Vehicle body scratch detection method and related equipment

Abstract

The invention belongs to the field of artificial intelligence and big data, is applied to the field of science and technology finance, and relates to a vehicle body scratch detection method which comprises the following steps: acquiring an initial vehicle body scratch detection model; convolution module de-weighting processing is carried out on the initial vehicle body scratch detection model based on a single-stage detector, and an initial vehicle body scratch detection model of a non-overlapping convolution module is obtained; training the initial vehicle body scratch detection model of the non-overlapping convolution module through the training data set to obtain a trained vehicle body scratch detection model; and carrying out vehicle body scratch detection on the vehicle image of the to-be-detected vehicle through the trained vehicle body scratch detection model to obtain a vehicle body scratch detection result of the to-be-detected vehicle. The invention further provides a vehicle body scratch detection device, computer equipment and a storage medium. The problem that the working efficiency is reduced due to the fact that an existing detection method is high in tiny scratch detection difficulty and low in detection accuracy is solved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a vehicle body scratch detection method and related equipment. Background Art

[0002] As a representative technical field of artificial intelligence, target detection has been applied in various industrial and consumer fields to improve production efficiency and meet consumer demand. Among many application fields, there are currently few mature solutions for detecting car body scratches, and mature detection technologies still have great potential in the field of detecting car body scratches that have not been explored and applied. As a common damage to automobiles, car body scratches occur frequently and are difficult to detect, resulting in long detection time. Rapid detection of car body scratches can be used as an auxiliary means for manual detection and assessment of vehicle damage, and can even be combined with other detection and assessment technologies to completely replace manual work, thereby greatly improving the efficiency of the customer's damage assessment process. The existing detection methods still have a high rate of missed detection for tiny scratches. Therefore, there is an urgent need for a car body scratch detection method to solve the problem that the existing detection methods have high difficulty in detecting tiny scratches and low detection accuracy, thereby reducing work efficiency. Summary of the invention

[0003] The purpose of the embodiments of the present application is to propose a vehicle body scratch detection method and related equipment to solve the problem that the existing detection methods have high difficulty in detecting tiny scratches and low detection accuracy, thereby reducing work efficiency. Its main purpose is to deduplicate the repeated convolution modules in the model based on a single-stage detector, delete the original repeated convolution modules while retaining the overall data connection structure of the model, and weight the deduplicated convolution modules to increase the feature attention of the scratches in the deduplicated convolution modules, enhance the detection of tiny scratches, effectively improve the model's detection accuracy for vehicle body scratches, reduce the required detection time, and improve work efficiency.

[0004] In order to solve the above technical problems, the present application provides a method for detecting scratches on a vehicle body, which adopts the following technical solution:

[0005] Acquire an initial vehicle body scratch detection model, wherein the initial vehicle body scratch detection model is obtained by training a pre-trained vehicle body scratch detection model using a training data set;

[0006] Based on the single-stage detector, the initial vehicle body scratch detection model is subjected to convolution module deduplication processing to obtain an initial vehicle body scratch detection model without overlapping convolution modules;

[0007] The initial vehicle body scratch detection model of the non-overlapping convolution module is trained by using the training data set to obtain a trained vehicle body scratch detection model;

[0008] The vehicle body scratch detection is performed on the vehicle image of the vehicle to be detected by using the trained vehicle body scratch detection model to obtain the vehicle body scratch detection result of the vehicle to be detected.

[0009] Furthermore, the obtaining of the initial vehicle body scratch detection model includes:

[0010] Obtaining a pre-trained vehicle body scratch detection model and a training data set, wherein the training data set includes a sample vehicle image and vehicle body scratch annotation data corresponding to the sample vehicle image;

[0011] The sample vehicle image is input into the pre-trained vehicle body scratch detection model for training, and the parameters of the pre-trained vehicle body scratch detection model are adjusted according to the vehicle body scratch annotation data corresponding to the sample vehicle image. After the training is completed, an initial vehicle body scratch detection model is obtained.

[0012] Furthermore, the single-stage detector is used to perform convolution module deduplication processing on the initial vehicle body scratch detection model to obtain an initial vehicle body scratch detection model without overlapping convolution modules, including:

[0013] Get the overlapping convolution modules in the initial car body scratch detection model;

[0014] The overlapping convolution modules are deduplicated based on a single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules.

[0015] Furthermore, the single-stage detector is used to perform deduplication processing on the overlapping convolution modules to obtain an initial vehicle body scratch detection model without overlapping convolution modules:

[0016] Get the initial weight matrix of the overlapping convolution modules in the initial car body scratch detection model;

[0017] Performing weight enhancement processing on the initial weight matrix to obtain a repeated convolution module after the weight enhancement processing;

[0018] Based on the single-stage detector, deduplication processing is performed on the repeated convolution modules after the weight enhancement processing, and after the processing is completed, an initial vehicle body scratch detection model without overlapping convolution modules is obtained.

[0019] Furthermore, the initial vehicle body scratch detection model of the non-overlapping convolution module is trained by the training data set to obtain a trained vehicle body scratch detection model, including:

[0020] The initial vehicle body scratch detection model of the non-overlapping convolution module is trained using the training data set. During the training process, the weight parameters of the convolution module are adjusted using the F-norm. After the training is completed, a trained vehicle body scratch detection model is obtained.

[0021] Furthermore, the F-norm is as follows:

[0022]

[0023] Among them, A is the initial weight matrix, B is the trained weight matrix, F is the F-norm of the matrix, m is the number of rows of the matrix, n is the number of columns of the matrix, a is the element corresponding to A, b is the element corresponding to B, a ij is the element in the i-th row and j-th column of matrix A, b ij is the element in the i-th row and j-th column of matrix B.

