A sample labeling method and device

By generating and automatically labeling sample images rendered by the three-dimensional model, the problem of high manual annotation cost is solved, automatic sample annotation is realized, the cost of neural network training is reduced, and detection accuracy and efficiency are improved.

CN113658345BActive Publication Date: 2024-12-06GUIZHOU YUAN YANG INTELLIGENT TECHNOLOGY CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202110947491.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-08-18
Publication Date
2024-12-06
Estimated Expiration
2041-08-18

AI Technical Summary

Technical Problem

In the prior art, a large number of manual labeling of sample images is required when the surface of the detection object is damaged, resulting in high labor cost for neural network training and high cost for professional testing equipment.

Method used

By acquiring the three-dimensional model, the sample image with damage is generated, and the damaged position is automatically marked according to the damaged position in the three-dimensional model, automatic sample annotation is realized and labor costs are reduced.

Benefits of technology

No need to collect or manually label sample images, effectively reducing the labor cost of training a damaged neural network and improving the accuracy and efficiency of detection results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113658345B_ABST
    Figure CN113658345B_ABST
Patent Text Reader

Abstract

The embodiments of the present application provide a sample annotation method and device. The method includes: obtaining a three-dimensional model of a sample object with damage; rendering the three-dimensional model to obtain a sample image of the sample object; determining the location of the damage in the sample image as the target location based on the location of the damage in the three-dimensional model; and annotating the target location in the sample image. A sample image with damage can be generated by rendering a three-dimensional model with damage, and the location of the damage in the sample image can be calculated based on the location of the damage in the three-dimensional model, thereby automatically completing the annotation of the sample image. There is no need to collect sample images, nor is there any need to manually annotate the sample images, so the labor cost required to train a neural network for detecting damage can be effectively reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of machine vision technology, and in particular to a sample labeling method and device. Background Art

[0002] For various reasons, the surface of some objects may be damaged, making them unable to be used normally. For example, the road surface may have cracks, potholes and other damage due to long-term weighing, affecting the flatness of the road surface and causing vehicles to bump when driving on the road. For another example, the pressure on the pipe wall and the temperature change may cause damage, affecting the airtightness and pressure resistance of the pipe, causing leakage during transportation.

[0003] Although professional testing equipment can be used to detect damage on an object, the cost of professional testing equipment is often high. Therefore, in order to reduce the hardware cost of detection, the related art can take an image of the surface of the object, and input the image into a pre-trained neural network for identifying damage, and determine whether the surface of the object is damaged based on the output of the neural network.

[0004] However, in order to improve the accuracy of the results output by the neural network, a large number of labeled sample images are needed to train the neural network, and the labeling of images is often done manually, resulting in high labor costs for training the neural network. Summary of the invention

[0005] The purpose of the embodiments of the present application is to provide a sample labeling method and device to realize automatic labeling of samples, thereby reducing the labor cost required for training a neural network for identifying damage. The specific technical solution is as follows:

[0006] In a first aspect of an embodiment of the present application, a sample labeling method is provided, the method comprising:

[0007] Obtaining a three-dimensional model of the sample object with damage;

[0008] Rendering the three-dimensional model to obtain a sample image of the sample object;

[0009] According to the damaged position in the three-dimensional model, determining the damaged position in the sample image as the target position;

[0010] The target position is marked in the sample image.

[0011] In a possible embodiment, obtaining a damaged three-dimensional model of the sample object includes:

[0012] constructing a three-dimensional model of the sample object without damage;

[0013] A damage effect is added to the outer surface of the three-dimensional model to obtain a three-dimensional model of the sample object with damage.

[0014] In a possible embodiment, adding a damage effect to the outer surface of the three-dimensional model includes:

[0015] Acquire a texture pattern for representing the texture generated on the outer surface of the sample object when it is damaged;

[0016] According to the texture pattern, a texture represented by the texture pattern is added to the outer surface of the three-dimensional model.

[0017] In a possible embodiment, determining the damaged position in the sample image as the target position according to the damaged position in the three-dimensional model includes:

[0018] Rendering the three-dimensional model using preset colors to obtain a color segmentation image;

[0019] According to the preset color, an area where the preset color is missing is determined in the color segmentation image as a target position.

[0020] In a possible embodiment, rendering the three-dimensional model to obtain a sample image of the sample object includes:

[0021] Setting the three-dimensional model in a preset virtual three-dimensional scene;

[0022] The preset virtual three-dimensional scene is rendered to obtain a sample image of the sample object.

[0023] In a second aspect of an embodiment of the present application, a sample labeling device is provided, the device comprising:

[0024] A model acquisition module, used to acquire a damaged three-dimensional model of the sample object;

[0025] A rendering module, used for rendering the three-dimensional model to obtain a sample image of the sample object;

[0026] A position determination module, used to determine the position where the damage exists in the sample image according to the position where the damage exists in the three-dimensional model, as the target position;

[0027] A marking module is used to mark the target position in the sample image.

[0028] In a possible embodiment, the model acquisition module acquires the damaged three-dimensional model of the sample object, including:

[0029] constructing a three-dimensional model of the sample object without damage;

[0030] A damage effect is added to the outer surface of the three-dimensional model to obtain a three-dimensional model of the sample object with damage.

[0031] In a possible embodiment, the model acquisition module adds a damage effect to the outer surface of the three-dimensional model, including:

[0032] Acquire a texture pattern for representing the texture generated on the outer surface of the sample object when it is damaged;

[0033] According to the texture pattern, a texture represented by the texture pattern is added to the outer surface of the three-dimensional model.

[0034] In a possible embodiment, the position determination module determines the position where the damage exists in the sample image as the target position according to the position where the damage exists in the three-dimensional model, including:

[0035] Rendering the three-dimensional model using preset colors to obtain a color segmentation image;

[0036] According to the preset color, an area where the preset color is missing is determined in the color segmentation image as a target position.

[0037] In a possible embodiment, the rendering module renders the three-dimensional model to obtain a sample image of the sample object, including:

[0038] Setting the three-dimensional model in a preset virtual three-dimensional scene;

[0039] The preset virtual three-dimensional scene is rendered to obtain a sample image of the sample object.

[0040] In a third aspect of the embodiments of the present application, an electronic device is provided, including:

[0041] Memory, used to store computer programs;

[0042] The processor is used to implement any method step described in the first aspect when executing the program stored in the memory.

[0043] In a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, any method step described in the first aspect is implemented.

[0044] Beneficial effects of the embodiments of the present application:

[0045] The sample annotation method and device provided in the embodiment of the present application can generate a sample image with damage by rendering a three-dimensional model with damage, and calculate the position of the damage in the sample image according to the position of the damage in the three-dimensional model, thereby automatically completing the annotation of the sample image. There is no need to collect sample images or manually annotate sample images, so the labor cost required for training a neural network for detecting damage can be effectively reduced.

[0046] Of course, implementing any product or method of the present application does not necessarily require achieving all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application, and for ordinary technicians in this field, other embodiments can also be obtained based on these drawings.

[0048] Figure 1 A schematic diagram of a flow chart of a sample labeling method provided in an embodiment of the present application;

[0049] Figure 2 A schematic diagram of a process of constructing a damaged three-dimensional model provided in an embodiment of the present application;

[0050] Figure 3 A schematic diagram of the structure of a sample labeling device provided in an embodiment of the present application;

[0051] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field based on the present application belong to the scope of protection of the present application.

[0053] In order to more clearly illustrate the sample annotation method provided in the embodiment of the present application, a possible application scenario of the sample annotation method provided in the embodiment of the present application will be exemplified below. It can be understood that the following example is only one possible application scenario of the sample annotation method provided in the embodiment of the present application. In other possible embodiments, the sample annotation method provided in the embodiment of the present application can also be applied to other possible application scenarios, and the following example does not impose any limitation on this.

[0054] For various reasons, the road surface may be damaged, that is, road defects occur. Road defect detection is an important task in urban governance. The current road detection method is to purchase millions of professional detection equipment vehicles for on-road detection. The use of deep learning for detection can greatly save costs, and only ordinary cameras are needed to complete the work. The deep learning solution requires a large amount of data for model training, and the collection of actual road defect data requires a lot of manpower and time. At the same time, the collected data needs to be manually calibrated, which also requires a lot of manpower costs.

[0055] Based on this, the present application embodiment provides a sample annotation method. The sample annotation method provided in the present application embodiment can be applied to any device with sample annotation capability. The sample annotation method provided in the present application embodiment can be referred to Figure 1 , Figure 1 The figure is a schematic diagram of a flow chart of a sample labeling method provided in an embodiment of the present application, which may include:

[0056] S101, obtaining a damaged three-dimensional model of a sample object.

[0057] S102, rendering the three-dimensional model to obtain a sample image of the sample object.

[0058] S103, determining the damaged position in the sample image according to the damaged position in the three-dimensional model as the target position.

[0059] S104, marking the target position in the sample image.

