Vehicle damage detection method, device and equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence and the field of financial science and technology, and discloses a vehicle damage detection method, device and equipment and a storage medium, and the method comprises the steps: obtaining a to-be-predicted vehicle damage image, and carrying out the image enhancement of the to-be-predicted vehicle damage image, and obtaining a standard vehicle damage image; extracting an initial image feature of the standard vehicle damage image, and performing multilayer deformable convolution on the initial image feature to obtain a layered convolution feature; performing feature fusion on the layered convolution features to obtain target features of the standard vehicle damage image; and calculating the vehicle damage degree corresponding to the standard vehicle damage image by using the target feature. According to the invention, the accuracy and efficiency of vehicle damage detection can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence technology and financial technology, and in particular to a vehicle damage detection method, device, electronic equipment and computer-readable storage medium. Background Art

[0002] Vehicle damage detection is a key link in many fields such as automobile insurance, accident handling, and vehicle maintenance. Traditional vehicle damage detection methods often rely on manual inspection, which is not only time-consuming and labor-intensive, but may also lead to inaccurate detection results due to human factors. With the continuous advancement of computer vision technology, automatic vehicle damage detection methods based on convolutional networks have gradually become a research hotspot.

[0003] The existing technology for vehicle damage detection mainly focuses on the architecture of convolutional neural networks and Transformer. These architectures detect vehicle damage by introducing residual connections, multi-scale feature extraction, and dense connections. However, this method is at a disadvantage in terms of long-range dependence and adaptive spatial aggregation. The convolution kernel does not have long-range dependence and adaptive spatial aggregation capabilities, the robustness of the features is also poor, and the detection accuracy is not high enough. Therefore, how to improve the accuracy of vehicle damage detection has become an urgent problem to be solved. Summary of the invention

[0004] The present invention provides a vehicle damage detection method, device and computer-readable storage medium, the main purpose of which is to solve the problems of insufficient accuracy and low detection efficiency of vehicle damage detection.

[0005] In a first aspect, to achieve the above-mentioned purpose, the present invention provides a vehicle damage detection method, comprising:

[0006] Acquire a vehicle damage image to be predicted, and perform image enhancement on the vehicle damage image to be predicted to obtain a standard vehicle damage image;

[0007] Extracting initial image features of the standard vehicle damage image, and performing multi-layer deformable convolution on the initial image features to obtain layered convolution features;

[0008] Performing feature fusion on the layered convolution features to obtain target features of the standard vehicle damage image;

[0009] The target feature is used to calculate the vehicle damage degree corresponding to the standard vehicle damage image.

[0010] In a second aspect, the present invention further provides a vehicle damage detection device, the device comprising:

[0011] An image enhancement module is used to obtain a vehicle damage image to be predicted, and perform image enhancement on the vehicle damage image to be predicted to obtain a standard vehicle damage image;

[0012] A deformable convolution module, used to extract initial image features of the standard vehicle damage image, and perform multi-layer deformable convolution on the initial image features to obtain layered convolution features;

[0013] A feature fusion module, used for performing feature fusion on the layered convolution features to obtain target features of the standard vehicle damage image;

[0014] The vehicle damage degree assessment module is used to calculate the vehicle damage degree corresponding to the standard vehicle damage image using the target feature.

[0015] In a third aspect, the present invention further provides an electronic device, the electronic device comprising:

[0016] at least one processor; and,

[0017] a memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the vehicle damage detection method described above.

[0019] In a fourth aspect, the present invention further provides a computer-readable storage medium, wherein at least one computer program is stored in the computer-readable storage medium, and the at least one computer program is executed by a processor in an electronic device to implement the above-mentioned vehicle damage detection method.

[0020] The present invention uses road monitoring or a vehicle driving recorder to take a vehicle damage image in real time as the vehicle damage image to be predicted, thereby ensuring the availability and timeliness of the obtained vehicle damage image, and performing image enhancement on the vehicle damage image to be predicted, thereby removing noise and interference in the vehicle damage image to be predicted, making the image clearer and sharper, thereby obtaining a standard vehicle damage image; extracting the initial image features of the standard vehicle damage image, thereby more accurately identifying and distinguishing the type and location of vehicle damage, performing multi-layer deformable convolution on the initial image features, fusing features at different levels, thereby making the model more robust in the face of different deformations and scale changes of the image, thereby improving the accuracy of damage detection; using convolution operations to perform feature fusion on layered convolution features, thereby capturing the spatial or temporal relationship between features, thereby improving the recognition and classification capabilities of vehicle damage images; performing convolution, pooling and activation processing on the target features, thereby converting the degree of damage into a specific numerical value or level, thereby making the prediction result more objective and comparable, thereby improving the efficiency of vehicle damage detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0022] Figure 1 is a schematic diagram of an application environment of a vehicle damage detection method in one embodiment of the present invention;

