damage assessment result includes a heaviest damaged component and a heaviest damaged category of the vehicle.
The first determining module 502 is specifically configured to:
inputting the multiple pictures to be recognized into the preset component segmentation model, and acquiring all damaged components contained in the multiple pictures to be recognized;
acquiring the area of each damaged part, and comparing the area of all the damaged parts;
and taking the damaged part with the largest area as a damaged main part of the vehicle.
The second determining module 503 is specifically configured to:
inputting the multiple pictures to be recognized into the preset component segmentation model to obtain a damage detection frame of each picture to be recognized;
determining a plurality of damage detection frames of the damaged main part according to the damaged main part and the damage detection frame of each picture to be identified.
The second determining module 503 is further specifically configured to:
determining a central point of each injury detection frame, and determining the position of the central point;
judging whether the position of the central point is positioned in the damaged main part;
if so, reserving a damage detection frame corresponding to the central point, and taking the corresponding damage detection frame as a damage detection frame of the damaged main part;
and if so, deleting the damage detection frame corresponding to the central point.
The obtaining module 501 is further configured to obtain the preset component segmentation model, and includes:
acquiring a plurality of historical vehicle loss pictures, wherein the historical vehicle loss pictures comprise vehicle loss pictures of all vehicle types on the market;
classifying and labeling the plurality of historical vehicle loss pictures according to components of the vehicle to obtain a component segmentation database, wherein the components of the vehicle comprise a fender, a vehicle door, a bumper and a hood;
taking the part segmentation database as a training set and a test set of vehicle part segmentation, and establishing a part segmentation model for analyzing the vehicle part segmentation effect;
and adjusting the learning rate, the training times and the optimization mode to obtain a part segmentation model with the best vehicle part segmentation effect as the preset part segmentation model.
The obtaining module 501 is further configured to obtain an obtaining manner of the preset flaw detection model, including:
acquiring a plurality of historical vehicle loss pictures, wherein the historical vehicle loss pictures comprise vehicle loss pictures of all vehicle types on the market;
performing rectangular marking on the plurality of historical vehicle damage pictures according to the damage degree of the vehicle to obtain a damage detection database, wherein the damage degree of the vehicle comprises scraping, sinking and missing;
taking the damage detection database as a training set and a test set for vehicle damage judgment, and establishing a damage detection model for analyzing the vehicle damage judgment effect;
and performing data enhancement on the picture data in the injury detection database, adjusting the learning rate, the training times and the optimization mode, and obtaining an injury detection model with the best vehicle injury judgment effect as the preset injury detection model.
For specific definition of the vehicle damage assessment device based on the deep learning model, the above definition of the vehicle damage assessment method based on the deep learning model can be referred to, and details are not repeated here. The modules in the vehicle damage assessment device based on the deep learning model can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 6. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external server through a network connection. The computer program is executed by a processor to implement a deep learning model-based vehicle impairment assessment method.
In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
acquiring a plurality of pictures to be identified of a vehicle, wherein the pictures to be identified are pictures reflecting the damaged condition of the vehicle;
determining a damaged main component of the vehicle according to the plurality of pictures to be recognized and a preset component segmentation model obtained through training, wherein the preset component segmentation model is a model with the best vehicle component segmentation effect;
determining a plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and a preset damage detection model obtained by training, wherein the preset damage detection model is a model with the best vehicle damage judgment effect;
and outputting a damage assessment result of the vehicle according to the damaged main component and the plurality of damage detection frames, wherein the damage assessment result comprises the heaviest damaged component and the heaviest damaged category of the vehicle.
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:
acquiring a plurality of pictures to be identified of a vehicle, wherein the pictures to be identified are pictures reflecting the damaged condition of the vehicle;
determining a damaged main component of the vehicle according to the plurality of pictures to be recognized and a preset component segmentation model obtained through training, wherein the preset component segmentation model is a model with the best vehicle component segmentation effect;
determining a plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and a preset damage detection model obtained by training, wherein the preset damage detection model is a model with the best vehicle damage judgment effect;
and outputting a damage assessment result of the vehicle according to the damaged main component and the plurality of damage detection frames, wherein the damage assessment result comprises the heaviest damaged component and the heaviest damaged category of the vehicle.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present invention, and are intended to be included within the scope of the present invention.
Claims (10)
1. A vehicle damage assessment method based on a deep learning model is characterized by comprising the following steps:
acquiring a plurality of pictures to be identified of a vehicle, wherein the pictures to be identified are pictures reflecting the damaged condition of the vehicle;
determining a damaged main component of the vehicle according to the plurality of pictures to be recognized and a preset component segmentation model obtained through training, wherein the preset component segmentation model is a model with the best vehicle component segmentation effect;
determining a plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and a preset damage detection model obtained through training, wherein the preset damage detection model is a model with the best vehicle damage judgment effect;
and outputting a damage assessment result of the vehicle according to the damaged main component and the plurality of damage detection frames, wherein the damage assessment result comprises the heaviest damaged component and the heaviest damaged category of the vehicle.