[0024] Furthermore, the vehicle image of the vehicle to be detected is subjected to body scratch detection by the trained body scratch detection model to obtain a body scratch detection result of the vehicle to be detected, including:

[0025] The trained vehicle body scratch detection model is used to perform feature extraction processing on the vehicle image of the vehicle to be detected, thereby obtaining the features of the vehicle to be detected;

[0026] Perform body scratch detection on the features of the vehicle to be detected to obtain a body scratch detection result of the vehicle to be detected.

[0027] In order to solve the above technical problems, the present application also provides a vehicle body scratch detection device, which adopts the following technical solution:

[0028] An acquisition module is used to acquire an initial vehicle body scratch detection model, wherein the initial vehicle body scratch detection model is obtained by training a pre-trained vehicle body scratch detection model using a training data set, wherein the training data set includes a sample vehicle image and vehicle body scratch annotation data corresponding to the sample vehicle image;

[0029] A deduplication processing module, used for performing convolution module deduplication processing on the initial vehicle body scratch detection model based on a single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules;

[0030] A training module, used for training the initial vehicle body scratch detection model of the non-overlapping convolution module using the training data set to obtain a trained vehicle body scratch detection model;

[0031] The scratch detection module is used to perform body scratch detection on the vehicle image of the vehicle to be detected by using a trained body scratch detection model to obtain a body scratch detection result of the vehicle to be detected.

[0032] 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:

[0033] The computer device includes a memory and a processor, the memory stores computer-readable instructions, and the processor implements the steps of the vehicle body scratch detection method described in any one of the embodiments of the present application when executing the computer-readable instructions.

[0034] 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:

[0035] The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps of the vehicle body scratch detection method described in any one of the embodiments of the present invention are implemented.

[0036] Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects: obtaining an initial vehicle body scratch detection model, wherein the initial vehicle body scratch detection model is obtained by training a pre-trained vehicle body scratch detection model through a training data set, wherein the training data set includes a sample vehicle image and vehicle body scratch annotation data corresponding to the sample vehicle image; performing convolution module deduplication processing on the initial vehicle body scratch detection model based on a single-stage detector to obtain an initial vehicle body scratch detection model of a non-overlapping convolution module; training the initial vehicle body scratch detection model of the non-overlapping convolution module through the training data set to obtain a trained vehicle body scratch detection model; performing vehicle body scratch detection on a vehicle image of a vehicle to be detected through the trained vehicle body scratch detection model to obtain a body scratch detection result of the vehicle to be detected. The present application deduplicates the repeated convolution modules in the model based on a single-stage detector, deletes the original repeated convolution modules while retaining the overall data connection structure of the model, and weights the deduplicated convolution modules to increase the feature attention of scratches in the deduplicated convolution modules, thereby enhancing the detection of subtle scratches. This solves the problem that the existing detection methods have high difficulty in detecting tiny scratches and low detection accuracy, thereby reducing work efficiency. It effectively improves the model's detection accuracy for car body scratches, reduces the required detection time, and improves work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] 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.

[0038] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0039] Figure 2is a flow chart of an embodiment of a vehicle body scratch detection method according to the present application;

[0040] Figure 3 yes Figure 2 A flowchart of a specific implementation of step S201;

[0041] Figure 4 yes Figure 2 A flowchart of a specific implementation of step S202;

[0042] Figure 5 yes Figure 4 A flowchart of a specific implementation method of step S2022;

[0043] Figure 6 yes Figure 2 A flowchart of a specific implementation method of step S204;

[0044] Figure 7 is a structural schematic diagram of a specific embodiment of a vehicle body scratch detection device according to the present application;

[0045] Figure 8 yes Figure 7 A schematic diagram of a structure of a specific implementation of the acquisition module 701;

[0046] Fig. 9 yes Figure 7 A structural diagram of a specific implementation of the deduplication processing module 702;

[0047] Fig.10 yes Fig. 9 A structural diagram of a specific implementation of the deduplication processing submodule 7022;

[0048] Fig.11 yes Figure 7 A schematic diagram of the structure of a specific implementation of the scratch detection module 704;

[0049] Fig.12 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0050] 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.

[0051] 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.

[0052] 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.

[0053] 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.

[0054] 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.

[0055] Terminal devices 101, 102, 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.

[0056] 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 .

[0057] It should be noted that the vehicle body scratch detection method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the vehicle body scratch detection device is generally arranged in the server / terminal device.

[0058] 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.

[0059] Continue to refer Figure 2 , shows a flow chart of an embodiment of a vehicle body scratch detection method according to the present application. The vehicle body scratch detection method comprises the following steps:

[0060] Step S201: Obtain an initial vehicle body scratch detection model.

[0061] In this embodiment, the vehicle body scratch detection method is executed on an electronic device (eg Figure 1 The server / terminal device shown in the figure can receive the vehicle body scratch detection request of the terminal device through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0062] The above-mentioned terminal devices may be terminal devices equipped with remote interaction functions such as personal assistants and intelligent customer service.

[0063] The above-mentioned vehicle body scratch detection method can be applied to online document detection, online document management and other businesses.

[0064] Specifically, the above electronic equipment can be used in different industry types such as medical target detection, property insurance target detection, financial target detection, store target detection, etc.

[0065] The above initial vehicle body scratch detection model is obtained by training the pre-trained vehicle body scratch detection model through the training data set.

[0066] The training data set includes sample vehicle images and vehicle body scratch annotation data corresponding to the sample vehicle images. The sample vehicle images include a large number of vehicle images, and each vehicle image corresponds to one or more vehicle body scratch annotation data. The vehicle body scratch annotation data may be the location, size, type, etc. of the scratches, and the vehicle body scratch annotation data is used to guide the model to learn how to identify different types of vehicle body scratches.

[0067] The above pre-trained model is a tool for identifying scratches on the surface of a vehicle. The scratches on the surface of a vehicle include but are not limited to irregular shapes, small dots, small lines, or even broken lines.

[0068] The above-mentioned initial vehicle body scratch detection model is a model obtained through learning of training data of a large number of training data sets. It is an initial vehicle body scratch detection model that can identify vehicle images and locate vehicle body scratches.

[0069] Step S202: performing convolution module deduplication processing on the initial vehicle body scratch detection model based on the single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules.

[0070] In this embodiment, the above-mentioned single-stage detector can be understood as a detector that directly predicts the position and category of the target in one detection without the need for a combination of multiple stages. The above-mentioned single-stage detector uses a neural network to simultaneously predict the position and category of the target, which has a faster speed and higher efficiency, such as the YOLO series detector.

[0071] The above convolutional module consists of one or more convolutional layers, each of which contains a series of convolution operations to extract local features of the input data. The purpose of the convolutional module is to extract more abstract, higher-level features from the input feature sequence. Each convolutional layer uses a different filter (or convolution kernel) to scan the input data to generate a new feature map. The convolutional layer extracts local features by sliding a small convolution kernel on the input sequence and integrates these features into a more abstract feature representation.

[0072] The above-mentioned deduplication of convolution modules can be understood as the process of deleting overlapping and repeated convolution modules in the initial body scratch detection model to ensure that each module appears only once, which can eliminate redundant information in the model and improve the efficiency and accuracy of the model. Deduplication is to delete all overlapping and repeated convolution modules of the model, but still retain the overall data connection structure of the model, thereby ensuring that the overall structure of the model is not destroyed, and at the same time it can also reduce the complexity of the model and improve the efficiency of the model.

[0073] The initial vehicle body detection model without repeated convolution modules can be understood as the final model obtained after the convolution module deduplication processing, in which there are no longer any overlapping convolution modules. The initial vehicle body detection model without repeated convolution modules can better capture the characteristics of vehicle body scratches and improve the accuracy of scratch detection.

[0074] It should be noted that by using a single-stage detector to perform convolution module deduplication processing on the initial body scratch detection model, the location and category information of the target can be directly predicted at one time, eliminating the complex candidate area generation and classification regression steps in multi-stage detection, thereby achieving higher detection efficiency and speed.

[0075] Step S203: training the initial vehicle body scratch detection model of the non-overlapping convolution module using the training data set to obtain a trained vehicle body scratch detection model.

[0076] In this embodiment, the training data set includes sample vehicle images and vehicle body scratch annotation data corresponding to the sample vehicle images.

[0077] The above training can be supervised training. Supervised training is a training method in machine learning. Its core is to use a set of data with known labels to train the model, and optimize the model parameters so that the model can predict the labels of new data or make decisions based on the characteristics of existing data. In the training process of supervised learning, first, the data set needs to be divided into a training set and a test set. The training set is used for model training, and the test set is used to evaluate the performance of the model. Then, the features useful to the model, namely the input features, are extracted from the original data. Then, the parameters of the model are adjusted by using the data of the training set through the optimization algorithm so that the model can fit the data well. Finally, the model is verified using the test set data to test the generalization ability of the model. After completing the training and verification, the initial body scratch detection model is obtained.

[0078] The above-trained vehicle body scratch detection model is a model obtained through learning of training data from a large number of training data sets. It is a vehicle body scratch detection model that can accurately identify and accurately locate vehicle body scratches in vehicle images or videos.

[0079] In one embodiment, all overlapping and repeated convolution modules of the model are deleted, and a vehicle body scratch detection model without overlapping convolution modules is trained. The deletion of all overlapping and repeated convolution modules of the model and the training of the vehicle body scratch detection model without overlapping convolution modules are executed cyclically until the verification effect meets the requirements, thereby obtaining a trained vehicle body scratch detection model. This application improves the model's attention to the features of scratches in each convolution module by continuously strengthening the convolutions that need to be paid attention to in the model to enhance the detection capability of subtle scratches. The trained vehicle body scratch detection model can detect vehicle body scratches more accurately and precisely, that is, there are fewer cases of missed detection and wrong detection, which greatly reduces the need for manual omission correction, thereby reducing the workload of manual detection and the required detection time.

[0080] Step S204: Perform body scratch detection on the vehicle image of the vehicle to be detected using the trained body scratch detection model to obtain a body scratch detection result of the vehicle to be detected.

[0081] In this embodiment, after obtaining the vehicle image of the vehicle to be detected, the electronic device may perform body scratch detection on the vehicle image of the vehicle to be detected by using a trained body scratch detection model, thereby obtaining a body scratch detection result of the vehicle to be detected.

[0082] The vehicle image of the vehicle to be detected includes an image of the vehicle body surface or a video of the vehicle body surface. It should be noted that if it is a video of the vehicle body surface, the video is first converted into an image sequence, and then each frame of the image is processed by image recognition to identify the image of the vehicle body to be detected.

[0083] The above-mentioned vehicle body scratch detection can be understood as a detection process of performing feature extraction and analysis on a vehicle image to determine whether there are vehicle body scratches in the image.

[0084] The above-mentioned vehicle body scratch detection results include information such as the scratch location, size, and type.

[0085] In one embodiment, the present application can be applied to the claims process of vehicle damage insurance. The present application can perform a body appearance inspection on the selected car to see if there are scratches, effectively reducing the required inspection time, speeding up the damage identification process, reducing the customer's review waiting time, and improving the claims efficiency of vehicle damage insurance.

[0086] In this embodiment, the present application uses a single-stage detector to focus on the body scratch detection model itself, improves the model's attention to small scratches, and makes the detection of body scratches more accurate and precise, that is, fewer missed detections and false detections occur. This greatly reduces the need for manual leak detection and filling, thereby reducing the workload of manual detection, reducing the required detection time, and improving work efficiency.

[0087] The present application obtains an initial body scratch detection model, performs convolution module deduplication processing on the initial body scratch detection model based on a single-stage detector, obtains an initial body scratch detection model of a non-overlapping convolution module, trains the initial body scratch detection model of the non-overlapping convolution module using a training data set, obtains a trained body scratch detection model, performs body scratch detection on a vehicle image of a to-be-detected vehicle using the trained body scratch detection model, thereby obtaining a body scratch detection result of the to-be-detected vehicle, and solves the problem that the existing detection methods have high difficulty in detecting tiny scratches and low detection accuracy, thereby reducing work efficiency.

[0088] Continue to refer Figure 3 , shown in Figure 2 A flowchart of a specific implementation of step S201 in FIG. 1 . Step S201 includes the following steps:

[0089] Step S2011, obtaining a pre-trained vehicle body scratch detection model and a training data set.

[0090] In this embodiment, the training data set includes sample vehicle images and vehicle body scratch annotation data corresponding to the sample vehicle images. The sample vehicle images include a large number of vehicle image images, and the vehicle images include vehicle body scratch images, vehicle body scratch videos, vehicle body scratch-free images, and vehicle body scratch-free videos. Each vehicle image corresponds to one or more vehicle body scratch annotation data, and each video corresponds to one or more vehicle body scratch annotation data. The vehicle body scratch annotation data may be the location, size, type, etc. of the scratches, and the vehicle body scratch annotation data is used to guide the model to learn how to identify different types of vehicle body scratches.

[0091] The above-mentioned pre-trained vehicle body scratch detection model is a tool for identifying scratches on the vehicle surface. The scratches on the vehicle surface include but are not limited to irregular, tiny dots, tiny lines, or even broken lines.

[0092] It should be noted that the training data set is used to guide model learning and optimize parameters.

[0093] Step S2012: input the sample vehicle image into the pre-trained vehicle body scratch detection model for training, and adjust the parameters of the pre-trained vehicle body scratch detection model according to the vehicle body scratch annotation data corresponding to the sample vehicle image. After the training is completed, an initial vehicle body scratch detection model is obtained.

[0094] In this embodiment, after obtaining the pre-trained vehicle body scratch detection model and the training data set, the electronic device can input the sample vehicle image into the pre-trained vehicle body scratch detection model for training, and adjust the parameters of the pre-trained vehicle body scratch detection model according to the vehicle body scratch annotation data corresponding to the sample vehicle image. After the training is completed, an initial vehicle body scratch detection model is obtained.

[0095] The above training can be supervised training, which is to train the model by providing labeled training data so that the model can predict the label or output value of unknown data.

[0096] During the model training process, the pre-trained body scratch detection model calculates the loss function based on the body scratch annotation data corresponding to the sample vehicle image, and uses the optimization algorithm to update the model parameters to minimize the loss function. The model parameters are optimized based on the minimum loss function. After the training is completed, the initial body scratch detection model is obtained.

[0097] The above-mentioned optimization algorithm can be a gradient descent method, etc.; the above-mentioned loss function can be a cross entropy loss function, a classification loss function, etc.

[0098] The above-mentioned initial vehicle body scratch detection model is a model obtained through learning of training data of a large number of training data sets. It is an initial vehicle body scratch detection model that can identify vehicle images and locate vehicle body scratches.

[0099] It should be noted that by continuously adjusting the model parameters, the model can better fit the training data set, thereby improving the accuracy of vehicle body scratch detection.

[0100] The present application obtains a pre-trained vehicle body scratch detection model and a training data set, inputs a sample vehicle image into the pre-trained vehicle body scratch detection model for training, and adjusts the parameters of the pre-trained vehicle body scratch detection model according to the vehicle body scratch annotation data corresponding to the sample vehicle image. After the training is completed, an initial vehicle body scratch detection model is obtained. By training with the training data set, the model can be better adapted to the needs of specific tasks and the detection accuracy can be improved.

[0101] Continue to refer Figure 4 , showing that in Figure 2 Flow chart of a specific implementation of step S202 in FIG. Step S202 specifically includes the following steps:

[0102] Step S2021: Obtain overlapping convolution modules in the initial vehicle body scratch detection model.

[0103] In this embodiment, the convolution module is composed of one or more convolution layers, each of which includes a series of convolution operations for extracting local features of input data.

[0104] Furthermore, during model training, different convolutional modules may learn different feature representations. In some cases, these convolutional modules may learn similar features or highly correlated feature representations.

[0105] Step S2022: De-duplication processing is performed on overlapping convolution modules based on a single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules.

[0106] In this embodiment, after obtaining the overlapping convolution modules in the initial vehicle body scratch detection model, the electronic device can deduplicate the overlapping convolution modules based on a single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules.

[0107] The above-mentioned single-stage detector can be understood as a detector that directly predicts the location and category of the target in one detection, without the need for a combination of multiple stages, such as the YOLO series of detectors. The single-stage detector can identify convolutional modules with similar or identical features.

[0108] The above deduplication process can be understood as the process of deleting overlapping and repeated convolution modules in the initial car body scratch detection model. Deduplication is to delete all overlapping and repeated convolution modules of the model, but still retain the overall data connection structure of the model, so as to ensure that the overall structure of the model is not destroyed, and at the same time, it can also reduce the complexity of the model and improve the efficiency of the model.

[0109] The above initial vehicle body detection model without repeated convolution modules can better capture the characteristics of vehicle body scratches and improve the accuracy of scratch detection.

[0110] The present application obtains overlapping convolution modules in the initial vehicle body scratch detection model, and deduplicates the overlapping convolution modules based on a single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules. This can better capture the characteristics of vehicle body scratches and improve the accuracy of scratch detection.

[0111] Continue to refer Figure 5 , showing that Figure 4 Flow chart of a specific implementation of step S2022 in FIG. Step S2022 specifically includes the following steps:

[0112] Step S20221: Obtain the initial weight matrix of the overlapping convolution modules in the initial vehicle body scratch detection model.