[0023] Figure 2 A schematic diagram of a process flow of a vehicle damage detection method provided by an embodiment of the present invention;

[0024] Figure 3 A schematic diagram of a module of a vehicle damage detection device provided by an embodiment of the present invention;

[0025] Figure 4 A schematic diagram of the structure of an electronic device of a vehicle damage detection device provided by an embodiment of the present invention;

[0026] Figure 5 It is another structural schematic diagram of an electronic device of a vehicle damage detection device provided by an embodiment of the present invention.

[0027] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the technical solution of the present disclosure, and to fully understand and implement how the present disclosure applies technical means to solve technical problems and achieve the corresponding technical effects, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only embodiments of a part of the present disclosure, not all of the embodiments. The embodiments of the present disclosure and the various features in the embodiments can be combined with each other without conflict, and the technical solutions formed are all within the scope of protection of the present disclosure. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present disclosure.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0030] An embodiment of the present application provides a vehicle damage detection method, and the execution subject of the vehicle damage detection method includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the device provided in the embodiment of the present application. In other words, the vehicle damage detection method can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0031] The vehicle damage detection method provided by the embodiment of the present invention can be applied in Figure 1In an application environment, the client communicates with the server through a network. The server can obtain the to-be-predicted vehicle damage image of vehicle damage detection through the client, perform image enhancement on the to-be-predicted vehicle damage image to obtain a standard vehicle damage image; extract the initial image features of the standard vehicle damage image, perform multi-layer deformable convolution on the initial image features to obtain layered convolution features; perform feature fusion on the layered convolution features to obtain target features of the standard vehicle damage image; use the target features to calculate the vehicle damage degree corresponding to the standard vehicle damage image, and feed back the vehicle damage detection result to the client. In the present invention, for vehicle damage detection, the present invention uses road monitoring or vehicle driving recorders to take vehicle damage images in real time as the vehicle damage images to be predicted, ensuring the availability and timeliness of the obtained vehicle damage images, performing image enhancement on the vehicle damage images to be predicted, removing noise and interference in the vehicle damage images to be predicted, making the images clearer and sharper, thereby obtaining standard vehicle damage images; extracting the initial image features of the standard vehicle damage images, more accurately identifying and distinguishing the types and locations of vehicle damage, performing multi-layer deformable convolution on the initial image features, fusing features at different levels, making the model more robust in the face of different deformations and scale changes of the image, thereby improving the accuracy of damage detection; using convolution operations to perform feature fusion on layered convolution features, it is possible to capture the spatial or temporal relationship between features, thereby improving the recognition and classification capabilities of vehicle damage images; performing convolution, pooling and activation processing on the target features, the degree of damage can be converted into specific values ​​or levels, making the prediction results more objective and comparable, and improving the efficiency of vehicle damage detection. Among them, the client can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers and portable wearable devices. The server side can be implemented by an independent server or a server cluster composed of multiple servers. The present invention is described in detail below through specific embodiments.

[0032] Reference Figure 2 , which is a flow chart of a vehicle damage detection method provided by an embodiment of the present invention.

[0033] In this embodiment, the vehicle damage detection method includes:

[0034] S1. Acquire a vehicle damage image to be predicted, and perform image enhancement on the vehicle damage image to be predicted to obtain a standard vehicle damage image.

[0035] In an embodiment of the present invention, the vehicle damage image to be predicted refers to image data formed by recording the damaged parts of the vehicle by means of photography or video recording; the vehicle damage image is an important basis for vehicle damage assessment and claims settlement, and can intuitively reflect the damage to the vehicle and provide accurate damage assessment information for insurance companies and repair shops; the image enhancement refers to an important technology in the field of digital image processing, which is a technology for improving its quality, readability and visualization; the present invention aims to improve the visual effect of the vehicle damage image to facilitate subsequent analysis and processing, and the image enhancement includes normalization processing, contrast stretching, Fourier transform, and low-pass filtering.

[0036] In an embodiment of the present invention, the vehicle damage image to be predicted can be captured in real time by road monitoring or a vehicle driving recorder, and the road monitoring and vehicle driving recorder can be remotely accessed in a preset video monitoring system to obtain the vehicle damage image.