2. The vehicle damage assessment method based on the deep learning model as claimed in claim 1, wherein the determining of the damaged main component of the vehicle according to the plurality of pictures to be recognized and the trained preset component segmentation model comprises:
inputting the multiple pictures to be recognized into the preset component segmentation model, and acquiring all damaged components contained in the multiple pictures to be recognized;
acquiring the area of each damaged part, and comparing the area of all the damaged parts;
and taking the damaged part with the largest area as a damaged main part of the vehicle.
3. The vehicle damage assessment method based on the deep learning model as claimed in claim 1, wherein the determining of the damaged main component of the vehicle according to the plurality of pictures to be recognized and the trained preset component segmentation model comprises:
inputting the multiple pictures to be recognized into the preset component segmentation model to obtain a damage detection frame of each picture to be recognized;
determining a plurality of damage detection frames of the damaged main part according to the damaged main part and the damage detection frame of each picture to be identified.
4. The deep learning model-based vehicle damage assessment method according to claim 3, wherein the determining a plurality of damage detection frames of the damaged main component according to the damaged main component and the damage detection frame of each of the pictures to be identified comprises:
determining a central point of each injury detection frame, and determining the position of the central point;
judging whether the position of the central point is positioned in the damaged main part;
if the position of the central point is located in the damaged main part, reserving a damage detection frame corresponding to the central point, and taking the corresponding damage detection frame as a damage detection frame of the damaged main part;
and if the position of the central point is not located in the damaged main part, deleting the damage detection frame corresponding to the central point.
5. The vehicle damage assessment method based on the deep learning model according to any one of claims 1 to 4, wherein the preset component segmentation model is obtained in a manner comprising:
acquiring a plurality of historical vehicle loss pictures, wherein the historical vehicle loss pictures comprise previous vehicle loss pictures of all vehicle types on the market;
classifying and labeling the plurality of historical vehicle loss pictures according to components of the vehicle to obtain a component segmentation database, wherein the components of the vehicle comprise a fender, a vehicle door, a bumper and a hood;
taking the part segmentation database as a training set and a test set of vehicle part segmentation, and establishing a part segmentation model for analyzing the vehicle part segmentation effect;
and adjusting the learning rate, the training times and the optimization mode to obtain a part segmentation model with the best vehicle part segmentation effect as the preset part segmentation model.
6. The deep learning model-based vehicle damage assessment method according to any one of claims 1-4, wherein the default damage detection model is obtained by a method comprising:
acquiring a plurality of historical vehicle loss pictures, wherein the historical vehicle loss pictures comprise previous vehicle loss pictures of all vehicle types on the market;
performing rectangular marking on the plurality of historical vehicle damage pictures according to the damage degree of the vehicle to obtain a damage detection database, wherein the damage degree of the vehicle comprises scraping, sinking and missing;
taking the damage detection database as a training set and a test set for vehicle damage judgment, and establishing a damage detection model for analyzing the vehicle damage judgment effect;
and performing data enhancement on the picture data in the injury detection database, adjusting the learning rate, the training times and the optimization mode, and obtaining an injury detection model with the best vehicle injury judgment effect as the preset injury detection model.
7. A vehicle damage assessment device based on a deep learning model is characterized by comprising:
the system comprises an acquisition module, a recognition module and a recognition module, wherein the acquisition module is used for acquiring a plurality of pictures to be recognized of a vehicle, and the pictures to be recognized are pictures reflecting the damaged condition of the vehicle;
the first determination module is used for determining a damaged main component of the vehicle according to the plurality of pictures to be recognized and a preset component segmentation model obtained through training, wherein the preset component segmentation model is a model with the best segmentation effect on vehicle components;
the second determining module is used for determining a plurality of damage detection frames of the damaged main component according to the plurality of pictures to be recognized and a preset damage detection model obtained through training, wherein the preset damage detection model is a model with the best vehicle damage judgment effect;
and the output module is used for outputting a damage assessment result of the vehicle according to the damaged main component and the damage detection frames, wherein the damage assessment result comprises the heaviest damaged component and the heaviest damaged category of the vehicle.
8. The deep learning model-based vehicle impairment assessment apparatus of claim 7, wherein the first determination module is specifically configured to:
inputting the multiple pictures to be recognized into the preset component segmentation model, and acquiring all damaged components contained in the multiple pictures to be recognized;
acquiring the area of each damaged part, and comparing the area of all the damaged parts;
and taking the damaged part with the largest area as a damaged main part of the vehicle.
