Road long-tail disease data synthesis method based on pavement degradation logic and chain evolution

Abstract

The application provides a road long-tail disease data synthesis method based on road surface degradation logic and chain evolution, and belongs to the technical field of computer vision and intelligent transportation. The method comprises the following steps: preprocessing an original road surface image containing long-tail diseases to obtain a core pixel block; determining a secondary crack area to generate a secondary crack mask; generating secondary disease pixel blocks from the host road surface image according to the secondary crack mask; combining virtual depth and self-occlusion shadow intensity compensation coefficients to correct the core pixel block to obtain a corrected core pixel block; using feathering masks, frequency domain and brightness alignment to realize consistent fusion of the synthetic disease blocks and the host road surface image to obtain a final synthesis image; and automatically synchronously generating a detection label based on the external contour of the feathering mask. The application solves the problems of the scarcity of serious disease samples and the lack of physical authenticity in traditional synthesis, and the generated samples conform to the road surface mechanics degradation law, thereby significantly improving the detection reliability of a deep learning model in a long-tail disease scene.

Description

Technical Field

[0001] This invention belongs to the field of computer vision and intelligent transportation technology, and in particular relates to a method for synthesizing road long-tail defects data based on road degradation logic and chain evolution. Background Technology

[0002] Automatic road defect identification is a key technology for the maintenance of transportation infrastructure. In the field of road maintenance and defect detection, road defect datasets are the core foundation for training defect identification models and realizing automated defect detection and hierarchical assessment. The completeness and authenticity of the samples directly determine the accuracy of the detection model and its practical application effect. However, in the road defect data currently collected, severe defects such as giant potholes have a low frequency of occurrence and exhibit a typical long-tail distribution, making it difficult for deep learning models to generalize due to insufficient samples. There is a general problem of scarce severe defect samples. In actual engineering scenarios, road defect datasets generally exhibit long-tail distribution characteristics. Samples of routine minor defects such as shallow fine lines and minor wear are easy to collect and plentiful. However, severe defects such as potholes, network cracks, and structural subsidence, as well as complex defects, are long-tail defects. Due to the timely handling of routine road maintenance, the actual number of effective samples that can be collected is extremely small, resulting in a serious sample imbalance problem. This directly leads to insufficient feature learning of long-tail defects by the detection model, significantly reducing the identification accuracy and making it difficult to meet the actual needs of rapid early warning and precise handling of high-risk defects in road maintenance.

[0003] To alleviate the imbalance in road damage samples, existing technologies often employ a cut-and-paste synthesis method to artificially expand scarce damage samples. This involves cropping the collected damage area from the original image and pasting it into a designated location on a normal road image to generate a synthetic sample. While this simple "cut-and-paste" technique incorporates perspective correction, it essentially treats damage as isolated image stickers. This method has two significant drawbacks: First, it ignores the logic of physical evolution. Real damage is often accompanied by secondary cracks caused by stress diffusion; isolated stickers cannot simulate the characteristics of real damage communities, resulting in a lack of physical realism in the synthetic sample. Second, environmental interaction is lacking. Existing methods cannot accurately reproduce the self-occluding shadows caused by depth inside potholes, leading to a distribution gap between the synthetic image and the real sample in the feature space. Summary of the Invention

[0004] The purpose of this invention is to provide a method for synthesizing long-tail road defects based on pavement degradation logic and chain evolution. By simulating stress influence zones and virtual depth fields, it achieves a three-in-one chain synthesis of "central defects - peripheral degradation - depth light and shadow", which significantly improves the physical realism of the synthesized samples. This solves the technical problems of scarcity of severe defect samples in existing road defect datasets and lack of physical realism in traditional clipping synthesis methods.

[0005] To solve the above-mentioned technical problems, the specific technical solution of the present invention is as follows:

[0006] A method for synthesizing long-tail road defect data based on pavement degradation logic and cascading evolution includes the following steps:

[0007] Step S1: Preprocess the original pavement image containing long-tail defects to obtain core pixel blocks;

[0008] Step S2: Based on the stress influence radius of the core pixel block, determine the secondary crack area and generate a secondary crack mask; the host road surface image generates secondary disease pixel blocks based on the secondary crack mask.