[0113] In this embodiment, weight is a mathematical expression of input data features during model learning, and determines how data is transferred between layers of the model and how it is transformed.

[0114] It should be noted that a convolution module consists of one or more convolution layers, which are used to extract features of the input image. When there are multiple identical convolution modules in the model, overlapping convolution modules will be generated. These modules may extract similar features on different input regions, so their weight matrices may have similarities. In order to reduce the amount of calculation and improve model performance, the initial weight matrices of these overlapping convolution modules can be obtained.

[0115] Step S20222: Perform weight enhancement processing on the initial weight matrix to obtain a repeated convolution module after weight enhancement processing.

[0116] In this embodiment, the above-mentioned weight strengthening process can be understood as the process of optimizing and adjusting the weight matrix. Specifically, the initial weight matrix can be optimized, and the weight matrix can be updated using optimization algorithms such as gradient descent method and stochastic gradient descent method to minimize the loss function, and the weight parameters can be updated according to the minimum loss function. By updating the weight parameters, the performance of the model can be improved. The purpose of the above-mentioned weight strengthening process is to enable the model to converge better during the training process and improve the performance and generalization ability of the model.

[0117] Step S20223: De-duplication processing is performed on the repeated convolution modules after the weight enhancement processing based on the single-stage detector. After the processing is completed, an initial vehicle body scratch detection model without overlapping convolution modules is obtained.

[0118] In this embodiment, after obtaining the repeated convolution modules after weight enhancement processing, the electronic device can use a single-stage detector to deduplicate the repeated convolution modules after weight enhancement processing. After the processing is completed, an initial vehicle body scratch detection model without overlapping convolution modules is obtained.

[0119] The above single-stage detector can identify convolutional modules with similar or identical features.

[0120] The above deduplication process deletes all overlapping and repeated convolution modules of the model, but still retains the overall data connection structure of the model.

[0121] The present application obtains the initial weight matrix of the overlapping convolution modules in the initial vehicle body scratch detection model, performs weight strengthening processing on the initial weight matrix, obtains the repeated convolution modules after the weight strengthening processing, and uses a single-stage detector to deduplicate the repeated convolution modules after the weight strengthening processing. After the processing is completed, an initial vehicle body scratch detection model without overlapping convolution modules is obtained, which can avoid model overfitting and improve the training speed and generalization ability of the model.

[0122] Optionally, step S203 includes training an initial vehicle body scratch detection model of a non-overlapping convolution module using a training data set. During the training process, the weight parameters of the convolution module are adjusted using an F-norm. After the training is completed, a trained vehicle body scratch detection model is obtained.

[0123] In this embodiment, the training data set includes sample vehicle images and vehicle body scratch annotation data corresponding to the sample vehicle images.

[0124] The above training can be supervised training, which is to train the model by providing labeled training data so that the model can predict the label or output value of unknown data.

[0125] The above-mentioned F-norm (Frobenius-norm) is a matrix norm, which is defined as the sum of the squares of the absolute values ​​of all elements in the matrix A.

[0126] It should be noted that the difference between the initial weight matrix of the convolution module and the trained weight matrix of the convolution module can be measured by using the F-norm, and according to the set threshold, for the weights whose F-norm between the initial weight matrix and the trained weight matrix is ​​higher than the set threshold, their weights are overlapped, which can enhance the model's attention to key points. The purpose of this is to enhance the feature extraction capability of high-weight local areas and make the model's understanding of the features of this step more diverse.

[0127] In this embodiment, the initial vehicle body scratch detection model of the non-overlapping convolution module is trained by using a training data set. During the training process, the weight parameters of the convolution module are adjusted by using the F-norm. After the training is completed, a trained vehicle body scratch detection model is obtained, which can enhance the feature extraction capability of high-weighted local areas during the training process, thereby improving the feature extraction capability and accuracy of the model.

[0128] Optionally, the F-norm is as follows:

[0129]

[0130] Among them, A is the initial weight matrix, B is the trained weight matrix, F is the F-norm of the matrix, m is the number of rows of the matrix, n is the number of columns of the matrix, a is the element corresponding to A, b is the element corresponding to B, a ij is the element in the i-th row and j-th column of matrix A and matrix B, b ij is the element in the i-th row and j-th column of matrix B.

[0131] In this embodiment, the F-norm is the norm distance between matrix A and matrix B, which measures the degree of difference between the two matrices.

[0132] In this embodiment, the above F-norm can be used to enable the model to better fit the data in the training data set, thereby improving the robustness and generalization ability of the model.

[0133] Continue to refer Figure 6 , showing Figure 2 Flow chart of a specific implementation manner of step S204 in FIG. Step S204 specifically includes the following steps:

[0134] Step S2041: Perform feature extraction processing on the vehicle image of the vehicle to be detected by using the trained vehicle body scratch detection model to obtain features of the vehicle to be detected.

[0135] In this embodiment, after obtaining the vehicle image of the vehicle to be detected, the electronic device may perform feature extraction processing on the vehicle image of the vehicle to be detected by using a trained vehicle body scratch detection model, thereby obtaining features of the vehicle to be detected.

[0136] The above-trained body scratch detection model is a body scratch detection model that can accurately identify and accurately locate body scratches in vehicle images or videos.

[0137] The vehicle image of the vehicle to be detected includes an image of the vehicle body surface or a video of the vehicle body surface. It should be noted that if it is a video of the vehicle body surface, the video is first converted into an image sequence, and then each frame of the image is processed by image recognition to identify the image of the vehicle body to be detected.

[0138] The above feature extraction process can be understood as the process of converting the input vehicle image into a set of numerical feature vectors, which can describe the properties and characteristics of the vehicle in the image. Feature extraction usually includes the extraction of features such as color, texture, and shape. Through feature extraction, the image of the vehicle to be detected can be converted into a set of numerical feature vectors, thereby providing input for subsequent scratch detection.