9. A computer arrangement comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, carries out the steps of the deep learning model based vehicle impairment method according to any one of claims 1 to 6.
10. A readable storage medium, in which a computer program is stored, which, when being executed by a processor, carries out the steps of the deep learning model based vehicle impairment method according to any one of claims 1 to 6.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010478087.9A CN111666995B (en) | 2020-05-29 | 2020-05-29 | Vehicle damage assessment method, device, equipment and medium based on deep learning model |
| PCT/CN2020/121777 WO2021135513A1 (en) | 2020-05-29 | 2020-10-19 | Deep learning model-based vehicle loss assessment method and apparatus, device and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010478087.9A CN111666995B (en) | 2020-05-29 | 2020-05-29 | Vehicle damage assessment method, device, equipment and medium based on deep learning model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111666995A true CN111666995A (en) | 2020-09-15 |
| CN111666995B CN111666995B (en) | 2024-04-09 |
Family
ID=72385208
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010478087.9A Active CN111666995B (en) | 2020-05-29 | 2020-05-29 | Vehicle damage assessment method, device, equipment and medium based on deep learning model |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111666995B (en) |
| WO (1) | WO2021135513A1 (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112801951A (en) * | 2021-01-15 | 2021-05-14 | 深圳市商汤科技有限公司 | Target component damage detection method, device, equipment and storage medium |
| CN112907576A (en) * | 2021-03-25 | 2021-06-04 | 平安科技（深圳）有限公司 | Vehicle damage grade detection method and device, computer equipment and storage medium |
| CN113033372A (en) * | 2021-03-19 | 2021-06-25 | 北京百度网讯科技有限公司 | Vehicle damage assessment method and device, electronic equipment and computer readable storage medium |
| WO2021135513A1 (en) * | 2020-05-29 | 2021-07-08 | 平安科技（深圳）有限公司 | Deep learning model-based vehicle loss assessment method and apparatus, device and medium |
| CN113177926A (en) * | 2021-05-11 | 2021-07-27 | 泰康保险集团股份有限公司 | Image detection method and device |
| CN113436175A (en) * | 2021-06-30 | 2021-09-24 | 平安科技（深圳）有限公司 | Method, device and equipment for evaluating segmentation quality of vehicle image and storage medium |
| CN115273044A (en) * | 2022-07-15 | 2022-11-01 | 哈尔滨市科佳通用机电股份有限公司 | Vehicle door damage fault identification and detection method based on improved graph convolution network |
| CN117671329A (en) * | 2023-11-14 | 2024-03-08 | 平安科技(上海)有限公司 | Vehicle damage analysis method, device, equipment and medium based on artificial intelligence |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114140430A (en) * | 2021-11-30 | 2022-03-04 | 北京比特易湃信息技术有限公司 | Vehicle damage reporting method based on deep learning |
| CN117671330B (en) * | 2023-11-14 | 2024-06-21 | 平安科技(上海)有限公司 | Vehicle damage assessment method, device, computer equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110569701A (en) * | 2018-12-29 | 2019-12-13 | 阿里巴巴集团控股有限公司 | computer-implemented vehicle damage assessment method and device |
| US20200074222A1 (en) * | 2018-08-31 | 2020-03-05 | Alibaba Group Holding Limited | System and method for performing image processing based on a damage assessment image judgement model |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107358596B (en) * | 2017-04-11 | 2020-09-18 | 阿里巴巴集团控股有限公司 | Vehicle loss assessment method and device based on image, electronic equipment and system |
| DE102017220027A1 (en) * | 2017-11-10 | 2019-05-16 | Ford Global Technologies, Llc | Method for damage control in motor vehicles |
| CN109635742A (en) * | 2018-12-13 | 2019-04-16 | 深源恒际科技有限公司 | A kind of subassembly damnification recognition method in vehicle image setting loss |
| CN111666995B (en) * | 2020-05-29 | 2024-04-09 | 平安科技（深圳）有限公司 | Vehicle damage assessment method, device, equipment and medium based on deep learning model |
-
2020
- 2020-05-29 CN CN202010478087.9A patent/CN111666995B/en active Active
- 2020-10-19 WO PCT/CN2020/121777 patent/WO2021135513A1/en active Application Filing
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200074222A1 (en) * | 2018-08-31 | 2020-03-05 | Alibaba Group Holding Limited | System and method for performing image processing based on a damage assessment image judgement model |
| CN110569701A (en) * | 2018-12-29 | 2019-12-13 | 阿里巴巴集团控股有限公司 | computer-implemented vehicle damage assessment method and device |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021135513A1 (en) * | 2020-05-29 | 2021-07-08 | 平安科技（深圳）有限公司 | Deep learning model-based vehicle loss assessment method and apparatus, device and medium |
