A dynamically evolving vehicle intelligent damage assessment method and system
Through the dynamically evolved intelligent vehicle loss determination method, multi-angle shooting, denoising, foreground extraction and deep neural network evaluation models, the problems of inefficiency and large error of traditional vehicle loss determination methods are solved, and the accurate assessment of vehicle damage is achieved and Quick loss determination improves the automation and accuracy of evaluation.
Patent Information
- Application Number
- CN202411012040.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-26
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-07-26
AI Technical Summary
Traditional vehicle damage determination methods are inefficient, subjective, and have large errors, and cannot provide accurate and rapid damage assessment, which affects the auto insurance claims process.
Using a dynamically evolved intelligent vehicle loss determination method, the vehicle damage is monitored and analyzed in real time through multi-angle shooting, denoising, foreground extraction, damage analysis and deep neural network evaluation models, and provides accurate damage assessment results.
It realizes accurate assessment and loss determination of vehicle damage, shortens loss determination time, reduces customer waiting time, and improves the degree of automation and accuracy of evaluation.
Smart Images

Figure CN118552532B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of intelligent assessment, and in particular to a dynamically evolving vehicle intelligent damage assessment method and system. Background Art
[0002] With the rapid development of modern science and technology, the automotive industry has undergone tremendous changes. The popularity of intelligent connected vehicles and the continuous advancement of autonomous driving technology have greatly improved the safety and convenience of vehicles. However, traffic accidents still inevitably occur, and vehicle damage assessment plays a vital role in the process of auto insurance claims. Traditional vehicle damage assessment methods mainly rely on manual experience, and there are problems such as low efficiency, strong subjectivity, and large errors. In recent years, with the rapid development of big data, artificial intelligence, and the Internet of Things, intelligent vehicle damage assessment methods have gradually attracted widespread attention. Summary of the invention
[0003] In view of this, the present invention proposes a dynamically evolving vehicle intelligent damage assessment method, which achieves accurate assessment and damage assessment of vehicle damage through real-time monitoring, data analysis and algorithm optimization. It not only focuses on static post-accident data, but also considers data changes before, during and after the accident, thereby providing more comprehensive and accurate damage assessment results, shortening the damage assessment time, and reducing customer waiting time.
[0004] To achieve the above object, the present invention provides a dynamically evolving vehicle intelligent damage assessment method, comprising the following steps:
[0005] S1: photographing the damaged vehicle at multiple angles to obtain a damaged image of the vehicle and corresponding shooting angles, and performing denoising and foreground extraction operations on the damaged image of the vehicle to obtain a damaged foreground image of the vehicle, wherein edge detection based on a hybrid operator is an implementation method of the foreground extraction operation;
[0006] S2: performing damage analysis and identification on the extracted damaged vehicle foreground image and segmenting the damaged area of the vehicle, wherein contour energy optimization is an implementation method of the damage analysis;
[0007] S3: performing regional segmentation on the undamaged vehicle image at the same shooting angle to obtain a standard region based on the segmented damaged region of the vehicle, and performing damage identification by comparing the segmented standard region and the damaged region to calculate the degree of depression of the damaged region of the vehicle, wherein the chaos degree calculation based on the gray level co-occurrence matrix is the implementation method of the damage identification;
[0008] S4: Construct a vehicle damage assessment model, which takes the number of pixels and the degree of depression in the damaged area of the vehicle obtained by segmentation as input and the vehicle damage assessment result as output, wherein a deep neural network is an implementation method of the vehicle damage assessment model.
[0009] As a further improvement method of the present invention:
[0010] Optionally, in step S1, performing denoising and foreground extraction operations on the damaged vehicle image to obtain the damaged vehicle foreground image includes:
[0011] S11: Perform fast non-local mean denoising on the damaged vehicle image to obtain a denoised damaged vehicle image. The specific process includes:
[0012] S11.1: converting the damaged vehicle image into a grayscale image and dividing the grayscale image into a plurality of image blocks of N*N size according to a preset image block length N;
[0013] S11.2: Expand each image block into an image block vector in row order, and construct a KD tree based on all the image block vectors;
[0014] S11.3: Expand each query image block into a query image block vector in row order, and search for the most similar K nodes on the KD tree. The image blocks corresponding to the K nodes obtained are the neighborhoods of the query image block. The node similarity calculation formula is:
[0015] ;
[0016] in:
[0017] and Represent the i-th and j-th image blocks respectively;
[0018] N represents the side length of the image block;
[0019] represents the Euclidean distance;
[0020] k represents the pixel index of the image block;
[0021] and Represent the kth pixel value in the i-th and j-th image blocks respectively;
[0022] S11.4: Calculate the similarity weights between image blocks according to the Euclidean distances between image blocks. The similarity weight calculation formula between image blocks is:
[0023] ;
[0024] in:
[0025] and represent the i-th and j-th image blocks respectively;
[0026] h represents the smoothing parameter;
[0027] S11.5: The calculated image block similarity weights are normalized to obtain normalized similarity weights, and the calculation formula is:
[0028] ;
[0029] in:
[0030] represents the neighborhood of the i-th image block;
[0031] S11.6: According to the similarity weight, similar blocks are weighted averaged to obtain denoised image blocks. The entire image is processed block by block to generate a denoised vehicle damaged image. The calculation formula is:
[0032] ;
[0033] in:
[0034] represents the i-th image block after denoising;
[0035] S12: performing foreground extraction on the denoised damaged vehicle image to obtain a damaged vehicle foreground image.
