CN117671330A - Vehicle damage assessment method, device, computer equipment and storage medium - Google Patents
Vehicle damage assessment method, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN117671330A CN117671330A CN202311517040.9A CN202311517040A CN117671330A CN 117671330 A CN117671330 A CN 117671330A CN 202311517040 A CN202311517040 A CN 202311517040A CN 117671330 A CN117671330 A CN 117671330A
- Authority
- CN
- China
- Prior art keywords
- image
- damage
- characteristic
- label
- identified
- 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 304
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000001514 detection method Methods 0.000 claims abstract description 20
- 238000000605 extraction Methods 0.000 claims description 43
- 230000006735 deficit Effects 0.000 claims description 40
- 238000004590 computer program Methods 0.000 claims description 17
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 230000003902 lesion Effects 0.000 description 10
- 230000006870 function Effects 0.000 description 7
- 208000027418 Wounds and injury Diseases 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 5
- 208000014674 injury Diseases 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000007781 pre-processing Methods 0.000 description 4
- 238000013145 classification model Methods 0.000 description 3
- 230000002349 favourable effect Effects 0.000 description 3
- 206010039203 Road traffic accident Diseases 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006748 scratching Methods 0.000 description 1
- 230000002393 scratching effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The application relates to the technical field of artificial intelligence and discloses a vehicle damage assessment method, a device, computer equipment and a storage medium, wherein an image to be identified of a vehicle to be damaged is obtained, and the image to be identified is an image containing at least one damage position of the vehicle to be damaged; performing damage detection on the image to be identified, and determining the number of damage class labels corresponding to the damage positions, the label names of the damage class labels and the probability of the label names; performing component recognition on the image to be recognized, and determining component mask information corresponding to the damage position; based on the number of damage class labels, the label names of the damage class labels, the probability of the label names and the component mask information, the damage assessment result of the damage position is output, the reliability of damage class identification of accident vehicles is improved, and further the damage condition of the vehicles is accurately identified.
Description
Technical Field
The invention relates to the technical fields of vehicle damage assessment, intelligent voice and artificial intelligence, in particular to a vehicle damage assessment method, a vehicle damage assessment device, computer equipment and a storage medium.
Background
With the progress of technology, vehicles become one of the indispensable transportation means in daily life, and the increase of vehicles increases the occurrence rate of traffic accidents, so that after purchasing vehicles, people purchase car insurance for vehicles, and when the vehicles have accidents, insurance companies usually survey the loss condition of the on-site insured vehicles at the first time, and the loss condition is taken as the basis of claim of the insurance companies.
Currently, it is generally required that an evaluator survey and judge an accident vehicle at a traffic accident occurrence site and classify pictures taken at the site to determine damage types of various parts of a vehicle body. However, since there are a plurality of injuries on some parts of the vehicle to be damaged, and there is a mixed dispute of the types of injuries, for example, there is a single injury on the vehicle door, the injury may be considered as scratch or may be considered as slight depression, so that the types of injuries cannot be accurately identified when classifying pictures, and the reliability of identifying the loss condition of the vehicle is low.
Disclosure of Invention
Based on the method, the device, the computer equipment and the storage medium for vehicle damage assessment are provided, the reliability of damage type identification of accident vehicles is improved, and further the method and the device are favorable for accurately identifying the loss condition of the vehicles.
In a first aspect, a method for vehicle damage assessment is provided, including:
acquiring an image to be identified of a vehicle to be damaged, wherein the image to be identified is an image containing at least one damaged position of the vehicle to be damaged;
performing damage detection on the image to be identified, and determining the number of damage class labels corresponding to the damage positions, the label names of the damage class labels and the probability of the label names;
performing component identification on the image to be identified, and determining component mask information corresponding to the damage position;
and outputting an impairment determination result of the impairment position based on the number of impairment category labels, the label names of the impairment category labels, the probability of the label names and the component mask information.
In a second aspect, there is provided a vehicle damage assessment device comprising:
the acquisition module is used for acquiring an image to be identified of the vehicle to be damaged, wherein the image to be identified is an image containing at least one damaged position of the vehicle to be damaged;
the first determining module is used for detecting damage to the image to be identified and determining the number of damage class labels corresponding to the damage position, the label names of the damage class labels and the probability of the label names;
The second determining module is used for carrying out component recognition on the image to be recognized and determining component mask information corresponding to the damage position;
and the output module is used for outputting the damage assessment result of the damage position based on the number of the damage category labels, the label names of the damage category labels, the probability of the label names and the component mask information.
Optionally, in some embodiments of the present application, the first determining module further includes:
the extraction sub-module is used for extracting the characteristics of the image to be identified to obtain a target characteristic image corresponding to the image to be identified;
the first determining submodule is used for carrying out feature classification on the target feature image and determining a feature classification result;
and the second determining submodule is used for determining the number of the damage category labels and the label names of the damage category labels according to the feature classification result.
Optionally, in some embodiments of the present application, the extraction sub-module further includes:
the output unit is used for carrying out grouping convolution on the image to be identified and outputting an initial characteristic image;
and the extraction unit is used for carrying out feature extraction on the initial feature image to obtain the target feature image.