[0037] In detail, the vehicle damage image can show the overall appearance of the accident vehicle, including information such as the license plate number, vehicle model, color, and the approximate situation of the vehicle damage, ensuring the accuracy and availability of the vehicle damage image for preliminary identification and assessment of vehicle damage.

[0038] In the embodiment of the present invention, the step of performing image enhancement on the vehicle damage image to be predicted to obtain a standard vehicle damage image includes:

[0039] Normalizing the vehicle damage image to be predicted to obtain a normalized vehicle damage image;

[0040] Performing image denoising on the normalized vehicle damage image to obtain a pollution-free vehicle damage image;

[0041] The pollution-free vehicle damage image is subjected to image adjustment to obtain a standard vehicle damage image.

[0042] In an embodiment of the present invention, the normalization processing refers to scaling the pixel values ​​corresponding to the vehicle damage image to be predicted to a range of 0 to 1 to eliminate the influence of image differences; the image denoising refers to improving the quality of the image by removing noise in the image and retaining useful image information.

[0043] In detail, the vehicle damage image to be predicted is normalized to obtain a normalized vehicle damage image and the image is denoised. The noise of the image is reduced by smoothing filtering, i.e., averaging the grayscale values ​​of adjacent pixels through a mean filter, to obtain an unpolluted vehicle damage image. The unpolluted vehicle damage image is adjusted, i.e., the corresponding enhancement parameters are adjusted. For example, the contrast of the vehicle damage image is adjusted to make the contrast between the damaged part and the surrounding area more obvious; the edge of the vehicle damage image is enhanced, i.e., sharpened, to make the damage outline clearer; the color of the vehicle damage image is adjusted to ensure that the color of the damaged part is truly reflected.

[0044] In the embodiment of the present invention, the step of normalizing the vehicle damage image to be predicted to obtain a normalized vehicle damage image includes:

[0045] Calculate the absolute value of the first difference between the minimum value and the maximum value of the pixel value of the vehicle damage image to be predicted respectively;

[0046] Respectively calculating the second absolute value of the difference between each pixel value and the minimum value in the vehicle damage image to be predicted, and dividing the first absolute value of the difference by the second absolute value of the difference to obtain a normalized value;

[0047] A normalized vehicle damage image is generated using the normalized value corresponding to each pixel value.

[0048] Specifically, by calculating the minimum and maximum values ​​in the vehicle damage image to be predicted, each pixel value is linearly transformed and scaled to a specified range, usually [0,1] or [-1,1]. The normalized value is (original value - minimum value) / (maximum value - minimum value). For example, in a vehicle damage image, if the original pixel value range is [0,255], then after minimum-maximum normalization, the pixel value will be scaled to the interval [0,1].

[0049] In detail, the present invention can eliminate the dimension effect through normalization processing, that is, different vehicle damage images may have different characteristics, and they are scaled to a smaller interval at a certain ratio, so that the difference between each pixel is more obvious, making different vehicle damage images contrastive.

[0050] In an embodiment of the present invention, a vehicle damage image is captured in real time by road monitoring or a vehicle driving recorder as the vehicle damage image to be predicted, thereby ensuring the availability and timeliness of the obtained vehicle damage image. Image enhancement is performed on the vehicle damage image to be predicted, and noise and interference in the vehicle damage image to be predicted can be removed to make the image clearer and sharper, thereby obtaining a standard vehicle damage image.

[0051] S2. Extracting initial image features of the standard vehicle damage image, and performing multi-layer deformable convolution on the initial image features to obtain layered convolution features.

[0052] In an embodiment of the present invention, the initial image features refer to image features including color features, texture features, shape features, etc. of the vehicle damage image; the multi-layer deformable convolution refers to stacking the convolution layers in the convolutional neural network according to a preset stacking rule to obtain a multi-layer stacked convolutional neural network, and performing hierarchical feature extraction using the initial image features of the multi-layer stacked convolutional neural network to obtain hierarchical convolution features.

[0053] In the embodiment of the present invention, the step of extracting the initial image features of the standard vehicle damage image includes:

[0054] Get the pre-built Initial Hybrid Convolutional Model;

[0055] Based on the multiple convolution kernels of the initial mixed convolution model, dividing the initial mixed convolution model into multiple mixed convolution models;

[0056] Based on the hybrid convolution model, convolution processing is performed on the standard vehicle damage image to obtain initial image features.