[0036] Optionally, in step S12, performing foreground extraction on the denoised damaged vehicle image to obtain a damaged vehicle foreground image includes:
[0037] S12.1: Calculate the mixed gradient amplitude for the denoised damaged vehicle image. The calculation formula is:
[0038] ;
[0039] in:
[0040] Represent the gradient amplitudes of Roberts operator, Canny operator and Sobel operator respectively;
[0041] , Represent the gradient amplitude weights of Roberts operator, Canny operator and Sobel operator respectively;
[0042] S12.2: performing threshold processing according to the mixed gradient amplitude, treating pixel points below the preset threshold as background points, and treating pixel points above the preset threshold as foreground points;
[0043] S12.3: According to the threshold processing result, mark all foreground point constituting areas in the denoised damaged vehicle image as the damaged vehicle foreground image.
[0044] Optionally, in step S2, performing damage analysis and identification on the extracted damaged vehicle foreground image and segmenting the damaged vehicle area includes:
[0045] S21: Randomly initialize polygonal contours to enclose the damaged area of the vehicle;
[0046] S22: constructing a contour energy function, and making the contour approach the target boundary by minimizing the energy, wherein the contour energy function includes internal energy and external energy, and the internal energy calculation formula is:
[0047] ;
[0048] in:
[0049] Represents the image contour area;
[0050] Represents the gray value of the image at position x;
[0051] Represents the gradient vector of the image at position x;
[0052] represents the Laplacian operator of the image at position x;
[0053] represents the smoothing parameter,
[0054] represents shape control parameters;
[0055] The external energy calculation formula is:
[0056] ;
[0057] in:
[0058] Represents the image contour area;
[0059] Represents the gray value of the image at position x;
[0060] represents the edge response function;
[0061] S23: Update the position of the contour turning point by minimizing the contour energy function. The calculation formula is:
[0062] ;
[0063] in:
[0064] represents the update step size;
[0065] Represents the gradient of the contour energy function to the contour turning point, where ;
[0066] Indicates the position of the turning point of the contour before updating;
[0067] Indicates the updated turning point position of the contour;
[0068] S24: iteratively execute step S23 until the maximum number of iterations is reached, and the contour position at this time is the segmented boundary of the damaged area of the vehicle.
[0069] Optionally, in step S3, comparing the segmented standard area and the damaged area to perform damage identification and calculate the degree of depression of the damaged area of the vehicle includes:
[0070] S31: Calculate the pixel-level difference between the segmented standard area and the damaged area to obtain a difference area image , the calculation formula is:
[0071] ;
[0072] in:
[0073] and Represent the pixel values of the standard area and the damaged area at the (x, y) position respectively;
[0074] Represents the pixel value of the difference area image at position (x,y);
[0075] S32: extracting texture features of the difference region image according to preset directions and pixel distances to obtain a grayscale co-occurrence matrix of the difference region image;
[0076] S33: Calculate the chaos degree of the obtained gray level co-occurrence matrix. The lower the chaos degree, the greater the difference between the standard area and the damaged area, and the more serious the depression. The calculated chaos degree is used as the depression degree. The chaos degree calculation formula is:
[0077] ;
[0078] in:
[0079] It indicates that the element value in the i-th row and j-th column of the gray-level co-occurrence matrix accounts for all matrix element values and probabilities.