Optionally, in some embodiments of the present application, the extraction unit further comprises:
the first extraction subunit is used for extracting the characteristics of the initial characteristic image and outputting a first characteristic image;
the second extraction subunit is used for extracting the characteristics of the first characteristic graph and outputting a second characteristic image;
the third extraction subunit is used for extracting the characteristics of the second characteristic graph and outputting a third characteristic image;
and the fourth extraction subunit is used for extracting the characteristics of the third characteristic image and outputting a target characteristic image.
Optionally, in some embodiments of the present application, the extraction unit further comprises:
a fifth extraction subunit, configured to obtain a first offset of the initial feature image, extract a first image feature of the initial feature image, perform deformable convolution on the first image feature and the first offset, and output a first feature image;
a sixth extraction subunit, configured to obtain a second offset of the first feature image, extract a second image feature of the first feature image, perform deformable convolution on the second image feature and the second offset, and output a second feature image;
A seventh extraction subunit, configured to obtain a third offset of the second feature image, extract a third image feature of the second feature image, perform deformable convolution on the third image feature and the third offset, and output a third feature image;
and the eighth extraction subunit is used for acquiring a fourth offset of the third characteristic image, extracting a fourth image characteristic of the third characteristic image, performing deformable convolution on the fourth image characteristic and the fourth offset, and outputting a target characteristic image.
Optionally, in some embodiments of the present application, the output module further includes:
a third determining sub-module for determining the number of damage category labels to determine whether the damage category label is greater than one;
the obtaining submodule is used for comparing the probability of each label name when the number of the damage class labels is larger than one, and selecting the label name corresponding to the maximum probability as a target label;
and the first output sub-module is used for taking the object tag and the part name in the part mask information as the damage assessment result and outputting the damage assessment result.
Optionally, in some embodiments of the present application, a third determining module is further included, where the third determining module is specifically configured to:
When the damage class label is determined to be one according to the number of the damage class labels, the label name is used as a target label; and taking the object tag and the part name in the part mask information as the damage assessment result and outputting the damage assessment result.
In a third aspect, a computer device is provided, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the vehicle impairment determination method described above when the computer program is executed.
In a fourth aspect, a computer readable storage medium is provided, the computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of the vehicle impairment method described above.
The application provides a vehicle damage assessment method, a device, computer equipment and a storage medium, wherein an image to be identified of a vehicle to be damaged is obtained, and the image to be identified is an image containing at least one damage position of the vehicle to be damaged; performing damage detection on the image to be identified, and determining the number of damage class labels corresponding to the damage positions, the label names of the damage class labels and the probability of the label names; performing component recognition on the image to be recognized, and determining component mask information corresponding to the damage position; and outputting an impairment determination result of the impairment position based on the number of impairment category labels, the label names of the impairment category labels, the probability of the label names and the component mask information. In the vehicle damage assessment scheme provided by the application, damage detection is carried out on the image to be identified of the vehicle to be assessed, the number of damage class labels corresponding to the damage position and the label names of the damage class labels are determined, the image to be identified is subjected to component identification, and the component mask information corresponding to the damage position is determined, so that damage assessment results of the damage position can be output according to the number of the damage class labels, the label names of the damage class labels, the probability of the label names and the component mask information, and the damage assessment results of the damage position can be output by combining the number of the damage class labels, the label names of the damage class labels, the probability of the label names and the component mask information corresponding to the damage position in the image to be identified.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is an application environment diagram of a vehicle damage assessment method provided in an embodiment of the present application;
FIG. 2 is a flow chart of a method for vehicle damage assessment provided in an embodiment of the present application;
fig. 3 is a block diagram of a vehicle damage assessment device according to an embodiment of the present application;
FIG. 4 is a first block diagram of a computer device according to an embodiment of the present application;
fig. 5 is a second structural block diagram of a computer device according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The vehicle damage assessment method based on artificial intelligence provided by the embodiment of the invention can be applied to an application environment as shown in fig. 1, wherein a client communicates with a server through a network. The method comprises the steps that a server side can obtain an image to be identified of a vehicle to be damaged through a client side, wherein the image to be identified is an image containing at least one damaged position of the vehicle to be damaged; performing damage detection on the image to be identified, and determining the number of damage class labels corresponding to the damage positions, the label names of the damage class labels and the probability of the label names; performing component recognition on the image to be recognized, and determining component mask information corresponding to the damage position; according to the method and the device, an assessment result of a damage position is output and fed back to a client based on the number of damage type labels, the label names of the damage type labels, the probability of the label names and the component mask information, and in the method and the device, damage detection is carried out on images to be identified of a vehicle to be assessed, the number of the damage type labels and the label names of the damage type labels corresponding to the damage position are determined, the components to be identified are identified, and the component mask information corresponding to the damage position is determined, so that the assessment result of the damage position can be output according to the number of the damage type labels, the label names of the damage type labels, the probability of the label names and the component mask information, the reliability of damage type identification of the vehicle to be assessed is improved, and further the identification of the loss condition of the vehicle is facilitated. The clients may be, but are not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices. The server may be implemented by a stand-alone server or a server cluster formed by a plurality of servers. The present invention will be described in detail with reference to specific examples.
Referring to fig. 2, fig. 2 is a schematic flow chart of a vehicle damage assessment method according to an embodiment of the present invention, where the method can be applied to a terminal or a server, and the embodiment is exemplified by a server. The vehicle damage assessment method comprises the following steps:
101: and acquiring an image to be identified of the vehicle to be damaged, wherein the image to be identified is an image containing at least one damaged position of the vehicle to be damaged.
The vehicle to be damaged may be a vehicle which needs an insurance company to determine the damage condition when an accident occurs to the vehicle to be protected, namely, an accident vehicle which needs the insurance company to damage.
For example, assuming that the body of the vehicle to be damaged is damaged, the damaged position on the body may be photographed by a photographing apparatus, thereby obtaining an image to be recognized.
102: and performing damage detection on the image to be identified, and determining the number of damage class labels corresponding to the damage positions, the label names of the damage class labels and the probability of the label names.
Wherein the damage class label is used for indicating the label of the damage class.
For example, if the image to be identified is an image corresponding to a damaged position of a body of the vehicle to be damaged, when damage detection is performed on the image to be identified, the number of damage class labels corresponding to the damaged position is determined to be 2, and the names of the damage class labels are scratch and dent.
Optionally, a pre-trained damage detection model may be used to detect damage to the image to be identified, and the number of damage class labels corresponding to the damage position, the label names of the damage class labels, and the probability of the label names are determined.
Wherein the damage detection model is obtained through deep learning.
After the image to be identified is obtained, the target feature image of the image to be identified can be obtained by extracting features of the image to be identified, and then the target feature image is subjected to feature classification, so that the number of damage class labels and the label names of the damage class labels can be determined, that is, optionally, in some embodiments, the step of performing damage detection on the image to be identified to determine the number of damage class labels corresponding to the damage position, the label names of the damage class labels and the probability of the label names can specifically include:
(11) Extracting features of the image to be identified to obtain a target feature image corresponding to the image to be identified;
the method comprises the steps of carrying out feature extraction on an image to be identified through a convolution network to obtain a target feature image corresponding to the image to be identified.
(12) Performing feature classification on the target feature image, and determining a feature classification result;
The target feature image can be input into an image feature classification model to perform feature classification, and feature classification results are determined. The feature classification result can be a label vector, the image feature classification model can be a model realized based on a normalized exponential function (softmax function), and in the application, the image feature classification model is used for carrying out multi-classification on the damage position in the image to be identified, so that the feature classification result is determined, the label name of the damage type label of the damage position to the drink can be prevented from being confused when the damage types are too similar, the label name of the damage type label of the damage position to the drink can be improved, the reliability of damage type identification of an accident vehicle can be improved, and the loss condition of the vehicle can be accurately identified.
(13) And determining the number of the damage class labels and the label names of the damage class labels according to the feature classification result.
Optionally, if the lesion location in the image to be identified includes two lesions, classifying the target feature image, and determining that the feature classification result is [ m1, m2,0, ], where 1 > m1 > 0,1 > m2 > 0, m1+m2=1, and m1, m2 represent probability values that the lesion label class belongs to a label name, based on which the number of lesion label class labels and label names of the lesion label class can be determined. For example, assuming that the damaged label category includes scratch, dent, fold, scratch, and scratch, for a damaged position in an image to be identified, the damaged position has a scratch probability value of 0.6, the damaged position has a dent probability value of 0.4, the damaged position has a fold probability value of 0, the damaged position has a scratch probability value of 0, the feature classification result is [0.6,0.4,0,0,0], the number of damaged label category corresponding to the damaged position is 2, and the label names of damaged label category corresponding to the damaged position are scratch and dent.
Optionally, if the lesion location in the image to be identified includes a lesion, when classifying the target feature image, the feature classification result is [1,0, 0..0 ], and 1 represents a probability value that the lesion label category belongs to a certain label name, based on which the number of lesion label categories and the label name of the lesion label category can be determined. For example, assuming that the damaged label category includes scratch, dent, fold, scratch, and scratch, for a damaged position in an image to be identified, the damaged position is a scratch having a probability value of 1, the damaged position is a dent having a probability value of 0, the damaged position is a fold having a probability value of 0, the damaged position is a scratch having a probability value of 0, the feature classification result is [1, 0], the number of damaged class labels corresponding to the damaged position is 1, and the label name of the damaged class label corresponding to the damaged position is scratch.
After the image to be identified is obtained, the image to be identified may be subjected to group convolution, an initial feature image is output, and then feature extraction is performed on the initial feature image to obtain a target feature image, that is, optionally, in some embodiments, the step of "performing feature extraction on the image to be identified to obtain a target feature image corresponding to the image to be identified" may specifically include:
(21) Carrying out grouping convolution on the images to be identified and outputting initial characteristic images;
optionally, through a convolution network with N branches, each branch includes a convolution layer with a convolution kernel 1*1, a convolution layer with a convolution kernel 3*3, and a convolution layer with a convolution kernel 1*1, where the step size of each convolution layer is 1, x branches are randomly selected from the N branches, the expansion convolution parameters of the x branches are set to 0, and the expansion convolution parameters of the remaining N-x branches are set to 1, so that the finally output initial feature image fuses different receptive field information, which is favorable for improving the accuracy of category identification of the damage category label, when the image to be identified is subjected to grouping convolution, each branch outputs a corresponding convolution result, and overlaps the convolution result of each branch with the input image to be identified, thereby outputting the initial feature image. Wherein N, x is a positive integer greater than 0. For example, assuming that the convolutional network has 32 branches, x may be 16, that is, 16 branches are randomly selected from the 32 branches, the expansion convolution parameters of the 16 branches are set to 0, and the expansion convolution parameters of the remaining 16 branches are set to 1, so that the finally output initial feature image fuses different receptive field information, which is beneficial to improving the accuracy of identifying the category of the damage category label.