[0057] The present invention can divide the initial mixed convolution model into multiple mixed convolution models with convolution kernels of different scales. Specifically, multiple parallel convolution kernel branches can be designed in a single convolution layer, each branch corresponding to a convolution kernel of a different size, and the initial mixed convolution model can be divided into multiple mixed convolution models. Each mixed convolution model performs a convolution operation on the vehicle damage image to obtain the initial image features.

[0058] Furthermore, by extracting the initial image features of the standard vehicle damage image, the type and location of vehicle damage can be more accurately identified and distinguished, thereby improving the accuracy of damage detection; image feature extraction can reduce the dimension of image data, remove redundant information, and make subsequent image processing and analysis more efficient.

[0059] In the embodiment of the present invention, the convolution processing is performed on the standard vehicle damage image based on the hybrid convolution model to obtain the initial image features, including:

[0060] Based on the one-dimensional convolution sub-model in the hybrid convolution model, the standard vehicle damage image is convolved in the time dimension to obtain image features after time convolution;

[0061] Based on the two-dimensional convolution sub-model in the hybrid convolution model, convolution processing is performed on the standard vehicle damage image in the spatial dimension to obtain image features after spatial convolution;

[0062] The image features after the temporal convolution and the image features after the spatial convolution are used as initial image features.

[0063] In detail, the initial hybrid convolution model includes two parts, namely a one-dimensional convolution sub-model and a two-dimensional convolution sub-model. For example, the initial hybrid convolution model may include a one-dimensional convolution sub-model in the time dimension, the convolution kernel size of the one-dimensional convolution sub-model is 3, and a two-dimensional convolution sub-model in the spatial dimension, the convolution kernel size of the two-dimensional convolution sub-model is 3x3. The two sub-models can realize the extraction of time features and avoid expensive calculations.

[0064] In the embodiment of the present invention, performing multi-layer deformable convolution on the initial image features to obtain layered convolution features includes:

[0065] According to the preset stacking rules, the parameters of different convolutional layers in the pre-built hybrid convolutional model are stacked to obtain convolutional layers with different parameters;

[0066] Performing convolution processing on the initial image features according to the convolution layers with different parameters to obtain image features corresponding to the convolution layers with different parameters;

[0067] The image features corresponding to the convolutional layers with different parameters are concatenated to obtain hierarchical convolutional features.

[0068] In detail, the stacking rule refers to stacking the parameters of different convolutional layers in the hybrid convolutional model, that is, the number of channels of the latter layer is determined by the number of channels of the previous layer, and the stacking mode is simplified to "AABA", that is, the number of channels of stages 1, 2 and 4 is the same and not greater than the number of channels of stage 3.

[0069] The present invention can adapt to the deformation of the vehicle in the image through deformable convolution, capture local features more accurately, and locate the damaged area more accurately through multi-layer convolution, thereby improving the positioning accuracy of damage detection.

[0070] In the embodiment of the present invention, the step of splicing the image features corresponding to the convolution layers with different parameters to obtain the layered convolution features includes:

[0071] According to the preset feature selection strategy, the image features to be retained are selected from the image features corresponding to the convolution layers with different parameters to obtain sub-features;

[0072] Based on a preset splicing order, the sub-features are spliced ​​to obtain a spliced ​​image feature, namely a hierarchical convolution feature.

[0073] In detail, the sub-features include image features corresponding to the initial vehicle damage image, including color features, texture features, shape features, etc. The image features to be retained are selected from the image features corresponding to the convolution layers with different parameters according to the feature selection strategy. The feature selection strategy performs feature selection based on the weight distribution of the convolution kernels in the convolution layers with different parameters. A larger weight indicates that the convolution kernel contributes more to the output of the image features, thereby selecting the corresponding image features as sub-features. The selected output features are spliced ​​along the channel dimension using the splicing operation in the deep learning framework to obtain the image features of each layer, namely, the hierarchical convolution features.

[0074] In an embodiment of the present invention, by extracting the initial image features of the standard vehicle damage image, the type and location of vehicle damage can be more accurately identified and distinguished. Multi-layer deformable convolution is performed on the initial image features to fuse features at different levels. This can make the model more robust when facing different deformations and scale changes of the image, thereby improving the accuracy of damage detection.

[0075] S3. Perform feature fusion on the layered convolution features to obtain target features of the standard vehicle damage image.

[0076] In an embodiment of the present invention, the feature fusion refers to merging or integrating different types of features to improve the performance and accuracy of the model. The feature fusion is an important method in the field of pattern recognition and data analysis, and is particularly widely used in fields such as computer vision and machine learning. Feature fusion can comprehensively utilize multiple features to achieve complementary advantages of multiple features, thereby obtaining more robust and accurate recognition or prediction results. In tasks such as image recognition and target detection, feature fusion can significantly improve the performance and generalization ability of the model.

[0077] In the implementation of the present invention, the step of fusing the layered convolution features to obtain the target features of the standard vehicle damage image includes:

[0078] The layered convolution feature is traversed in a sliding window manner through a pre-constructed convolution kernel to obtain a pixel value corresponding to the layered convolution feature;

[0079] Multiplying the pixel value in the convolution kernel by the pixel value corresponding to the layered convolution feature and summing them up to obtain the pixel value corresponding to the target feature;

[0080] The pixel values ​​corresponding to the target features are mapped to a preset target feature map to obtain the target features.

[0081] In detail, the convolution kernel is used to perform convolution operations with the matrix of the layered convolution features. The parameters such as the size, element value and step size of the convolution kernel will affect the result of the convolution operation. When designing the convolution kernel, it is usually adjusted according to the characteristics of the specific task and data set to extract the most effective features.

[0082] Furthermore, the sliding window traversal is a way to implement the convolution operation. The convolution kernel moves on the layered convolution feature like a sliding window. Each movement will be multiplied and summed with the feature value in the window to obtain a new pixel value. This new pixel value corresponds to a point on the target feature map. Finally, all pixel values ​​obtained by the convolution operation are mapped to the preset target feature map according to the position on the original feature map to obtain the target feature.

[0083] In an embodiment of the present invention, a convolution operation is used to perform feature fusion on layered convolution features, so that a richer and more semantic feature representation can be captured. By fusing features at different levels, the spatial or temporal relationship between features can be captured, thereby improving the recognition and classification capabilities of vehicle damage images.

[0084] S4. Calculate the vehicle damage degree corresponding to the standard vehicle damage image using the target feature.

[0085] In the embodiment of the present invention, the process of using the target feature to calculate the vehicle damage degree corresponding to the standard vehicle damage image includes a complex process of target feature preparation, classifier design and training, target feature classification, and result evaluation. In this process, it is necessary to comprehensively use a variety of technical means such as image processing technology, machine learning algorithms, and data analysis methods to achieve a rapid and accurate assessment of the vehicle damage degree.

[0086] In the embodiment of the present invention, the step of calculating the vehicle damage degree corresponding to the standard vehicle damage image by using the target feature includes:

[0087] Pooling and activating the target features to obtain processed target features;

[0088] The processed target features are fully connected to obtain the vehicle damage degree corresponding to the standard vehicle damage image.

[0089] In detail, the present invention classifies the degree of vehicle damage by convolving, pooling and activating the target features to obtain processed target features. The pooling process can reduce the spatial size (i.e., height and width) of the image features, thereby reducing the amount of calculation and the number of parameters, while extracting important features and introducing nonlinear factors through the activation function to obtain more accurate image features.

[0090] Furthermore, the pooling process is usually followed by the convolution process to further reduce the spatial size of the data (for text processing, it can be understood as reducing the dimension of the feature vector), thereby reducing the amount of calculation and avoiding overfitting; according to the pooling process, by retaining important features and discarding redundant information, a higher-level feature representation is obtained, which is more robust and effective for subsequent parsing tasks.

[0091] Furthermore, after convolution and pooling, the activation processing usually applies a nonlinear activation function (such as ReLU) to increase the nonlinear ability of the data and enable learning of more complex patterns; the ReLU activation function can set all negative values ​​to 0 and keep positive values ​​unchanged. The ReLU activation function not only speeds up the calculation process (because it does not involve complex exponential operations), but also alleviates the gradient vanishing problem and helps data convergence.

[0092] In the embodiment of the present invention, the full connection can map the target feature to the feature space corresponding to the degree of vehicle damage, and convert it into a feature vector or value that can reflect the degree of vehicle damage.

[0093] In an embodiment of the present invention, the target features are convolved, pooled, and activated to make the prediction results more objective and comparable; the entire prediction process can be highly automated without the need to manually check each image one by one, especially when processing a large number of vehicle damage images, which can significantly shorten the evaluation cycle and improve vehicle damage detection efficiency.

[0094] The present invention uses road monitoring or a vehicle driving recorder to take a vehicle damage image in real time as the vehicle damage image to be predicted, thereby ensuring the availability and timeliness of the obtained vehicle damage image, and performing image enhancement on the vehicle damage image to be predicted, thereby removing noise and interference in the vehicle damage image to be predicted, making the image clearer and sharper, thereby obtaining a standard vehicle damage image; extracting the initial image features of the standard vehicle damage image, thereby more accurately identifying and distinguishing the type and location of vehicle damage, performing multi-layer deformable convolution on the initial image features, fusing features at different levels, thereby making the model more robust in the face of different deformations and scale changes of the image, thereby improving the accuracy of damage detection; using convolution operations to perform feature fusion on layered convolution features, thereby capturing the spatial or temporal relationship between features, thereby improving the recognition and classification capabilities of vehicle damage images; performing convolution, pooling and activation processing on the target features, thereby converting the degree of damage into a specific numerical value or level, thereby making the prediction result more objective and comparable, thereby improving the efficiency of vehicle damage detection.

[0095] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0096] In an embodiment of the present disclosure, a vehicle damage detection device is provided, which corresponds one-to-one to the vehicle damage detection method in the above embodiment. Figure 3 As shown, the vehicle damage detection device 100 can be installed in an electronic device. According to the functions to be implemented, the vehicle damage detection device 100 includes an image enhancement module 101, a deformable convolution module 102, a feature fusion module 103, and a vehicle damage assessment module 104. The functional modules are described in detail as follows:

[0097] The image enhancement module 101 is used to obtain a vehicle damage image to be predicted, and perform image enhancement on the vehicle damage image to be predicted to obtain a standard vehicle damage image;

[0098] The deformable convolution module 102 is used to extract the initial image features of the standard vehicle damage image, and perform multi-layer deformable convolution on the initial image features to obtain layered convolution features;

[0099] A feature fusion module 103 is used to perform feature fusion on the layered convolution features to obtain target features of the standard vehicle damage image;

[0100] The vehicle damage degree assessment module 104 is used to calculate the vehicle damage degree corresponding to the standard vehicle damage image using the target feature.

[0101] In one embodiment, the image enhancement module 101 is specifically configured to:

[0102] Normalizing the vehicle damage image to be predicted to obtain a normalized vehicle damage image;

[0103] Performing image denoising on the normalized vehicle damage image to obtain a pollution-free vehicle damage image;

[0104] The pollution-free vehicle damage image is subjected to image adjustment to obtain a standard vehicle damage image.

[0105] In one embodiment, the image enhancement module 101 is specifically configured to:

[0106] The step of normalizing the vehicle damage image to be predicted to obtain a normalized vehicle damage image includes:

[0107] Calculate the absolute value of the first difference between the minimum value and the maximum value of the pixel value of the vehicle damage image to be predicted respectively;

[0108] Respectively calculating the second absolute value of the difference between each pixel value and the minimum value in the vehicle damage image to be predicted, and dividing the first absolute value of the difference by the second absolute value of the difference to obtain a normalized value;

[0109] A normalized vehicle damage image is generated using the normalized value corresponding to each pixel value.

[0110] In one embodiment, the deformable convolution module 102 is specifically configured to:

[0111] Get the pre-built Initial Hybrid Convolutional Model;

[0112] Based on the multiple convolution kernels of the initial mixed convolution model, dividing the initial mixed convolution model into multiple mixed convolution models;

[0113] Based on the hybrid convolution model, convolution processing is performed on the standard vehicle damage image to obtain initial image features.

[0114] In one embodiment, the deformable convolution module 102 is specifically configured to:

[0115] The method of performing convolution processing on the standard vehicle damage image based on the hybrid convolution model to obtain initial image features includes:

[0116] Based on the one-dimensional convolution sub-model in the hybrid convolution model, the standard vehicle damage image is convolved in the time dimension to obtain image features after time convolution;

[0117] Based on the two-dimensional convolution sub-model in the hybrid convolution model, convolution processing is performed on the standard vehicle damage image in the spatial dimension to obtain image features after spatial convolution;

[0118] The image features after the temporal convolution and the image features after the spatial convolution are used as initial image features.

[0119] In one embodiment, the deformable convolution module 102 is specifically configured to:

[0120] According to the preset stacking rules, the parameters of different convolutional layers in the pre-built hybrid convolutional model are stacked to obtain convolutional layers with different parameters;

[0121] Performing convolution processing on the initial image features according to the convolution layers with different parameters to obtain image features corresponding to the convolution layers with different parameters;

[0122] The image features corresponding to the convolutional layers with different parameters are concatenated to obtain hierarchical convolutional features.

[0123] In one embodiment, the deformable convolution module 102 is specifically configured to:

[0124] According to the preset feature selection strategy, the image features to be retained are selected from the image features corresponding to the convolution layers with different parameters to obtain sub-features;

[0125] Based on a preset splicing order, the sub-features are spliced ​​to obtain a spliced ​​image feature, namely a hierarchical convolution feature.