[0080] Optionally, constructing a vehicle damage assessment model in step S4 includes:
[0081] S41: constructing a vehicle damage assessment model based on a deep neural network, and establishing a nonlinear mapping relationship between the number of pixels in the damaged area of the vehicle, the degree of dents and the vehicle damage assessment result, wherein the vehicle damage assessment model is implemented based on a multi-layer perceptron, and includes an input layer, a hidden layer and an output layer, wherein the input layer receives the number of pixels and the degree of dents in the damaged area of the vehicle obtained by segmentation as input features, the hidden layer is used to extract features, and the output layer is used to predict the vehicle damage assessment result,
[0082] The hidden layer calculation formula is:
[0083] ;
[0084] The output layer calculation formula is:
[0085] ;
[0086] in:
[0087] Indicates the number of pixels in the segmented damaged area of the vehicle;
[0088] Indicates the degree of vehicle denting;
[0089] and Represent the weight matrices of the hidden layer and the output layer respectively;
[0090] , The bias terms σ representing the hidden layer and the output layer respectively represent the Sigmoid activation function;
[0091] represents the hidden feature vector;
[0092] S42: Optimize and solve the constructed vehicle damage assessment model to obtain optimal model parameters, and instantiate the model according to the model parameters to perform vehicle damage assessment.
[0093] Optionally, in step S42, the constructed vehicle damage assessment model is optimized to obtain optimal model parameters, and the model is instantiated according to the model parameters to perform vehicle damage assessment, including:
[0094] S42.1: Randomly initialize the variable x to be optimized, the auxiliary variable z and the Lagrange multiplier u;
[0095] S42.2: Iteratively update variables and parameters, where the update calculation formula of the variable x to be optimized is:
[0096] ;
[0097] in:
[0098] represents the variable x updated at the k+1th iteration;
[0099] represents the model objective function;
[0100] represents the regularization parameter;
[0101] and They represent the variable z and multiplier u updated at the kth iteration respectively;
[0102] The update calculation formula of the auxiliary variable z is:
[0103] ;
[0104] in:
[0105] represents the variable z updated at the k+1th iteration;
[0106] Represents the constraint penalty term in the objective function;
[0107] The multiplier u is updated according to the updated variables to be optimized and auxiliary variables. The calculation formula is:
[0108] ;
[0109] in:
[0110] represents the multiplier u updated at the k+1th iteration;
[0111] and They represent the variable x to be optimized and the auxiliary variable z updated at the k+1th iteration respectively;
[0112] S42.3: Check whether the stopping criteria are met, the stopping criteria include reaching the maximum number of iterations and the objective function change being less than a threshold. If met, the variables to be optimized are output as the optimal parameters of the vehicle damage assessment model, otherwise return to S42.2 to continue execution.
[0113] In order to solve the above problems, the present invention provides a dynamically evolving vehicle intelligent damage assessment method system, characterized in that a damaged vehicle foreground extraction module is used to shoot the damaged vehicle at multiple angles to obtain a damaged vehicle image and corresponding shooting angles, and to perform denoising and foreground extraction operations on the damaged vehicle image to obtain a damaged vehicle foreground image;
[0114] The damage analysis and identification module is used to perform damage analysis and identification on the extracted damaged foreground image of the vehicle and segment the damaged area of the vehicle, and segment the undamaged vehicle image taken at the same shooting angle according to the segmented damaged area of the vehicle to obtain the standard area, and compare the segmented standard area with the damaged area to perform damage identification and calculate the degree of depression of the damaged area of the vehicle;
[0115] The vehicle damage assessment device is used to construct a vehicle damage assessment model and obtain a vehicle damage assessment result based on the number of pixels and the degree of depression in the damaged area of the vehicle obtained by segmentation.
[0116] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:
[0117] A memory storing at least one instruction;
[0118] Communication interface, enabling electronic equipment to communicate; and
[0119] A processor executes instructions stored in the memory to implement the dynamically evolving vehicle intelligent damage assessment method described above.
[0120] In order to solve the above-mentioned problem, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned dynamically evolving vehicle intelligent damage assessment method.
[0121] Compared with the prior art, the present invention proposes a dynamically evolving vehicle intelligent damage assessment method, which has the following advantages:
[0122] First, this scheme proposes a dynamically evolving vehicle intelligent damage assessment method, which combines multi-angle shooting, denoising, foreground extraction, damage analysis, damage identification and vehicle damage assessment methods to form a complete vehicle damage analysis and assessment process, which improves the degree of automation and accuracy of vehicle damage analysis and assessment. When performing foreground extraction on the image to obtain the damaged foreground image of the vehicle, this scheme uses edge detection based on a hybrid operator as the main method of foreground extraction. This method can effectively extract edge information in the image and highlight the outline of the damaged area of the vehicle.
[0123] At the same time, this solution adopts the method of contour energy optimization for damage analysis and identification. By optimizing the contour energy, it automatically finds the damaged contour of the vehicle and further segments the damaged area of the vehicle, thereby providing more accurate analysis results of the damaged area. It also adopts the chaos calculation based on the grayscale co-occurrence matrix to quantify the degree of chaos in the structure of the damaged area of the vehicle by analyzing the relationship between the grayscale levels of the image, thereby providing objective damage identification results.
[0124] In addition, this scheme proposes a vehicle damage assessment model, which takes the number of pixels (damaged area) and the degree of dents (damaged depth) in the damaged area of the vehicle as input, performs vehicle damage assessment through a deep neural network, establishes nonlinear mapping associations, and improves the accuracy and generalization ability of vehicle damage assessment. BRIEF DESCRIPTION OF THE DRAWINGS
[0125] Figure 1 A schematic diagram of a process flow of a dynamically evolving vehicle intelligent damage assessment method provided by an embodiment of the present invention;
[0126] Figure 2 A functional module diagram of a dynamically evolving vehicle intelligent damage assessment method system provided by an embodiment of the present invention, wherein 100 is a dynamically evolving vehicle intelligent damage assessment method system, 101 is a damaged vehicle foreground extraction module, 102 is a damage analysis and identification module, and 103 is a vehicle damage assessment device;
[0127] Figure 3 A schematic diagram of the structure of an electronic device for implementing a vehicle intelligent damage assessment method for dynamic evolution provided by an embodiment of the present invention, 1 electronic device, 10 processor, 11 memory, 12 program, 13 communication interface;
[0128] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0129] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0130] The embodiment of the present application provides a dynamically evolving vehicle intelligent damage assessment method. The execution subject of the dynamically evolving vehicle intelligent damage assessment method includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the dynamically evolving vehicle intelligent damage assessment method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. Example 1
[0131] S1: photographing the damaged vehicle at multiple angles to obtain a damaged image of the vehicle and corresponding shooting angles, and performing denoising and foreground extraction operations on the damaged image of the vehicle to obtain a damaged foreground image of the vehicle, wherein edge detection based on a hybrid operator is an implementation method of the foreground extraction operation.
[0132] In step S1, the damaged vehicle image is subjected to denoising and foreground extraction operations to obtain a damaged vehicle foreground image, including:
[0133] S11: Perform fast non-local mean denoising on the damaged vehicle image to obtain a denoised damaged vehicle image. The specific process includes:
[0134] S11.1: converting the damaged vehicle image into a grayscale image and dividing the grayscale image into a plurality of image blocks of N*N size according to a preset image block length N;
[0135] S11.2: Expand each image block into an image block vector in row order, and construct a KD tree based on all the image block vectors;
[0136] S11.3: Expand each query image block into a query image block vector in row order, and search for the most similar K nodes on the KD tree. The image blocks corresponding to the K nodes obtained are the neighborhoods of the query image block. The node similarity calculation formula is:
[0137] ;
[0138] in:
[0139] and Represent the i-th and j-th image blocks respectively;
[0140] N represents the side length of the image block;
[0141] represents the Euclidean distance;
[0142] k represents the pixel index of the image block;
[0143] and Represent the kth pixel value in the i-th and j-th image blocks respectively;
[0144] S11.4: Calculate the similarity weights between image blocks according to the Euclidean distances between image blocks. The similarity weight calculation formula between image blocks is:
[0145] ;
[0146] in:
[0147] and Represent the i-th and j-th image blocks respectively;
[0148] h represents the smoothing parameter;
[0149] S11.5: The calculated image block similarity weights are normalized to obtain normalized similarity weights, and the calculation formula is:
[0150] ;
[0151] in:
[0152] represents the neighborhood of the i-th image block;
[0153] S11.6: According to the similarity weight, similar blocks are weighted averaged to obtain denoised image blocks. The entire image is processed block by block to generate a denoised vehicle damaged image. The calculation formula is:
[0154]
[0155] in:
[0156] represents the i-th image block after denoising;
[0157] S12: performing foreground extraction on the denoised damaged vehicle image to obtain a damaged vehicle foreground image.
[0158] In the step S12, foreground extraction is performed on the denoised damaged vehicle image to obtain a damaged vehicle foreground image, including:
[0159] S12.1: Calculate the mixed gradient amplitude for the denoised damaged vehicle image. The calculation formula is:
[0160] ;
[0161] in:
[0162] Represent the gradient amplitudes of Roberts operator, Canny operator and Sobel operator respectively;
[0163] , Represent the gradient amplitude weights of Roberts operator, Canny operator and Sobel operator respectively;
[0164] S12.2: performing threshold processing according to the mixed gradient amplitude, treating pixel points below the preset threshold as background points, and treating pixel points above the preset threshold as foreground points;
[0165] S12.3: According to the threshold processing result, mark all foreground point constituting areas in the denoised damaged vehicle image as the damaged vehicle foreground image.
[0166] S2: performing damage analysis and identification on the extracted damaged vehicle foreground image and segmenting the damaged vehicle area, wherein contour energy optimization is an implementation method of the damage analysis.
[0167] In step S2, the damaged foreground image of the extracted vehicle is subjected to damage analysis and identification, and the damaged area of the vehicle is segmented, including:
[0168] S21: Randomly initialize polygonal contours to enclose the damaged area of the vehicle;
[0169] S22: constructing a contour energy function, and making the contour approach the target boundary by minimizing the energy, wherein the contour energy function includes internal energy and external energy, and the internal energy calculation formula is:
[0170] ;
[0171] in:
[0172] Represents the image contour area;
[0173] Represents the gray value of the image at position x;
[0174] Represents the gradient vector of the image at position x;
[0175] represents the Laplacian operator of the image at position x;
[0176] represents the smoothing parameter,
[0177] represents shape control parameters;
[0178] The external energy calculation formula is:
[0179] ;
[0180] in:
[0181] Represents the image contour area;
[0182] Represents the gray value of the image at position x;
[0183] represents the edge response function;
[0184] S23: Update the position of the contour turning point by minimizing the contour energy function. The calculation formula is:
[0185] ;
[0186] in:
[0187] represents the update step size;
[0188] Represents the gradient of the contour energy function to the contour turning point, where ;
[0189] Indicates the position of the turning point of the contour before updating;
[0190] Indicates the updated turning point position of the contour;
[0191] S24: iteratively execute step S23 until the maximum number of iterations is reached, and the contour position at this time is the segmented boundary of the damaged area of the vehicle.
[0192] S3: According to the damaged area of the vehicle obtained by segmentation, the image of the undamaged vehicle at the same shooting angle is segmented to obtain a standard area, and damage identification is performed by comparing the standard area and the damaged area obtained by segmentation to calculate the degree of depression of the damaged area of the vehicle, wherein the chaos degree calculation based on the gray level co-occurrence matrix is the implementation method of the damage identification.
[0193] In the step S3, the damage identification is performed based on the comparison between the segmented standard area and the damaged area to calculate the degree of depression of the damaged area of the vehicle, including:
[0194] S31: Calculate the pixel-level difference between the segmented standard area and the damaged area to obtain a difference area image , the calculation formula is:
[0195] ;
[0196] in:
[0197] and Represent the pixel values of the standard area and the damaged area at the (x, y) position respectively;
[0198] Represents the pixel value of the difference area image at position (x,y);
[0199] S32: extracting texture features of the difference region image according to preset directions and pixel distances to obtain a grayscale co-occurrence matrix of the difference region image;
[0200] S33: Calculate the chaos degree of the obtained gray level co-occurrence matrix. The lower the chaos degree, the greater the difference between the standard area and the damaged area, and the more serious the depression. The calculated chaos degree is used as the depression degree. The chaos degree calculation formula is:
[0201] ;
[0202] in:
[0203] It indicates that the element value in the i-th row and j-th column of the gray-level co-occurrence matrix accounts for all matrix element values and probabilities.
[0204] S4: Construct a vehicle damage assessment model, which takes the number of pixels and the degree of depression in the damaged area of the vehicle obtained by segmentation as input and the vehicle damage assessment result as output, wherein a deep neural network is an implementation method of the vehicle damage assessment model.
[0205] The vehicle damage assessment model is constructed in step S4, including:
[0206] S41: constructing a vehicle damage assessment model based on a deep neural network, and establishing a nonlinear mapping relationship between the number of pixels in the damaged area of the vehicle, the degree of dents and the vehicle damage assessment result, wherein the vehicle damage assessment model is implemented based on a multi-layer perceptron, and includes an input layer, a hidden layer and an output layer, wherein the input layer receives the number of pixels and the degree of dents in the damaged area of the vehicle obtained by segmentation as input features, the hidden layer is used to extract features, and the output layer is used to predict the vehicle damage assessment result,
[0207] The hidden layer calculation formula is:
[0208] ;
[0209] The output layer calculation formula is:
[0210] ;
[0211] in:
[0212] Indicates the number of pixels in the segmented damaged area of the vehicle;
[0213] Indicates the degree of vehicle denting;
[0214] and Represent the weight matrices of the hidden layer and the output layer respectively;
[0215] , The bias terms σ representing the hidden layer and the output layer respectively represent the Sigmoid activation function;
[0216] represents the hidden feature vector;
[0217] S42: Optimize and solve the constructed vehicle damage assessment model to obtain optimal model parameters, and instantiate the model according to the model parameters to perform vehicle damage assessment.
[0218] In the step S42, the constructed vehicle damage assessment model is optimized to obtain optimal model parameters, and the model is instantiated according to the model parameters to perform vehicle damage assessment, including:
[0219] S42.1: Randomly initialize the variable x to be optimized, the auxiliary variable z and the Lagrange multiplier u;
[0220] S42.2: Iteratively update variables and parameters, where the update calculation formula of the variable x to be optimized is:
[0221] ;
[0222] in:
[0223] represents the variable x updated at the k+1th iteration;
[0224] represents the model objective function;
[0225] represents the regularization parameter;
[0226] and They represent the variable z and multiplier u updated at the kth iteration respectively;
[0227] The update calculation formula of the auxiliary variable z is:
[0228] ;
[0229] in:
[0230] represents the variable z updated at the k+1th iteration;
[0231] Represents the constraint penalty term in the objective function;
[0232] The multiplier u is updated according to the updated variables to be optimized and auxiliary variables. The calculation formula is:
[0233] ;
[0234] in:
[0235] represents the multiplier u updated at the k+1th iteration;
[0236] and They represent the variable x to be optimized and the auxiliary variable z updated at the k+1th iteration respectively;
[0237] S42.3: Check whether the stopping criteria are met, the stopping criteria include reaching the maximum number of iterations and the objective function change being less than a threshold. If met, the variables to be optimized are output as the optimal parameters of the vehicle damage assessment model, otherwise return to S42.2 to continue execution. Example 2
[0238] like Figure 2 As shown, it is a functional module diagram of a dynamically evolving vehicle intelligent damage assessment method system provided by an embodiment of the present invention, which can implement the dynamically evolving vehicle intelligent damage assessment method in Example 1.
[0239] The dynamically evolving vehicle intelligent damage assessment method system 100 of the present invention can be installed in an electronic device. According to the functions to be implemented, the dynamically evolving vehicle intelligent damage assessment method system can include a damaged vehicle foreground extraction module 101, a damage analysis and identification module 102, and a vehicle damage assessment device 103. The module of the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.
[0240] The damaged vehicle foreground extraction module 101 is used to shoot the damaged vehicle at multiple angles to obtain a damaged vehicle image and corresponding shooting angles, and to perform denoising and foreground extraction operations on the damaged vehicle image to obtain a damaged vehicle foreground image;
[0241] The damage analysis and identification module 102 is used to perform damage analysis and identification on the extracted damaged vehicle foreground image and segment the damaged area of the vehicle, and segment the undamaged vehicle image taken at the same shooting angle to obtain the standard area based on the segmented damaged vehicle area, and compare the segmented standard area with the damaged area to perform damage identification and calculate the degree of depression of the damaged area of the vehicle;
[0242] The vehicle damage assessment device 103 is used to construct a vehicle damage assessment model and obtain a vehicle damage assessment result according to the number of pixels and the degree of depression in the damaged area of the vehicle obtained by segmentation.
[0243] In detail, each module in the dynamically evolving vehicle intelligent damage assessment method system 100 in the embodiment of the present invention adopts the same method as above when used. Figure 1 The dynamically evolving vehicle intelligent damage assessment method described in the invention has the same technical means and can produce the same technical effects, so I will not go into details here. Example 3
[0244] like Figure 3 The figure is a schematic diagram of the structure of an electronic device for implementing a dynamically evolving vehicle intelligent damage assessment method provided by one embodiment of the present invention.
[0245] The electronic device 1 may include a processor 10 , a memory 11 , a communication interface 13 and a bus, and may also include a computer program stored in the memory 11 and executable on the processor 10 , such as a program 12 .
[0246] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. In some embodiments, the memory 11 may be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 11 may also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 11 may not only be used to store application software and various types of data installed in the electronic device 1, such as the code of the program 12, etc., but also be used to temporarily store data that has been output or is to be output.
[0247] In some embodiments, the processor 10 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect the various components of the entire electronic device. It runs or executes the programs or modules stored in the memory 11 (such as the program 12 for realizing the dynamically evolving vehicle intelligent damage assessment method), and calls the data stored in the memory 11 to execute various functions of the electronic device 1 and process data.
[0248] The communication interface 13 may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices, and to achieve connection and communication between internal components of the electronic device.
[0249] The bus may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.
[0250] Figure 3 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 3 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.
[0251] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to various components. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.
[0252] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.
[0253] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.
[0254] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.
[0255] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.
[0256] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A dynamically evolving vehicle intelligent damage assessment method, characterized in that: The method comprises: S1: photographing the damaged vehicle at multiple angles to obtain a damaged image of the vehicle and corresponding shooting angles, and performing denoising and foreground extraction operations on the damaged image of the vehicle to obtain a damaged foreground image of the vehicle, wherein edge detection based on a hybrid operator is an implementation method of the foreground extraction operation; S2: performing damage analysis and identification on the extracted damaged vehicle foreground image and segmenting the damaged area of the vehicle, wherein contour energy optimization is an implementation method of the damage analysis; S3: performing regional segmentation on the undamaged vehicle image at the same shooting angle to obtain a standard region based on the segmented damaged region of the vehicle, and performing damage identification by comparing the segmented standard region and the damaged region to calculate the degree of depression of the damaged region of the vehicle, wherein the chaos degree calculation based on the gray level co-occurrence matrix is the implementation method of the damage identification; S4: constructing a vehicle damage assessment model, wherein the model takes the number of pixels and the degree of depression in the damaged area of the vehicle obtained by segmentation as input and takes the vehicle damage assessment result as output, wherein a deep neural network is an implementation method of the vehicle damage assessment model; The damaged foreground image of the vehicle is extracted for damage analysis and identification and segmentation of the damaged area of the vehicle, including: S21: Randomly initialize polygonal contours to enclose the damaged area of the vehicle; S22: constructing a contour energy function, and minimizing the energy to make the contour approach the target boundary, wherein the contour energy function includes internal energy and external energy, and the calculation formula of the internal energy is: ; in: Represents the image contour area; Represents the gray value of the image at position x; Represents the gradient vector of the image at position x; represents the Laplacian operator of the image at position x; represents the smoothing parameter, represents shape control parameters; The calculation formula of the external energy is: ; in: Represents the image contour area; Represents the gray value of the image at position x; represents the edge response function; S23: Update the position of the contour turning point by minimizing the contour energy function. The calculation formula is: ; in: represents the update step size; Represents the gradient of the contour energy function to the contour turning point, where ; Indicates the position of the turning point of the contour before updating; Indicates the updated turning point position of the contour; S24: iteratively execute step S23 until the maximum number of iterations is reached, and the contour position at this time is the segmented boundary of the damaged area of the vehicle.
2. A dynamically evolving vehicle intelligent damage assessment method as claimed in claim 1, characterized in that: In step S1, the damaged vehicle image is subjected to denoising and foreground extraction operations to obtain a damaged vehicle foreground image, including: S11: Perform fast non-local mean denoising on the damaged vehicle image to obtain a denoised damaged vehicle image. The specific process includes: S11.1: converting the damaged vehicle image into a grayscale image and dividing the grayscale image into a plurality of image blocks of N*N size according to a preset image block length N; S11.2: Expand each image block into an image block vector in row order, and construct a KD tree based on all the image block vectors; S11.3: Expand each query image block into a query image block vector in row order, and search for the most similar K nodes on the KD tree. The image blocks corresponding to the K nodes obtained are the neighborhoods of the query image block. The node similarity calculation formula is: ; in: and Represent the i-th and j-th image blocks respectively; N represents the side length of the image block; represents the Euclidean distance; k represents the pixel index of the image block; and Represent the kth pixel value in the i-th and j-th image blocks respectively; S11.4: Calculate the similarity weights between image blocks according to the Euclidean distances between image blocks. The similarity weight calculation formula between image blocks is: ; in: and Represent the i-th and j-th image blocks respectively; h represents the smoothing parameter; S11.5: The calculated image block similarity weights are normalized to obtain normalized similarity weights, and the calculation formula is: in: represents the neighborhood of the i-th image block; S11.6: According to the similarity weight, similar blocks are weighted averaged to obtain denoised image blocks. The entire image is processed block by block to generate a denoised vehicle damaged image. The calculation formula is: in: represents the i-th image block after denoising; S12: performing foreground extraction on the denoised damaged vehicle image to obtain a damaged vehicle foreground image.
3. A dynamically evolving vehicle intelligent damage assessment method as claimed in claim 2, characterized in that: In the step S12, foreground extraction is performed on the denoised damaged vehicle image to obtain a damaged vehicle foreground image, including: S12.1: Calculate the mixed gradient amplitude for the denoised damaged vehicle image. The calculation formula is: in: Represent the gradient amplitudes of Roberts operator, Canny operator and Sobel operator respectively; , Represent the gradient amplitude weights of Roberts operator, Canny operator and Sobel operator respectively; S12.2: performing threshold processing according to the mixed gradient amplitude, treating pixel points below the preset threshold as background points, and treating pixel points above the preset threshold as foreground points; S12.3: According to the threshold processing result, mark all foreground point constituting areas in the denoised damaged vehicle image as the damaged vehicle foreground image.
4. A dynamically evolving vehicle intelligent damage assessment method as claimed in claim 1, characterized in that: In the step S3, the damage identification is performed based on the comparison between the segmented standard area and the damaged area to calculate the degree of depression of the damaged area of the vehicle, including: S31: Calculate the pixel-level difference between the segmented standard area and the damaged area to obtain a difference area image , the calculation formula is: in: and Represent the pixel values of the standard area and the damaged area at the (x, y) position respectively; Represents the pixel value of the difference area image at position (x,y); S32: extracting texture features of the difference region image according to preset directions and pixel distances to obtain a grayscale co-occurrence matrix of the difference region image; S33: Calculate the chaos degree of the obtained gray level co-occurrence matrix. The lower the chaos degree, the greater the difference between the standard area and the damaged area, and the more serious the depression. The calculated chaos degree is used as the depression degree. The chaos degree calculation formula is: in: It indicates that the element value in the i-th row and j-th column of the gray-level co-occurrence matrix accounts for all matrix element values and probabilities.
5. A dynamically evolving vehicle intelligent damage assessment method as claimed in claim 1, characterized in that: The vehicle damage assessment model is constructed in step S4, including: S41: constructing a vehicle damage assessment model based on a deep neural network, and establishing a nonlinear mapping relationship between the number of pixels in the damaged area of the vehicle, the degree of dents and the vehicle damage assessment result, wherein the vehicle damage assessment model is implemented based on a multi-layer perceptron, and includes an input layer, a hidden layer and an output layer, wherein the input layer receives the number of pixels and the degree of dents in the damaged area of the vehicle obtained by segmentation as input features, the hidden layer is used to extract features, and the output layer is used to predict the vehicle damage assessment result, The hidden layer calculation formula is: The output layer calculation formula is: in: Indicates the number of pixels in the segmented damaged area of the vehicle; Indicates the degree of vehicle denting; and Represent the weight matrices of the hidden layer and the output layer respectively; , The bias terms σ representing the hidden layer and the output layer respectively represent the Sigmoid activation function; represents the hidden feature vector; S42: Optimize and solve the constructed vehicle damage assessment model to obtain optimal model parameters, and instantiate the model according to the model parameters to perform vehicle damage assessment.
6. A dynamically evolving vehicle intelligent damage assessment method as claimed in claim 5, characterized in that: In the step S42, the constructed vehicle damage assessment model is optimized to obtain optimal model parameters, and the model is instantiated according to the model parameters to perform vehicle damage assessment, including: S42.1: Randomly initialize the variable x to be optimized, the auxiliary variable z and the Lagrange multiplier u; S42.2: Iteratively update variables and parameters, where the update calculation formula of the variable x to be optimized is: ; in: represents the variable x updated at the k+1th iteration; represents the model objective function; represents the regularization parameter; and They represent the variable z and multiplier u updated at the kth iteration respectively; The update calculation formula of the auxiliary variable z is: ; in: represents the variable z updated at the k+1th iteration; Represents the constraint penalty term in the objective function; The multiplier u is updated according to the updated variables to be optimized and auxiliary variables. The calculation formula is: in: represents the multiplier u updated at the k+1th iteration; and They represent the variable x to be optimized and the auxiliary variable z updated at the k+1th iteration respectively; S42.3: Check whether the stopping criteria are met, the stopping criteria include reaching the maximum number of iterations and the objective function change being less than a threshold. If so, the variables to be optimized are output as the optimal parameters of the vehicle damage assessment model, otherwise, return to S42.2 to continue execution.
7. A dynamically evolving vehicle intelligent damage assessment system, characterized in that: The system comprises: A damaged vehicle foreground extraction module is used to shoot the damaged vehicle at multiple angles to obtain a damaged vehicle image and corresponding shooting angles, and to perform denoising and foreground extraction operations on the damaged vehicle image to obtain a damaged vehicle foreground image; The damage analysis and identification module is used to perform damage analysis and identification on the extracted damaged foreground image of the vehicle and segment the damaged area of the vehicle, and segment the undamaged vehicle image taken at the same shooting angle according to the segmented damaged area of the vehicle to obtain the standard area, and compare the segmented standard area with the damaged area to perform damage identification and calculate the degree of depression of the damaged area of the vehicle; A vehicle damage assessment device is used to construct a vehicle damage assessment model, and obtain a vehicle damage assessment result based on the number of pixels and the degree of depression in the damaged area of the vehicle obtained by segmentation, so as to realize a dynamically evolving vehicle intelligent damage assessment method as described in any one of claims 1-6.