(22) And extracting the features of the initial feature image to obtain a target feature image.
Optionally, feature extraction may be performed on the initial feature image through a convolutional network to obtain the target feature image. For example, the convolution network includes a preprocessing layer and a bottleneck layer, the initial feature image is input into the preprocessing layer to perform preprocessing so as to perform dimension reduction processing on the initial feature image, for example, assuming that the number of channels, the height and the width of the initial feature image are (3, 224, 224), after the initial feature image passes through the preprocessing layer, the output of the number of channels, the height and the width of the initial feature image are (64, 56, 56) is obtained, and then the preprocessed initial feature image is input into the bottleneck layer to perform feature sampling for a plurality of times, and then the target feature image is output.
The initial feature image may be subjected to feature extraction for multiple times to obtain a target feature image, which is favorable for finding the most representative feature for the damage position, and reducing redundant information between features, so as to improve accuracy and reliability of damage category identification, that is, optionally, in some embodiments, the step of "extracting the initial feature image to obtain the target feature image" may specifically include:
(31) Extracting features of the initial feature image, and outputting a first feature image;
alternatively, a convolutional network may be used to perform feature extraction on the initial feature image and output the first feature image.
(32) Extracting the characteristics of the first characteristic graph and outputting a second characteristic graph;
alternatively, a convolutional network may be used to perform feature extraction on the first feature image and output a second feature image.
(33) Extracting the characteristics of the second characteristic graph and outputting a third characteristic image;
alternatively, a convolutional network may be used to perform feature extraction on the second feature image and output a third feature image.
(34) And extracting the characteristics of the third characteristic image and outputting a target characteristic image.
Alternatively, a convolution network may be used to perform feature extraction on the third feature image, and output the target feature image.
For example, the feature extraction may be performed on the initial feature image based on a convolutional network formed by 3 bottleneck layers to obtain a first feature image, where the 3 bottleneck layers are sequentially one bottleneck layer (BTNK 2) with the same number of input and output channels and two bottleneck layers (BTNK 1) with different numbers of input and output channels; the first characteristic image can be subjected to characteristic extraction based on a convolution network formed by 4 bottleneck layers to obtain a second characteristic image, wherein the 4 bottleneck layers are sequentially one bottleneck layer with the same number of input and output channels and three bottleneck layers with different numbers of input and output channels; the second characteristic image can be subjected to characteristic extraction based on a convolution network formed by 6 bottleneck layers to obtain a third characteristic image, wherein the 6 bottleneck layers are sequentially one bottleneck layer with the same number of input and output channels and five bottleneck layers with different numbers of input and output channels; the third feature image may be extracted based on a convolutional network formed by 3 bottleneck layers, where the 3 bottleneck layers are bottleneck layers with the same number of input and output channels and bottleneck layers with different numbers of input and output channels, so as to obtain a target feature image.
The method comprises the steps of obtaining a target feature image by introducing deformable convolution, and being beneficial to more accurately identifying the damage complex deformation feature in the image feature, thereby improving the number of damage class labels corresponding to the damage identification position and the label names of the damage class labels, namely, optionally, in some embodiments, carrying out feature extraction on the initial feature image to obtain the target feature image, wherein the method specifically comprises the following steps:
(41) Acquiring a first offset of an initial feature image, extracting a first image feature of the initial feature image, performing deformable convolution on the first image feature and the first offset, and outputting the first feature image;
(42) Acquiring a second offset of the first characteristic image, extracting a second image characteristic of the first characteristic image, performing deformable convolution on the second image characteristic and the second offset, and outputting the second characteristic image;
(43) Acquiring a third offset of the second characteristic image, extracting a third image characteristic of the second characteristic image, performing deformable convolution on the third image characteristic and the third offset, and outputting the third characteristic image;
(44) And acquiring a fourth offset of the third characteristic image, extracting a fourth image characteristic of the third characteristic image, performing deformable convolution on the fourth image characteristic and the fourth offset, and outputting a target characteristic image.
For example, a first offset of an initial feature image can be obtained through a convolutional neural network, a first image feature of the initial feature image is extracted based on a convolutional network formed by 3 bottleneck layers, then the first offset and the first image feature are input into a deformable convolutional network to perform deformable convolution, and the first feature image is output, wherein the 3 bottleneck layers are sequentially a bottleneck layer (BTNK 2) with the same number of input and output channels and a bottleneck layer (BTNK 1) with two different numbers of input and output channels; acquiring a second offset of the first characteristic image through a convolutional neural network, extracting a second image characteristic of the first characteristic image based on a convolutional network formed by 4 bottleneck layers, and then inputting the second offset and the second image characteristic into a deformable convolutional network to perform deformable convolution, and outputting the second characteristic image, wherein the 4 bottleneck layers are sequentially a bottleneck layer with the same number of input channels and a bottleneck layer with different numbers of three input channels and output channels; extracting a third offset of the second characteristic image through a convolutional neural network, extracting a third image characteristic of the second characteristic image based on a convolutional network formed by 6 bottleneck layers, and then inputting the third offset and the third image characteristic into a deformable convolutional network to perform deformable convolution, and outputting the third characteristic image, wherein the 6 bottleneck layers are sequentially one bottleneck layer with the same number of input channels and five bottleneck layers with different numbers of input channels and output channels; the fourth offset of the third characteristic image can be obtained through a convolutional neural network, the fourth image characteristic of the third characteristic image is extracted based on a convolutional network formed by 3 bottleneck layers, then the fourth offset and the fourth image characteristic are input into a deformable convolutional network to perform deformable convolution, and a target characteristic image is output, wherein the 3 bottleneck layers are the bottleneck layers with the same input and output channels and the bottleneck layers with different input and output channels, and the target characteristic image is obtained.