[0126] In one embodiment, the feature fusion module 103 is specifically used to:

[0127] The step of fusing the layered convolution features to obtain target features of the standard vehicle damage image includes:

[0128] The layered convolution feature is traversed in a sliding window manner through a pre-constructed convolution kernel to obtain a pixel value corresponding to the layered convolution feature;

[0129] Multiplying the pixel value in the convolution kernel by the pixel value corresponding to the layered convolution feature and summing them up to obtain the pixel value corresponding to the target feature;

[0130] The pixel values ​​corresponding to the target features are mapped to a preset target feature map to obtain the target features.

[0131] In one embodiment, the vehicle damage assessment module 104 is specifically configured to:

[0132] Pooling and activating the target features to obtain processed target features;

[0133] The processed target features are fully connected to obtain the vehicle damage degree corresponding to the standard vehicle damage image.

[0134] In the present invention, for vehicle damage detection, a vehicle damage image is first captured in real time through road monitoring or a vehicle driving recorder as a vehicle damage image to be predicted, ensuring the availability and timeliness of the obtained vehicle damage image, and performing image enhancement on the vehicle damage image to be predicted, which can remove noise and interference in the vehicle damage image to be predicted, making the image clearer and sharper, thereby obtaining a standard vehicle damage image; extracting the initial image features of the standard vehicle damage image can more accurately identify and distinguish the type and location of vehicle damage, and performing multi-layer deformable convolution on the initial image features, fusing features at different levels, which can make the model more robust when facing different deformations and scale changes of the image, thereby improving the accuracy of damage detection; using convolution operations to perform feature fusion on layered convolution features, it can capture the spatial or temporal relationship between features, thereby improving the recognition and classification capabilities of vehicle damage images; performing convolution, pooling and activation processing on the target features can convert the degree of damage into a specific value or level, making the prediction result more objective and comparable, and improving the efficiency of vehicle damage detection. The specific definition of the vehicle damage detection device can refer to the definition of the vehicle damage detection method above, which will not be repeated here. Each module in the above vehicle damage detection device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0135] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, it implements the functions or steps on the service side of a vehicle damage detection method.

[0136] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 5As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of the client side of a vehicle damage detection method.

[0137] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0138] Acquire a vehicle damage image to be predicted, and perform image enhancement on the vehicle damage image to be predicted to obtain a standard vehicle damage image;

[0139] Extracting initial image features of the standard vehicle damage image, and performing multi-layer deformable convolution on the initial image features to obtain layered convolution features;

[0140] Performing feature fusion on the layered convolution features to obtain target features of the standard vehicle damage image;

[0141] The target feature is used to calculate the vehicle damage degree corresponding to the standard vehicle damage image.

[0142] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and apparatuses can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0143] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0144] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0145] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0146] In some implementations of this embodiment, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that when the computer program is executed by a processor, the steps of the method described in the above embodiment are implemented.

[0147] The readable storage medium of the present invention stores a computer program, and when the computer program is executed by a processor of an electronic device, the computer program can achieve:

[0148] Acquire a vehicle damage image to be predicted, and perform image enhancement on the vehicle damage image to be predicted to obtain a standard vehicle damage image;

[0149] Extracting initial image features of the standard vehicle damage image, and performing multi-layer deformable convolution on the initial image features to obtain layered convolution features;

[0150] Performing feature fusion on the layered convolution features to obtain target features of the standard vehicle damage image;

[0151] The target feature is used to calculate the vehicle damage degree corresponding to the standard vehicle damage image.

[0152] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0153] The computer-readable storage medium may also store at least one computer executable program / instruction, which may be, for example, a computer-readable instruction. The computer-readable storage medium includes, but is not limited to, for example, a volatile memory and / or a non-volatile memory. The volatile memory may include, for example, a random access memory (RAM) and / or a cache memory (cache), etc. The computer-readable storage medium may include, for example, a read-only memory (ROM), a hard disk, a flash memory, etc. For example, a non-transitory computer-readable storage medium may be connected to a computing device such as a computer, and then, when the computing device runs the computer-readable instructions stored on the computer-readable storage medium, the various methods described above may be performed.

[0154] In addition, the computer device may also include (but not limited to) a data bus, an input / output (I / O) bus, a display, and input / output devices (eg, keyboard, mouse, speaker, etc.), etc.

[0155] The processor may communicate with external devices via an I / O bus via a wired or wireless network.