[0009] Step S3: Combine virtual depth and self-occlusion shadow intensity compensation coefficient to correct the core pixel block and obtain the corrected core pixel block;

[0010] Step S4: Use feathering mask and frequency domain and brightness alignment to achieve consistent fusion of the synthesized disease block and the host road surface image to obtain the final synthesized image;

[0011] Step S5: Automatically and synchronously generate detection labels based on the outer contour of the feathered mask.

[0012] Further, step S1 includes the following steps:

[0013] Step S11: Use an image semantic segmentation and annotation tool to extract the mask from the original road surface image containing long-tail defects, and obtain the binary mask of the central defect.

[0014] Step S12: Perform pixel-level logical AND operation or mask mapping cropping on the binary mask and the original road surface image to obtain core pixel blocks that retain only the texture information of the defects.

[0015] Further, step S2 includes the following steps:

[0016] Step S21: Based on the core pixel block, introduce the stress diffusion coefficient to characterize the looseness of the roadbed, calculate the stress influence radius, and the circumference formed by the stress influence radius is the stress influence boundary.

[0017] Step S22: Determine the secondary crack region based on the geometric equivalent radius and stress influence radius of the core pixel block, and generate a secondary crack mask within the secondary crack region;

[0018] Step S23: The host road surface image generates secondary disease pixel blocks based on the secondary crack mask.

[0019] Further, step S3 includes the following steps:

[0020] Step S31: Calculate the virtual depth at any point within the secondary crack region;

[0021] Step S32: Solve for the ambient light direction vector of the host road surface image;

[0022] Step S33: Calculate the self-occlusion shadow intensity compensation coefficient at any point inside the stress-affected boundary;

[0023] Step S34: Correct the core pixel block using the self-occlusion shadow intensity compensation coefficient to obtain the corrected core pixel block.

[0024] Further, step S4 includes the following steps:

[0025] Step S41: Spatially superimpose and combine the corrected core pixel block with the secondary disease pixel block to obtain a composite disease block;

[0026] Step S42: Perform Gaussian blurring on the binary mask of the central lesion and the mask of the secondary crack to obtain the feathered mask;

[0027] Step S43: Determine the disease implantation area in the host road surface image, extract the original texture features of the disease implantation area in the host road surface image, and align the frequency and brightness of the synthesized disease block with the host road surface image to obtain the final synthesized image.

[0028] Furthermore, in step S21, the formula for calculating the stress influence radius is:

[0029]

[0030] in, Indicates the radius of stress influence; Represents the geometric equivalent radius of the core pixel block; This represents the stress diffusion coefficient.

[0031] Further, in step S23, the secondary disease pixel block is generated in the following manner:

[0032]

[0033] in, Represents the secondary disease pixel block; The aging texture of the host road surface image; This is the contrast attenuation factor; It represents the Hadamardi (or Hadama) stack; This represents the secondary crack mask.

[0034] Furthermore, in step S33,

[0035] The self-occlusion shadow intensity compensation coefficient at any point inside the stress-affected boundary is calculated as follows:

[0036]

[0037] in, Represents any point within the stress-affected boundary. The self-occlusion shadow intensity compensation coefficient; The illumination elevation angle, representing the ambient light direction vector; Indicates the shadow depth attenuation coefficient; This represents the virtual depth of any point within the secondary crack region.

[0038] Furthermore, in step S34, the self-occlusion shadow intensity compensation coefficient is corrected for the core pixel block in the following manner:

[0039]

[0040] in, This indicates the corrected core pixel block; Indicates the core pixel block; It represents the Hadamardi (or Hadama) stack.

[0041] Furthermore, in step S43, the final composite image is synthesized in the following manner:

[0042]

[0043] in, This represents the final composite image; This represents the Hadamard product operation; Indicates the feathering mask; Represents the aging texture of the host road surface image; Represents the global brightness alignment factor; This indicates a synthetic diseased block.

[0044] Compared with the prior art, the present invention has the following beneficial technical effects:

[0045] 1) This invention overcomes the limitations of traditional synthesis methods that lack physical evolution logic, significantly improving the physical realism of pavement defect samples. Existing technologies typically treat defects as isolated image stickers, ignoring the community characteristics of pavement defects. This invention constructs a stress-affected zone (SIZ) and generates secondary defects, using stress attenuation logic to generate secondary cracks and texture distortion around the core defect. This chain evolution mechanism of "center defect - peripheral degradation" conforms to the laws of pavement mechanical degradation, making the synthesized samples more closely resemble the aging state of real pavement in terms of texture structure.