[0060] By selecting this embodiment, a sample image with damage can be generated by rendering a 3D model with damage, and the position of the damage in the sample image can be calculated according to the position of the damage in the 3D model, thereby automatically completing the labeling of the sample image. There is no need to collect sample images or manually label the sample images, so the labor cost required for training a neural network for detecting damage can be effectively reduced.

[0061] Among them, in S101, the damaged three-dimensional model can be constructed by the execution subject of the sample annotation method provided in the embodiment of the present application, or it can be constructed by other electronic devices other than the execution subject. How to construct the damaged three-dimensional model will be described in detail below and will not be repeated again.

[0062] The sample objects may be different types of objects depending on the application scenario, including but not limited to road surfaces, pipe walls, etc. And depending on the sample objects, the damage in the 3D model may be different types of damage, but the damage in the 3D model should be the damage that may exist in the sample object. For example, assuming that the sample object may not have damage of a specific texture due to material reasons, the damage in the 3D model should not be the damage of the specific texture.

[0063] Taking the sample object as a road surface as an example, the damage in the three-dimensional model may include but is not limited to the following texture damage: block cracks, strip cracks, and potholes.

[0064] In S102, the rendering method may be different depending on the application scenario. For example, it may be raster rendering or rendering based on ray tracing technology. In a possible embodiment, in order to improve the quality of the rendered sample image, a rendering method based on ray tracing technology may be used to render the three-dimensional model.

[0065] It is understandable that in the actual working process, the input image of the neural network used to detect damage may include other objects in addition to the target object to be detected. For example, taking the target object as the road surface, the input image may also include objects such as vehicles and manhole covers.

[0066] Therefore, in order to make the sample image closer to the image input during the actual working process and to make the trained neural network for detecting damage more accurate, in a possible embodiment, the three-dimensional model can be set in a preset virtual three-dimensional scene, and the preset virtual three-dimensional scene can be rendered to obtain a sample image of the sample object.

[0067] In addition to the damaged 3D model of the sample object, the preset virtual 3D scene may also include 3D models or textures of other objects other than the sample object, such as 3D models of vehicles, 3D models of manhole covers, textures of road surface materials, textures of road markings, etc. In addition, the light of the preset virtual 3D scene and the material of the 3D model of each object may be set according to actual needs or user experience.

[0068] In S103, it can be understood that since the three-dimensional model is known, the location where damage exists in the three-dimensional model can be determined from the three-dimensional model, and rendering the three-dimensional model can be regarded as a mapping from the three-dimensional model to the two-dimensional image. Therefore, when the rendering method is known, in theory, the location where damage exists in the three-dimensional model can be mapped from the three-dimensional model to the two-dimensional image according to the mapping, and the location to which the location where damage exists is mapped is the location where damage exists in the sample image.

[0069] Exemplarily, in a possible embodiment, the three-dimensional model may be rendered using preset colors to obtain a color segmentation image, and based on the preset colors, an area lacking the preset colors is determined in the color segmentation image as the target area.

[0070] The preset color may be any color, such as green, blue, red, etc. Moreover, for an embodiment in which a 3D model is set in a preset virtual 3D scene and the preset virtual 3D scene is rendered to obtain a sample image of a sample object, in order to distinguish the 3D model of the sample object from the 3D models or textures of other objects in the preset virtual 3D scene, different identifiers may be set for the 3D models or textures of different objects and rendered in different colors.

[0071] For example, assuming that the sample object is a road surface, and in addition to the three-dimensional model of the road surface, there are also three-dimensional models of manhole covers and vehicles in the preset three-dimensional scene, then when rendering to obtain a color segmentation image, the three-dimensional model of the road surface may be rendered using a first preset color, the three-dimensional model of the manhole cover may be rendered using a second preset color, and the three-dimensional model of the vehicle may be rendered using a third preset color, wherein the first preset color, the second preset color, and the third preset color may be any three different colors. In the obtained color segmentation image, the road surface, the manhole cover, and the vehicle are different colors, respectively, so the road surface, the manhole cover, and the vehicle can be distinguished from the color segmentation image based on color.

[0072] After rendering the color segmentation image, the color and position of the damaged object in the color segmentation image can be determined based on the identification of the damaged object, and based on the position of the damaged object in the color segmentation image, the area where the color of the damaged object in the color segmentation image is missing can be determined as the target position. For example, if the damaged object is a road surface, and the color of the road surface in the color segmentation image is a first preset color, then the area where the first preset color is missing can be determined from the area where the road surface is located as the target position. The method of determining the area where the color is missing can be different depending on the application scenario. For example, the minimum rectangular area surrounding all the pixels of the missing color can be calculated as the target position, or all the pixels of the missing color can be clustered according to a preset clustering method, and the cluster center can be used as the target position.

[0073] In order to more clearly illustrate the sample annotation method provided in the embodiment of the present application, the construction of the three-dimensional model will be described below. Figure 2 , Figure 2 The figure is a flow chart of a three-dimensional model construction method provided in an embodiment of the present application, which may include:

[0074] S201, constructing a three-dimensional model of the sample object without damage.

[0075] The three-dimensional model without damage may refer to a theoretical model of the sample object when no damage occurs. For example, taking the sample object as a road surface, the three-dimensional model without damage may be a plane model. For another example, taking the sample object as a pipe wall, the three-dimensional model without damage may be a cylindrical model.

[0076] S202, adding a damage effect to the outer surface of the three-dimensional model to obtain a three-dimensional model of the sample object with damage.

[0077] The added damage effect should be a damage effect corresponding to the damage that may exist in the sample object, and only one damage effect may be added, or multiple different damage effects may be added. In a possible embodiment, a texture pattern for representing the outer surface of the sample object when it is damaged may be obtained, and a texture represented by the texture pattern may be added to the outer surface of the three-dimensional model according to the texture pattern.

[0078] For example, assuming that it is necessary to increase the block crack effect, the acquired texture pattern can be a two-dimensional lattice image. The method of generating the two-dimensional lattice image can be different depending on the application scenario. For example, a two-dimensional lattice image can be randomly generated by a Voronoi image algorithm, and the user can control the generated two-dimensional lattice image by adjusting the relevant parameters of the Voronoi image algorithm according to actual needs or experience.

[0079] For another example, assuming that a crack effect needs to be added, the obtained texture pattern can be a curve. The method of generating the curve can be different depending on the application scenario. For example, a straight line can be converted into a randomly shaped curve through one-dimensional Oerlin noise and fractal algorithm image algorithm.

[0080] For another example, assuming that a hole effect needs to be added, the acquired texture pattern can be a binary texture. The method of generating the binary texture can be different depending on the application scenario. For example, an image containing a binary texture can be generated by a two-dimensional Perlin noise plus a fractal algorithm, and the binary texture can be obtained by intercepting the middle area of ​​the image.

[0081] See also Figure 3 , Figure 3 The figure is a schematic diagram of a structure of a sample labeling device provided in an embodiment of the present application, which may include:

[0082] The model acquisition module 301 is used to acquire a three-dimensional model of the sample object with damage;

[0083] A rendering module 302, used for rendering the three-dimensional model to obtain a sample image of the sample object;

[0084] A position determination module 303, configured to determine the position of the damage in the sample image according to the position of the damage in the three-dimensional model as the target position;

[0085] The marking module 304 is used to mark the target position in the sample image.

[0086] In a possible embodiment, the model acquisition module acquires the damaged three-dimensional model of the sample object, including:

[0087] constructing a three-dimensional model of the sample object without damage;

[0088] A damage effect is added to the outer surface of the three-dimensional model to obtain a three-dimensional model of the sample object with damage.

[0089] In a possible embodiment, the model acquisition module adds a damage effect to the outer surface of the three-dimensional model, including:

[0090] Acquire a texture pattern for representing the texture generated on the outer surface of the sample object when it is damaged;

[0091] According to the texture pattern, a texture represented by the texture pattern is added to the outer surface of the three-dimensional model.

[0092] In a possible embodiment, the position determination module determines the position where the damage exists in the sample image as the target position according to the position where the damage exists in the three-dimensional model, including:

[0093] Rendering the three-dimensional model using preset colors to obtain a color segmentation image;

[0094] According to the preset color, an area where the preset color is missing is determined in the color segmentation image as a target position.

[0095] In a possible embodiment, the rendering module renders the three-dimensional model to obtain a sample image of the sample object, including:

[0096] Setting the three-dimensional model in a preset virtual three-dimensional scene;

[0097] The preset virtual three-dimensional scene is rendered to obtain a sample image of the sample object.

[0098] The present application also provides an electronic device, such as Figure 4 As shown, including:

[0099] Memory 401, used for storing computer programs;

[0100] The processor 402 is used to execute the program stored in the memory 401 to implement the following steps:

[0101] Obtaining a three-dimensional model of the sample object with damage;

[0102] Rendering the three-dimensional model to obtain a sample image of the sample object;

[0103] According to the damaged position in the three-dimensional model, determining the damaged position in the sample image as the target position;

[0104] The target position is marked in the sample image.