[0156] In one embodiment, the at least one computer executable instruction may also be compiled into or constitute a software product / computer program product, wherein one or more computer executable instructions are executed by a processor to perform the various functions and / or method steps in the embodiments described in the present technology.

[0157] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0158] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0159] In the embodiments provided in the present disclosure, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the above-mentioned module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0160] It should be noted that in the present disclosure, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element limited by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0161] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

[0162] It should be noted that if software tools or components other than those of our company appear in the embodiments of the present application, they are only used for illustrative purposes and do not represent actual use.

Claims

1. A vehicle damage detection method, characterized in that: The method comprises: Acquire a vehicle damage image to be predicted, and perform image enhancement on the vehicle damage image to be predicted to obtain a standard vehicle damage image; Extracting initial image features of the standard vehicle damage image, and performing multi-layer deformable convolution on the initial image features to obtain layered convolution features; Performing feature fusion on the layered convolution features to obtain target features of the standard vehicle damage image; The target feature is used to calculate the vehicle damage degree corresponding to the standard vehicle damage image.

2. A vehicle damage detection method as claimed in claim 1, characterized in that: The step of performing image enhancement on the vehicle damage image to be predicted to obtain a standard vehicle damage image includes: Normalizing the vehicle damage image to be predicted to obtain a normalized vehicle damage image; Performing image denoising on the normalized vehicle damage image to obtain a pollution-free vehicle damage image; The pollution-free vehicle damage image is subjected to image adjustment to obtain a standard vehicle damage image.

3. A vehicle damage detection method as claimed in claim 2, characterized in that: The step of normalizing the vehicle damage image to be predicted to obtain a normalized vehicle damage image includes: Calculate the absolute value of the first difference between the minimum value and the maximum value of the pixel value of the vehicle damage image to be predicted respectively; Respectively calculating the second absolute value of the difference between each pixel value and the minimum value in the vehicle damage image to be predicted, and dividing the first absolute value of the difference by the second absolute value of the difference to obtain a normalized value; A normalized vehicle damage image is generated using the normalized value corresponding to each pixel value.

4. A vehicle damage detection method as claimed in claim 1, characterized in that: The extracting of initial image features of the standard vehicle damage image comprises: Get the pre-built Initial Hybrid Convolutional Model; Based on the multiple convolution kernels of the initial mixed convolution model, dividing the initial mixed convolution model into multiple mixed convolution models; Based on the hybrid convolution model, convolution processing is performed on the standard vehicle damage image to obtain initial image features.

5. A vehicle damage detection method as claimed in claim 1, characterized in that: The performing multi-layer deformable convolution on the initial image features to obtain layered convolution features includes: According to the preset stacking rules, the parameters of different convolutional layers in the pre-built hybrid convolutional model are stacked to obtain convolutional layers with different parameters; Performing convolution processing on the initial image features according to the convolution layers with different parameters to obtain image features corresponding to the convolution layers with different parameters; The image features corresponding to the convolutional layers with different parameters are concatenated to obtain hierarchical convolutional features.

6. A vehicle damage detection method as claimed in claim 1, characterized in that: The step of fusing the layered convolution features to obtain target features of the standard vehicle damage image includes: The layered convolution feature is traversed in a sliding window manner through a pre-constructed convolution kernel to obtain a pixel value corresponding to the layered convolution feature; Multiplying the pixel value in the convolution kernel by the pixel value corresponding to the layered convolution feature and summing them up to obtain the pixel value corresponding to the target feature; The pixel values ​​corresponding to the target features are mapped to a preset target feature map to obtain the target features.

7. A vehicle damage detection method as claimed in claim 1, characterized in that: The calculating the vehicle damage degree corresponding to the standard vehicle damage image by using the target feature includes: Pooling and activating the target features to obtain processed target features; The processed target features are fully connected to obtain the vehicle damage degree corresponding to the standard vehicle damage image.

8. A vehicle damage detection device, characterized in that: The device comprises: An image enhancement module is used to obtain a vehicle damage image to be predicted, and perform image enhancement on the vehicle damage image to be predicted to obtain a standard vehicle damage image; A deformable convolution module, used to extract initial image features of the standard vehicle damage image, and perform multi-layer deformable convolution on the initial image features to obtain layered convolution features; A feature fusion module, used for performing feature fusion on the layered convolution features to obtain target features of the standard vehicle damage image; The vehicle damage degree assessment module is used to calculate the vehicle damage degree corresponding to the standard vehicle damage image using the target feature.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute a vehicle damage detection method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a vehicle damage detection method as described in any one of claims 1 to 7 is implemented.

Images