[0139] The above-mentioned features of the vehicle to be detected can reflect information such as whether there are scratches on the vehicle surface and the nature of the scratches.

[0140] Step S2042: Perform body scratch detection on the features of the vehicle to be detected to obtain a body scratch detection result of the vehicle to be detected.

[0141] In this embodiment, after obtaining the characteristics of the vehicle to be detected, the electronic device can perform body scratch detection on the characteristics of the vehicle to be detected, thereby obtaining a body scratch detection result of the vehicle to be detected.

[0142] The above-mentioned vehicle body scratch detection can be understood as a detection process of analyzing the characteristics of the vehicle to be detected to detect whether there are scratches on the vehicle surface.

[0143] The above-mentioned vehicle body scratch detection results include information such as the scratch location, size, and type.

[0144] The present application uses a trained body scratch detection model to perform feature extraction processing on a vehicle image of a vehicle to be detected to obtain features of the vehicle to be detected, and performs body scratch detection on the features of the vehicle to be detected to obtain a body scratch detection result of the vehicle to be detected, thereby improving detection efficiency and accuracy and reducing the time and labor costs required for manual detection.

[0145] In this embodiment, the present application performs deduplication processing on the repeated convolution modules in the model based on a single-stage detector, deletes the original repeated convolution modules and retains the overall data connection structure of the model, and weights the deduplicated convolution modules to increase the feature attention of scratches in the deduplicated convolution modules, so as to detect the scratches on the vehicle body more accurately and precisely, that is, there are fewer missed detections and false detections, which greatly reduces the need for manual leak detection and filling, thereby reducing the workload of manual detection and the required detection time. The present application can be extended to other common damage conditions of automobiles, such as cracking, perforation, etc., which can improve the recall rate and accuracy of detection to a certain extent, and effectively improve the work efficiency of vehicle body scratch detection.

[0146] In this embodiment, the non-Company software tools or components that appear in the embodiments of this application are merely examples and do not represent actual use.

[0147] This embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses 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.

[0148] The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning. This application can be applied to the fields of artificial intelligence and big data, thereby promoting the construction of smart cities.

[0149] This application belongs to the field of smart cities, and this solution can promote the construction of smart cities.

[0150] 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).

[0151] 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.

[0152] Further references Figure 7 , as a response to the above Figure 2 The present application provides an embodiment of a vehicle body scratch detection device, and the vehicle body scratch detection device embodiment is Figure 2 Corresponding to the embodiment of the vehicle body scratch detection method shown, the vehicle body scratch detection device can be specifically applied to various electronic devices.

[0153] like Figure 7 As shown, the vehicle body scratch detection device 700 described in this embodiment includes: an acquisition module 701, a deduplication processing module 702, a training module 703 and a scratch detection module 704. Among them:

[0154] An acquisition module 701 is used to acquire an initial vehicle body scratch detection model, where the initial vehicle body scratch detection model is obtained by training a pre-trained vehicle body scratch detection model using a training data set;

[0155] A deduplication processing module 702 is used to perform convolution module deduplication processing on the initial vehicle body scratch detection model based on a single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules;

[0156] The training module 703 is used to train the initial vehicle body scratch detection model of the non-overlapping convolution module using the training data set to obtain a trained vehicle body scratch detection model.

[0157] The scratch detection module 704 is used to perform body scratch detection on the vehicle image of the vehicle to be detected by using the trained body scratch detection model to obtain a body scratch detection result of the vehicle to be detected.

[0158] In this embodiment, an initial body scratch detection model is obtained, and convolution module deduplication processing is performed on the initial body scratch detection model based on a single-stage detector to obtain an initial body scratch detection model of a non-overlapping convolution module. The initial body scratch detection model of the non-overlapping convolution module is trained using a training data set to obtain a trained body scratch detection model. Body scratch detection is performed on a vehicle image of a vehicle to be inspected using the trained body scratch detection model to obtain a body scratch detection result of the vehicle to be inspected. This solves the problem that the existing detection methods have high difficulty in detecting tiny scratches and low detection accuracy, thereby reducing work efficiency.

[0159] See also Figure 8 ,for Figure 7 The schematic diagram of the structure of a specific implementation of the acquisition module 701 is shown in FIG. 7 , where the acquisition module 701 includes a first acquisition submodule 7011 and a training submodule 7012.

[0160] in:

[0161] A first acquisition submodule 7011 is used to acquire a pre-trained vehicle body scratch detection model and a training data set, wherein the training data set includes a sample vehicle image and vehicle body scratch annotation data corresponding to the sample vehicle image;

[0162] The training submodule 7012 is used to train the pre-trained vehicle body scratch detection model by inputting the sample vehicle image into the pre-trained vehicle body scratch detection model, and adjust the parameters of the pre-trained vehicle body scratch detection model according to the vehicle body scratch annotation data corresponding to the sample vehicle image. After the training is completed, an initial vehicle body scratch detection model is obtained.

[0163] In this embodiment, a pre-trained vehicle body scratch detection model and a training data set are obtained, and a sample vehicle image is input into the pre-trained vehicle body scratch detection model for training. The parameters of the pre-trained vehicle body scratch detection model are adjusted according to the vehicle body scratch annotation data corresponding to the sample vehicle image. After the training is completed, an initial vehicle body scratch detection model is obtained. Through training with the training data set, the model can be better adapted to the needs of specific tasks and the detection accuracy can be improved.

[0164] See also Fig. 9 ,for Figure 7 The deduplication processing module 702 is a schematic diagram of a specific implementation scheme, wherein the deduplication processing module 702 includes a second acquisition submodule 7021 and a deduplication processing submodule 7022.

[0165] The second acquisition submodule 7021 is used to acquire the overlapping convolution modules in the initial vehicle body scratch detection model;

[0166] The deduplication processing submodule 7022 is used to perform deduplication processing on the overlapping convolution modules based on a single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules.

[0167] In this embodiment, overlapping convolution modules in the initial vehicle body scratch detection model are obtained, and the overlapping convolution modules are deduplicated based on a single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules, which can better capture the characteristics of vehicle body scratches and improve the accuracy of scratch detection.

[0168] See also Fig.10 ,for Fig. 9 The deduplication processing submodule 7022 is a schematic diagram of a specific implementation scheme, wherein the deduplication processing submodule 7022 includes an acquisition unit 70221, a weight enhancement processing unit 70222, and a deduplication processing unit 70223.

[0169] An acquisition unit 70221 is used to acquire an initial weight matrix of overlapping convolution modules in an initial vehicle body scratch detection model;

[0170] A weight enhancement processing unit 70222 is used to perform weight enhancement processing on the initial weight matrix to obtain a repeated convolution module after the weight enhancement processing;

[0171] The deduplication processing unit 70223 is used to perform deduplication processing on the repeated convolution modules after the weight enhancement processing based on the single-stage detector. After the processing is completed, an initial vehicle body scratch detection model without overlapping convolution modules is obtained.

[0172] In this embodiment, by obtaining the initial weight matrix of the overlapping convolution modules in the initial vehicle body scratch detection model, and performing weight strengthening processing on the initial weight matrix, the repeated convolution modules after the weight strengthening processing are obtained, and the repeated convolution modules after the weight strengthening processing are deduplicated using a single-stage detector. After the processing is completed, the initial vehicle body scratch detection model without overlapping convolution modules is obtained, which can avoid model overfitting and improve the training speed and generalization ability of the model.

[0173] Optionally, the training module 703 is also used to train the initial vehicle body scratch detection model of the non-overlapping convolution module through the training data set. During the training process, the weight parameters of the convolution module are adjusted through the F-norm. After the training is completed, a trained vehicle body scratch detection model is obtained.

[0174] In this embodiment, the initial vehicle body scratch detection model of the non-overlapping convolution module is trained by using a training data set. During the training process, the weight parameters of the convolution module are adjusted by using the F-norm. After the training is completed, a trained vehicle body scratch detection model is obtained, which can enhance the feature extraction capability of high-weighted local areas during the training process, thereby improving the feature extraction capability and accuracy of the model.

[0175] Optionally, the F-norm is as follows:

[0176]

[0177] Among them, A is the initial weight matrix, B is the trained weight matrix, F is the F-norm of the matrix, m is the number of rows of the matrix, n is the number of columns of the matrix, a is the element corresponding to A, b is the element corresponding to B, a ij is the element in the i-th row and j-th column of matrix A, b ij is the element in the i-th row and j-th column of matrix B.

[0178] In this embodiment, the F-norm is as follows:

[0179]

[0180] Among them, A is the initial weight matrix, B is the trained weight matrix, F is the F-norm of the matrix, m is the number of rows of the matrix, n is the number of columns of the matrix, a is the element corresponding to A, b is the element corresponding to B, a ij is the element in the i-th row and j-th column of matrix A and matrix B, b ij is the element in the i-th row and j-th column of matrix B, which can enable the model to better fit the data in the training dataset, thereby improving the robustness and generalization ability of the model.

[0181] See also Fig.11 ,for Figure 7 The scratch detection module 704 is a schematic diagram of a specific implementation scheme, wherein the scratch detection module 704 includes a feature extraction submodule 7041 and a scratch detection submodule 7042.

[0182] The feature extraction submodule 7041 is used to perform feature extraction processing on the vehicle image of the vehicle to be detected by using the trained vehicle body scratch detection model to obtain the features of the vehicle to be detected;

[0183] The scratch detection submodule 7042 is used to perform body scratch detection on the features of the vehicle to be detected to obtain a body scratch detection result of the vehicle to be detected.

[0184] In this embodiment, feature extraction is performed on the vehicle image of the vehicle to be detected by using a trained body scratch detection model to obtain features of the vehicle to be detected, and body scratch detection is performed on the features of the vehicle to be detected to obtain a body scratch detection result of the vehicle to be detected, thereby improving detection efficiency and accuracy and reducing the time and labor costs required for manual detection.

[0185] To solve the above technical problems, the present application also provides a computer device. Fig.12 , Fig.12 This is a basic structural block diagram of the computer device in this embodiment.

[0186] The computer device 12 includes a memory 121, a processor 122, and a network interface 123 that are interconnected and communicated through a system bus. It should be noted that the figure only shows a computer device 12 with components 121-123, 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 (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0187] 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.

[0188] The memory 121 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.

[0189] In some embodiments, the memory 121 may be an internal storage unit of the computer device 12, such as a hard disk or memory of the computer device 12. In other embodiments, the memory 121 may also be an external storage device of the computer device 12, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 12. Of course, the memory 121 may also include both the internal storage unit of the computer device 12 and its external storage device. In this embodiment, the memory 121 is generally used to store the operating system and various application software installed on the computer device 12, such as computer-readable instructions of the vehicle body scratch detection method, etc. In addition, the memory 121 may also be used to temporarily store various types of data that have been output or are to be output. The processor 122 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 122 is generally used to control the overall operation of the computer device 12. In this embodiment, the processor 122 is used to run computer-readable instructions or process data stored in the memory 121, such as computer-readable instructions for running the vehicle body scratch detection method.

[0190] The network interface 123 may include a wireless network interface or a wired network interface. The network interface 123 is generally used to establish a communication connection between the computer device 12 and other electronic devices.

[0191] In this embodiment, the impact data of the system in the knowledge base can be used to generate a car body scratch detection affected by the target system to be tested for the user, so that the user can intuitively obtain the impact range of the target system to be tested. According to the impact range of the target system to be tested, the impact of the change can be evaluated more accurately and quickly, irrelevant interference points can be eliminated, and regression use cases can be streamlined to avoid missing impact points in the early stage of the evaluation, resulting in the subsequent development of various systems. When they are jointly debugged, it is found that some associations have not been processed. In the mildest case, it will take up development time to supplement the logic. In the most serious case, the solution design may need to be overturned and redone, thereby improving the efficiency of project development.

[0192] The present application also provides another embodiment, namely, 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 at least one processor to enable the at least one processor to perform the steps of the vehicle body scratch detection method as described above.

[0193] In this embodiment, the impact data of the system in the knowledge base can be used to generate a car body scratch detection affected by the target system to be tested for the user, so that the user can intuitively obtain the impact range of the target system to be tested. According to the impact range of the target system to be tested, the impact of the change can be evaluated more accurately and quickly, irrelevant interference points can be eliminated, and regression use cases can be streamlined to avoid missing impact points in the early stage of the evaluation, resulting in the subsequent development of various systems. When they are jointly debugged, it is found that some associations have not been processed. In the mildest case, it will take up development time to supplement the logic. In the most serious case, the solution design may need to be overturned and redone, thereby improving the efficiency of project development.

[0194] 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.

[0195] 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. A method for detecting scratches on a vehicle body, characterized in that: The steps include: Acquire an initial vehicle body scratch detection model, wherein the initial vehicle body scratch detection model is obtained by training a pre-trained vehicle body scratch detection model using a training data set; Based on the single-stage detector, the initial vehicle body scratch detection model is subjected to convolution module deduplication processing to obtain an initial vehicle body scratch detection model without overlapping convolution modules; The initial vehicle body scratch detection model of the non-overlapping convolution module is trained by using the training data set to obtain a trained vehicle body scratch detection model; The vehicle body scratch detection is performed on the vehicle image of the vehicle to be detected by using the trained vehicle body scratch detection model to obtain the vehicle body scratch detection result of the vehicle to be detected.

2. The vehicle body scratch detection method according to claim 1, characterized in that: The obtaining of the initial vehicle body scratch detection model comprises: Obtaining a pre-trained vehicle body scratch detection model and a training data set, wherein the training data set includes a sample vehicle image and vehicle body scratch annotation data corresponding to the sample vehicle image; The sample vehicle image is input into the pre-trained vehicle body scratch detection model for training, and the parameters of the pre-trained vehicle body scratch detection model are adjusted according to the vehicle body scratch annotation data corresponding to the sample vehicle image. After the training is completed, an initial vehicle body scratch detection model is obtained.

3. The vehicle body scratch detection method according to claim 2, characterized in that: The method of performing convolution module deduplication processing on the initial vehicle body scratch detection model based on the single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules includes: Get the overlapping convolution modules in the initial car body scratch detection model; The overlapping convolution modules are deduplicated based on a single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules.

4. The vehicle body scratch detection method according to claim 3, characterized in that: The single-stage detector is used to perform deduplication processing on the overlapping convolution modules to obtain an initial vehicle body scratch detection model without overlapping convolution modules: Get the initial weight matrix of the overlapping convolution modules in the initial car body scratch detection model; Performing weight enhancement processing on the initial weight matrix to obtain a repeated convolution module after the weight enhancement processing; Based on the single-stage detector, deduplication processing is performed on the repeated convolution modules after the weight enhancement processing, and after the processing is completed, an initial vehicle body scratch detection model without overlapping convolution modules is obtained.

5. The vehicle body scratch detection method according to claim 4, characterized in that: The initial vehicle body scratch detection model of the non-overlapping convolution module is trained by the training data set to obtain a trained vehicle body scratch detection model, including: The initial vehicle body scratch detection model of the non-overlapping convolution module is trained using the training data set. During the training process, the weight parameters of the convolution module are adjusted using the F-norm. After the training is completed, a trained vehicle body scratch detection model is obtained.

6. The vehicle body scratch detection method according to claim 5, characterized in that: The F-norm is as follows: Among them, A is the initial weight matrix, B is the trained weight matrix, F is the F-norm of the matrix, m is the number of rows of the matrix, n is the number of columns of the matrix, a is the element corresponding to A, b is the element corresponding to B, a ih is the element in the i-th row and j-th column of matrix A, b ij is the element in the i-th row and j-th column of matrix B.

7. The vehicle body scratch detection method according to claim 6, characterized in that: The method of performing body scratch detection on the vehicle image of the vehicle to be detected by using the trained body scratch detection model to obtain the body scratch detection result of the vehicle to be detected includes: The trained vehicle body scratch detection model is used to perform feature extraction processing on the vehicle image of the vehicle to be detected, thereby obtaining the features of the vehicle to be detected; Perform body scratch detection on the features of the vehicle to be detected to obtain a body scratch detection result of the vehicle to be detected.

8. A vehicle body scratch detection device, characterized in that: include: An acquisition module is used to acquire an initial vehicle body scratch detection model, wherein the initial vehicle body scratch detection model is obtained by training a pre-trained vehicle body scratch detection model using a training data set, wherein the training data set includes a sample vehicle image and vehicle body scratch annotation data corresponding to the sample vehicle image; A deduplication processing module, used for performing convolution module deduplication processing on the initial vehicle body scratch detection model based on a single-stage detector to obtain an initial vehicle body scratch detection model without overlapping convolution modules; A training module, used for training the initial vehicle body scratch detection model of the non-overlapping convolution module using the training data set to obtain a trained vehicle body scratch detection model; The scratch detection module is used to perform body scratch detection on the vehicle image of the vehicle to be detected by using a trained body scratch detection model to obtain a body scratch detection result of the vehicle to be detected.

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 vehicle body scratch detection method as claimed in 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 the processor, the steps of the vehicle body scratch detection method according to any one of claims 1 to 7 are implemented.

Images