[0105] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0106] The communication interface is used for communication between the above electronic device and other devices.

[0107] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0108] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0109] In another embodiment provided in the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned sample labeling methods are implemented.

[0110] In another embodiment provided in the present application, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any sample labeling method in the above embodiments.

[0111] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0112] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants 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, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0113] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of the device, electronic device, computer-readable storage medium, and computer program product, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0114] The above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application are included in the protection scope of the present application.

Claims

1. A sample labeling method, characterized in that: The method comprises: Obtaining a three-dimensional model of the sample object with damage; Rendering the three-dimensional model to obtain a sample image of the sample object; According to the damaged position in the three-dimensional model, determining the damaged position in the sample image as the target position; marking the target position in the sample image; The step of obtaining a damaged three-dimensional model of the sample object comprises: Constructing a three-dimensional model of the sample object without damage, wherein the three-dimensional model without damage refers to a theoretical model of the sample object when no damage occurs; Adding a damage effect to the outer surface of the three-dimensional model to obtain a three-dimensional model of the sample object with damage, wherein the damage effect added to the outer surface of the three-dimensional model is at least one, and the damage effect is a damage effect corresponding to the damage that may exist in the sample object, and the damage effect includes a block crack effect, a strip crack effect, and a pothole effect; Adding a damage effect to the outer surface of the three-dimensional model includes: Randomly generating a texture pattern for representing the texture generated on the outer surface of the sample object when it is damaged by an algorithm, wherein the texture pattern includes a two-dimensional lattice image, a curve, and a binary texture; According to the texture pattern, adding a texture represented by the texture pattern to the outer surface of the three-dimensional model; The step of determining the damaged position in the sample image as the target position according to the damaged position in the three-dimensional model includes: Rendering the three-dimensional model using preset colors to obtain a color segmentation image; According to the preset color, an area where the preset color is missing is determined in the color segmentation image as a target position.

2. The method according to claim 1, characterized in that The rendering of the three-dimensional model to obtain a sample image of the sample object includes: Setting the three-dimensional model in a preset virtual three-dimensional scene; The preset virtual three-dimensional scene is rendered to obtain a sample image of the sample object.

3. A sample labeling device, characterized in that: The device comprises: A model acquisition module, used to acquire a damaged three-dimensional model of the sample object; A rendering module, used for rendering the three-dimensional model to obtain a sample image of the sample object; A position determination module, used to determine the position where the damage exists in the sample image according to the position where the damage exists in the three-dimensional model, as the target position; A marking module, used for marking the target position in the sample image; The model acquisition module acquires a damaged three-dimensional model of the sample object, including: Constructing a three-dimensional model of the sample object without damage, wherein the three-dimensional model without damage refers to a theoretical model of the sample object when no damage occurs; Adding a damage effect to the outer surface of the three-dimensional model to obtain a three-dimensional model of the sample object with damage, wherein the damage effect added to the outer surface of the three-dimensional model is at least one, and the damage effect is a damage effect corresponding to the damage that may exist in the sample object, and the damage effect includes a block crack effect, a strip crack effect, and a pothole effect; The model acquisition module adds a damage effect to the outer surface of the three-dimensional model, including: randomly generating a texture pattern through an algorithm to represent the texture produced on the outer surface of the sample object when it is damaged; According to the texture pattern, adding a texture represented by the texture pattern to the outer surface of the three-dimensional model; The position determination module determines the damaged position in the sample image as the target position according to the damaged position in the three-dimensional model, including: Rendering the three-dimensional model using preset colors to obtain a color segmentation image; According to the preset color, an area where the preset color is missing is determined in the color segmentation image as a target position.

4. The device according to claim 3, characterized in that The rendering module renders the three-dimensional model to obtain a sample image of the sample object, including: Setting the three-dimensional model in a preset virtual three-dimensional scene; The preset virtual three-dimensional scene is rendered to obtain a sample image of the sample object.

Citation Information

Patent Citations

  • Image data generation method and device

    CN110428388A

  • Image training sample generation method and device

    CN111680758A

  • Steel rail defect image generation method based on 3D model and point cloud processing

    CN113240790A

  • Way to generate tight 2d bounding boxes for autonomous driving labeling

    US20210150226A1

  • Component search and repair device and method for maintaining partial damage of component, and 3D printing-based component maintenance system

    WO2019117450A1