Crack image segmentation method and device based on damage mechanics, and electronic equipment

Abstract

The application discloses a crack image segmentation method and device based on damage mechanics, and electronic equipment, comprising: obtaining an original crack image and performing local pretreatment, and extracting a crack geometric feature matrix; the nonlinear mapping is carried out to form an initial yield stress field in a two-dimensional structured grid, a displacement boundary condition is dynamically applied to cause local tensile stress; based on the continuous medium damage mechanics model, the low yield stress area element accumulates damage and stiffness degradation occurs, when the damage reaches the critical value, the element is deleted to form a failure element, and the physical expansion of the crack is characterized; the current window failure element coordinates are extracted, and the crack propagation direction is determined by linear fitting, which is used as the initial state and load condition of the next local tracking window, and the failure element coordinates of all local windows are integrated to obtain a complete continuous crack segmentation path. The method is suitable for accurate detection and continuous topology extraction of complex structure surface fuzzy, fine and discontinuous cracks.

Description

Technical Field

[0001] This application relates to the interdisciplinary field of computer vision and computational mechanics, and in particular to a crack image segmentation method and device, and electronic equipment based on damage mechanics. Background Technology

[0002] Cracks are the most common manifestation of early-stage damage in engineering structures, and their formation and propagation typically reflect changes in the internal stress state or material properties of the structure. If not detected and effectively addressed in a timely manner, cracks may further induce problems such as leakage, corrosion, fatigue propagation, and local instability, thereby affecting the safety and service life of the structure. Therefore, crack detection is a crucial foundation for structural health monitoring and safety assessment, and a key link in achieving intelligent operation and maintenance of engineering structures.

[0003] In recent years, with the rapid development of computer vision and deep learning technologies, image-based crack detection methods have gradually become an important research direction in the field of intelligent nondestructive testing due to their advantages such as non-contact operation, high efficiency, low cost, and ease of automated deployment. However, in real engineering environments, cracks, especially early-stage fine cracks, typically have characteristics such as narrow width, low contrast, blurred edges, and local discontinuity. They are also easily affected by complex background textures, lighting variations, shadow occlusion, water contamination, and imaging noise, leading to a high degree of coupling between crack features and background information, thus significantly increasing the difficulty of crack detection. Therefore, how to achieve accurate identification and complete extraction of fine cracks in complex scenes has become a key and challenging problem in current image crack detection research. Existing technologies and their limitations: 1. Traditional image processing algorithms (such as Otsu thresholding and Canny operator): These methods mainly rely on pixel-level features or local gradient information for analysis. They are sensitive to image noise and uneven lighting, and are prone to missed detections, false detections, or path breaks when processing low-contrast, blurred, or physically discontinuous fine cracks, making it difficult to guarantee the physical continuity of the extracted crack edges.

[0004] 2. Deep learning models (such as UNet++, SegFormer, etc.): Although they possess strong feature extraction and mapping capabilities, they are essentially data-driven statistical learning methods, lacking explicit constraints on prior physical laws such as crack generation and propagation. Furthermore, they rely on large amounts of high-quality labeled data, resulting in high training and computational costs, and their generalization ability and robustness under cross-scenario or zero-shot conditions are often limited.

[0005] 3. Fracture mechanics-based segmentation methods (such as LCPM - Local Crack Propagation Method): This type of research attempts to map image gradients to physical grooves in a three-dimensional finite element model and uses the Extended Finite Element Method (XFEM) to simulate crack propagation for segmentation. Its main limitations in practical applications are as follows: (1) Large computational overhead: XFEM requires the introduction of enriched degrees of freedom (EDF) to characterize the crack surface. Combining the three-dimensional geometric model with implicit nonlinear solution increases the dimension of the system stiffness matrix and the computational time of iterative solution. (2) Limited model convergence stability: Directly mapping image grayscale to three-dimensional groove depth can easily produce irregular geometric boundaries, which may lead to non-physical local stress singularities, increasing the risk that the implicit solution process cannot converge. (3) High dependence on boundary conditions: It is necessary to pre-determine a clear geometric crack as an initial defect in the local model. When processing real engineering images with complex backgrounds, the tracking continuity of the algorithm is easily affected by the fluctuation of local stress field and may be interrupted. Summary of the Invention

[0006] The purpose of this application is to provide a crack image segmentation method, device, and electronic device based on damage mechanics, in order to solve the problems of low detection accuracy and dependence on training data in existing computer vision technology when processing blurry, fine, and discontinuous cracks, as well as the limitations of existing mechanical image segmentation methods (such as 3D XFEM) caused by the use of implicit nonlinear solutions and complex 3D geometric updates, such as increased degrees of freedom, large computational resource consumption, and poor model convergence performance.

[0007] According to a first aspect of the embodiments of this application, a crack image segmentation method based on damage mechanics is provided, comprising: S1: Obtain the original crack image to be detected; S2: Perform local preprocessing on the original crack image within the current local tracking window to obtain a local feature matrix containing crack geometric feature information; S3: Nonlinearly map the local feature matrix to the initial yield stress field in a two-dimensional structured grid, wherein the potential crack target area in the image is mapped to a low yield stress area and the background area is mapped to a high yield stress area. S4: Based on the current crack tip and its propagation direction, dynamically apply time-modulated symmetric displacement boundary conditions to induce a local tensile stress field in the two-dimensional structured mesh; S5: Based on the continuous medium damage mechanics model, the elements in the low yield stress zone are driven to accumulate damage and undergo stiffness degradation through damage initiation criteria and damage evolution rules. When the damage variable of the element reaches the critical threshold, the element is removed from the mesh to form a failed element. The local stress concentration effect caused by the geometric discontinuity formed by the failed element is used to achieve the bridging of the adjacent high yield stress zone to characterize the physical propagation of the crack. S6: Extract the coordinates of the failure elements within the current local tracking window, determine the crack propagation direction through linear fitting, and map the coordinates of the terminal failure elements to the initial state and load conditions of the next local tracking window. Iterate through steps S2 to S6 until the termination condition is met, integrate the coordinates of the failure elements in all local windows, and obtain the final continuous crack segmentation path.

[0008] According to a second aspect of the embodiments of this application, a crack image segmentation apparatus based on damage mechanics is provided, comprising: The acquisition module is used to acquire the original crack image to be detected; The preprocessing module is used to perform local preprocessing on the original crack image within the current local tracking window to obtain a local feature matrix containing crack geometric feature information; The mapping module is used to nonlinearly map the local feature matrix into an initial yield stress field in a two-dimensional structured grid, wherein the potential crack target area in the image is mapped to a low yield stress area and the background area is mapped to a high yield stress area. A boundary condition loading module is used to dynamically apply time-modulated symmetric displacement boundary conditions based on the current crack tip and its propagation direction to induce a local tensile stress field in the two-dimensional structured mesh. The damage evolution solution module is used to drive the accumulation of damage and stiffness degradation of the elements in the low yield stress zone based on the damage mechanics model of continuous medium through damage initiation criteria and damage evolution rules. When the damage variable of the element reaches the critical threshold, the element is removed from the mesh to form a failed element. The local stress concentration effect caused by the geometric discontinuity formed by the failed element is used to achieve the bridging of the adjacent high yield stress zone to characterize the physical propagation of cracks. The window iteration and path integration module is used to extract the coordinates of the failure elements within the current local tracking window, determine the crack propagation direction through linear fitting, and map the coordinates of the terminal failure elements to the initial state and load conditions of the next local tracking window. Iteratively execute the preprocessing module to the window iteration and path integration module until the termination condition is met, integrate the coordinates of the failure elements of all local windows, and obtain the final continuous crack segmentation path.

[0009] According to a third aspect of the embodiments of this application, an electronic device is provided, comprising: 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 method as described in the first aspect.

[0010] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the steps of the method as described in the first aspect.

[0011] The technical solutions provided by the embodiments of this application may include the following beneficial effects: 1. Reduce the dimensionality of the computational system and optimize solution efficiency; This method replaces the three-dimensional XFEM with a two-dimensional plane stress finite element model, avoiding the EDF introduced by characterizing the three-dimensional crack surface and significantly reducing the dimension of the system stiffness matrix. Simultaneously, it transforms the implicit nonlinear solution into an element reduction process based on explicit dynamics, reducing the risk of non-physical stress singularities and solution non-convergence caused by irregular local geometric mappings, thus improving overall computational efficiency while maintaining the integrity of the mechanical partitioning logic.

[0012] 2. To achieve discontinuous bridging based on the mechanical stress concentration mechanism; This method does not rely on simple image pixel connectivity for feature inference, but rather processes physical discontinuities in images based on fracture mechanics mechanisms. When the macroscopic trajectory formed by successive failure units approaches the discontinuous region of high yield stress, the stress concentration effect at the crack tip caused by geometric discontinuity will significantly amplify the local true stress, causing the high yield stress unit to reach the damage evolution threshold based on fracture energy and fail, thereby achieving the bridging of the fracture path through physical and mechanical evolution.

[0013] 3. Not dependent on data-driven approaches, adaptable to cross-scenario applications; The segmentation process of this invention is driven by objective physical constitutive models and continuous medium damage mechanics rules, without relying on pre-labeled image datasets for statistical fitting or model training. When facing engineering scenarios with different material backgrounds and lighting conditions, it objectively maps local feature matrices to physical fields and utilizes mechanical mechanisms to maintain the extraction of crack topology, demonstrating good cross-scene versatility.

[0014] 4. Supports continuous solving of large-sized images; To address the issues of high computational cost and the need for pre-setting global cracks when processing large images using traditional fracture mechanics methods, this invention, in addition to the core mechanical segmentation framework, incorporates local computational window sliding and coordinate fitting techniques. This mechanism dynamically provides coherent displacement boundary conditions and loading directions for the local finite element model, supporting the continuous advancement of the mechanical failure segmentation process in complex, large images.

[0015] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] Figure 1 This is a flowchart illustrating a crack image segmentation method based on damage mechanics according to an exemplary embodiment.

[0018] Figure 2 This is a Frangi response structure diagram illustrated according to an exemplary embodiment, wherein (a) is the original image, (b) is the corresponding local grayscale image, (c) is the two-dimensional planar diagram of the Frangi structure response after the above processing, and (d) is the three-dimensional visualization diagram of the Frangi structure response.

[0019] Figure 3 This is a diagram illustrating the preprocessing results according to an exemplary embodiment.

[0020] Figure 4 This is a three-dimensional visualization of a nonlinear mapping of a feature matrix according to an exemplary embodiment, where (a) is the local feature matrix numerical distribution and (b) is the physical yield stress field distribution generated by the corresponding nonlinear mapping.

[0021] Figure 5 This is a load distribution and orientation diagram illustrated according to an exemplary embodiment.

[0022] Figure 6 This is an equivalent plastic strain contour plot shown according to an exemplary embodiment.

[0023] Figure 7 This is an exemplary embodiment illustrating the process of element stiffness degradation and damage evolution.

[0024] Figure 8This is a diagram illustrating the effect of cell deletion according to an exemplary embodiment, wherein (a) is the state of the model cell state variable in the initial non-degenerate stage, (b) is the process state of the model cell state variable changing to zero as damage accumulates, and (c) is the failure state where the cell state variable drops to zero and is forcibly deleted.

[0025] Figure 9 This is a diagram illustrating the fuzzy crack reduction process according to an exemplary embodiment.

[0026] Figure 10 This is a diagram illustrating the process of discontinuous crack connection according to an exemplary embodiment.

[0027] Figure 11 This is a schematic diagram of a local iteration according to an exemplary embodiment.

[0028] Figure 12 This is a diagram illustrating a local path visualization process according to an exemplary embodiment.

[0029] Figure 13 This is a diagram illustrating the results of road surface crack segmentation according to an exemplary embodiment.

[0030] Figure 14 This is a diagram illustrating the result of wall crack segmentation according to an exemplary embodiment.

[0031] Figure 15 This is a schematic diagram of a crack image segmentation device based on damage mechanics, according to an exemplary embodiment. Detailed Implementation

[0032] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application.

[0033] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0034] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0035] Figure 1 This is a flowchart illustrating a crack image segmentation method based on damage mechanics according to an exemplary embodiment, such as... Figure 1 As shown, the method may include the following steps: S1: Obtain the original crack image to be detected; Specifically, a two-dimensional original engineering image containing the surface of the target structure is acquired through an image acquisition device, which serves as the initial data input for subsequent local feature extraction and mechanical evolution modeling.

[0036] S2: Perform local preprocessing on the original crack image within the current local tracking window to obtain a local feature matrix containing crack geometric feature information; including the following sub-steps: S21: Perform median filtering on the original crack image within the current local tracking window to estimate background illumination and texture distribution; Specifically, in the image to be detected, the first local image is selected near the crack initiation point, and the global starting point and initial tip are selected to define the first propagation direction. This provides initial boundary conditions for the mechanical model.

[0037] For the input image Applying medium filtering to estimate low-frequency illumination and texture distribution in the background: (1) in, The corresponding coordinates are obtained by median filtering estimation. The grayscale values ​​of low-frequency background illumination and texture distribution at that location; The median operator; This is a two-dimensional local sliding window for median filtering; For the local sliding window Pixel space offset coordinates within; This is the grayscale value of the input image at the corresponding offset position within the local sliding window.

[0038] The purpose of this step is to utilize the nonlinear smoothing properties of median filtering within the local spatial neighborhood, causing small-area, fine-grained crack pixels to be replaced by surrounding background pixels during median calculation, thereby estimating a background matrix that only includes macroscopic illumination and surface texture. The advantage is that it obtains a baseline image without local crack features, providing a computational basis for subsequent image differencing to separate background interference and uneven illumination.

[0039] S22: Perform image subtraction on the median-filtered image to extract image details; Specifically, crack details are extracted through image subtraction, and a bilateral filter is used to smooth noise while preserving crack edges: (2) In the formula, The difference image containing crack detail features, obtained after image subtraction, is located at pixel coordinates. The grayscale value at that location.

[0040] The purpose of image subtraction is to eliminate common information about global illumination and macroscopic texture by subtracting the original image from the background reference image. This operation significantly highlights subtle crack signals in the original image, improving the contrast of the target area.

[0041] S23: Perform bilateral filtering on the subtractive image to smooth noise and preserve crack edges; Specifically, a bilateral filter is used to smooth noise while preserving crack edges: (3) In the formula, The output image obtained after bilateral filtering is in pixel coordinates The grayscale value at that location; and The coordinates of the target pixel in the input image (i.e., the detail image obtained after image subtraction) are respectively. and adjacent pixel coordinates The grayscale value at that location; For The set of pixels in the two-dimensional local spatial neighborhood centered on the image; Let be the spatial Gaussian kernel function based on spatial physical distance; ||·|| is the Euclidean distance between two points; It is a Gaussian kernel function with a range based on the difference in pixel grayscale values; These are the normalized weighting coefficients.

[0042] The purpose of introducing a bilateral filter is to simultaneously consider the spatial distance between pixels and the difference in grayscale values ​​for weighted smoothing. Its advantage is that it can effectively filter out isolated local noise points while preserving the grayscale step characteristics between the crack area and the background area, avoiding edge blurring caused by conventional smoothing filters, and providing accurate geometric boundary localization for subsequently mapping image features to finite element model properties.

[0043] S24: Perform pixel-by-pixel Gaussian smoothing on the bilaterally filtered image to construct the Hessian matrix; Specifically, a multi-scale Frangi filter based on Hessian eigenvalues ​​is used to extract deep features from the bilaterally filtered image. For two-dimensional images... Its scale Gaussian smoothed Hessian matrix Defined as: (4) In the formula, For two-dimensional images at scale Next pixel coordinates The corresponding Gaussian smoothed Hessian matrix at that location; It is a matrix of second-order partial derivative operators; The input is a two-dimensional image matrix after pre-filtering. It is a two-dimensional spatial convolution operator; The standard deviation is The two-dimensional Gaussian kernel function.

[0044] The purpose of constructing the second-order partial derivative Hessian matrix is ​​to extract curvature variation features within the pixel neighborhood. Its advantage is that it can objectively quantify the directionality and continuity of the local structure of an image from a mathematical perspective, providing a geometric criterion for distinguishing tubular crack features from speckled noise.

[0045] S25: Perform multi-scale Frangi linear filtering on the Hessian matrix to enhance the linear structure features, and generate the local feature matrix through nonlinear mapping; Specifically, by solving for the eigenvalues ​​of the Hessian matrix, a blob measure and structural strength are introduced to construct the Frangi linear filter response function. The maximum response value of each pixel is extracted within a defined multi-scale space to generate a linear structure response map. Then, a power-law exponential nonlinear mapping function is used to normalize and transform the response values ​​to a preset numerical range, generating a local feature matrix that serves as the input for subsequent mechanical boundaries.

[0046] set up The two eigenvalues ​​are and And they are arranged according to their absolute values. To distinguish linear structures from other morphologies such as spots and background, two geometric metrics are introduced: spot measure. With structural strength

[0047] (5) At a single scale, the response function of the Frangi linear filter The structure is as follows: (6) in, and These are constant parameters controlling the sensitivity to speckle and background noise, respectively. To accommodate fine cracks of different widths, the maximum response at multiple scales is taken within a given scale space, and the final line structure response map is calculated. : (7) In the formula, For finding the maximum value operator; The scaling parameters are for Gaussian smoothing. For a pre-defined discrete multi-scale spatial set; It should be pointed out that A high value clearly indicates that the pixel more closely resembles a slender, crack-like structure. As a purely image-based feature indicator, It only represents the geometric similarity of local textures and is not equivalent to the actual physical depth or the actual crack width. For example... Figure 2 As shown, (a) is the original image, (b) is the corresponding local grayscale image, (c) is the corresponding two-dimensional planar image of the Frangi structure response after the above processing, and (d) is the three-dimensional visualization of the Frangi structure response.

[0048] Introducing the power-law exponent A nonlinear mapping is performed to transform the response map into the local characteristic matrix required for subsequent mechanical fields. : (8) in, and These are the preset upper and lower limits of the equivalent feature; and These are the maximum and minimum response values ​​in the line structure response graph, respectively. and These are the preset upper and lower limits of the equivalent feature, respectively.

[0049] Through preprocessing, the local feature matrix Strictly bound to to Between. It needs to be clarified that this numerical range serves only as an "equivalent geometric feature index," representing the probability that a local pixel belongs to a crack. For example... Figure 3 As shown, this is the local feature matrix obtained after the above preprocessing of the original local grayscale image. .

[0050] The purpose of constructing linear filters and performing nonlinear mappings is to enhance linear structures and suppress isotropic noise through eigenvalue analysis. The advantage of converting it into a local feature matrix is ​​that it standardizes the purely visual geometric similarity features into a defined numerical range, establishing a unified dimensionless interface for subsequent conversion into finite element physical and mechanical properties.

[0051] S3: Nonlinearly map the local feature matrix to an initial yield stress field in a two-dimensional structured mesh, wherein the potential crack target region in the image is mapped to a low yield stress region, and the background region is mapped to a high yield stress region; including the following sub-steps: S31: In accordance with the principle of one pixel corresponding to one finite element mesh unit, a two-dimensional orthogonal structured finite element mesh model is established within the local tracking window, and the pixel coordinates are mapped to the corresponding node coordinates; Specifically, following the principle of "one pixel corresponds to one finite element mesh unit," a two-dimensional orthogonal structured mesh is established within the local window. Pixel coordinates The corresponding node coordinates are: (9) in, The physical size of a pixel. The height is the local window height. This model uses a four-node plane stress-reduced integral element (CPS4R).

[0052] The purpose of adopting this mesh construction principle is to achieve a lossless geometric transformation from image pixel space to computational mechanics physical space. Its advantage is that it eliminates the complex mesh generation and local refinement processes of traditional mechanics modeling, ensuring a complete mapping between image feature distribution and the mechanical carrier.

[0053] S32: The eigenvalues ​​in the local feature matrix are discretized and mapped to physical yield stress using a nonlinear mapping function, wherein the potential crack target area in the image is mapped to a low yield stress area and the background area is mapped to a high yield stress area. Specifically, a nonlinear Gamma function is used to map the eigenvalues ​​in the local characteristic matrix to physical yield stresses, and the mapping formula is as follows: (10) In the formula, Local feature matrix Non-repeating eigenvalues ​​in; and These are the minimum and maximum values ​​in the local feature matrix, respectively. The initial yield stress lower limit set for the crack region; The upper limit of the initial yield stress in the background region; This is a non-linear mapping exponent. For example... Figure 4 The diagram shows a three-dimensional visualization. (a) shows the numerical distribution of the local feature matrix, and (b) shows the distribution of the physical yield stress field generated by the corresponding nonlinear mapping.

[0054] The purpose of introducing nonlinear Gamma mapping is to differentiate the physical properties between the potential crack zone and the background zone. This allows the target crack region to be precisely defined as a low yield stress zone, ensuring that under global loading, plastic deformation can be stably concentrated in this target region.

[0055] S33: Generate section properties based on the discretized physical yield stress and assign them to the corresponding elements in the two-dimensional orthogonal structured finite element mesh model to form an initial yield stress field; Specifically, the finite element solver generates corresponding section properties (uniformly set as plane stress thickness) based on the discretized yield stress values ​​and assigns them to four-node plane stress reduced integral elements (CPS4R) with corresponding eigenvalues. Through this mapping, the numerical distribution of the eigenvalue matrix is ​​transformed into a physical yield stress field, causing the crack target region (low yield stress zone) to preferentially undergo plastic yielding under load. All elements in the model use the same linear elastic parameters: the elastic modulus is set to... Poisson's ratio is set to The two-dimensional orthogonal structured finite element mesh model adopts an ideal elastoplastic constitutive model in the plastic stage. The ideal elastoplastic constitutive model (i.e., strain-free hardening) is as follows: (11) In the formula, The current Mises equivalent stress of the element, Assign the initial yield stress to the element for mapping. The equivalent plastic strain is used. This constitutive model ensures that the elements do not harden after reaching the initial yield stress; under local tensile loads, low-yield-stress elements will continue to accumulate plastic strain until the damage initiation threshold is met, thereby triggering the subsequent stiffness degradation and element reduction process.

[0056] The purpose of employing an ideal elastoplastic constitutive model is to ensure that the equivalent plastic strain of low-yield-stress elements can continue to accumulate after reaching the initial yield state. This avoids local stress redistribution caused by material hardening effects and prevents the plastic zone from spreading to the surrounding high-yield-stress background region, thereby ensuring that subsequent damage evolution and element culling are confined to the potential crack region mapped by the image feature map.

[0057] S4: Based on the current crack tip and its propagation direction, dynamically apply time-modulated symmetric displacement boundary conditions to induce a local tensile stress field in the two-dimensional structured mesh; including the following sub-steps: S41: Based on the current crack tip and its propagation direction, calculate the normal vector perpendicular to the propagation direction, and set the center coordinates of the two loading regions in front of the current crack tip; Specifically, at the crack tip A local opposing displacement boundary condition is introduced at the beginning. The algorithm first calculates the boundary condition perpendicular to the crack propagation trend. normal vector A localized Mode-I (opening type) tearing force field is formed in front of the crack tip, and the center coordinates of the loading regions on both sides are set: (12) In the formula, and These are the center coordinates of the loading regions on both sides in front of the crack tip; These are the spatial coordinates of the current crack tip; The set loading area offset distance parameter; It is the normal vector perpendicular to the crack propagation trend.

[0058] The purpose of setting up loading zones on both sides is to precisely construct a pure opening-type (Mode-I) tearing mechanical environment in front of the crack tip. This ensures that the energy driving crack propagation can be directionally delivered to the intended failure region.

[0059] S42: Based on the normal vector and center coordinates, extract the mesh nodes in the loading areas on both sides to construct a local tension node set; Specifically, using the center coordinates of the two loading areas respectively and Using the geometric center as the reference point and combining it with the pre-defined characteristic dimensions of the loading range, all finite element mesh nodes falling within this geometric range are extracted, thereby constructing sets of locally tensioned nodes for applying local counter-displacement. and .

[0060] The purpose of extracting a specific range of nodes for loading is to avoid non-physical stress singularities caused by single-point loading. This has the advantage of making the local load distribution more uniform and stable, reducing the risk of numerical oscillations in the subsequent solution process.

[0061] S43: Apply local counter-loads to the local tensile node set using symmetrical displacement boundary conditions to generate a stable local Mode-I tearing stress field in front of the current crack tip, which serves as a local tensile stress field. Specifically, a local opposing load is applied using displacement boundary conditions (Displacement BC) to ensure the convergence of the solution for the element stiffness degradation and pruning process. For the set... Apply For sets Apply In the formula, The preset maximum displacement loading amplitude must be large enough to drive the element damage evolution until the pruning threshold is reached. For example... Figure 5 As shown, the load is symmetrically distributed at the crack tip. Forward, and in a direction perpendicular to .

[0062] The purpose of introducing displacement boundary conditions is to drive mechanical evolution by specifying the deformation of local tension nodes. This avoids computational divergence that can easily occur when elements are subjected to force-controlled loading during damage evolution and stiffness degradation stages, ensuring the numerical convergence of the forced element reduction process.

[0063] S5: Based on the continuum damage mechanics model, damage accumulation and stiffness degradation of elements in the low-yield stress zone of the two-dimensional structured mesh are driven by damage initiation criteria and damage evolution rules. When the damage variable of an element reaches a critical threshold, the element is removed from the mesh to form a failed element. The local stress concentration effect caused by the geometric discontinuity formed by the failed element is used to bridge the adjacent high-yield stress zone, thus characterizing the physical propagation of the crack; including the following sub-steps: S51: Based on the continuous medium damage mechanics model, the finite element solution is performed using explicit dynamic analysis steps, and a tough damage mechanism is introduced. When the equivalent plastic strain of the element reaches the critical value, the damage initiation criterion is triggered. Specifically, this model establishes a total duration of The model employs an explicit dynamics step. Utilizing the time integral operator of the central difference method, this step effectively handles severe mesh damage, transient element pruning, and complex stress wave propagation. During the stress field redistribution induced by local loads, the extremely low yield stress weak elements mapped by the image crack feature map will reach the yield critical point first. The model introduces a ductile damage mechanism, defining state variables to track damage accumulation. Damage formally initiates when the cumulative equivalent plastic strain of the element reaches a critical threshold. For equivalent plastic strain, This is the critical equivalent plastic strain at which damage begins to occur in the material, and its value depends on the stress triaxiality. With equivalent plastic strain rate Once a unit's Once the value reaches 1, the unit enters the damage evolution stage. For example... Figure 6 As shown in the figure, the contour plot on the right illustrates the distribution of equivalent plastic strain (PEEQ). It can be seen that the crack region reaches the yield zero point first due to the extremely low yield stress. The gray area represents the PEEQ having reached or exceeded the critical threshold. This signifies that the elements in this region have officially met the ductile damage initiation criterion and are about to enter the stiffness degradation evolution stage.

[0064] The purpose of introducing the equivalent plastic strain triggering mechanism is to accurately capture the irreversible deformation accumulation of low yield stress elements under tensile field. Its advantage is that it provides an objective physical criterion, allowing regions with the most significant image features to preferentially enter the damage evolution state.

[0065] S52: After the damage initiation criterion is triggered, the damage evolution stage is entered. The damage evolution rule based on equivalent plastic displacement is adopted to control the stiffness of the element to degrade linearly with the accumulation of damage variables. Specifically, after entering the damage evolution stage, the material stiffness of the element begins to irreversibly degrade, and the load-bearing capacity of the target element gradually weakens, but the element remains within the model. A scalar damage variable is introduced. (SDEG), whose value gradually accumulates from 0 to approach 1. It adopts a method based on equivalent plastic displacement. The law of damage evolution: ,in, The evolution rate of the scalar damage variable; This represents the equivalent rate of change of plastic displacement. The geometric feature length of the finite element; This is the equivalent plastic strain rate; This is the ultimate equivalent plastic displacement at which the material completely fails (this critical value is calculated from a preset material fracture energy). For example... Figure 7 As shown, SDEG exhibits a continuous spatial gradient distribution from 0 to 1 in the crack region, which closely matches the evolution law of continuous medium damage mechanics.

[0066] The purpose of employing a damage evolution rule based on equivalent plastic displacement and preset material fracture energy is to calculate the energy dissipation of the element during stiffness degradation based on the material fracture energy. Its advantage is that it eliminates the dependence of the finite element mesh size on the energy dissipation calculation results, allowing the damage evolution process of the element to be controlled by the fracture energy parameters.

[0067] S53: When the damage variable of the unit reaches the critical threshold, the state variable of the unit is forcibly set to zero, and the unit is permanently removed from the two-dimensional structured mesh to form the failure unit. The geometric discontinuity formed by the failure unit induces a local stress concentration effect in front of the crack tip, so that the damage variable of the unit in the adjacent high yield stress zone reaches the critical threshold, thereby realizing the bridging of the high yield stress zone to characterize the physical propagation of the crack. Specifically, as the local tensile load continues to do work, when At that time, the explicit dynamics solver will include the element state variables STATUS By forcibly setting the value from 1 to 0, the cell is permanently removed from the computational mesh, forming a physical crack path. For example... Figure 8 As shown, when the damage variable tends to 1, the solver forcibly removes failed elements. STATUS (Reduced to 0), a clear macroscopic physical fracture surface spontaneously formed in the originally continuous finite element mesh. The process from (a) to (b) is the model element's... STATUS The transition from state 1 to 0, (c) is... STATUS The state of a cell with a state of 0 (i.e., the blue area) is that which has been removed.

[0068] When faced with an ambiguous crack pattern, minute fluctuations in strength are insufficient to suppress the continuous accumulation of plastic strain, and the damage trajectory can continue to evolve in accordance with the complex bending morphology of the weak zone. For example... Figure 9 As shown, for a fuzzy crack, its yield stress still has a certain difference from the background yield stress. Therefore, following the "principle of minimum energy dissipation", the element will still be removed along the correct target region.

[0069] When faced with a discontinuous crack, a significant stress singularity will occur in front of the macroscopic "crack tip" composed of continuously subtracted elements. The stress field distribution at the tip follows the following rules: In the formula Stress intensity factor The distance from the tip, and The dimensionless angular distribution function collectively determines the directional stress concentration effect in the local region at the crack tip. This intense stress concentration, caused by geometric discontinuities, exponentially amplifies the actual stress level in front of the crack tip, sufficient to overcome the yield threshold of the discontinuous element. This forces the high-stress discontinuous element to eventually reach the damage evolution threshold and be forcibly purged, thus spontaneously achieving physical bridging of the discontinuous region. Figure 10 As shown, in the case of crack discontinuity, the elements with weak yield stress are eliminated first. Under the stress singularity caused by geometric discontinuity, the local stress of the high-strength discontinuous elements that have not yet failed is amplified sharply and reaches the damage evolution condition, thus spontaneously realizing the physical bridging of the discontinuous region.

[0070] The purpose of forced pruning of execution units is to explicitly generate realistic geometric fracture boundaries in a continuous mesh. Its significant benefit is that the crack tips generated after pruning can induce a real local stress concentration effect, thereby amplifying the local stress and naturally overcoming and crossing the discontinuous background region of high yield stress. Thus, spontaneous physical connectivity of discontinuous cracks can be achieved without pre-setting artificial connection rules.

[0071] S6: Extract the coordinates of the failed elements within the current local tracking window, determine the crack propagation direction through linear fitting, and map the coordinates of the terminal failed element to the initial state and load conditions of the next local tracking window. Iterate through steps S2 to S6 until the termination condition is met. Integrate the coordinates of the failed elements from all local windows to obtain the final continuous crack segmentation path; including the following sub-steps: S61: Read the state variables of the unit to filter out the deleted units, and extract the centroid coordinates of the failed unit that is farthest from the local calculation start point in the current local tracking window; Specifically, read the unit state variables ( STATUS ), filter out The deleted unit. Calculate its geometric center. The centroid of the unit furthest from the starting point is taken as the final crack tip. The physical coordinates are then inversely converted to global pixel coordinates of the image. (13) In the formula, and These represent the column and row coordinates of the pixel in the original global crack image after the transformation; The physical size corresponding to a set pixel This represents the total pixel height of the current local tracking window; and These are the pixel column offset coordinates and row offset coordinates of the reference point in the global original image for the current local tracking window.

[0072] The purpose of extracting the coordinates of the furthest failure element is to determine the location of the crack tip generated in the current local calculation step. Its advantage is that it accurately maps the local mechanical results back to the global pixel coordinate system, providing an accurate geometric reference for subsequent iterations.

[0073] S62: Establish the centroid coordinates as the current crack tip, and use them as the geometric center of the next local tracking window; Specifically, the centroid coordinates of the failure element furthest from the local calculation start point are extracted from the above steps, and this coordinate is established as the current crack tip, serving as the geometric center of the next local tracking window. The purpose of establishing a new geometric center is to ensure that the local tracking window always moves correctly with the crack propagation. Its advantage is that it limits the mesh range of the finite element calculation and reduces the computational load of a single solution.

[0074] S63: Based on the current crack tip, backtrack to extract the coordinates of the previous failed unit, and use the least squares method to perform local linear fitting of the spatial coordinates of the previous failed unit to determine the crack propagation direction; Specifically, backtracking extraction The direction angle was calculated by using local linear fitting with a set of discrete tail points using the least squares method. Introducing an exponential smoothing mechanism to predict the next expansion direction: (14) In the formula, This represents the iteration step number of the local tracking window; For the predicted next local tracking window (i.e. the first...) The extension direction angle of the step; For the current number The fitting direction angle calculated within the local tracking window; For the previous calculation step (i.e. the first step) Step) Input of historical extended direction angle; The set exponential smoothing weighting coefficient ( This is used to adjust the contribution ratio of historical paths and the currently fitted path to the prediction direction. The smoothed angle... Then it is transformed into the first The predicted propagation direction unit vector of the step This provides a precise loading direction reference for the next tracking window.

[0075] The purpose of using the least squares method for local linear fitting is to calculate the crack propagation direction using the coordinates of multiple preceding failed elements. Its advantage is that it reduces the direction calculation deviation caused by discrete mesh elements, providing a stable load direction for the next calculation step.

[0076] S64: Using the geometric center of the next local tracking window as a reference, back a specified length along the crack propagation direction, and map the end crack segment as a pre-fabricated crack into the initial state of the next local tracking window; Specifically, along Directional backtracking by a specified length The end crack segment is mapped as a pre-fabricated crack into the current mesh. For example... Figure 11 As shown, in the first k After the local tracking window mechanical solution is completed, the algorithm uses the coordinates of the failure element truncated at the end of the previous step as a reference to establish the first step. k +1 step: Track the geometric center of the new tracking window and the prefabricated crack to realize the process logic of shifting the computational domain. Then, repeat the above mechanical solution and update process in the new window until no new failure elements are generated.

[0077] The purpose of backtracking a pre-cast crack by a specified length is to use the end crack of the previous local tracking window as the initial crack of the new window. This has the advantage of transmitting load location information, ensuring correct load application and stable iteration.

[0078] S65: Determine whether the termination condition is met. If not, iterate through steps S2 to S64 until the tracing process terminates. Then, concatenate the coordinates of the failed units obtained in each iteration into global coordinates and map them into a binary mask. Specifically, the algorithm stitches together the local failure unit point clouds obtained from each iteration using global coordinates and maps them to a binary mask. The purpose of this global coordinate stitching and mapping is to uniformly transform the failure unit coordinates obtained from each local iteration back to the pixel coordinate system of the original image. Its advantage is that it integrates the segmented local calculation results to generate the crack segmentation result corresponding to the entire original image.

[0079] S66: Use connected component area filtering technology to remove isolated noise points in the global binary mask, perform skeletonization processing on the retained core components and prune short branches to obtain the final continuous crack segmentation path; Specifically, connected component area filtering is used to remove scattered isolated noise points, retaining only the core components with the largest area contribution. The centroid coordinates of the failed units are inversely mapped back to the global pixel space of the original crack image according to their physical dimensions. Inverse coloring is then performed using the mapped coordinates, naturally reconstructing a visualized crack path with true width (1-2 pixels) and rough edge features using the geometric shape of the failed units. For example... Figure 12 As shown in the figure, this illustrates the transformation process of inversely mapping the discrete failure element point cloud in a two-dimensional mechanical mesh model to the global pixel coordinate system and reconstructing it into a binary mask pixel path. Skeletonization is performed on the locked region, pruning short branches of non-physical length, and outputting the final continuous path. Figures 13-14The results shown demonstrate the global continuous crack segmentation path extracted by this method when processing different material surfaces and cracks with features such as fineness, fuzziness, and discontinuity.

[0080] The purpose of using connected component filtering and skeletonization is to eliminate redundant branches and isolated noise points caused by numerical fluctuations in local mechanics solutions. Its advantage is that it refines the final output and removes unrealistic micro-branches.

[0081] As can be seen from the above embodiments, this application proposes a crack image segmentation method, apparatus, and electronic device based on damage mechanics. By deeply integrating computer vision feature extraction with the damage mechanics mechanism of continuous media, this application effectively overcomes many limitations of existing technologies in crack detection of complex engineering images, specifically solving the following technical problems and achieving significant technical effects: 1. Reduce the dimensionality of the computational system and optimize solution efficiency. Existing fracture mechanics-based segmentation methods (such as the 3D extended finite element method, XFEM) introduce enriched degrees of freedom to characterize the 3D crack surface, increasing the dimensionality of the system stiffness matrix and computational overhead. Implicit nonlinear solutions are prone to stress singularities when facing irregular image mapping boundaries, leading to model non-convergence. This application nonlinearly maps the image feature matrix to an initial yield stress field of a 2D mesh, replacing the 3D XFEM with a 2D plane stress finite element model, thus avoiding the introduction of 3D enriched degrees of freedom and reducing the dimensionality of the computational system. Simultaneously, an explicit dynamic analysis step combined with a forced element pruning mechanism replaces the implicit nonlinear solution. This mechanism avoids stress singularities and solution divergence caused by local mapping irregularities, improving the algorithm's running efficiency and convergence stability while maintaining the integrity of the physical and mechanical logic.

[0082] 2. Discontinuous bridging based on mechanical stress concentration mechanism Existing image processing algorithms that rely on pixel gradients are prone to path breakage when dealing with low-contrast cracks or cracks with physical discontinuities; existing mechanical tracking methods are susceptible to cessation of tracking due to fluctuations in local stress fields. This application introduces a continuum damage mechanics model to drive evolution. When the tip of a macroscopic crack composed of failure elements approaches a discontinuous region in the image exhibiting high yield stress, the scheme utilizes the stress concentration effect induced by geometric discontinuities to amplify the local true stress. This mechanism enables high-strength elements in the discontinuous region to satisfy the damage evolution criterion based on fracture energy and undergo failure pruning, achieving the bridging and continuous extraction of crack breakpoints.

[0083] 3. Not data-driven, adaptable to cross-scenario applications Existing deep learning-based segmentation models rely on data-driven statistical fitting and image-annotated data, resulting in high training costs and limited generalization ability across different scenarios. The crack segmentation evolution process in this application is driven by constitutive relations (such as initial yield and equivalent plastic strain accumulation) and damage evolution rules. The algorithm extracts local geometric features from the current image and performs mechanical field mapping, without introducing an image dataset for model training. This method extracts crack morphology through the aforementioned physical mechanism when facing different lighting conditions and structural surfaces (such as cement pavements and walls), demonstrating cross-scenario versatility.

[0084] 4. Supports continuous solving of large-sized images. Existing fracture mechanics methods consume significant global computational resources when directly processing large-scale engineering images, and require pre-setting global cracks as initial boundary defects. This application designs a local tracking window and a load condition update module. By extracting the centroid coordinates of the current local window failure element, the least squares method is used for linear fitting to predict the propagation direction, and the boundary conditions of the local loading displacement are dynamically updated. This local sliding window iteration and coordinate system update mechanism provides the initial state and loading direction for mechanical segmentation, enabling the failure segmentation process to advance in the global image, and generating a continuous segmentation path through global coordinate stitching.

[0085] Corresponding to the aforementioned embodiments of the crack image segmentation method based on damage mechanics, this application also provides embodiments of a crack image segmentation apparatus based on damage mechanics.

[0086] Figure 15 This is a block diagram of a crack image segmentation device based on damage mechanics, according to an exemplary embodiment. (Refer to...) Figure 15 The device includes: Acquisition module 1 is used to acquire the original crack image to be detected; Preprocessing module 2 is used to perform local preprocessing on the original crack image within the current local tracking window to obtain a local feature matrix containing crack geometric feature information; Mapping module 3 is used to nonlinearly map the local feature matrix into an initial yield stress field in a two-dimensional structured grid, wherein the potential crack target area in the image is mapped to a low yield stress area and the background area is mapped to a high yield stress area. Boundary condition loading module 4 is used to dynamically apply time-modulated symmetric displacement boundary conditions based on the current crack tip and its propagation direction, so as to induce a local tensile stress field in the two-dimensional structured mesh. The damage evolution solution module 5 is used to drive the accumulation of damage and stiffness degradation of the elements in the low yield stress zone of the two-dimensional structured mesh based on the continuous medium damage mechanics model through the damage initiation criterion and damage evolution rules. When the damage variable of the element reaches the critical threshold, the element is removed from the mesh to form a failed element. The local stress concentration effect caused by the geometric discontinuity formed by the failed element is used to achieve the bridging of the adjacent high yield stress zone to characterize the physical propagation of the crack. The window iteration and path integration module 6 is used to extract the coordinates of the failure elements in the current local tracking window, determine the crack propagation direction through linear fitting, and map the coordinates of the terminal failure elements to the initial state and load conditions of the next local tracking window. Iteratively execute the preprocessing module to the window iteration and path integration module until the termination condition is met, integrate the coordinates of the failure elements of all local windows, and obtain the final continuous crack segmentation path.

[0087] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0088] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0089] Accordingly, this application also provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; and when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the crack image segmentation method based on damage mechanics as described above.

[0090] Accordingly, this application also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the crack image segmentation method based on damage mechanics as described above.

[0091] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only.

[0092] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A crack image segmentation method based on damage mechanics, characterized in that, include: S1: Obtain the original image of the crack to be detected; S2: Perform local preprocessing on the original crack image within the current local tracking window to obtain a local feature matrix containing crack geometric feature information; S3: Nonlinearly map the local feature matrix to the initial yield stress field in a two-dimensional structured grid, wherein the potential crack target area in the image is mapped to a low yield stress area and the background area is mapped to a high yield stress area. S4: Based on the current crack tip and its propagation direction, dynamically apply time-modulated symmetric displacement boundary conditions to induce a local tensile stress field in the two-dimensional structured mesh; S5: Based on the continuous medium damage mechanics model, the elements in the low yield stress zone are driven to accumulate damage and undergo stiffness degradation through damage initiation criteria and damage evolution rules. When the damage variable of the element reaches the critical threshold, the element is removed from the mesh to form a failed element. The local stress concentration effect caused by the geometric discontinuity formed by the failed element is used to achieve the bridging of the adjacent high yield stress zone to characterize the physical propagation of the crack. S6: Extract the coordinates of the failure elements within the current local tracking window, determine the crack propagation direction through linear fitting, and map the coordinates of the terminal failure elements to the initial state and load conditions of the next local tracking window. Iterate through steps S2 to S6 until the termination condition is met, integrate the coordinates of the failure elements in all local windows, and obtain the final continuous crack segmentation path.

2. The method according to claim 1, characterized in that, Within the current local tracking window, the original crack image is locally preprocessed to obtain a local feature matrix containing crack geometric feature information, including: S21: Perform median filtering on the original crack image within the current local tracking window to estimate background illumination and texture distribution; S22: Perform image subtraction on the median-filtered image to extract image details; S23: Perform bilateral filtering on the subtractive image to smooth noise and preserve crack edges; S24: Perform pixel-by-pixel Gaussian smoothing on the bilaterally filtered image to construct the Hessian matrix; S25: Perform multi-scale Frangi linear filtering on the Hessian matrix to enhance the linear structure features, and generate the local feature matrix through nonlinear mapping.

3. The method according to claim 1 or 2, characterized in that, The local characteristic matrix is ​​nonlinearly mapped to an initial yield stress field in a two-dimensional structured mesh, including: S31: In accordance with the principle of one pixel corresponding to one finite element mesh unit, a two-dimensional orthogonal structured finite element mesh model is established within the local tracking window, and the pixel coordinates are mapped to the corresponding node coordinates; S32: The eigenvalues ​​in the local feature matrix are discretized and mapped to physical yield stress using a nonlinear mapping function, wherein the potential crack target area in the image is mapped to a low yield stress area and the background area is mapped to a high yield stress area. S33: Generate section properties based on the discretized physical yield stress and assign them to the corresponding elements in the two-dimensional orthogonal structured finite element mesh model to form an initial yield stress field.

4. The method according to claim 3, characterized in that, The two-dimensional orthogonal structured finite element mesh model adopts an ideal elastic-plastic constitutive model in the plastic stage, so that the elements accumulate equivalent plastic strain without strain hardening after reaching the initial yield stress under load.

5. The method according to claim 1, characterized in that, Based on the current crack tip and its propagation direction, a time-modulated symmetric displacement boundary condition is dynamically applied to induce a local tensile stress field in the two-dimensional structured mesh, including the following sub-steps: S41: Based on the current crack tip and its propagation direction, calculate the normal vector perpendicular to the propagation direction, and set the center coordinates of the two loading regions in front of the current crack tip; S42: Based on the normal vector and center coordinates, extract the mesh nodes in the loading areas on both sides to construct a local tension node set; S43: Apply local opposing loads to the local tensile node set using symmetrical displacement boundary conditions to generate a stable local Mode-I tearing stress field in front of the current crack tip, which serves as a local tensile stress field.

6. The method according to claim 1, characterized in that, Based on the continuum damage mechanics model, damage accumulation and stiffness degradation of elements in the low yield stress zone are driven by damage initiation criteria and damage evolution rules. When the damage variable of an element reaches a critical threshold, the element is removed from the mesh to form a failed element. The local stress concentration effect caused by the geometric discontinuity formed by the failed element is used to bridge the adjacent high yield stress zone, thereby characterizing the physical propagation of the crack. The process includes the following sub-steps: S51: Based on the continuous medium damage mechanics model, the finite element solution is performed using explicit dynamic analysis steps, and a tough damage mechanism is introduced. When the equivalent plastic strain of the element reaches the critical value, the damage initiation criterion is triggered. S52: After the damage initiation criterion is triggered, the damage evolution stage is entered. The damage evolution rule based on equivalent plastic displacement is adopted to control the stiffness of the low yield stress zone element to degrade linearly with the accumulation of damage variables. S53: When the damage variable of the unit reaches the critical threshold, the state variable of the unit is forcibly set to zero, and the unit is permanently removed from the two-dimensional structured mesh to form the failed unit. The geometric discontinuity formed by the failed unit induces a local stress concentration effect in front of the crack tip, so that the damage variable of the unit in the adjacent high yield stress zone reaches the critical threshold, thereby realizing the bridging of the high yield stress zone to characterize the physical propagation of the crack.

7. The method according to claim 1, characterized in that, Extract the coordinates of the failed elements within the current local tracking window, determine the crack propagation direction through linear fitting, and map the coordinates of the terminal failed element to the initial state and load conditions of the next local tracking window. Iterate through steps S2 to S6 until the termination condition is met. Integrate the coordinates of the failed elements from all local windows to obtain the final continuous crack segmentation path, including the following sub-steps: S61: Read the state variables of the unit to filter out the units that have been removed, and extract the centroid coordinates of the failed unit that is farthest from the local calculation start point in the current local tracking window; S62: Establish the centroid coordinates as the current crack tip, and use them as the geometric center of the next local tracking window; S63: Based on the current crack tip, backtrack to extract the coordinates of the previous failed unit, and use the least squares method to perform local linear fitting of the spatial coordinates of the previous failed unit to determine the crack propagation direction; S64: Using the geometric center of the next local tracking window as a reference, back a specified length along the crack propagation direction, and map the end crack segment as a pre-fabricated crack into the initial state of the next local tracking window; S65: Determine whether the termination condition is met. If not, iterate through steps S2 to S64 until the tracking process terminates. Then, concatenate the coordinates of the failed units obtained in each iteration into global coordinates and map them into a binary mask. S66: Use connected component area filtering technology to remove isolated noise points in the global binary mask, perform skeletonization processing on the retained core components and prune short branches to obtain the final continuous crack segmentation path.

8. A crack image segmentation device based on damage mechanics, characterized in that, include: The acquisition module is used to acquire the original crack image to be detected; The preprocessing module is used to perform local preprocessing on the original crack image within the current local tracking window to obtain a local feature matrix containing crack geometric feature information; The mapping module is used to nonlinearly map the local feature matrix into an initial yield stress field in a two-dimensional structured grid, wherein the potential crack target area in the image is mapped to a low yield stress area and the background area is mapped to a high yield stress area. A boundary condition loading module is used to dynamically apply time-modulated symmetric displacement boundary conditions based on the current crack tip and its propagation direction to induce a local tensile stress field in the two-dimensional structured mesh. The damage evolution solution module is used to drive the accumulation of damage and stiffness degradation of the elements in the low yield stress zone based on the damage mechanics model of continuous medium through damage initiation criteria and damage evolution rules. When the damage variable of the element reaches the critical threshold, the element is removed from the mesh to form a failed element. The local stress concentration effect caused by the geometric discontinuity formed by the failed element is used to achieve the bridging of the adjacent high yield stress zone to characterize the physical propagation of cracks. The window iteration and path integration module is used to extract the coordinates of the failure elements within the current local tracking window, determine the crack propagation direction through linear fitting, and map the coordinates of the terminal failure elements to the initial state and load conditions of the next local tracking window. Iteratively execute the preprocessing module to the window iteration and path integration module until the termination condition is met, integrate the coordinates of the failure elements of all local windows, and obtain the final continuous crack segmentation path.

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 implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the method as described in any one of claims 1-7.

Images