103: and carrying out component identification on the image to be identified, and determining component mask information corresponding to the damage position.
Based on a pre-trained CNN segmentation network model, such as a deep Lab network model, the image to be identified is subjected to component identification, and component mask information corresponding to the damage position is determined.
Wherein the part mask information includes part names of parts to which the damage positions belong.
104: and outputting an impairment determination result of the impairment position based on the number of impairment category labels, the label names of the impairment category labels, the probability of the label names and the component mask information.
The impairment determination result can be output through at least one mode of voice, characters and the like.
The method comprises the steps that the part name where the damage position is located can be determined according to part mask information, the number of damage class labels can be used for determining whether the damage position in an image to be identified corresponds to a plurality of damage class labels, and the label name of the damage class label can be used for determining the damage class corresponding to the damage position.
The damage assessment method comprises the steps of carrying out splicing according to the number of damage category labels, the label formation of the damage category labels, the probability of label names and the part names, and therefore obtaining damage assessment results. Illustratively, the impairment results may be [ number: 2; part name: a vehicle body; tag name: scratching; probability of scratch: 0.6; tag name: a recess; probability of dishing: 0.57].
In the case that the number of damage class labels is greater than 1, comparing the probability of each label name, thereby determining a target label, and further determining a damage result, that is, optionally, in some embodiments, the step of outputting a damage result of the damage position based on the number of damage class labels, the label names of the damage class labels, the probability of the label names, and the component mask information may specifically include:
(51) Determining the number of damage class labels to determine whether the damage class label is greater than one;
(52) When the number of the damage class labels is larger than one, comparing the probability of each label name, and selecting the label name corresponding to the maximum probability as a target label;
(53) And taking the part names in the target label and the part mask information as the damage assessment result and outputting the damage assessment result.
For example, assuming that the number of damage class labels is two, label names are scratch and dent, respectively, wherein the probability of scratch is 0.65 and the probability of dent is 0.61, then scratch is taken as a target label, and the target label and the part name are spliced, thereby obtaining an impairment result, that is, the impairment result is expressed as [ part name: a vehicle body; tag name: scratch ].
When the number of the damage class labels is one, the label name can be directly used as the target label, so as to determine the damage result, that is, optionally, in some embodiments, the step of outputting the damage result of the damage position based on the number of the damage class labels, the label name of the damage class label, the probability of the label name and the component mask information specifically may include:
(61) When the damage class label is determined to be one according to the number of the damage class labels, the label name is taken as a target label;
(62) And taking the part names in the target label and the part mask information as the damage assessment result and outputting the damage assessment result.
For example, assuming that the number of damage class labels is one and the label name is scratch, the scratch is taken as a target label, and the target label and the part name are spliced, thereby obtaining an impairment result, that is, the impairment result is expressed as [ part name: a vehicle body; tag name: scratch ].
The above is the vehicle damage assessment flow of the present application.
As above, the present application provides a method, an apparatus, a computer device, and a storage medium for vehicle damage assessment, where an image to be identified of a vehicle to be damaged is obtained, and the image to be identified is an image including at least one damaged position of the vehicle to be damaged; performing damage detection on the image to be identified, and determining the number of damage class labels corresponding to the damage positions, the label names of the damage class labels and the probability of the label names; performing component recognition on the image to be recognized, and determining component mask information corresponding to the damage position; and outputting an impairment determination result of the impairment position based on the number of impairment category labels, the label names of the impairment category labels, the probability of the label names and the component mask information. In the vehicle damage assessment scheme provided by the application, damage detection is carried out on the image to be identified of the vehicle to be assessed, the number of damage class labels corresponding to the damage position and the label names of the damage class labels are determined, the image to be identified is subjected to component identification, and the component mask information corresponding to the damage position is determined, so that damage assessment results of the damage position can be output according to the number of the damage class labels, the label names of the damage class labels, the probability of the label names and the component mask information, and the damage assessment results of the damage position can be output by combining the number of the damage class labels, the label names of the damage class labels, the probability of the label names and the component mask information corresponding to the damage position in the image to be identified.
It should be understood that the sequence number of each step in the foregoing embodiment does not mean that the execution sequence of each process should be determined by the function and the internal logic, and should not limit the implementation process of the embodiment of the present invention.
In an embodiment, a vehicle damage assessment device is provided, and the vehicle damage assessment device corresponds to the vehicle damage assessment method in the embodiment, and specific implementation details can be referred to in the method embodiment. Referring to fig. 3, the vehicle damage assessment device includes an acquisition module 201, a first determination module 202, a second determination module 203, and an output module 204, where each functional module is described in detail as follows:
the acquiring module 201 is configured to acquire an image to be identified of a vehicle to be damaged, where the image to be identified is an image including at least one damaged position of the vehicle to be damaged;
the first determining module 202 is configured to perform damage detection on an image to be identified, and determine the number of damage class labels corresponding to the damage position, the label names of the damage class labels, and the probability of the label names;
a second determining module 203, configured to perform component recognition on the image to be recognized, and determine component mask information corresponding to the damage position;
And an output module 204, configured to output a damage determination result of the damage location based on the number of damage category labels, the label names of the damage category labels, the probability of the label names, and the component mask information.