| CN112801951A (en) * | 2021-01-15 | 2021-05-14 | 深圳市商汤科技有限公司 | Target component damage detection method, device, equipment and storage medium |
| CN113033372B (en) * | 2021-03-19 | 2023-08-18 | 北京百度网讯科技有限公司 | Vehicle damage assessment method, device, electronic equipment and computer readable storage medium |
| CN113033372A (en) * | 2021-03-19 | 2021-06-25 | 北京百度网讯科技有限公司 | Vehicle damage assessment method and device, electronic equipment and computer readable storage medium |
| CN112907576A (en) * | 2021-03-25 | 2021-06-04 | 平安科技（深圳）有限公司 | Vehicle damage grade detection method and device, computer equipment and storage medium |
| CN112907576B (en) * | 2021-03-25 | 2024-02-02 | 平安科技（深圳）有限公司 | Vehicle damage grade detection method and device, computer equipment and storage medium |
| CN113177926A (en) * | 2021-05-11 | 2021-07-27 | 泰康保险集团股份有限公司 | Image detection method and device |
| CN113177926B (en) * | 2021-05-11 | 2023-11-14 | 泰康保险集团股份有限公司 | Image detection method and device |
| CN113436175B (en) * | 2021-06-30 | 2023-08-18 | 平安科技（深圳）有限公司 | Method, device, equipment and storage medium for evaluating vehicle image segmentation quality |
| WO2023273296A1 (en) * | 2021-06-30 | 2023-01-05 | 平安科技（深圳）有限公司 | Vehicle image segmentation quality evaluation method and apparatus, device, and storage medium |
| CN113436175A (en) * | 2021-06-30 | 2021-09-24 | 平安科技（深圳）有限公司 | Method, device and equipment for evaluating segmentation quality of vehicle image and storage medium |
| CN115273044A (en) * | 2022-07-15 | 2022-11-01 | 哈尔滨市科佳通用机电股份有限公司 | Vehicle door damage fault identification and detection method based on improved graph convolution network |
| CN117671329A (en) * | 2023-11-14 | 2024-03-08 | 平安科技(上海)有限公司 | Vehicle damage analysis method, device, equipment and medium based on artificial intelligence |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111666995B (en) | 2024-04-09 |
| WO2021135513A1 (en) | 2021-07-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111666995B (en) | Vehicle damage assessment method, device, equipment and medium based on deep learning model | |
| US11823365B2 (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 | |
| US20210375032A1 (en) | Photo deformation techniques for vehicle repair analysis | |
| US11443288B2 (en) | Automatic assessment of damage and repair costs in vehicles | |
| CN111680746B (en) | Vehicle damage detection model training, vehicle damage detection method, device, equipment and medium | |
| WO2021114809A1 (en) | Vehicle damage feature detection method and apparatus, computer device, and storage medium | |
| CA3174691A1 (en) | Human face fuzziness detecting method, device, computer equipment and storage medium | |
| CN110706261A (en) | Vehicle violation detection method and device, computer equipment and storage medium | |
| CN110807491A (en) | License plate image definition model training method, definition detection method and device | |
| CN116485779B (en) | Adaptive wafer defect detection method and device, electronic equipment and storage medium | |
| EP4131061A1 (en) | Vehicle loss assessment method, vehicle loss assessment apparatus, and electronic device using same | |
| CN112668462B (en) | Vehicle damage detection model training, vehicle damage detection method, device, equipment and medium | |
| CN111325256A (en) | Vehicle appearance detection method and device, computer equipment and storage medium | |
| CN110781887A (en) | License plate screw detection method and device and computer equipment | |
| US20220036132A1 (en) | Semantic image segmentation for cognitive analysis of physical structures | |
| CN110728680A (en) | Automobile data recorder detection method and device, computer equipment and storage medium | |
| CN112241705A (en) | Target detection model training method and target detection method based on classification regression | |
| CN117173145A (en) | Method, device, equipment and storage medium for detecting surface defects of power equipment | |
| CN110751623A (en) | Joint feature-based defect detection method, device, equipment and storage medium | |
| CN112446402A (en) | Loading rate identification method and device, computer equipment and storage medium | |
| CN115984759A (en) | Substation switch state identification method and device, computer equipment and storage medium | |
| CN111242070A (en) | Target object detection method, computer device, and storage medium | |
| CN110991331A (en) | Detection method and device for construction site, computer equipment and storage medium | |
| CN113516161B (en) | Risk early warning method for tunnel constructors |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40032357 Country of ref document: HK |
|
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |