Vehicle surface repair method, electronic device, and storage medium

Abstract

This application provides a vehicle surface repair method, electronic device, and storage medium, belonging to the field of automotive surface repair technology. The method includes: acquiring damaged morphological data of a damaged area of ​​a vehicle to be repaired; acquiring mirror morphological data of a mirror area corresponding to the damaged area; constructing a damaged surface model based on the damaged morphological data and a mirror surface model based on the mirror morphological data; performing mirror processing on the mirror surface model to obtain a reference surface model; performing a difference operation on the damaged surface model and the reference surface model to obtain a three-dimensional repair model; determining putty printing parameters based on the three-dimensional repair model, and performing putty printing on the damaged area based on the putty printing parameters. This application can effectively improve the process accuracy and repair efficiency of automotive surface repair.

Description

Technical Field

[0001] This application relates to the field of automotive surface repair technology, and more particularly to a vehicle surface repair method, electronic device, and storage medium. Background Technology

[0002] In the field of automotive repair, repairing deformation of the vehicle body surface caused by a collision is a routine and important technical step. Traditional repair processes typically include the following steps: First, the shape of the damaged area is assessed manually using a combination of visual inspection and simple measuring tools. Then, mechanical stretching or tapping is used to roughly restore the metal surface. Finally, residual dents are filled by hand with putty, and after sanding, touch-up painting is performed.

[0003] However, in traditional repair processes, the thickness of the putty (cream putty) application relies mainly on manual experience and judgment, lacking scientific calculation basis. This method is easily affected by factors such as the worker's skill level and visual errors, resulting in insufficient geometric matching between the repaired surface and the original vehicle body.

[0004] Therefore, there is an urgent need to develop a new method for automotive surface repair that can overcome the low precision defects of traditional repair processes. Summary of the Invention

[0005] The purpose of this application is to provide a vehicle surface repair method, electronic device, and storage medium to solve the above-mentioned problems.

[0006] To achieve the above objectives, in a first aspect, this application proposes a method for repairing vehicle surfaces, the method comprising: acquiring damage morphology data of the damaged portion of the vehicle to be repaired; and

[0007] Obtain the mirror image morphology data of the mirror image area corresponding to the damaged area;

[0008] A damaged surface model is constructed based on the damaged morphology data, and a mirror surface model is constructed based on the mirror morphology data.

[0009] The mirror surface model is mirrored to obtain the reference surface model;

[0010] Perform a difference operation on the damaged surface model and the reference surface model to obtain a three-dimensional repair model formed by the space enclosed by the two surfaces;

[0011] Based on the three-dimensional repair model, putty printing parameters are determined, and putty is printed on the damaged area based on the putty printing parameters.

[0012] In some implementations, prior to acquiring damage morphology data of the damaged portion of the vehicle to be repaired, the following steps are included:

[0013] Acquire the first image of the damaged surface of the vehicle to be repaired;

[0014] Extract the edge contours of the first captured image, and apply morphological operations to enhance the edge contours to obtain the contour features of the supposedly damaged region; and / or

[0015] Convert the first captured image to a grayscale image, and based on the grayscale image, determine the texture features of the region to be damaged; and / or

[0016] The captured image is converted to a preset color space, and the color features of the supposedly damaged area are extracted based on the color space.

[0017] Based on at least one of the contour features, texture features, and color features of the supposedly damaged region, the input features of the pre-trained vehicle damage recognition model are determined;

[0018] The input features are input into the vehicle damage recognition model to obtain the segmentation result of the suspected damaged area output by the vehicle damage recognition model based on the input features;

[0019] Based on the segmentation results, the damaged parts of the vehicle to be repaired are determined.

[0020] In some embodiments, before obtaining the mirror image topography data of the mirror region corresponding to the damaged region, the following steps are included:

[0021] Using the main axis of symmetry of the vehicle to be repaired as a reference, determine the mirror image corresponding to the damaged surface of the vehicle to be repaired;

[0022] A second image of the mirror surface is acquired, wherein the shooting points of the second image and the first image are mirror-symmetrical about the main axis of symmetry of the vehicle to be repaired.

[0023] The second captured image is mirror-compared with the first captured image that identifies the damaged area;

[0024] Based on the comparison results, the mirror image of the damaged area is determined.

[0025] In some implementations, prior to acquiring damage morphology data of the damaged portion of the vehicle to be repaired, the following steps are included:

[0026] Acquire the full-vehicle 3D scan data of the vehicle to be repaired, and extract the point cloud data of the damaged surface of the vehicle to be repaired from the full-vehicle 3D scan data;

[0027] Extract the geometric features of the point cloud data, including curvature, normal vector, and edge points;

[0028] The geometric features are mapped onto a two-dimensional feature image to form a geometric feature map;

[0029] The geometric feature map is converted into a multidimensional feature vector, which is then used as the model input for the pre-trained vehicle damage discrimination model.

[0030] The multidimensional feature vector is input into the vehicle damage discrimination model to obtain the probability distribution map of each point cloud data belonging to the damaged part, which is output by the vehicle damage discrimination model based on the multidimensional feature vector.

[0031] Based on the probability distribution map, the damaged parts of the vehicle to be repaired are determined, and the damage morphology data of the damaged parts are extracted from the three-dimensional scanning data of the whole vehicle.

[0032] In some implementations, determining the damaged parts of the vehicle to be repaired based on the probability distribution map includes:

[0033] Based on a preset damage threshold, the probability distribution map is converted into a binary segmentation mask, which includes damaged marker values ​​and undamaged marker values.

[0034] Based on the damage marker value, the damaged parts of the vehicle to be repaired are determined.

[0035] In some embodiments, obtaining the mirror image morphology data of the mirror region corresponding to the damaged region includes:

[0036] Based on the vehicle's 3D scanning data, a 3D data model of the vehicle is generated;

[0037] Determine the location coordinates of the damaged part in the three-dimensional data model of the vehicle;

[0038] Based on the geometric symmetry of the vehicle to be repaired, the mirror image of the damaged part is determined in the three-dimensional data model of the whole vehicle;

[0039] The mirror image data of the mirror area is extracted from the three-dimensional scanning data of the whole vehicle.

[0040] In some implementations, determining the putty printing parameters based on the three-dimensional repair model includes:

[0041] Based on the key features of the three-dimensional repair model, the three-dimensional repair model is divided into several sub-regions;

[0042] The geometric complexity and thickness variation of each sub-region are calculated separately, and the corresponding printing parameter combination is determined based on the geometric complexity and thickness variation. The printing parameter combination includes nozzle diameter, material output speed, printing path planning and layer printing parameters.

[0043] Based on the combination of printing parameters corresponding to each sub-region, the putty printing parameters corresponding to the damaged area are determined.

[0044] In some implementations, determining the corresponding combination of printing parameters based on the geometric complexity and the thickness variation includes:

[0045] The nozzle diameter is determined based on the geometric complexity, wherein the nozzle diameter is inversely proportional to the geometric complexity;

[0046] Based on the thickness variation, determine the layer printing parameters;

[0047] The output speed is determined based on the layered printing parameters;

[0048] Based on the nozzle diameter, the discharge speed, and the layered printing parameters, a printing path is planned.

[0049] Secondly, this application proposes an electronic device, comprising:

[0050] One or more processors;

[0051] A memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to perform the vehicle surface repair method as described above.

[0052] Thirdly, this application proposes a storage medium storing executable instructions that, when executed by a processor, cause the processor to perform the vehicle surface repair method as described above.

[0053] Compared with the prior art, the beneficial effects of this application include:

[0054] Firstly, by acquiring the morphological data of the damaged area and its mirror image, high-precision damaged surface models and reference surface models are generated using 3D scanning technology. This method avoids visual errors inherent in traditional manual operations, ensuring a high degree of matching between the repair surface of the damaged area and the original vehicle body geometry. Secondly, by performing difference operations on the damaged surface model and the reference surface model, the 3D repair model can be accurately determined, avoiding the drawback of lacking scientific calculation basis for the thickness of putty application. This provides accurate data support for subsequent putty printing, further improving repair accuracy. Attached Figure Description

[0055] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation on the scope of this application.

[0056] Figure 1 This is a flowchart illustrating a vehicle surface repair method in one embodiment;

[0057] Figure 2 This is a detailed flowchart of a vehicle surface repair method in one embodiment;

[0058] Figure 3 This is a flowchart illustrating a vehicle surface repair method in another embodiment;

[0059] Figure 4 This is a partial flowchart of a vehicle surface repair method in another embodiment;

[0060] Figure 5 This is a flowchart illustrating a vehicle surface repair method in yet another embodiment;

[0061] Figure 6 This is a schematic diagram of the electronic device involved in the vehicle surface repair method in the embodiments of this application. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0063] All terms used in this application (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0064] For example, the terms "first," "second," etc., used in this application may be used herein to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish one element from another. For instance, without departing from the scope of this application, a first captured image may be referred to as a second captured image, and similarly, a second captured image may be referred to as a first captured image. Both the first captured image and the second captured image are captured images, but they are not the same captured image.

[0065] For example, the terms "comprising" or "including" used in this application indicate the presence of features, steps, operations and / or components, but do not exclude the presence or addition of one or more other features, steps, operations or components.

[0066] As mentioned earlier, in traditional repair processes, the thickness of the putty (cream) application relies primarily on manual judgment and lacks scientific calculation. This method is easily affected by factors such as worker skill level and visual errors, leading to insufficient geometric matching between the repaired surface and the original vehicle body. Therefore, there is an urgent need to develop a new automotive surface repair method that can overcome the low-precision defects of traditional repair processes. To this end, this application proposes a vehicle surface repair method, electronic device, and storage medium that can effectively improve the process precision and repair efficiency of automotive surface repair.

[0067] like Figure 1 As shown in the embodiments of this application, a vehicle surface repair method includes the following steps:

[0068] Step S10: Obtain damage morphology data of the damaged parts of the vehicle to be repaired.

[0069] In this embodiment, the damaged area refers to the area on the surface of the vehicle to be repaired that has been damaged due to collision, scratches, or other reasons. Damage morphology data refers to data used to describe the surface geometry of the damaged area, typically acquired using a 3D scanning device, and includes point cloud data or mesh data. The 3D scanning device can be a handheld, fixed, or robot-mounted scanning head.

[0070] In one implementation, a high-precision 3D scanner can be used to scan the damaged area to obtain damage morphology data.

[0071] In another implementation, a high-precision 3D scanner can be used to scan the entire vehicle to be repaired beforehand to obtain 3D scan data of the entire vehicle. Damage morphology data corresponding to the damaged areas can then be selected from the 3D scan data.

[0072] In step S10, precise morphological data of the damaged area is obtained through 3D scanning, providing a scientific basis for subsequent repair. Compared with traditional manual visual inspection, 3D scanning avoids visual errors and improves the accuracy of the data.

[0073] Step S20: Obtain the mirror image morphology data of the mirror image region corresponding to the damaged region.

[0074] In this embodiment, the mirror image area refers to the undamaged area that is symmetrical to the damaged area, based on the symmetry of the vehicle design. For example, when the left front door is damaged, the right front door is the mirror image area. Mirror image topography data refers to data used to describe the surface geometry of the mirror image area.

[0075] In one implementation, a high-precision 3D scanner can be used to scan the mirrored area to obtain mirrored topographic data.

[0076] In another implementation, damaged topographic data corresponding to the mirrored portion can be extracted from the vehicle's 3D scan data. Specifically, a 3D data model of the vehicle is generated based on the vehicle's 3D scan data. The position coordinates of the damaged portion are determined in the vehicle's 3D data model. Based on the geometric symmetry of the vehicle to be repaired, the mirrored portion corresponding to the damaged portion is determined in the vehicle's 3D data model. The mirrored topographic data of the mirrored portion is extracted from the vehicle's 3D scan data.

[0077] In step S20, the morphological data of the mirrored part provides a scientific basis for the repair, ensuring that the repaired surface is consistent with the original vehicle body.

[0078] Step S30: Construct a damaged surface model based on the damaged morphology data.

[0079] Step S40: Construct a mirror surface model based on the mirror topography data.

[0080] In this embodiment, the damaged surface model refers to a three-dimensional mathematical model constructed based on the topographic data of the damaged area, used to describe the geometric features of the damaged region. The mirrored surface model refers to a three-dimensional mathematical model constructed based on the topographic data of the mirrored area, used to describe the geometric features of the undamaged area.

[0081] In some implementations, a damaged surface model can be generated by importing the damaged morphology data of the damaged area into modeling software (such as Geomagic, CATIA, etc.). Similarly, a mirrored surface model can be generated by importing the mirrored morphology data of the mirrored area into the software.

[0082] Step S50: Perform mirroring on the mirrored surface model to obtain the reference surface model.

[0083] In this embodiment, the reference surface model refers to the three-dimensional data model of the mirrored surface model after mirroring, which is used as a reference standard for repairing damaged parts.

[0084] In some implementations, a reference surface model is obtained by flipping the mirror surface model from its original position to the target position corresponding to the damaged part through mathematical transformation (such as reflection transformation).

[0085] In step S50, by mirroring the mirrored surface model, a reference surface model corresponding to the damaged area can be accurately generated, ensuring that the repaired surface is consistent with the original vehicle body geometry. Furthermore, using the method of this embodiment, there is no need to redesign the reference surface model; modeling can be completed directly using the data and symmetry of the mirrored area, saving time and resources.

[0086] Step S60: Perform a difference operation on the damaged surface model and the reference surface model to obtain a three-dimensional repair model formed by the space enclosed by the two surfaces.

[0087] In this embodiment, the core of the difference operation is to calculate the difference region between the damaged surface model and the reference surface model. The mathematical formula can be expressed as: 3D repair model = reference surface model - damaged surface model, where "-" represents a Boolean subtraction operation, that is, removing the part that overlaps with the damaged surface model from the reference surface model, and the remaining part is the 3D repair model.

[0088] In step S60, the geometric difference between the damaged area and the reference surface can be accurately calculated through difference operations, thereby clearly defining the three-dimensional repair model. Automated calculation to generate the three-dimensional repair model avoids errors inherent in traditional manual judgment and significantly improves repair efficiency.

[0089] Step S70: Based on the three-dimensional repair model, determine the putty printing parameters, and perform putty printing on the damaged area based on the putty printing parameters.

[0090] In this embodiment, putty printing parameters refer to the printing equipment parameters used in the process of repairing damaged areas, including nozzle diameter, feed rate, printing path planning, and layer printing parameters. Nozzle diameter refers to the aperture size of the print head, affecting the amount of material dispensed per pass and resolution. Smaller nozzle diameters are suitable for high-precision printing, while larger nozzle diameters are suitable for quickly covering large areas. Feed rate refers to the rate at which material is output during printing. Too fast a feed rate may result in uneven material accumulation, while too slow a rate reduces efficiency. Printing path planning refers to the path design of the nozzle movement during printing. Layer printing parameters include the thickness of the material layer deposited in each print and the number of layers printed. Thinner material layer thickness improves accuracy but increases printing time; thicker material layer thickness improves efficiency but sacrifices some accuracy.

[0091] In some implementations, the 3D repair model can be directly imported into a 3D putty printing device with depth vision. The 3D putty printing device converts the 3D repair model into a series of thin layers, each layer being a 2D image. The 3D putty printing device deposits material layer by layer to ultimately form a 3D object. The 3D putty printing device can automatically select appropriate putty printing parameters, such as nozzle diameter, output speed, printing path planning, and layer thickness, based on the complexity and size of the 3D repair model.

[0092] In some implementations, the electronic device of this application embodiment can determine putty printing parameters based on the geometric complexity and thickness variation of the three-dimensional repair model. Geometric complexity describes the complexity of the surface shape and typically includes indicators such as curvature variation and the number of surface features. Thickness variation corresponds to the putty thickness required for repair or printing at different locations of the damaged part.

[0093] In some implementations, the putty printing parameters may include several combinations of printing parameters. This is because different regions have different geometric complexities and thickness variations, necessitating the application of different combinations of printing parameters when printing on different areas of the damaged site. Putty is then printed on different areas of the damaged site based on these combinations of printing parameters.

[0094] In some implementations, such as Figure 2 As shown, step S70 includes:

[0095] Step S71: Based on the key features of the three-dimensional repair model, the three-dimensional repair model is divided into several sub-regions.

[0096] In this embodiment, key features refer to parts of the 3D restoration model that have significant geometric or structural characteristics. These features typically include, but are not limited to, edges, sharp corners, uneven areas, and areas with high curvature. Key features are the basis for identifying and dividing sub-regions, ensuring that the geometric characteristics of each sub-region are relatively consistent. A sub-region refers to a logically independent part divided from the 3D restoration model. The geometric features within each sub-region have a certain degree of consistency to facilitate putty printing.

[0097] In some implementations, edge detection, curvature analysis, and feature extraction algorithms can be used to identify and label key features of the 3D restoration model. Based on the labeled key features, the 3D restoration model is divided into several logically independent sub-regions. The division principles can be: regions with similar curvature are grouped into the same sub-region; and / or regions with consistent surface features are grouped into the same sub-region; and / or regions with similar geometric complexity are grouped into the same sub-region. Here, surface features refer to specific morphological features on a curved surface, such as grooves, protrusions, planes, and smooth transitions.

[0098] In step S71, by dividing the region into sub-regions based on key features, precise repair of the geometric features of different regions can be achieved, significantly improving the repair quality. The printing parameters for each sub-region can be more personalized, avoiding the incompatibility issues caused by globally uniform parameters. Furthermore, reasonable sub-region division makes subsequent geometric complexity calculations and thickness change measurements simpler and more efficient, reducing unnecessary repetitive calculations and data processing.

[0099] Step S72: Calculate the geometric complexity and thickness variation of each sub-region, and determine the corresponding combination of printing parameters based on the geometric complexity and thickness variation.

[0100] In this embodiment, the printing parameter combination refers to the printing equipment parameters used in the process of repairing each sub-region, including nozzle diameter, output speed, printing path planning, layer printing parameters, etc.

[0101] In some implementations, the nozzle diameter is determined based on the geometric complexity, wherein the nozzle diameter is inversely proportional to the geometric complexity. Layer printing parameters are determined based on the thickness variation; the output speed is determined based on the layer printing parameters. A printing path is planned based on the nozzle diameter, the output speed, and the layer printing parameters.

[0102] Specifically, for areas with high geometric complexity, choosing a smaller nozzle diameter (such as Φ1) can ensure clear details; for areas with low complexity, choosing a larger nozzle diameter (such as Φ5) can improve efficiency.

[0103] The nozzle diameter determines the width of a single discharge, directly affecting the path spacing (i.e., path density). When the single discharge width is narrow, a smaller path spacing (higher path density) should be set to ensure uniform material coverage. When the single discharge width is wide, the path spacing can be appropriately increased (lower path density). In other words, the nozzle diameter is directly proportional to the single discharge width and the path spacing, and inversely proportional to the path density. The discharge speed also affects the amount of material extruded per unit time, thus indirectly influencing the choice of path spacing.

[0104] For areas with significant thickness variations, a thinner material layer is chosen to ensure surface quality and detail. For areas with minimal thickness variations, a thicker material layer is chosen to improve efficiency. In other words, the material layer thickness is inversely proportional to the thickness variation, and also inversely proportional to the number of printing layers.

[0105] In some implementations, the basic path spacing can be determined based on the nozzle diameter, and the basic path spacing and the nozzle diameter can be set according to a preset ratio (such as a value between 80% and 100%); the layer printing parameters can be determined based on the thickness variation, and the layer printing parameters include the material layer thickness and the number of printing layers for a single layer; the printing path planning can be determined based on the basic path spacing and the number of printing layers.

[0106] In some embodiments, the basic path spacing is determined based on the nozzle diameter; layer printing parameters are determined based on the thickness variation, including the thickness of a single-layer material and the number of printing layers; the output speed is determined based on the thickness of the single-layer material and is used as a reference speed. During putty printing, the real-time output speed is monitored to see if it exceeds the reference speed; when the real-time output speed exceeds the reference speed, the basic path spacing is narrowed to avoid voids; when the real-time output speed is lower than the reference speed, the basic path spacing is widened. The magnitude of the narrowing and widening adjustments is positively correlated with the ratio of the real-time output speed to the reference speed. Based on the adjusted path spacing and the number of printing layers, a printing path plan is determined. In this embodiment, when the output speed is higher than the reference speed, the nozzle outputs more material, which may lead to localized accumulation or material waste if the path spacing is not adjusted; conversely, if the output speed is lower than the reference speed, it may result in insufficient filling, voids, or a loose structure. By dynamically narrowing or widening the path spacing, it can cope with changes in real-time output speed, ensuring the stability of print quality.

[0107] Step S73: Based on the combination of printing parameters corresponding to each sub-region, determine the putty printing parameters corresponding to the damaged part.

[0108] In this embodiment, the putty printing parameters are composed of a combination of printing parameters for each sub-region, including but not limited to key parameters such as nozzle diameter, layer thickness, discharge speed, and printing path planning. These parameters are configured independently according to the geometric characteristics (such as complexity and height difference) and functional requirements of the sub-region.

[0109] In some implementations, based on the printing parameter combinations corresponding to each sub-region, the key parameters of each sub-region are mapped onto the 3D repair model corresponding to the damaged area to generate a key parameter distribution map. Based on the key parameter distribution map, transitional problem areas that negatively affect the smoothness of the printing effect are identified. For the transitional problem areas, the key parameters corresponding to the transitional problem areas are dynamically adjusted to obtain the adjusted putty printing parameters for the entire damaged area. Here, the transitional problem areas refer to adjacent sub-regions where the key parameter differences exceed a preset difference threshold.

[0110] In some implementations, global parameter optimization can be performed on all sub-regions of the damaged area based on the parameter difference adjustment results, so as to obtain the optimized putty printing parameters corresponding to the entire damaged area.

[0111] In some implementations, the global optimization step may be to simulate the printing process based on the optimized putty printing parameters corresponding to the damaged area using finite element analysis or computer-aided engineering tools, and verify whether the simulated printing process meets the printing effect smoothness target; if it does not meet the printing effect smoothness target, the step of calculating the parameter difference value between adjacent sub-regions is repeated until the printing effect smoothness target is achieved.

[0112] In the vehicle surface repair method proposed in this application, firstly, by acquiring the morphological data of the damaged area and its mirror image, a high-precision damaged surface model and a reference surface model are generated using 3D scanning technology. This method avoids visual errors in traditional manual operations and ensures a high degree of matching between the repair surface of the damaged area and the original vehicle geometry. Secondly, by performing a difference operation on the damaged surface model and the reference surface model, the 3D repair model can be accurately determined, avoiding the drawback of lacking scientific calculation basis for the thickness of putty application. This provides accurate data support for subsequent putty printing and further improves the repair accuracy.

[0113] In one embodiment, such as Figure 3 As shown, before step S10, the following may be included:

[0114] Step A10: Obtain the first image of the damaged surface of the vehicle to be repaired.

[0115] A first image is obtained by taking a picture of the damaged surface of the vehicle to be repaired using a high-resolution camera or other imaging device. The first image is a color image (RGB image).

[0116] Step A21: Extract the edge contour of the first captured image and apply morphological operations to enhance the edge contour to obtain the contour features of the supposedly damaged region.

[0117] Edge detection algorithms (such as Canny edge detection) are used to extract the edge contours of the damaged area from the first captured image; the edge contours are then enhanced by morphological operations (such as dilation, erosion, etc.) to remove noise and strengthen the contour features of the supposedly damaged area. The contour features can capture the shape information of the damaged area and help distinguish between damaged and undamaged areas.

[0118] Step A22: Convert the first captured image into a grayscale image, and determine the texture features of the area to be damaged based on the grayscale image.

[0119] The first captured image is converted from color to grayscale. Local Binary Pattern (LBP) or other texture analysis methods are used to extract texture features from the suspected damaged area. Texture features reflect changes in the surface material, such as subtle variations in scratches or cracks. This is because damaged areas, due to scratches, dents, etc., exhibit significantly different texture features compared to smooth, undamaged areas. By extracting these texture features, the suspected damaged area can be identified.

[0120] Step A23: Convert the captured image to a preset color space, and extract the color features of the supposedly damaged area based on the color space.

[0121] The first captured image is converted from the RGB color space to the HSV or LAB color space, and color features of the damaged area are extracted based on this color space. By extracting color features, the damaged area can be located more accurately, especially in cases of damage with significant color changes (such as rust, fading, etc.).

[0122] It should be noted that there is no strict order of execution between steps S21 to S23, and not all steps need to be executed. In practice, at least one of steps S21, S22, and S23 can be selected to be executed according to the actual application scenario to meet specific technical requirements or optimize the processing flow.

[0123] Step A30: Determine the input features of the pre-trained vehicle damage recognition model based on at least one of the contour features, texture features, and color features of the allegedly damaged area.

[0124] At least one of the contour features extracted in step A21, the texture features extracted in step A22, and the color features extracted in step A23 is integrated to form the input features of the vehicle damage recognition model. High-quality input data is provided to the vehicle damage recognition model to ensure that the model can accurately predict damaged areas.

[0125] The vehicle damage recognition model is an image segmentation model based on deep learning techniques (such as convolutional neural networks, CNNs). Before step A30, a labeled dataset containing various vehicle damage images is obtained. This dataset labels the location, type (e.g., scratches, dents), and severity of damage for each vehicle damage image. The vehicle damage images can be obtained from real-world photography or generated through simulation. The labeled dataset is used to train a first preset model, and during training, the parameters of the first preset model are adjusted using a backpropagation algorithm until the error between the predicted result and the actual annotation is minimized, thus obtaining the vehicle damage recognition model.

[0126] Step A40: Input the input features into the vehicle damage recognition model to obtain the segmentation result of the suspected damaged area output by the vehicle damage recognition model based on the input features.

[0127] In this embodiment, the input features generated in step A30 are input into a pre-trained vehicle damage recognition model. The model outputs the segmentation result of the damaged area based on the input features. The segmentation result can be presented in the form of a probability map or a binary mask. In the probability map, each pixel corresponds to a probability value, representing the likelihood that the pixel belongs to the damaged area. The higher the probability value, the more likely the pixel is to be located in the damaged area. In the binary mask, each pixel is marked as "damaged" or "undamaged," represented by white (or 1) and black (or 0), respectively.

[0128] Step A50: Based on the segmentation results, determine the damaged parts of the vehicle to be repaired.

[0129] Based on the damaged areas marked in the segmentation results, the damaged parts of the vehicle to be repaired can be determined by extracting the pixel coordinates or bounding boxes of the damaged areas.

[0130] The vehicle surface repair method proposed in this application significantly improves the accuracy of damaged area identification by comprehensively utilizing contour, texture, and color features, combined with the powerful capabilities of deep learning models. Furthermore, this method automates the entire process from image acquisition to damaged area localization. The overall solution boasts advantages of high precision, high efficiency, and low cost, and can be widely applied in vehicle repair-related fields.

[0131] Based on this embodiment, such as Figure 4 As shown, this application proposes another embodiment, which further includes the following step before step S20:

[0132] Step A60: Using the main axis of symmetry of the vehicle to be repaired as a reference, determine the mirror image corresponding to the damaged surface of the vehicle to be repaired.

[0133] In this embodiment, the main axis of symmetry of the vehicle to be repaired typically refers to the vehicle's longitudinal centerline (i.e., the central axis from front to back), which is the basic axis of symmetry in most vehicle designs. The corresponding mirror plane is determined based on the location of the damaged surface. For example, if the damaged surface is located on the left front door, the mirror plane is the right front door; if the damaged surface is located on the left side of the roof, the mirror plane is the right side of the roof.

[0134] Step A70: Obtain a second image of the mirror surface, wherein the shooting points of the second image and the first image are mirror-symmetrical about the main axis of symmetry of the vehicle to be repaired.

[0135] After determining the mirror plane, a second image of the mirror plane is acquired. To ensure comparability between the second and first images, the shooting points of the second image and the first image are mirror images of each other about the vehicle's main axis of symmetry. For example, if the first image was taken from a 45° angle from the front left side of the vehicle, the second image should be taken from a 45° angle from the front right side of the vehicle.

[0136] In addition, the shooting parameters (such as focal length and resolution) of the two images should be kept as consistent as possible to reduce errors caused by different shooting conditions.

[0137] Step A80: Compare the second captured image with the first captured image that has identified the damaged area.

[0138] The first image, which identifies or marks the damaged area, is mirrored to obtain a third image, which is then compared with the second image.

[0139] In some implementations, the second and third images need to be aligned and corrected before comparison. This step includes adjusting the angle, scale, and position of the images to ensure that the two images are compared in the same coordinate system.

[0140] Specifically, image registration techniques (such as feature point matching methods) can be used to align the second and third captured images. For example, key feature points (such as headlights, door handles, etc.) in the first and third captured images can be extracted and aligned to the same reference frame using feature point matching.

[0141] Step A90: Based on the comparison results, determine the mirror image of the damaged area.

[0142] Based on the comparison results, the damaged part in the third image is determined to be the mirror image of the damaged part in the second image, and the mirror image is the undamaged part.

[0143] The vehicle surface repair method proposed in this application makes full use of the vehicle's symmetry and provides a scientific basis for the repair of damaged parts through mirror processing and image comparison.

[0144] In one embodiment, such as Figure 5 As shown, prior to step S10, the following may also be included:

[0145] Step B10: Obtain the full-vehicle 3D scan data of the vehicle to be repaired, and extract the point cloud data of the damaged surface of the vehicle to be repaired from the full-vehicle 3D scan data.

[0146] A high-precision 3D laser scanner or structured light scanning device is used to perform a comprehensive scan of the entire vehicle to obtain complete 3D scan data and generate a 3D model of the vehicle. The first position of the damaged surface where the damaged part is located in the 3D model of the vehicle is determined, and the point cloud data corresponding to the first position in the 3D model of the vehicle is obtained.

[0147] Step B20: Extract the geometric features of the point cloud data, including curvature, normal vector, and edge points.

[0148] Geometric features, including curvature, normal vectors, and edge points, are extracted from point cloud data. Changes in curvature and normal vectors can effectively reflect surface deformation, while edge points can be used to identify regions with abrupt shape changes.

[0149] Step B30: Map the geometric features onto a two-dimensional feature image to form a geometric feature map.

[0150] Geometric features (such as curvature, normals, edges, etc.) are projected onto a two-dimensional plane to form a geometric feature map. This geometric feature map can be multi-channel, with each channel corresponding to a different geometric feature. For example, one channel represents curvature, another channel represents the X component of the normal vector, and a third channel represents the Y component of the normal vector, etc.

[0151] Because 3D point cloud data is high-dimensional and unstructured, direct processing is quite complex. Step B30 constructs a geometric feature map, which can better capture and represent the geometric information in the point cloud, making subsequent classification or segmentation tasks more accurate.

[0152] Step B40: Convert the geometric feature map into a multidimensional feature vector, which is used as the model input for the pre-trained vehicle damage discrimination model.

[0153] In this embodiment, multidimensional feature vectors are the standard input format for deep learning models (vehicle damage discrimination models), which can fully express the information in the geometric feature maps. The geometric feature maps are encoded into multidimensional feature vectors and used as input to the vehicle damage discrimination model.

[0154] Step B40 achieves efficient conversion from image data (geometric feature map) to numerical data (multidimensional feature vector) through feature vectorization.

[0155] Step B50: Input the multidimensional feature vector into the vehicle damage discrimination model to obtain a probability distribution map of each point cloud data point output by the vehicle damage discrimination model based on the multidimensional feature vector, indicating that each point cloud data point belongs to a damaged part.

[0156] The multidimensional feature vector is input into the vehicle damage discrimination model, which then performs classification or segmentation tasks based on the input feature vector. For each point cloud, the vehicle damage discrimination model outputs a probability value indicating whether it is damaged or undamaged. These probability values ​​can form a probability distribution map corresponding to the original point cloud data. The position of each point cloud in the probability distribution map corresponds to its position in the original point cloud data, and its value represents the probability that the point cloud belongs to a damaged area.

[0157] Step B60: Based on the probability distribution map, determine the damaged parts of the vehicle to be repaired.

[0158] Based on the probability distribution map, the point cloud distribution area with a damage probability value greater than a preset damage threshold is determined as the damaged part of the vehicle to be repaired, so as to extract the damage morphology data of the damaged part from the whole vehicle 3D scanning data.

[0159] In the vehicle surface repair method proposed in this application, the processing flow can be simplified and mature deep learning models can be fully utilized by mapping geometric features to a two-dimensional feature map. Using multi-dimensional feature vectors as input ensures that the model receives rich geometric information, while the probability distribution map as output provides a clear indication of whether each point cloud is damaged.

[0160] Based on this embodiment, step B60 further includes:

[0161] Step B61: Based on a preset damage threshold, the probability distribution map is converted into a binary segmentation mask, which includes damaged marker values ​​and undamaged marker values.

[0162] Based on a pre-defined damage threshold (e.g., 0.5), the probability value of each point cloud in the probability distribution map is converted into a binary segmentation mask. For example, if the probability value of a point cloud is greater than or equal to the pre-defined damage threshold (e.g., 0.5), it is marked as a "damaged value" (usually represented by 1). If the probability value of a point cloud is less than the pre-defined threshold, it is marked as an "undamaged value" (usually represented by 0).

[0163] Step B62: Based on the damage marker value, determine the damaged parts of the vehicle to be repaired.

[0164] The specific damaged area is determined based on the damaged marker value (i.e. the point with a value of 1) in the binary segmentation mask.

[0165] In the vehicle surface repair method proposed in this application, the complex probability distribution map is simplified into an intuitive binary segmentation mask, reducing the complexity of data processing. By setting a threshold, continuous probability values ​​are converted into discrete binary segmentation masks, ensuring accurate positioning of damaged areas.

[0166] In one embodiment, a computer-readable storage medium is provided having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the steps in the above method embodiments.

[0167] In one embodiment, an electronic device is also provided, including one or more processors; and a memory storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors perform the steps in the above method embodiments.

[0168] In one embodiment, such as Figure 6 The diagram illustrates the structure of an electronic device used to implement an embodiment of this application. The electronic device 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage portion 608 into a random access memory (RAM) 603. The RAM 603 also stores various programs and data required for the operation of the electronic device 600. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0169] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.

[0170] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer-readable medium carrying instructions that, in such embodiments, can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the instructions are executed by central processing unit (CPU) 601, the various method steps described in this application are performed.

[0171] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0172] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of this application and form different embodiments. For example, in the foregoing claims, any of the claimed embodiments can be used in any combination. The information disclosed in this background section is intended only to enhance the understanding of the general background of this application and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art.

Claims

1. A method for repairing vehicle surfaces, characterized in that, The method includes: Obtain damage morphology data of the damaged parts of the vehicle to be repaired; and Obtain the mirror image morphology data of the mirror image area corresponding to the damaged area; A damaged surface model is constructed based on the damaged morphology data, and a mirror surface model is constructed based on the mirror morphology data. The mirror surface model is mirrored to obtain the reference surface model; Perform a difference operation on the damaged surface model and the reference surface model to obtain a three-dimensional repair model formed by the space enclosed by the two surfaces; Based on the three-dimensional repair model, putty printing parameters are determined, and putty is printed on the damaged area based on the putty printing parameters.

2. The vehicle surface repair method according to claim 1, characterized in that, Before acquiring the damage morphology data of the damaged parts of the vehicle to be repaired, the following steps are included: Acquire the first image of the damaged surface of the vehicle to be repaired; Extract the edge contours of the first captured image, and apply morphological operations to enhance the edge contours to obtain the contour features of the supposedly damaged region; and / or Convert the first captured image to a grayscale image, and based on the grayscale image, determine the texture features of the region to be damaged; and / or The captured image is converted to a preset color space, and the color features of the supposedly damaged area are extracted based on the color space. Based on at least one of the contour features, texture features, and color features of the supposedly damaged region, the input features of the pre-trained vehicle damage recognition model are determined; The input features are input into the vehicle damage recognition model to obtain the segmentation result of the suspected damaged area output by the vehicle damage recognition model based on the input features; Based on the segmentation results, the damaged parts of the vehicle to be repaired are determined.

3. The vehicle surface repair method according to claim 2, characterized in that, Before obtaining the mirror image morphology data of the mirror region corresponding to the damaged area, the following steps are included: Using the main axis of symmetry of the vehicle to be repaired as a reference, determine the mirror image corresponding to the damaged surface of the vehicle to be repaired; A second image of the mirror surface is acquired, wherein the shooting points of the second image and the first image are mirror-symmetrical about the main axis of symmetry of the vehicle to be repaired. The second captured image is mirror-compared with the first captured image that identifies the damaged area; Based on the comparison results, the mirror image of the damaged area is determined.

4. The vehicle surface repair method according to claim 1, characterized in that, Before acquiring the damage morphology data of the damaged parts of the vehicle to be repaired, the following steps are included: Acquire the full-vehicle 3D scan data of the vehicle to be repaired, and extract the point cloud data of the damaged surface of the vehicle to be repaired from the full-vehicle 3D scan data; Extract the geometric features of the point cloud data, including curvature, normal vector, and edge points; The geometric features are mapped onto a two-dimensional feature image to form a geometric feature map; The geometric feature map is converted into a multidimensional feature vector, which is then used as the model input for the pre-trained vehicle damage discrimination model. The multidimensional feature vector is input into the vehicle damage discrimination model to obtain the probability distribution map of each point cloud data belonging to the damaged part, which is output by the vehicle damage discrimination model based on the multidimensional feature vector. Based on the probability distribution map, the damaged parts of the vehicle to be repaired are determined, and the damage morphology data of the damaged parts are extracted from the three-dimensional scanning data of the whole vehicle.

5. The vehicle surface repair method according to claim 4, characterized in that, The step of determining the damaged parts of the vehicle to be repaired based on the probability distribution map includes: Based on a preset damage threshold, the probability distribution map is converted into a binary segmentation mask, which includes damaged marker values ​​and undamaged marker values. Based on the damage marker value, the damaged parts of the vehicle to be repaired are determined.

6. The vehicle surface repair method according to claim 4, characterized in that, The step of obtaining the mirror image morphology data of the mirror image region corresponding to the damaged region includes: Based on the vehicle's 3D scanning data, a 3D data model of the vehicle is generated; Determine the location coordinates of the damaged part in the three-dimensional data model of the vehicle; Based on the geometric symmetry of the vehicle to be repaired, the mirror image of the damaged part is determined in the three-dimensional data model of the whole vehicle; The mirror image data of the mirror area is extracted from the three-dimensional scanning data of the whole vehicle.

7. The vehicle surface repair method according to any one of claims 1 to 6, characterized in that, The determination of putty printing parameters based on the three-dimensional repair model includes: Based on the key features of the three-dimensional repair model, the three-dimensional repair model is divided into several sub-regions; The geometric complexity and thickness variation of each sub-region are calculated separately, and the corresponding printing parameter combination is determined based on the geometric complexity and thickness variation. The printing parameter combination includes nozzle diameter, material output speed, printing path planning and layer printing parameters. Based on the combination of printing parameters corresponding to each sub-region, the putty printing parameters corresponding to the damaged area are determined.

8. The vehicle surface repair method according to claim 7, characterized in that, The step of determining the corresponding combination of printing parameters based on the geometric complexity and the thickness variation includes: The nozzle diameter is determined based on the geometric complexity, wherein the nozzle diameter is inversely proportional to the geometric complexity; Based on the thickness variation, determine the layer printing parameters; The output speed is determined based on the layered printing parameters; Based on the nozzle diameter, the discharge speed, and the layered printing parameters, a printing path is planned.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors perform the vehicle surface repair method as described in any one of claims 1 to 8.

10. A storage medium, characterized in that, The storage medium stores executable instructions that, when executed by a processor, cause the processor to perform the vehicle surface repair method as described in any one of claims 1 to 8.

Images