Optionally, in some embodiments, the first determining module further comprises:
the extraction sub-module is used for extracting the characteristics of the image to be identified to obtain a target characteristic image corresponding to the image to be identified;
the first determining submodule is used for carrying out feature classification on the target feature image and determining a feature classification result;
and the second determining submodule is used for determining the number of the damage category labels and the label names of the damage category labels according to the feature classification result.
Optionally, in some embodiments, the extraction sub-module further comprises:
the output unit is used for carrying out grouping convolution on the image to be identified and outputting an initial characteristic image;
and the extraction unit is used for extracting the characteristics of the initial characteristic image to obtain a target characteristic image.
Optionally, in some embodiments, the extraction unit further comprises:
the first extraction subunit is used for extracting the characteristics of the initial characteristic image and outputting a first characteristic image;
the second extraction subunit is used for extracting the characteristics of the first characteristic graph and outputting a second characteristic graph;
The third extraction subunit is used for extracting the characteristics of the second characteristic graph and outputting a third characteristic image;
and the fourth extraction subunit is used for extracting the characteristics of the third characteristic image and outputting a target characteristic image.
Optionally, in some embodiments, the extraction unit further comprises:
a fifth extraction subunit, configured to obtain a first offset of the initial feature image, extract a first image feature of the initial feature image, perform deformable convolution on the first image feature and the first offset, and output a first feature image;
a sixth extraction subunit, configured to obtain a second offset of the first feature image, extract a second image feature of the first feature image, perform deformable convolution on the second image feature and the second offset, and output a second feature image;
a seventh extraction subunit, configured to obtain a third offset of the second feature image, extract a third image feature of the second feature image, perform deformable convolution on the third image feature and the third offset, and output a third feature image;
and the eighth extraction subunit is used for acquiring a fourth offset of the third characteristic image, extracting a fourth image characteristic of the third characteristic image, performing deformable convolution on the fourth image characteristic and the fourth offset, and outputting a target characteristic image.
Optionally, in some embodiments, the output module further comprises:
a third determining sub-module for determining the number of damage category labels to determine whether the damage category label is greater than one;
the acquisition sub-module is used for comparing the probability of each label name when the number of the damage class labels is larger than one, and selecting the label name corresponding to the maximum probability as a target label;
and the first output sub-module is used for taking the target label and the part name in the part mask information as the damage assessment result and outputting the damage assessment result.
Optionally, in some embodiments, a third determining module is further included, and the third determining module is specifically configured to:
when the damage class label is determined to be one according to the number of the damage class labels, the label name is taken as a target label; and taking the part names in the target label and the part mask information as the damage assessment result and outputting the damage assessment result.
In one embodiment, a computer device is provided, which may be a server, and the internal structure of which may be as shown in fig. 4. The computer device includes a processor, a memory, a network interface, and a database 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 includes non-volatile and/or volatile storage media and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface of the computer device is for communicating with an external client via a network connection. The computer program is executed by a processor to perform functions or steps of a vehicle impairment determination method service side.
In one embodiment, a computer device is provided, which may be a client, the internal structure of which may be as shown in FIG. 5. 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 includes a non-volatile 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 the operating system and computer programs in the non-volatile storage media. The network interface of the computer device is for communicating with an external server via a network connection. The computer program is executed by a processor to implement functions or steps of a vehicle impairment method client side.
In one embodiment, a computer device is provided comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
acquiring an image to be identified of a vehicle to be damaged, wherein the image to be identified is an image containing at least one damaged position of the vehicle to be damaged; performing damage detection on the image to be identified, and determining the number of damage class labels corresponding to the damage positions, the label names of the damage class labels and the probability of the label names; performing component recognition on the image to be recognized, and determining component mask information corresponding to the damage position; and outputting an impairment determination result of the impairment position based on the number of impairment category labels, the label names of the impairment category labels, the probability of the label names and the component mask information.
In this embodiment, the number of damage class labels and the label names of the damage class labels corresponding to the damage positions are determined by performing damage detection on the image to be identified of the vehicle to be damaged, the image to be identified is subjected to component identification, and component mask information corresponding to the damage positions is determined, so that the damage determination result of the damage positions can be output according to the number of the damage class labels, the label names of the damage class labels, the probability of the label names and the component mask information, and the damage determination result of the damage positions can be output by combining the number of the damage class labels, the label names of the damage class labels, the probability of the label names and the component mask information corresponding to the damage positions in the image to be identified, thereby improving the reliability of damage class identification of the accident vehicle and further being beneficial to accurately determining the damage condition of the vehicle.
In one embodiment, a computer readable storage medium is provided, the computer readable storage medium storing a computer program which when executed by a processor performs the steps of:
acquiring an image to be identified of a vehicle to be damaged, wherein the image to be identified is an image containing at least one damaged position of the vehicle to be damaged; performing damage detection on the image to be identified, and determining the number of damage class labels corresponding to the damage positions, the label names of the damage class labels and the probability of the label names; performing component recognition on the image to be recognized, and determining component mask information corresponding to the damage position; and outputting an impairment determination result of the impairment position based on the number of impairment category labels, the label names of the impairment category labels, the probability of the label names and the component mask information.