[0046] 2) This invention addresses the issues of "flat sticker-like" appearance and inconsistent lighting in traditional synthetic images, enhancing visual integration. To address the inability of existing methods to reproduce the self-occluding shadows generated by depth within potholes, this invention establishes a lighting and shadow adaptive model based on a virtual depth field. By constructing a normalized virtual depth and combining it with the ambient light direction vector of the host road image, the intensity of the self-occluding shadow inside the pothole is dynamically calculated. This ensures that the shadow inside the pothole dynamically changes with depth and illumination angle, achieving a visual leap from "two-dimensional plane" to "three-dimensional perception," and guaranteeing a high degree of consistency in lighting and shadow characteristics between the synthetic pothole and the host background.

[0047] 3) This invention effectively alleviates the "long-tail effect" caused by the scarcity of severe disease samples, significantly improving the reliability and generalization ability of the detection model. By generating high-quality, highly diverse synthetic data, this invention solves the problem of insufficient samples for severe diseases such as giant craters. Experiments show that after introducing the synthetic data of this invention, the average detection accuracy (mAP@0.5) of deep learning models (such as YOLOv8) is significantly improved (e.g., from 72.1% to 76.5%), while the false alarm rate is reduced by more than 17%. This proves that the samples generated by this invention can effectively fill the distribution gap in the feature space and significantly enhance the detection performance of the model in complex long-tail scenarios. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart of the road defect data synthesis method based on pavement degradation logic and chain evolution of the present invention.

[0050] Figure 2 This is a schematic diagram of the evolution logic of the stress-affected zone (SIZ) of the present invention.

[0051] Figure 3 This is a schematic diagram comparing the feature maps of the traditional copy-paste method and the synthesis effect of the present invention.

[0052] Figure 4 This diagram illustrates the improvement in detection accuracy on the model after data synthesis. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] This invention proposes a method for synthesizing long-tail road defect data based on pavement degradation logic and cascading evolution, such as... Figure 1 As shown, the method includes the following steps:

[0055] Step S1: Preprocess the original road surface image containing long-tail defects to obtain core pixel blocks.

[0056] Step S11: Use an image semantic segmentation and annotation tool to extract the mask from the original road surface image containing long-tail defects, and obtain the binary mask of the central defect.

[0057] Specifically, the input data consists of raw pavement images containing long-tail defects (such as giant potholes) collected in real-world scenarios. Image semantic segmentation and annotation tools (such as LabelMe or CVAT) are used as mask extraction tools to perform pixel-level contour annotation on the defect areas in the raw pavement images, outputting a binary mask describing the geometry of the defects. The pixel value of the diseased area is 1, and the pixel value of the background is 0.

[0058] Step S12: Perform pixel-level logical AND operation or mask mapping cropping on the binary mask and the original road surface image to obtain core pixel blocks that retain only the texture information of the defects.

[0059] Specifically, the binary mask Pixel-level logical AND operations or mask mapping are performed with the original road surface image to crop the image, thereby filtering out background noise and extracting core pixel blocks that retain only the texture information of the road surface. The area where the core pixel block is located represents the central disease.

[0060] Step S2: Based on the stress influence radius of the core pixel block, determine the secondary crack area and generate a secondary crack mask; the host road surface image generates secondary disease pixel blocks based on the secondary crack mask.

[0061] This invention is based on the principle of pavement mechanical degradation and simulates the cascading degradation characteristics around the disease. Its evolution logic is as follows: Figure 2 As shown. The disease does not exist in isolation, but rather originates from a central disease and generates a destructive field in all directions. The specific implementation process is as follows:

[0062] Step S21: Based on the core pixel block, introduce the stress diffusion coefficient to characterize the looseness of the roadbed, calculate the stress influence radius, and the circumference formed by the stress influence radius is the stress influence boundary.

[0063] Specifically, the total area of ​​non-zero pixels representing the diseased region in the binary mask of the central disease is calculated. According to the formula Obtain the geometric equivalent radius of the core pixel block Introducing the stress diffusion coefficient, which characterizes the degree of roadbed looseness. (usually set) The stress influence radius is calculated using the following formula:

[0064]

[0065] in, Indicates the radius of stress influence; Represents the geometric equivalent radius of the core pixel block; This represents the stress diffusion coefficient.

[0066] The circumference formed by the stress influence radius is the stress influence boundary.

[0067] Step S22: Determine the secondary crack region based on the geometric equivalent radius and stress influence radius of the core pixel block, and generate a secondary crack mask within the secondary crack region.

[0068] Specifically, using the geometric equivalent radius of the core pixel block For inner diameter and stress influence radius The annular region with an outer diameter This is the secondary crack region, also known as the stress-affected zone (SIZ). Within the secondary crack region, fractal geometry algorithms (such as random walk or fractal growth models) are used to simulate the random propagation characteristics of the cracks, generating a secondary crack mask with a natural branching structure. .

[0069] Step S23: The host road surface image generates secondary disease pixel blocks based on the secondary crack mask.

[0070] Specifically, the host road surface image refers to the road image to be synthesized, which is used to implant target defect features and provide original texture background and ambient light field information. To ensure that the generated cracks blend naturally with the road surface background, following the aging texture logic and stress evolution logic of the host road surface image, secondary defect pixel blocks are generated through texture transfer. The texture transfer calculation formula is as follows:

[0071]

[0072] in, Represents the secondary disease pixel block; The aging texture of the host road surface image; This represents the Hadamard product, which is a pixel-wise matrix multiplication operation used to precisely fill the texture into the mask area; This is the contrast attenuation factor, and its value range is usually [value range missing]. .

[0073] The contrast attenuation factor is used to reduce the brightness and contrast of local areas, thereby simulating the texture distortion and aging of road materials under long-term stress compression and weathering (such as oil stains, blackening, etc.), ensuring that the synthesized secondary diseases have a physical realism.

[0074] Step S3: Combine virtual depth and self-occlusion shadow intensity compensation coefficient to correct the core pixel block and obtain the corrected core pixel block.

[0075] Step S31: Calculate the virtual depth at any point within the secondary crack region.

[0076] Specifically, the Euclidean distance from any point within the secondary crack region to the stress influence boundary is calculated, and the shortest Euclidean distance is normalized to... The interval is used to obtain any point. virtual depth ;in, This represents the x-coordinate of any point within the secondary crack region. This represents the ordinate of any point within the secondary crack region, with the center point having the greatest depth.

[0077] Generated virtual depth The larger the value, the farther the location is from the stress influence boundary (the edge of the pit), that is, the deeper the simulated physical depth, thus presenting a bowl-shaped topology structure with "shallow edges and deep center".

[0078] Step S32: Solve for the ambient light direction vector of the host road surface image.

[0079] Specifically, in the illumination consistency processing, this invention adopts an automatic calculation strategy based on physical geometric features: First, it identifies vertical reference objects (such as signposts) in the host road image and their shadow projection areas on the road surface; by establishing a projection vector from the root of the vertical reference object to the vertex of the shadow end, its reverse direction is determined as the horizontal azimuth angle of the ambient light. Subsequently, the geometric trigonometric ratio between the shadow length and the height of the reference object (i.e., Estimate the angle of illumination elevation. This allows for the accurate calculation of the ambient light direction vector. This ensures that the direction of the shadows cast by the synthetic disease is physically consistent with that of the real scene.

[0080] Step S33: Calculate the self-occlusion shadow intensity compensation coefficient at any point inside the stress-affected boundary.

[0081] The self-occlusion shadow intensity compensation coefficient at any point inside the stress-affected boundary is calculated as follows:

[0082]

[0083] in, Represents any point within the stress-affected boundary. The self-occlusion shadow intensity compensation coefficient; The illumination elevation angle, representing the ambient light direction vector; This represents the shadow depth attenuation factor, typically ranging from 2.0 to 5.0. The shadow depth attenuation factor is used to adjust the rate at which shadow intensity decreases with increasing depth. The larger the value is set, the more drastic the light attenuation on the simulated pit's inner wall, resulting in a stronger sense of depth and three-dimensionality.

[0084] Step S34: Correct the core pixel block using the self-occlusion shadow intensity compensation coefficient to obtain the corrected core pixel block.

[0085] Specifically, the self-occlusion shadow intensity compensation coefficient corrects the core pixel block in the following way:

[0086]

[0087] in, This indicates the corrected core pixel block; It represents the Hadamardi (or Hadama) stack.

[0088] The above steps ensure that the shadow inside the lesion changes dynamically with depth and lighting angle, eliminating the "flat sticker" feel.

[0089] Step S4: Use feathering mask and frequency domain and brightness alignment to achieve consistent fusion of the synthesized disease block and the host road surface image to obtain the final synthesized image.

[0090] Step S41: Spatially superimpose and combine the corrected core pixel block with the secondary disease pixel block to obtain a synthetic disease block.

[0091] Specifically, the corrected core pixel block With secondary disease pixel blocks Spatial overlay and combination are performed to generate a complete disease texture image, i.e., synthesized disease blocks. Its synthesis formula is as follows:

[0092]

[0093] Synthetic disease blocks It includes information on the depth of the pit and details of the surrounding cracks.

[0094] Step S42: Perform Gaussian blurring on the binary mask of the central lesion and the mask of the secondary crack to obtain the feathered mask.

[0095] To address the jagged edges caused by traditional clipping methods, a feathered mask is obtained by applying Gaussian blurring to the binary mask of the central lesion and the mask of secondary cracks. Feathering mask The pixel values ​​transition smoothly and continuously between 0 and 1, which is used to define the blending ratio between the synthetic disease and the host road surface, thereby simulating the natural scattering effect of light and achieving a natural blending of the disease edge with the background.

[0096] Step S43: Determine the disease implantation area in the host road surface image, extract the original texture features of the disease implantation area in the host road surface image, and align the frequency and brightness of the synthesized disease block with the host road surface image to obtain the final synthesized image.

[0097] In the host pavement image, the disease implantation area is determined, i.e., the selected spatial coverage area for fusing disease blocks. High-frequency features of aging texture are extracted from the host pavement image, and feathering masks are used. As an Alpha blending channel, a global luminance alignment coefficient is introduced. By modulating the high-frequency texture components, the synthesized diseased block and the host road surface image are simultaneously aligned in terms of frequency detail and overall brightness, resulting in the final synthesized image. The calculation formula is as follows:

[0098]

[0099] in, This represents the final composite image; This represents the Hadamard product operation.

[0100] The formula is passed and The complementary weighting achieves a seamless "feathering" transition between the edge of the defect and the road background. At the same time, the coefficients ensure the consistency of the overall brightness and tone between the synthesized area and the surrounding road surface, thereby generating high-quality training samples.

[0101] Step S5: Automatically and synchronously generate detection labels based on the outer contour of the feathered mask.

[0102] Specifically, for the feathering mask The non-zero pixel region is scanned to calculate the smallest bounding rectangle that can completely cover the diseased area, which is then used as the annotation box. The vertex coordinates of the annotation box are then obtained. , It is the x-coordinate of the leftmost edge of the smallest bounding rectangle in the image. It is the ordinate of the top edge of the smallest bounding rectangle in the image; these two values ​​are combined. This represents the coordinates of the top-left corner vertex of the rectangle. It is the x-coordinate of the rightmost edge of the smallest bounding rectangle in the image. It is the ordinate of the bottom edge of the smallest bounding rectangle in the image; these two values ​​are combined. This represents the coordinates of the bottom right corner vertex of the rectangle. Finally, the vertex coordinates of the annotation box are automatically converted into YOLO or VOC format annotation files to obtain the labels, thus realizing data synthesis and label generation.

[0103] In practical applications, this invention will use core pixel blocks Stress diffusion coefficient Set to a random value between 1.5 and 2.5 to simulate foundation degradation of varying intensities.

[0104] Figure 3 This paper presents a comparative analysis of the visual effects and feature space of the synthesized sample of this invention and the traditional copy-paste method. The images show that the traditional method exhibits obvious pixel abrupt changes and inconsistencies in lighting at the edges of the potholes, and the feature distribution is disconnected from the real sample. In contrast, the sample generated by this invention not only has natural edge transitions, but the self-occluding shadows inside the potholes perfectly match the direction of the ambient light on the road surface, and the feature distribution highly overlaps with the real pothole sample. Figure 4 As shown, after introducing the synthetic data of this invention, the average detection accuracy (mAP@0.5) of YOLOv8 increased from 72.1% to 76.5%. Experimental results further demonstrate that the chain-type disease samples generated by this invention can reduce the false positive rate of the model by more than 17% when dealing with scarce giant potholes, significantly improving the detection reliability and generalization ability of the model in long-tailed complex scenarios.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention 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; and these 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 the present invention.

Claims

1. A method for synthesizing road long-tail disease data based on road surface degradation logic and coevolution, characterized in that, Includes the following steps: Step S1: Preprocess the original pavement image containing long-tail defects to obtain core pixel blocks; Step S2: Based on the stress influence radius of the core pixel block, determine the secondary crack region and generate the secondary crack mask; The host road surface image generates secondary disease pixel blocks based on the secondary crack mask; Step S3: Combine virtual depth and self-occlusion shadow intensity compensation coefficient to correct the core pixel block and obtain the corrected core pixel block; Step S4: Use feathering mask and frequency domain and brightness alignment to achieve consistent fusion of the synthesized disease block and the host road surface image to obtain the final synthesized image; Step S5: Automatically and synchronously generate detection labels based on the outer contour of the feathered mask; Step S2 includes the following steps: Step S21: Based on the core pixel block, introduce the stress diffusion coefficient to characterize the looseness of the roadbed, calculate the stress influence radius, and the circumference formed by the stress influence radius is the stress influence boundary. Step S22: Determine the secondary crack region based on the geometric equivalent radius and stress influence radius of the core pixel block, and generate a secondary crack mask within the secondary crack region; Step S23: The host road surface image generates secondary disease pixel blocks based on the secondary crack mask; Step S3 includes the following steps: Step S31: Calculate the virtual depth at any point within the secondary crack region; Step S32: Solve for the ambient light direction vector of the host road surface image; Step S33: Calculate the self-occlusion shadow intensity compensation coefficient at any point inside the stress-affected boundary; Step S34: Correct the core pixel block using the self-occlusion shadow intensity compensation coefficient to obtain the corrected core pixel block; In step S33, the self-occlusion shadow intensity compensation coefficient at any point inside the stress-affected boundary is calculated as follows: in, Represents any point within the stress-affected boundary. The self-occlusion shadow intensity compensation coefficient; The illumination elevation angle, representing the ambient light direction vector; Indicates the shadow depth attenuation coefficient; This represents the virtual depth of any point within the secondary crack region.

2. The method for synthesizing road long-tail defects data based on pavement degradation logic and cascading evolution according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Use an image semantic segmentation and annotation tool to extract the mask from the original road surface image containing long-tail defects, and obtain the binary mask of the central defect. Step S12: Perform pixel-level logical AND operation or mask mapping cropping on the binary mask and the original road surface image to obtain core pixel blocks that retain only the texture information of the defects.

3. The method for synthesizing road long-tail defects data based on pavement degradation logic and cascading evolution according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Spatially superimpose and combine the corrected core pixel block with the secondary disease pixel block to obtain a composite disease block; Step S42: Perform Gaussian blurring on the binary mask of the central lesion and the mask of the secondary crack to obtain the feathered mask; Step S43: Determine the disease implantation area in the host road surface image, extract the original texture features of the disease implantation area in the host road surface image, and align the frequency and brightness of the synthesized disease block with the host road surface image to obtain the final synthesized image.

4. The method for synthesizing road long-tail defects data based on pavement degradation logic and cascading evolution according to claim 1, characterized in that, In step S21, the formula for calculating the stress influence radius is: in, Indicates the radius of stress influence; Represents the geometric equivalent radius of the core pixel block; This represents the stress diffusion coefficient.

5. The method for synthesizing road long-tail defects data based on pavement degradation logic and cascading evolution according to claim 1, characterized in that, In step S23, the secondary disease pixel block is generated in the following manner: in, Represents the secondary disease pixel block; The aging texture of the host road surface image; This is the contrast attenuation factor; It represents the Hadamardi (or Hadama) stack; This represents the secondary crack mask.

6. The method for synthesizing road long-tail defects data based on pavement degradation logic and cascading evolution according to claim 1, characterized in that, In step S34, the self-occlusion shadow intensity compensation coefficient is corrected for the core pixel block in the following manner: in, This indicates the corrected core pixel block; Indicates the core pixel block; It represents the Hadamardi (or Hadama) stack.

7. The method for synthesizing road long-tail defects data based on pavement degradation logic and cascading evolution according to claim 3, characterized in that, In step S43, the final composite image is synthesized in the following manner: in, This represents the final composite image; This represents the Hadamard product operation; Indicates the feathering mask; Represents the aging texture of the host road surface image; Represents the global brightness alignment factor; This indicates a synthetic diseased block.

Images