Image amplification method, device and equipment and storage medium thereof
Image samples are generated through vehicle modeling and collision simulation, and diffusion model fine-tuning technology is used to solve the problem of insufficient image samples for vehicle business models, achieving the adequacy of accident image samples and the efficiency of model training.
Patent Information
- Application Number
- CN202411183285.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-11-26
AI Technical Summary
In the prior art, the image samples of the vehicle business model are insufficient, which seriously affects the training progress of the vehicle business model and cannot cover various traffic accident scenarios.
By obtaining the reconstruction modeling parameters of the target vehicle, vehicle modeling and collision simulation are carried out, batch collision images are generated, and the diffusion model is fine-tuned using the real accident image. The fine-tuned diffusion model is used to diffusion generation of the batch collision images, and obtain More image samples.
It ensures the adequacy of accident image samples during vehicle accident handling and claims model training, and improves the efficiency and accuracy of model training.
Smart Images

Figure CN119027592A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of image processing technology, and is applied to image sample amplification scenarios, and in particular to an image amplification method, device, equipment, and storage medium thereof. Background Art
[0002] As the number of private cars increases, insurance companies are involved in more and more extensive vehicle businesses, such as vehicle insurance business, vehicle insurance claims business, vehicle maintenance business, etc. In the field of vehicle insurance business processing, such as vehicle insurance claims business and vehicle maintenance business, these businesses often require vehicle image information and accident image information after the accident.
[0003] At present, although relevant parties have collected a certain number of vehicle accident pictures, there are still too few existing vehicle accident pictures to train the auto insurance claim model to maturity and obtain a more intelligent accident attribution model. The collection after each accident often fails to meet the training requirements. Moreover, collecting and labeling vehicle accident pictures also takes a lot of time, which seriously affects the progress of model training. The collected and labeled vehicle accident pictures often cannot cover various traffic accident scenes, resulting in insufficient image samples of the vehicle business model, which seriously affects the training progress of the vehicle business model. Summary of the invention
[0004] The purpose of the embodiments of the present application is to propose an image augmentation method, device, equipment and storage medium thereof to solve the problem that the image samples of the vehicle business model in the prior art are insufficient, which seriously affects the training progress of the vehicle business model.
[0005] In order to solve the above technical problems, the present application provides an image amplification method, which adopts the following technical solutions:
[0006] An image amplification method comprises the following steps:
[0007] Acquire the reconstructed modeling parameters of the target vehicle, wherein the reconstructed modeling parameters of the target vehicle include spatial parameter information and physical parameter information of the vehicle and all components of the vehicle;
[0008] Inputting the reconstruction modeling parameters into a preset simulation modeling component to perform reconstruction modeling of the target vehicle to obtain a vehicle modeling body;
[0009] Setting collision parameters of the vehicle modeling body to perform vehicle driving collision simulation;
[0010] Batch collision images are obtained by repeatedly adjusting collision parameters;
[0011] Inputting the batch collision images and the real accident images of the target vehicle collected in advance into a preset diffusion model;
[0012] Fine-tuning the preset diffusion model using the real accident image;
[0013] The batch collision images are diffusely generated using the fine-tuned diffusion model to obtain all generated images output by the diffusion model.
[0014] Furthermore, before executing the step of obtaining the reconstruction modeling parameters of the target vehicle, the method further includes:
[0015] A parameter configuration form is pre-transmitted to a target receiving end in a data stream transmission manner, wherein the target receiving end includes a form parsing end that pre-establishes a data transmission relationship with the simulation modeling component, and the parameter configuration form includes spatial parameter information and physical parameter information of the vehicle and all components of the vehicle, wherein the spatial parameter information includes three-dimensional shape information, size information, and thickness information, and the physical parameter information includes material information, weight information, and stamping resistance information;
[0016] The step of obtaining the reconstruction modeling parameters of the target vehicle specifically includes:
[0017] The parameter configuration form is formatted and parsed by the form parsing end to obtain spatial parameter information in a unified representation format and physical parameter information in a unified representation format.
[0018] Furthermore, the preset simulation modeling component includes B l ender three-dimensional graphics image software. Before executing the step of inputting the reconstruction modeling parameters into the preset simulation modeling component, performing the target vehicle reconstruction modeling, and obtaining the vehicle modeling body, the method further includes:
[0019] Importing a target parsing template into the simulation modeling component, wherein the target parsing template is adapted to a preset transmission template;
[0020] Importing the assembly connection relationship between all components of the vehicle into the simulation modeling component as a configuration relationship;
[0021] The step of inputting the reconstruction modeling parameters into a preset simulation modeling component, performing the reconstruction modeling of the target vehicle, and obtaining a vehicle modeling body specifically includes:
[0022] Using the preset transmission template to send the spatial parameter information in the unified representation format and the physical parameter information in the unified representation format to the simulation modeling component;
[0023] Performing transmission content analysis on the preset transmission template through the target analysis template to obtain the reconstruction modeling parameters;
[0024] Based on the reconstruction modeling parameters and the assembly connection relationship, the target vehicle is reconstructed and modeled to obtain the vehicle modeling body, and the physical type of the vehicle modeling body is set to a rigid body.
[0025] Furthermore, the step of setting the collision parameters of the vehicle modeling body and performing vehicle driving collision simulation specifically includes:
[0026] Extracting the collision parameters from a preset collision parameter configuration interface, wherein the preset collision parameter configuration interface is a human-computer interaction interface, and the collision parameters include the speed of the target vehicle at the moment of collision, the deceleration after the collision, the collision area, and the spatial parameter information and physical parameter information of the target collision body, the speed at the moment of collision, and the collision area;
[0027] Determine whether the preset command monitoring component receives the collision command;
[0028] If the instruction monitoring component does not receive the collision instruction, continue to perform instruction monitoring;
[0029] If the instruction monitoring component receives the collision instruction, it simulates the vehicle driving collision according to the collision parameters, and obtains the collision video at the moment of the collision and before the target vehicle stops after the collision, and obtains all collision images generated by the collision by frame processing of the collision video.
[0030] Furthermore, before executing the step of obtaining batch collision images by repeatedly adjusting the collision parameters, the method further includes:
[0031] Obtaining the fine-tuning change functions and limiting conditions to which all collision parameters pre-set in the collision parameter configuration interface are respectively subject;
[0032] The step of obtaining batch collision images by repeatedly adjusting the collision parameters specifically includes:
[0033] Step 501, automatically adjusting the collision parameters of the last collision according to the collision parameters of the last collision, the fine-tuning change functions and the limiting conditions to which all the collision parameters are respectively subject, so as to obtain the collision parameters of this collision;
[0034] Step 502, simulating a vehicle driving collision according to the collision parameters of this collision, and obtaining a collision video of the moment of this collision and before the target vehicle stops after the collision, and obtaining all collision images generated by this collision by performing frame processing on the collision video;
[0035] Step 503 , repeatedly executing steps 501 to 502 to obtain collision images of a target number of times, until the target number of times reaches a preset number threshold, stopping the repetition, and using the collision images of the target number of times as the batch collision images.
[0036] Furthermore, the step of fine-tuning the preset diffusion model using the real accident image specifically includes:
[0037] The forward noise processing component of the diffusion model is used to perform forward noise processing on all real accident images respectively, and the corresponding noisy images after each noise processing and the number of noise processing times corresponding to each noisy image are obtained, wherein the image noise introduced during the forward noise processing obeys Gaussian distribution;
[0038] Filter out images whose image resolution meets a preset resolution threshold from all noisy images, and construct a noisy image set;
[0039] According to the number of denoising processes corresponding to all the denoised images in the denoised image set, a majority selection is performed, and the number of denoising processes after the majority selection is set as the optimal value of the forward denoising process number of the preset diffusion model, thereby completing the fine-tuning of the preset diffusion model.
[0040] Furthermore, the step of using the fine-tuned diffusion model to perform diffusion generation on the batch collision images to obtain all generated images output by the diffusion model specifically includes:
[0041] Using the forward noise processing component of the diffusion model to perform forward noise processing on all batch collision images respectively;
[0042] Determining whether the number of noise addition processing times of each collision image by the forward noise addition processing component reaches the optimal number of forward noise addition processing times;
[0043] If the number of times the forward noise processing component performs noise processing on the current collision image reaches the optimal number of forward noise processing times, all noise processing images corresponding to the current collision image are obtained, and an instruction to stop performing noise processing on the current collision image is sent to the forward noise processing component;
[0044] If the number of noise addition processing times of the current collision image by the forward noise addition processing component has not reached the optimal number of forward noise addition processing times, then the noise addition processing is continued on the noise addition image corresponding to the current collision image based on the current number of noise additions until the number of noise addition processing times of the current collision image by the forward noise addition processing component reaches the optimal number of forward noise addition processing times, then all noise addition processing images corresponding to the current collision image are obtained, and an instruction to stop noise addition processing on the current collision image is sent to the forward noise addition processing component.
[0045] In order to solve the above technical problems, the embodiment of the present application also provides an image amplification device, which adopts the following technical solution:
[0046] An image amplification device, comprising:
[0047] A reconstruction modeling parameter acquisition module is used to acquire reconstruction modeling parameters of a target vehicle, wherein the reconstruction modeling parameters of the target vehicle include spatial parameter information and physical parameter information of the vehicle and all components of the vehicle;
[0048] A target vehicle modeling processing module is used to input the reconstruction modeling parameters into a preset simulation modeling component to perform reconstruction modeling of the target vehicle to obtain a vehicle modeling body;
[0049] A vehicle collision simulation module is used to set the collision parameters of the vehicle modeling body and perform vehicle driving collision simulation;
[0050] A batch collision image acquisition module is used to obtain batch collision images by repeatedly adjusting collision parameters;
[0051] An image input module, used for inputting the batch collision images and the real accident images of the target vehicle collected in advance into a preset diffusion model;
[0052] A diffusion model fine-tuning module, used for fine-tuning the preset diffusion model through the real accident image;
[0053] The image diffusion generation module is used to perform diffusion generation on the batch collision images using the fine-tuned diffusion model to obtain all generated images output by the diffusion model.
[0054] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:
[0055] A computer device comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the above-mentioned image augmentation method when executing the computer-readable instructions.
[0056] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:
[0057] A computer-readable storage medium stores computer-readable instructions, which implement the steps of the image augmentation method described above when executed by a processor.
[0058] Compared with the prior art, the embodiments of the present application have the following beneficial effects:
[0059] The image amplification method described in the embodiment of the present application obtains the reconstruction modeling parameters of the target vehicle; inputs the reconstruction modeling parameters into the simulation modeling component to reconstruct the target vehicle and obtain a vehicle modeling body; sets the collision parameters of the vehicle modeling body to simulate vehicle driving collision; obtains batch collision images by repeatedly adjusting the collision parameters; fine-tunes the diffusion model through real accident images; and uses the fine-tuned diffusion model to diffusely generate batch collision images to obtain all generated images output by the diffusion model. In the application fields of vehicle accident handling and vehicle claims, the combination of physical simulation components, collision simulation and diffusion models for image acquisition and regeneration ensures the sufficiency of accident image samples when training related models. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0061] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;
[0062] Figure 2 is a flow chart of an embodiment of the image amplification method according to the present application;
[0063] Figure 3 yes Figure 2 A flowchart of a specific embodiment of step 202 is shown;
[0064] Figure 4 yes Figure 2 A flowchart of a specific embodiment of step 203 is shown;
[0065] Figure 5 yes Figure 2 A flowchart of a specific embodiment of step 204 is shown;
[0066] Figure 6 yes Figure 2 A flowchart of a specific embodiment of step 206 is shown;
[0067] Figure 7 yes Figure 2 A flowchart of a specific embodiment of step 207 is shown;
[0068] Figure 8 is a schematic structural diagram of an embodiment of an image amplification device according to the present application;
[0069] Fig. 9 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION
[0070] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.
[0071] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0072] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.
[0073] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.
[0074] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.
[0075] The terminal devices 101, 102, and 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.
[0076] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .
[0077] It should be noted that the image amplification method provided in the embodiment of the present application is generally executed by a server, and accordingly, the image amplification device is generally disposed in the server.
[0078] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.
[0079] Continue to refer Figure 2 , shows a flow chart of an embodiment of an image amplification method according to the present application. The image amplification method comprises the following steps:
[0080] Step 201, obtaining reconstruction modeling parameters of the target vehicle.
[0081] Among them, the reconstructed modeling parameters of the target vehicle include spatial parameter information and physical parameter information of the vehicle and all its components. The spatial parameter information includes three-dimensional shape information, size information, and thickness information; the physical parameter information includes material information, weight information, and stamping resistance information.
[0082] By obtaining the spatial parameter information and physical parameter information of the target vehicle and all components of the target vehicle in advance, it is convenient to subsequently reconstruct and model the target vehicle to obtain a 3D vehicle modeling body.
[0083] In this embodiment, before executing the step of obtaining the reconstructed modeling parameters of the target vehicle, the method also includes: using a data stream transmission method to pre-transmit a parameter configuration form to the target receiving end, wherein the target receiving end includes a form parsing end that pre-establishes a data transmission relationship with the simulation modeling component, and the parameter configuration form contains spatial parameter information and physical parameter information of the vehicle and all its components.
[0084] Specifically, the spatial parameter information and physical parameter information of the target vehicle and all components of the target vehicle are pre-organized into the parameter configuration form, and the data content in the parameter configuration form is transmitted to the target receiving end using a data stream transmission method, so as to facilitate subsequent direct parsing of the data content in the parameter configuration form to obtain the reconstructed modeling parameters of the target vehicle.
[0085] In this embodiment, the step of obtaining the reconstructed modeling parameters of the target vehicle specifically includes: formatting and parsing the parameter configuration form through the form parsing end to obtain spatial parameter information in a unified representation format and physical parameter information in a unified representation format.
[0086] Specifically, by formatting and parsing the parameter configuration form, spatial parameter information in a unified representation format and physical parameter information in a unified representation format are obtained respectively. In this embodiment, the spatial parameter information and the physical parameter information may adopt different representation formats or the same representation format. However, only for the spatial parameter information or the physical parameter information, a unified data representation format is adopted for formatting and parsing, which facilitates data search and calling during subsequent reconstruction and modeling of the target vehicle, saving a certain amount of screening time.
[0087] Step 202: input the reconstruction modeling parameters into a preset simulation modeling component, perform reconstruction modeling of the target vehicle, and obtain a vehicle modeling body.
[0088] In this embodiment, the preset simulation modeling component includes Blackender three-dimensional graphics and image software. Before executing the step of inputting the reconstruction modeling parameters into the preset simulation modeling component, performing the target vehicle reconstruction modeling, and obtaining the vehicle modeling body, the method also includes: importing the target parsing template into the simulation modeling component, wherein the target parsing template is adapted to the preset transmission template; and importing the assembly connection relationship between all components of the vehicle as a configuration relationship into the simulation modeling component.
[0089] By importing the target parsing template that is compatible with the preset transmission template into the simulation modeling component, it is convenient for the simulation modeling component to parse the data content in the transmission template, and importing the assembly connection relationship between all components of the vehicle as a configuration relationship into the simulation modeling component, it is convenient for the simulation modeling component to reconstruct and model the target vehicle according to the assembly connection relationship, perform modeling control and improve modeling efficiency.
[0090] Specifically, Blackender is an open source 3D modeling and rendering software. Users can use Blackender to perform various physical simulation operations. For example, collision detection and reaction force simulation can provide more realistic simulation scenes and simulation effects. By directly using simulation modeling components to build a vehicle modeling body, it is convenient to directly use Blackender's three-dimensional graphics and image software to perform target vehicle collision simulation in the future, and obtain batches of collision images. It can be used in the simulation of collision detection scenes for new cars, as well as in the training of classification and prediction models related to vehicle business, and in the simulation generation of accident images when there are fewer actual vehicle accident images, thus ensuring the adequacy of the image sample volume.
[0091] Continue to refer Figure 3 , Figure 3 yes Figure 2 The flowchart of a specific embodiment of step 202 shown includes:
[0092] Step 301, using the preset transmission template to send the spatial parameter information in the unified representation format and the physical parameter information in the unified representation format to the simulation modeling component;
[0093] Specifically, a preset transmission template is used to send the spatial parameter information in a unified representation format and the physical parameter information in a unified representation format to the simulation modeling component. In the subsequent actual analysis, the cache position of the transmitted data content can be quickly located according to the identification information of the transmission template, which facilitates subsequent accurate analysis.
[0094] Step 302, performing transmission content analysis on the preset transmission template through the target analysis template to obtain the reconstruction modeling parameters;
[0095] Specifically, the transmission content of the preset transmission template is parsed through the target parsing template to obtain the reconstruction modeling parameters. Since the target parsing template is compatible with the transmission template, rapid parsing can be performed.
[0096] Step 303: Based on the reconstruction modeling parameters and the assembly connection relationship, the target vehicle is reconstructed and modeled to obtain the vehicle modeling body, and the physical type of the vehicle modeling body is set to a rigid body.
[0097] By reconstructing and modeling the target vehicle through the reconstructed modeling parameters and the assembly connection relationship, both modeling control can be performed and modeling efficiency can be relatively improved.
[0098] Step 203, setting collision parameters of the vehicle modeling body to perform vehicle driving collision simulation.
[0099] Continue to refer Figure 4 , Figure 4 yes Figure 2 The flowchart of a specific embodiment of step 203 shown includes:
[0100] Step 401, extracting the collision parameters from a preset collision parameter configuration interface;
[0101] Among them, the preset collision parameter configuration interface is a human-computer interaction interface, and the collision parameters include the speed of the target vehicle at the moment of collision, the deceleration after the collision, the collision area, and the spatial parameter information and physical parameter information of the target collision body, the speed at the moment of collision, and the collision area.
[0102] Specifically, the target collision body refers to an object that collides with the target vehicle, including dynamic objects with speed, such as other vehicles, or static objects without speed, such as objects that move with the collision, such as a stationary vehicle, or objects that do not move with the collision, such as a mountain.
[0103] Step 402, determining whether the preset instruction monitoring component receives a collision instruction;
[0104] In this embodiment, the preset instruction monitoring component includes a human-computer interaction instruction monitoring component. For example, after configuring the corresponding collision parameters, it is necessary to click the instruction to start the collision simulation. The "click" is the "human-computer interaction instruction".
[0105] Step 403, if the instruction monitoring component does not receive the collision instruction, continue to perform instruction monitoring;
[0106] Step 404, if the instruction monitoring component receives the collision instruction, it simulates the vehicle driving collision according to the collision parameters, and obtains the collision video at the moment of this collision and before the target vehicle stops after the collision, and obtains all collision images generated by this collision by frame processing of the collision video.
[0107] Specifically, since there is a dynamic change process between objects at the moment of collision and before the target vehicle stops after the collision, it is considered to obtain a collision video, and obtain all collision images generated by this collision by frame processing on the collision video, so that multiple collision images can be obtained for one collision, enriching the vehicle accident image and ensuring the adequacy of the image samples.
[0108] Step 204 , obtaining batch collision images by repeatedly adjusting the collision parameters.
[0109] In this embodiment, before executing the step of obtaining batch collision images by repeatedly adjusting the collision parameters, the method further includes: obtaining fine-tuning change functions and limiting conditions to which all collision parameters pre-set in the collision parameter configuration interface are respectively subject.
[0110] Specifically, the fine-tuning variation function and limiting conditions to which all collision parameters are respectively subject, taking the collision speed of a vehicle as an example, assuming that the simulated vehicle is in a high-speed normal driving state, the driving speed limiting condition is set to 60km / h-120km / h, and the speed limiting condition after the collision is set to 120km / h-0km / h. The fine-tuning variation function to which the speed is subject after the collision, for example: after the collision, the speed gradually decreases according to a certain acceleration until it decreases to 0. Therefore, all collision parameters can be freely adjusted within the limiting conditions corresponding to all collision parameters and according to the fine-tuning variation functions to which all collision parameters are respectively subject, so that each time a collision simulation is performed, the collision effect produced is different, thereby obtaining a large number of collision images.
[0111] Continue to refer Figure 5 , Figure 5 yes Figure 2 The flowchart of a specific embodiment of step 204 shown includes:
[0112] Step 501, automatically adjusting the collision parameters of the last collision according to the collision parameters of the last collision, the fine-tuning change functions and the limiting conditions to which all the collision parameters are respectively subject, so as to obtain the collision parameters of this collision;
[0113] Specifically, the collision parameters at the time of the last collision are automatically adjusted according to the collision parameters at the time of the last collision, the fine-tuning change function to which all the collision parameters respectively obey, and the limiting conditions, thereby avoiding manual and continuous adjustment of the collision parameters, which consumes a lot of manpower and material resources. The use of an automated adjustment method is more intelligent and improves the overall completion efficiency of the collision simulation.
[0114] Step 502, simulating a vehicle driving collision according to the collision parameters of this collision, and obtaining a collision video of the moment of this collision and before the target vehicle stops after the collision, and obtaining all collision images generated by this collision by performing frame processing on the collision video;
[0115] Step 503 , repeatedly executing steps 501 to 502 to obtain collision images of a target number of times, until the target number of times reaches a preset number threshold, stopping the repetition, and using the collision images of the target number of times as the batch collision images.
[0116] Specifically, a repeated iteration method may be used, for example, the collision simulation is performed 100 times, and the collision images of these 100 times are obtained by repeatedly executing step 501 to step 502, thereby ensuring the sufficiency of the image samples.
[0117] Step 205: input the batch collision images and the pre-collected real accident images of the target vehicle into a preset diffusion model.
[0118] Step 206: fine-tune the preset diffusion model using the real accident image.
[0119] Continue to refer Figure 6 , Figure 6 yes Figure 2 The flowchart of a specific embodiment of step 206 shown includes:
[0120] Step 601, using the forward noise processing component of the diffusion model to perform forward noise processing on all real accident images respectively, to obtain the corresponding noisy image after each noise processing and the number of noise processing times corresponding to each noisy image, wherein the image noise introduced during the forward noise processing obeys Gaussian distribution;
[0121] Specifically, the diffusion model includes a diffusion probability model based on DDPM (Denoising Diffusion Probalistic Models, DDPM), and its diffusion principle includes a forward denoising process and a reverse denoising process. The forward denoising process is used to enrich the original sample volume, and the reverse denoising process is used to identify the noisy samples. Since the embodiment of the present application only performs image amplification generation, only the forward denoising processing component in the diffusion model can be used to generate a noisy image.
[0122] Step 602, selecting images whose image resolutions meet a preset resolution threshold from all noisy images, and constructing a noisy image set;
[0123] Specifically, since the image resolution will become lower and lower each time image noise is added, it is not conducive to the training of the model and the subsequent classification prediction work. Therefore, in this embodiment, when amplifying the image sample, it is set that the image resolution generated by the amplification must meet the preset resolution threshold, thereby avoiding blindly amplifying the image multiple times and reducing the amount of useless images introduced.
[0124] Step 603, performing a majority selection according to the number of denoising processes corresponding to all the denoised images in the denoised image set, setting the number of denoising processes after the majority selection as the optimal value of the forward denoising process number of the preset diffusion model, and completing fine-tuning of the preset diffusion model.
[0125] By means of majority selection, the optimal number of forward denoising processing times of the preset diffusion model is screened out, that is, the optimal number of image amplifications using the diffusion model is selected. Combined with actual vehicle accident images, blindly performing multiple image amplifications is avoided and the amount of useless images introduced is reduced.
[0126] The preset diffusion model is fine-tuned using real accident images to ensure that the fine-tuned diffusion model is more in line with the actual real accident images being processed. The fine-tuning in this embodiment mainly adjusts the optimal number of forward noisy processes. Of course, other parameters of the diffusion model can also be fine-tuned, such as the convolution kernel size and size of the neural network layer used in the forward noisy process.
[0127] Step 207 , using the fine-tuned diffusion model to perform diffusion generation on the batch collision images, to obtain all generated images output by the diffusion model.
[0128] Continue to refer Figure 7 , Figure 7 yes Figure 2 The flowchart of a specific embodiment of step 207 shown includes:
[0129] Step 701, using the forward noise processing component of the diffusion model to perform forward noise processing on all batch collision images respectively;
[0130] Step 702, determining whether the number of noise addition processing times of each collision image by the forward noise addition processing component reaches the optimal number of forward noise addition processing times;
[0131] Step 703: if the number of times the forward noise processing component performs noise processing on the current collision image reaches the optimal number of times the forward noise processing component performs noise processing, then all noise processing images corresponding to the current collision image are obtained, and an instruction to stop performing noise processing on the current collision image is sent to the forward noise processing component;
[0132] Step 704: if the number of noise addition processing times of the current collision image by the forward noise addition processing component has not reached the optimal number of forward noise addition processing times, then continue to perform noise addition processing on the noise addition image corresponding to the current collision image based on the current number of noise addition processing times, until the number of noise addition processing times of the current collision image by the forward noise addition processing component reaches the optimal number of forward noise addition processing times, then obtain all the noise addition processing images corresponding to the current collision image, and send an instruction to the forward noise addition processing component to stop performing noise addition processing on the current collision image.
[0133] In this embodiment, since the optimal number of forward noise addition processing times of the diffusion model has been adjusted before, it is only necessary to perform forward noise addition processing on the batch collision images obtained by the collision simulation according to the optimal number of forward noise addition processing times, so that the image samples can be amplified again, thereby further ensuring the adequacy of the image sample volume. Especially in the field of vehicle accident claims prediction, the physical simulation components, collision simulation and diffusion models are combined for image acquisition and regeneration, thereby ensuring the adequacy of accident image samples when training the claims model.
[0134] This application obtains the reconstruction modeling parameters of the target vehicle; inputs the reconstruction modeling parameters into the simulation modeling component to reconstruct the target vehicle and obtain the vehicle modeling body; sets the collision parameters of the vehicle modeling body to simulate vehicle driving collision; obtains batch collision images by repeatedly adjusting the collision parameters; fine-tunes the diffusion model through real accident images; uses the fine-tuned diffusion model to diffusely generate batch collision images and obtain all generated images output by the diffusion model. In the application fields of vehicle accident handling and vehicle claims, the combination of physical simulation components, collision simulation and diffusion models for image acquisition and regeneration ensures the sufficiency of accident image samples when training related models.
[0135] The embodiments of the present application can acquire and process relevant data based on image processing technology. Among them, image processing (Artificial Intelligence, AI) is the theory, method, technology and application system of using digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
[0136] Image processing basic technologies generally include sensors, dedicated image processing chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. Image processing software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.
[0137] In the embodiment of the present application, the reconstruction modeling parameters of the target vehicle are obtained; the reconstruction modeling parameters are input into the simulation modeling component to reconstruct the target vehicle and obtain the vehicle modeling body; the collision parameters of the vehicle modeling body are set to simulate the vehicle driving collision; batch collision images are obtained by repeatedly adjusting the collision parameters; the diffusion model is fine-tuned through real accident images; the batch collision images are diffusely generated using the fine-tuned diffusion model to obtain all generated images output by the diffusion model. In the application fields of vehicle accident handling and vehicle claims, the physical simulation component, collision simulation and diffusion model are combined for image acquisition and regeneration to ensure the sufficiency of accident image samples when training related models.
[0138] Further references Figure 8 , as a response to the above Figure 2 In order to realize the method shown in the figure, the present application provides an embodiment of an image amplification device, and the device embodiment is Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.
[0139] like Figure 8 As shown, the image amplification device 800 described in this embodiment includes: a reconstruction modeling parameter acquisition module 801, a target vehicle modeling processing module 802, a vehicle collision simulation module 803, a batch collision image acquisition module 804, an image input module 805, a diffusion model fine-tuning module 806 and an image diffusion generation module 807. Among them:
[0140] The reconstruction modeling parameter acquisition module 801 is used to acquire the reconstruction modeling parameters of the target vehicle, wherein the reconstruction modeling parameters of the target vehicle include the spatial parameter information and physical parameter information of the vehicle and all components of the vehicle;
[0141] The target vehicle modeling processing module 802 is used to input the reconstruction modeling parameters into a preset simulation modeling component to perform the reconstruction modeling of the target vehicle to obtain a vehicle modeling body;
[0142] The vehicle collision simulation module 803 is used to set the collision parameters of the vehicle modeling body and perform vehicle driving collision simulation;
[0143] A batch collision image acquisition module 804 is used to obtain batch collision images by repeatedly adjusting collision parameters;
[0144] An image input module 805, used for inputting the batch collision images and the real accident images of the target vehicle collected in advance into a preset diffusion model;
[0145] A diffusion model fine-tuning module 806, configured to fine-tune the preset diffusion model using the real accident image;
[0146] The image diffusion generation module 807 is used to perform diffusion generation on the batch collision images using the fine-tuned diffusion model to obtain all generated images output by the diffusion model.
[0147] This application obtains the reconstruction modeling parameters of the target vehicle; inputs the reconstruction modeling parameters into the simulation modeling component to reconstruct the target vehicle and obtain the vehicle modeling body; sets the collision parameters of the vehicle modeling body to simulate vehicle driving collision; obtains batch collision images by repeatedly adjusting the collision parameters; fine-tunes the diffusion model through real accident images; uses the fine-tuned diffusion model to diffusely generate batch collision images and obtain all generated images output by the diffusion model. In the application fields of vehicle accident handling and vehicle claims, the combination of physical simulation components, collision simulation and diffusion models for image acquisition and regeneration ensures the sufficiency of accident image samples when training related models.
[0148] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).
[0149] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.
[0150] To solve the above technical problems, the present application also provides a computer device. Fig. 9 , Fig. 9 This is a basic structural block diagram of the computer device in this embodiment.
[0151] The computer device 9 includes a memory 9a, a processor 9b, and a network interface 9c that are interconnected through a system bus. It should be noted that the figure only shows a computer device 9 having components 9a-9c, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.
[0152] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.
[0153] The memory 9a includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 9a can be an internal storage unit of the computer device 9, such as a hard disk or memory of the computer device 9. In other embodiments, the memory 9a can also be an external storage device of the computer device 9, such as a plug-in hard disk equipped on the computer device 9, a smart memory card (Smart Med i aCard, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 9a can also include both the internal storage unit of the computer device 9 and its external storage device. In this embodiment, the memory 9a is generally used to store the operating system and various application software installed on the computer device 9, such as a computer-readable instruction of an image amplification method, etc. In addition, the memory 9a can also be used to temporarily store various types of data that have been output or are to be output.
[0154] The processor 9b may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 9b is generally used to control the overall operation of the computer device 9. In this embodiment, the processor 9b is used to run the computer-readable instructions stored in the memory 9a or process data, such as computer-readable instructions for running the image amplification method.
[0155] The network interface 9c may include a wireless network interface or a wired network interface. The network interface 9c is generally used to establish a communication connection between the computer device 9 and other electronic devices.
[0156] The computer device proposed in this embodiment belongs to the field of image processing technology and is applied to image sample augmentation scenarios. This application obtains the reconstruction modeling parameters of the target vehicle; inputs the reconstruction modeling parameters into the simulation modeling component to reconstruct the target vehicle and obtain a vehicle modeling body; sets the collision parameters of the vehicle modeling body to simulate vehicle driving collision; obtains batch collision images by repeatedly adjusting the collision parameters; fine-tunes the diffusion model through real accident images; uses the fine-tuned diffusion model to diffusely generate batch collision images and obtain all generated images output by the diffusion model. In the application fields of vehicle accident processing and vehicle claims, the combination of physical simulation components, collision simulation and diffusion models for image acquisition and regeneration ensures the sufficiency of accident image samples when training related models.
[0157] The present application also provides another embodiment, that is, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by a processor to enable the processor to perform the steps of the image amplification method as described above.
[0158] The computer-readable storage medium proposed in this embodiment belongs to the field of image processing technology and is applied to image sample augmentation scenarios. This application obtains the reconstruction modeling parameters of the target vehicle; inputs the reconstruction modeling parameters into the simulation modeling component to reconstruct the target vehicle and obtain a vehicle modeling body; sets the collision parameters of the vehicle modeling body to simulate vehicle driving collision; obtains batch collision images by repeatedly adjusting the collision parameters; fine-tunes the diffusion model through real accident images; uses the fine-tuned diffusion model to diffusely generate batch collision images and obtain all generated images output by the diffusion model. In the application fields of vehicle accident processing and vehicle claims, the physical simulation component, collision simulation and diffusion model are combined for image acquisition and regeneration to ensure the sufficiency of accident image samples when training related models.
[0159] 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 application, 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), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.
[0160] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.
Claims
1. An image amplification method, characterized in that: The steps include: Acquire the reconstructed modeling parameters of the target vehicle, wherein the reconstructed modeling parameters of the target vehicle include spatial parameter information and physical parameter information of the vehicle and all components of the vehicle; Inputting the reconstruction modeling parameters into a preset simulation modeling component to perform reconstruction modeling of the target vehicle to obtain a vehicle modeling body; Setting collision parameters of the vehicle modeling body to perform vehicle driving collision simulation; Batch collision images are obtained by repeatedly adjusting collision parameters; Inputting the batch collision images and the real accident images of the target vehicle collected in advance into a preset diffusion model; Fine-tuning the preset diffusion model using the real accident image; The batch collision images are diffusely generated using the fine-tuned diffusion model to obtain all generated images output by the diffusion model.
2. The image amplification method according to claim 1, characterized in that: Before executing the step of obtaining the reconstruction modeling parameters of the target vehicle, the method further includes: A parameter configuration form is pre-transmitted to a target receiving end in a data stream transmission manner, wherein the target receiving end includes a form parsing end that pre-establishes a data transmission relationship with the simulation modeling component, and the parameter configuration form includes spatial parameter information and physical parameter information of the vehicle and all components of the vehicle, wherein the spatial parameter information includes three-dimensional shape information, size information, and thickness information, and the physical parameter information includes material information, weight information, and stamping resistance information; The step of obtaining the reconstruction modeling parameters of the target vehicle specifically includes: The parameter configuration form is formatted and parsed by the form parsing end to obtain spatial parameter information in a unified representation format and physical parameter information in a unified representation format.
3. The image amplification method according to claim 2, characterized in that: The preset simulation modeling component includes Blender three-dimensional graphics software. Before executing the step of inputting the reconstruction modeling parameters into the preset simulation modeling component, performing the target vehicle reconstruction modeling, and obtaining the vehicle modeling body, the method further includes: Importing a target parsing template into the simulation modeling component, wherein the target parsing template is adapted to a preset transmission template; Importing the assembly connection relationship between all components of the vehicle into the simulation modeling component as a configuration relationship; The step of inputting the reconstruction modeling parameters into a preset simulation modeling component, performing the reconstruction modeling of the target vehicle, and obtaining a vehicle modeling body specifically includes: Using the preset transmission template to send the spatial parameter information in the unified representation format and the physical parameter information in the unified representation format to the simulation modeling component; Performing transmission content analysis on the preset transmission template through the target analysis template to obtain the reconstruction modeling parameters; Based on the reconstruction modeling parameters and the assembly connection relationship, the target vehicle is reconstructed and modeled to obtain the vehicle modeling body, and the physical type of the vehicle modeling body is set to a rigid body.
4. The image amplification method according to claim 1, characterized in that: The step of setting the collision parameters of the vehicle modeling body and performing vehicle driving collision simulation specifically includes: Extracting the collision parameters from a preset collision parameter configuration interface, wherein the preset collision parameter configuration interface is a human-computer interaction interface, and the collision parameters include the speed of the target vehicle at the moment of collision, the deceleration after the collision, the collision area, and the spatial parameter information and physical parameter information of the target collision body, the speed at the moment of collision, and the collision area; Determine whether the preset command monitoring component receives the collision command; If the instruction monitoring component does not receive the collision instruction, continue to perform instruction monitoring; If the instruction monitoring component receives the collision instruction, it simulates the vehicle driving collision according to the collision parameters, and obtains the collision video at the moment of the collision and before the target vehicle stops after the collision, and obtains all collision images generated by the collision by frame processing of the collision video.
5. The image amplification method according to claim 4, characterized in that: Before executing the step of obtaining batch collision images by repeatedly adjusting the collision parameters, the method further includes: Obtaining the fine-tuning change functions and limiting conditions to which all collision parameters pre-set in the collision parameter configuration interface are respectively subject; The step of obtaining batch collision images by repeatedly adjusting the collision parameters specifically includes: Step 501, automatically adjusting the collision parameters of the last collision according to the collision parameters of the last collision, the fine-tuning change functions and the limiting conditions to which all the collision parameters are respectively subject, so as to obtain the collision parameters of this collision; Step 502, simulating a vehicle driving collision according to the collision parameters of this collision, and obtaining a collision video of the moment of this collision and before the target vehicle stops after the collision, and obtaining all collision images generated by this collision by performing frame processing on the collision video; Step 503 , repeatedly executing steps 501 to 502 to obtain collision images of a target number of times, until the target number of times reaches a preset number threshold, stopping the repetition, and using the collision images of the target number of times as the batch collision images.
6. The image amplification method according to claim 1, characterized in that: The step of fine-tuning the preset diffusion model using the real accident image specifically includes: The forward noise processing component of the diffusion model is used to perform forward noise processing on all real accident images respectively, and the corresponding noisy images after each noise processing and the number of noise processing times corresponding to each noisy image are obtained, wherein the image noise introduced during the forward noise processing obeys Gaussian distribution; Filter out images whose image resolution meets a preset resolution threshold from all noisy images, and construct a noisy image set; According to the number of denoising processes corresponding to all the denoised images in the denoised image set, a majority selection is performed, and the number of denoising processes after the majority selection is set as the optimal value of the forward denoising process number of the preset diffusion model, thereby completing the fine-tuning of the preset diffusion model.
7. The image amplification method according to claim 6, characterized in that: The step of using the fine-tuned diffusion model to diffusely generate the batch collision images to obtain all generated images output by the diffusion model specifically includes: Using the forward noise processing component of the diffusion model to perform forward noise processing on all batch collision images respectively; Determining whether the number of noise addition processing times of each collision image by the forward noise addition processing component reaches the optimal number of forward noise addition processing times; If the number of times the forward noise processing component performs noise processing on the current collision image reaches the optimal number of forward noise processing times, all noise processing images corresponding to the current collision image are obtained, and an instruction to stop performing noise processing on the current collision image is sent to the forward noise processing component; If the number of noise addition processing times of the current collision image by the forward noise addition processing component has not reached the optimal number of forward noise addition processing times, then the noise addition processing is continued on the noise addition image corresponding to the current collision image based on the current number of noise additions until the number of noise addition processing times of the current collision image by the forward noise addition processing component reaches the optimal number of forward noise addition processing times, then all noise addition processing images corresponding to the current collision image are obtained, and an instruction to stop noise addition processing on the current collision image is sent to the forward noise addition processing component.
8. An image amplification device, characterized in that: include: A reconstruction modeling parameter acquisition module is used to acquire reconstruction modeling parameters of a target vehicle, wherein the reconstruction modeling parameters of the target vehicle include spatial parameter information and physical parameter information of the vehicle and all components of the vehicle; A target vehicle modeling processing module is used to input the reconstruction modeling parameters into a preset simulation modeling component to perform reconstruction modeling of the target vehicle to obtain a vehicle modeling body; A vehicle collision simulation module is used to set the collision parameters of the vehicle modeling body and perform vehicle driving collision simulation; A batch collision image acquisition module is used to obtain batch collision images by repeatedly adjusting collision parameters; An image input module, used for inputting the batch collision images and the real accident images of the target vehicle collected in advance into a preset diffusion model; A diffusion model fine-tuning module, used for fine-tuning the preset diffusion model through the real accident image; The image diffusion generation module is used to perform diffusion generation on the batch collision images using the fine-tuned diffusion model to obtain all generated images output by the diffusion model.
9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the image amplification method according to any one of claims 1 to 7 when executing the computer-readable instructions.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the image augmentation method according to any one of claims 1 to 7 are implemented.