In this embodiment, the number of damage class labels and the label names of the damage class labels corresponding to the damage positions are determined by performing damage detection on the image to be identified of the vehicle to be damaged, the image to be identified is subjected to component identification, and component mask information corresponding to the damage positions is determined, so that the damage determination result of the damage positions can be output according to the number of the damage class labels, the label names of the damage class labels, the probability of the label names and the component mask information, and the damage determination result of the damage positions can be output by combining the number of the damage class labels, the label names of the damage class labels, the probability of the label names and the component mask information corresponding to the damage positions in the image to be identified, thereby improving the reliability of damage class identification of the accident vehicle and further being beneficial to accurately determining the damage condition of the vehicle.
It should be noted that, the functions or steps implemented by the computer readable storage medium or the computer device may correspond to the relevant descriptions of the server side and the client side in the foregoing method embodiments, and are not described herein for avoiding repetition.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile 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), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions.
The above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the 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 scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention, and are intended to be included in the scope of the present invention.
Claims (10)
1. A method of vehicle damage assessment, the method comprising:
acquiring an image to be identified of a vehicle to be damaged, wherein the image to be identified is an image containing at least one damaged position of the vehicle to be damaged;
Performing damage detection on the image to be identified, and determining the number of damage class labels corresponding to the damage positions, the label names of the damage class labels and the probability of the label names;
performing component identification on the image to be identified, and determining component mask information corresponding to the damage position;
and outputting an impairment determination result of the impairment position based on the number of impairment category labels, the label names of the impairment category labels, the probability of the label names and the component mask information.
2. The method for vehicle damage assessment according to claim 1, wherein the performing damage detection on the image to be identified, and determining the number of damage class labels corresponding to the damage location, the label names of the damage class labels, and the probability of the label names, includes:
extracting features of the image to be identified to obtain a target feature image corresponding to the image to be identified;
performing feature classification on the target feature image, and determining a feature classification result;
and determining the number of the damage class labels and the label names of the damage class labels according to the feature classification result.
3. The method for vehicle damage assessment according to claim 2, wherein the feature extraction of the image to be identified to obtain a target feature image corresponding to the image to be identified includes:
Carrying out grouping convolution on the image to be identified and outputting an initial characteristic image;
and extracting the characteristics of the initial characteristic image to obtain the target characteristic image.
4. The method for vehicle damage assessment according to claim 3, wherein said performing feature extraction on said initial feature image to obtain said target feature image comprises:
extracting the characteristics of the initial characteristic image and outputting a first characteristic image;
extracting the characteristics of the first characteristic graph and outputting a second characteristic image;
extracting the characteristics of the second characteristic graph and outputting a third characteristic image;
and extracting the characteristics of the third characteristic image and outputting a target characteristic image.
5. The method for vehicle damage assessment according to claim 3, wherein said performing feature extraction on said initial feature image to obtain said target feature image comprises:
acquiring a first offset of the initial feature image, extracting a first image feature of the initial feature image, performing deformable convolution on the first image feature and the first offset, and outputting a first feature image;
acquiring a second offset of the first characteristic image, extracting a second image characteristic of the first characteristic image, performing deformable convolution on the second image characteristic and the second offset, and outputting a second characteristic image;
Acquiring a third offset of the second characteristic image, extracting a third image characteristic of the second characteristic image, performing deformable convolution on the third image characteristic and the third offset, and outputting a third characteristic image;
and acquiring a fourth offset of the third characteristic image, extracting a fourth image characteristic of the third characteristic image, performing deformable convolution on the fourth image characteristic and the fourth offset, and outputting a target characteristic image.
6. The vehicle damage assessment method according to any one of claims 1 to 5, wherein the outputting of the damage assessment result of the damage location based on the number of damage category labels, the label names of the damage category labels, the probability of the label names, and the component mask information includes:
determining the number of damage class labels to determine whether the damage class label is greater than one;
when the number of the damage class labels is larger than one, comparing the probability of each label name, and selecting the label name corresponding to the maximum probability as a target label;
and taking the object tag and the part name in the part mask information as the damage assessment result and outputting the damage assessment result.
7. The vehicle damage assessment method according to claim 6, further comprising:
when the damage class label is determined to be one according to the number of the damage class labels, the label name is used as a target label;
and taking the object tag and the part name in the part mask information as the damage assessment result and outputting the damage assessment result.
8. A vehicle damage assessment device, comprising:
the acquisition module is used for acquiring an image to be identified of the vehicle to be damaged, wherein the image to be identified is an image containing at least one damaged position of the vehicle to be damaged;
the first determining module is used for detecting damage to the image to be identified and determining the number of damage class labels corresponding to the damage position, the label names of the damage class labels and the probability of the label names;
the second determining module is used for carrying out component recognition on the image to be recognized and determining component mask information corresponding to the damage position;
and the output module is used for outputting the damage assessment result of the damage position based on the number of the damage category labels, the label names of the damage category labels, the probability of the label names and the component mask information.
9. Computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the vehicle impairment estimation method according to any one of claims 1 to 7 when the computer program is executed by the processor.
10. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the vehicle impairment estimation method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311517040.9A CN117671330B (en) | 2023-11-14 | 2023-11-14 | Vehicle damage assessment method, device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311517040.9A CN117671330B (en) | 2023-11-14 | 2023-11-14 | Vehicle damage assessment method, device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117671330A true CN117671330A (en) | 2024-03-08 |
| CN117671330B CN117671330B (en) | 2024-06-21 |
Family
ID=90074331
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311517040.9A Active CN117671330B (en) | 2023-11-14 | 2023-11-14 | Vehicle damage assessment method, device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117671330B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110163301A (en) * | 2019-05-31 | 2019-08-23 | 北京金山云网络技术有限公司 | A kind of classification method and device of image |
| 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 |
| CN114219757A (en) * | 2021-11-08 | 2022-03-22 | 华南理工大学 | Vehicle intelligent loss assessment method based on improved Mask R-CNN |
| CN114723945A (en) * | 2022-04-07 | 2022-07-08 | 平安科技（深圳）有限公司 | Vehicle damage detection method and device, electronic equipment and storage medium |
| CN114898155A (en) * | 2022-05-18 | 2022-08-12 | 平安科技（深圳）有限公司 | Vehicle damage assessment method, device, equipment and storage medium |
| CN115147702A (en) * | 2022-05-31 | 2022-10-04 | 深圳市麦谷科技有限公司 | YOLACT + + based vehicle damage identification method, device, equipment and medium |
| US20230289887A1 (en) * | 2021-07-30 | 2023-09-14 | Swiss Reinsurance Company Ltd. | Optical Fraud Detector for Automated Detection Of Fraud In Digital Imaginary-Based Automobile Claims, Automated Damage Recognition, and Method Thereof |
-
2023
- 2023-11-14 CN CN202311517040.9A patent/CN117671330B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110163301A (en) * | 2019-05-31 | 2019-08-23 | 北京金山云网络技术有限公司 | A kind of classification method and device of image |
| 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 |
| US20230289887A1 (en) * | 2021-07-30 | 2023-09-14 | Swiss Reinsurance Company Ltd. | Optical Fraud Detector for Automated Detection Of Fraud In Digital Imaginary-Based Automobile Claims, Automated Damage Recognition, and Method Thereof |
| CN114219757A (en) * | 2021-11-08 | 2022-03-22 | 华南理工大学 | Vehicle intelligent loss assessment method based on improved Mask R-CNN |
| CN114723945A (en) * | 2022-04-07 | 2022-07-08 | 平安科技（深圳）有限公司 | Vehicle damage detection method and device, electronic equipment and storage medium |
| CN114898155A (en) * | 2022-05-18 | 2022-08-12 | 平安科技（深圳）有限公司 | Vehicle damage assessment method, device, equipment and storage medium |
| CN115147702A (en) * | 2022-05-31 | 2022-10-04 | 深圳市麦谷科技有限公司 | YOLACT + + based vehicle damage identification method, device, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117671330B (en) | 2024-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021135499A1 (en) | Damage detection model training and vehicle damage detection methods, device, apparatus, and medium | |
| US11244435B2 (en) | Method and apparatus for generating vehicle damage information | |
| CN111191539B (en) | Certificate authenticity verification method and device, computer equipment and storage medium | |
| CN111651992A (en) | Named entity labeling method and device, computer equipment and storage medium | |
| WO2021114809A1 (en) | Vehicle damage feature detection method and apparatus, computer device, and storage medium | |
| CN111368696A (en) | Dangerous chemical transport vehicle illegal driving behavior detection method and system based on visual cooperation | |
| CN114663871A (en) | Image recognition method, training method, device, system and storage medium | |
| CN113837151A (en) | Table image processing method and device, computer equipment and readable storage medium | |
| CN114357174B (en) | Code classification system and method based on OCR and machine learning | |
| CN112668462A (en) | Vehicle loss detection model training method, vehicle loss detection device, vehicle loss detection equipment and vehicle loss detection medium | |
| CN112232336A (en) | Certificate identification method, device, equipment and storage medium | |
| CN112036488A (en) | Event identification method, device and equipment based on image identification and storage medium | |
| CN114332883A (en) | Invoice information identification method and device, computer equipment and storage medium | |
| CN117671330B (en) | Vehicle damage assessment method, device, computer equipment and storage medium | |
| CN112396060A (en) | Identity card identification method based on identity card segmentation model and related equipment thereof | |
| CN110909733A (en) | Template positioning method and device based on OCR picture recognition and computer equipment | |
| CN114359918A (en) | Bill information extraction method and device and computer equipment | |
| CN113516148A (en) | Image processing method, device and equipment based on artificial intelligence and storage medium | |
| CN113901817A (en) | Document classification method and device, computer equipment and storage medium | |
| CN113221907A (en) | Vehicle part segmentation method, device, equipment and storage medium | |
| CN113449716B (en) | Field positioning and classifying method, text image recognition method, device and equipment | |
| CN113239128B (en) | Data pair classification method, device, equipment and storage medium based on implicit characteristics | |
| KR102529552B1 (en) | Method for monitoring authentication based on video call using neural network and system for the method | |
| CN117197498A (en) | Vehicle flaw similarity recognition method and system | |
| CN117058086A (en) | Automatic generation method, device, storage medium and equipment for vehicle risk assessment loss list |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |