Highway pavement microcrack detection system based on improved swarm algorithm

Abstract

The invention discloses a highway pavement microcrack detection system based on an improved swarm algorithm, and the system comprises an image collection module which is used for outputting a standardized highway pavement image; the image preprocessing module is used for generating a preprocessed highway pavement image; the parameter optimization module is used for performing global search and local refinement search on the preprocessed highway pavement image based on a two-stage bitter fish optimization algorithm model; the image segmentation module is used for generating a preliminary micro-crack candidate region map; the feature extraction module is used for forming a depth feature map; the region reconstruction module is used for outputting a final highway pavement micro-crack region segmentation result; and the detection and output module is used for carrying out candidate region detection and positioning on the highway pavement micro-crack region segmentation result and outputting a final highway pavement micro-crack detection result containing the micro-crack position and region information. According to the method, the response capability to a weak edge is improved, a crack topological structure consistency function is fused in a local refining stage, and the detection structure reduction precision of a micro-crack region is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of crack detection, and particularly to a highway pavement micro-crack detection system based on an improved swarm algorithm. Background Art

[0002] With the acceleration of the urbanization process and the improvement of the transportation intensity, micro-crack early diseases are likely to occur in highway pavements during long-term high-load use. Although they are difficult to detect in the initial stage, if not discovered and treated in time, they are extremely likely to evolve into more serious structural damages, thereby affecting traffic safety and road life. Therefore, carrying out research on high-precision and high-efficiency highway pavement micro-crack detection technology has become a key task in the field of road maintenance and management.

[0003] Currently, the detection methods for highway pavement micro-cracks mainly include manual inspection, laser scanning, infrared thermal imaging, and image processing technology paths. Among them, the manual inspection method is limited by subjective judgment and is not suitable for large-scale rapid detection; although laser scanning and thermal imaging have certain automation capabilities, their equipment costs are high, the operation is complex, and the resolution is limited when detecting fine cracks, resulting in problems of missed detection or false detection. Therefore, the image processing method based on computer vision has gradually become a research hotspot.

[0004] However, the existing image processing-based detection methods usually rely on fixed threshold segmentation, edge detection, or shallow feature extraction, and it is difficult to cope with the problems of low contrast, weak edges, and irregular shapes of micro-crack images in complex road scenes. For example, traditional Canny or Sobel operators are prone to generating false edges when dealing with random noise or uneven illumination in images, resulting in inaccurate crack positioning; the fixed threshold strategy is difficult to adapt to the diversity of different image qualities and crack morphologies. In addition, most image segmentation methods lack analysis means for the structural connectivity and topological consistency of micro-cracks, and are prone to problems such as fracture, adhesion, or missing areas, affecting the integrity and accuracy of the final detection results.

[0005] Therefore, there is an urgent need for a new micro-crack detection method that combines intelligent optimization algorithms and deep image feature modeling capabilities to effectively improve the stability, accuracy, and practicality of detection. Summary of the Invention

[0006] An object of the present invention is to propose a highway pavement micro-crack detection system based on an improved swarm algorithm. The present invention improves the response ability to weak edges and integrates a crack topological structure consistency function in the local refinement stage, improving the detection structure reduction accuracy of the micro-crack area.

[0007] A highway pavement micro-crack detection system based on an improved swarm algorithm according to an embodiment of the present invention includes:

[0008] An image acquisition module, which is used to acquire road surface images, complete standardization processing, and output standardized road surface images;

[0009] An image preprocessing module, which is used to perform noise suppression and edge enhancement operations on the standardized road surface images to generate preprocessed road surface images;

[0010] A parameter optimization module, which performs global search and local refinement search on the preprocessed road surface images based on the two-stage bitter fish optimization algorithm model, and outputs the final fine image segmentation parameters;

[0011] An image segmentation module, which takes the preprocessed road surface images and the final fine image segmentation parameters as inputs, performs preliminary microcrack region image segmentation, and generates a preliminary microcrack candidate region map;

[0012] A feature extraction module, which inputs the preliminary microcrack candidate region map into a multi-scale dilated convolution module, generates a multi-scale feature response map, and performs channel attention mechanism and feature splicing processing to form a depth feature map;

[0013] A region reconstruction module, which performs adaptive threshold binarization and morphological reconstruction on the depth feature map, and outputs the final road surface microcrack region segmentation result;

[0014] A detection and output module, which performs candidate region detection and positioning on the road surface microcrack region segmentation result, judges and filters the road surface microcrack region by using region contour analysis, and outputs the final road surface microcrack detection result including the microcrack position and region information.

[0015] A method for detecting road surface microcracks based on an improved swarm algorithm, which is applied to a road surface microcrack detection system based on an improved swarm algorithm, and includes the following steps:

[0016] S1. Acquire road surface images and perform standard processing to form standardized road surface images;

[0017] S2. Preprocess the standardized road surface images to generate preprocessed road surface images;

[0018] S3. Use the two-stage bitter fish optimization algorithm model constructed based on the preprocessed road surface images, initialize the optimization population in the global search stage, and evaluate the image segmentation parameters according to the fitness function to obtain coarse-grained image segmentation parameters;

[0019] S4. Use the coarse-grained image segmentation parameters to enter the local refinement search stage of the two-stage bitter fish optimization algorithm model, and obtain the final fine image segmentation parameters through local perturbation and regulation;

[0020] S5. Take the preprocessed road surface image and the final fine image segmentation parameters as inputs, perform image segmentation operations, and generate a preliminary microcrack candidate region map;

[0021] S6. Input the preliminary microcrack candidate region map into the multi-scale hollow convolution module, generate a multi-scale feature response map, and perform channel attention mechanism and feature splicing processing to form a depth feature map;

[0022] S7. Perform adaptive threshold binaryzation processing and morphological reconstruction on the depth feature map to eliminate residual noise and repair crack continuity, and output the final road surface microcrack region segmentation result;

[0023] S8. Perform candidate region detection and positioning on the road surface microcrack region segmentation result, use region contour analysis to judge and screen the road surface microcrack region, and output the final road surface microcrack detection result containing the microcrack position and region information.

[0024] Optionally, S1 includes the following steps:

[0025] S11. Set the road surface image acquisition frequency, obtain the road surface image through the road surface image acquisition device, and the resolution of the road surface image is R w ×R h , where R w represents the width of the road surface image, and R h represents the height of the road surface image;

[0026] S12. Perform size normalization processing on the collected original road surface image, and uniformly scale different-sized road surface images to the standard size R s = W s ×H s , where R s represents the standard road surface image size, and W s and H s are the unified width and height of the standard road surface image respectively;

[0027] S13. Perform a color space conversion operation on the size-normalized road surface image, and convert the RGB road surface image to a grayscale space road surface image;

[0028] S14. Perform road surface image enhancement processing on the grayscale road surface image to improve the resolvability of microcracks in low-contrast regions, obtain the enhanced road surface image I enh , and use the enhanced road surface image as the standardized road surface image.

[0029] Optionally, S2 includes the following steps:

[0030] S21. Based on the standardized road surface image I enh , a Gaussian filter is used to perform preliminary noise suppression processing on the road surface image to suppress the random noise generated during the environmental acquisition process, and a Gaussian smoothed road surface image I gauss is generated;

[0031] S22. For the Gaussian smoothed road surface image, the filtering result is regulated by using the gradient direction of the road surface image and the brightness difference between adjacent pixels to generate an edge-preserved road surface image;

[0032] S23. Perform multi-channel filtering processing on the edge-preserved road surface image, and respectively construct guiding filter channels in different directions {I θ}, where θ ∈ {0°, 45°, 90°, 135°}. Each direction filtering channel enhances the micro-crack linear structure through directional response and generates a set of direction-enhanced road surface images {I dir,θ};

[0033] S24. Perform response fusion processing on the multi-channel direction-enhanced road surface image set {I dir,θ}, and calculate the fused road surface image I fused by using a weighted superposition strategy;

[0034] S25. Perform edge detection processing on the fused road surface image, and use the Sobel operator to calculate the gradient response values G x (x, y) and G y (x, y) of the road surface image in the horizontal and vertical directions to obtain the edge response map I edge of the road surface image;

[0035] S26. Perform pixel-level fusion processing on the edge response map I edge and the Gaussian smoothed road surface image I gauss to generate a preprocessed road surface image I pre containing noise suppression and edge enhancement features:

[0036]

[0037] Among them, θ1 is a noise modulation factor used to enhance the noise suppression effect of the Gaussian smoothed road surface image I gauss , θ2 is an attenuation constant used to regulate the attenuation rate of the influence of the exponential function on the Gaussian smoothed road surface image I gauss in different brightness regions, and θ3 is an edge enhancement scaling factor used to magnify the edge response map I edgeThe contribution is to retain the edge details of microcracks during the fusion process. θ4 is the edge sensitivity adjustment factor, which adjusts the non-linear mapping of the edge response value through the hyperbolic tangent function to distinguish between low-response and high-response regions. μ is the edge threshold parameter that sets the reference threshold for the edge response. Only when the edge response map I edge (x, y) exceeds the edge threshold parameter, the edge information is enhanced. exp(·) and tanh(·) represent the exponential function and the hyperbolic tangent function respectively.

[0038] Optionally, S3 includes the following steps:

[0039] S31. Construct a two-stage bitter fish optimization algorithm model including a global search stage and a local refinement search stage, and initialize the bitter fish optimization population The bitter fish optimization individual X i represents a set of preprocessing parameters for highway pavement image segmentation, including the binary threshold T for the preliminary segmentation of the microcrack region i , the filter kernel size K i and the feature enhancement weight parameter λ i , where N is the size of the bitter fish optimization population;

[0040] S32. For the characteristics of highway pavement microcrack region segmentation detection, based on the bitter fish optimization individual X i construct a fitness function F(X i ):

[0041]

[0042] Among them, is the between-class variance between the microcrack region and the background region in the preprocessed highway pavement image under the action of the preprocessing highway pavement image segmentation parameters X i , which measures the segmentation discrimination of the microcrack region. E edge (X i ) is the edge saliency of the preprocessed highway pavement image under the action of the preprocessing highway pavement image segmentation parameters X i , which is used to evaluate the clarity of the microcrack edge. R conn (X i ) is the connectivity index of the microcrack region in the preprocessed highway pavement image under the action of the preprocessing highway pavement image segmentation parameters X i , which is used to evaluate the continuity of the microcrack region. α1, α2, and α3 are the corresponding weight coefficients;

[0043] S33. In the global search stage of the two-stage bitter fish optimization, calculate the current optimal bitter fish optimization individual position i and the average position of the bitter fish optimization population according to the fitness function F(X And introduce a crack sensitivity factor

[0044]

[0045] wherein, is the image edge saliency corresponding to the current optimal bitter fish optimized individual, is the average value of the edge saliencies of all bitter fish optimized individuals in the current bitter fish optimization population, and β is the crack sensitivity adjustment coefficient;

[0046] S34. Use the crack sensitivity factor to update the positions of the bitter fish optimized individuals:

[0047]

[0048] wherein, respectively represent the positions of the bitter fish optimized individuals at the t-th and (t + 1)-th iterations, and r2 and r1 are random factors;

[0049] S35. Set the crack structure convergence threshold in the global search stage, and calculate the fitness value of the current optimal bitter fish optimized individual according to the fitness function. When the fitness value of the current optimal bitter fish optimized individual is greater than or equal to the crack structure convergence threshold, terminate the iteration in the global search stage and obtain the coarse-grained image segmentation parameter X for the preliminary segmentation of microcracks on the highway pavement global , otherwise return to step S33 to continue the global search process.

[0050] Optionally, S4 includes the following steps:

[0051] S41. In the local refinement search stage of the two-stage bitter fish optimization algorithm model, construct a set of local bitter fish optimized individuals where M << N, and each local bitter fish optimized individual X′ i is generated by adding a perturbation vector Δ i to the coarse-grained image segmentation parameter:

[0052]

[0053] wherein, represents a multi-dimensional Gaussian perturbation distribution with a mean of 0 and a covariance matrix of Σ loc for fine-grained sampling of the local search space;

[0054] S42. Based on the preprocessed highway pavement image I pre construct a microcrack topology graph G=(V, E), where V represents the set of crack candidate points extracted by edge detection, and E represents the set of crack structure edges connected by the local gradient direction and the distance threshold. Combine the topological structure to construct a crack region consistency function C(X′i , G), for evaluating the matching degree between the current micro-crack segmentation result and the true topological structure of the micro-cracks:

[0055]

[0056] where R(X′ i ) represents the micro-crack segmentation region obtained under the parameter X′ i , θ uv is the direction consistency angle of the crack structure edge (u, v), is an indicator function for judging whether the endpoints of the crack structure edge are simultaneously in the candidate region;

[0057] S43. Construct a crack region consistency function based on the topological structure to build a composite fitness function to evaluate the locally optimized individuals of bitter fish:

[0058] F local (X′ i ) = γ1·F(X′ i ) + γ2·C(X′ i , G);

[0059] where F(X′ i ) is the fitness function of the locally optimized individuals of bitter fish, and γ1, γ2 are the weight coefficients of the global search stage and the local refinement search stage;

[0060] S44. Apply the bitter fish optimization update strategy to the locally optimized individual X′ i , set the update direction of the local search to the vector difference between the current locally optimal bitter fish optimization individual X′ best and the current bitter fish optimization individual position X′ i , and introduce a dynamic step factor to finely control the search process. The dynamic step factor is dynamically attenuated according to the composite fitness function F local (X′ i ) and the preset target composite fitness function F target , and the step size gradually decreases as the individual fitness improves;

[0061] S45. When the termination condition is satisfied, that is, the fitness value of the current locally optimal bitter fish optimization individual is greater than or equal to the local crack structure convergence threshold, output the final fine image segmentation parameter X final = {T final , K final , λ final} and complete the local search of the two-stage bitter fish optimization.

[0062] Optionally, the S5 includes the following steps:

[0063] S51. Based on the preprocessed road surface image I pre and the final fine image segmentation parameter X final , perform the preliminary image segmentation operation of the road surface microcracks to generate the preliminary segmentation image I seg :

[0064]

[0065] wherein, I seg (x,y) represents the pixel value at the coordinate (x,y) after the preliminary image segmentation, and T final is the optimal binary threshold in the final fine image segmentation parameter, which is used to distinguish the microcrack region from the non-crack region;

[0066] S52. Based on the filter kernel size K in the final fine image segmentation parameter final perform the morphological closing operation on the preliminary segmentation image I seg to generate the preliminary microcrack candidate region map I with enhanced connectivity crack :

[0067]

[0068] wherein, and respectively represent the morphological dilation and erosion operations, and the size of the structural element is determined by the final fine image segmentation parameter K final and is used to smooth the interference and fracture regions in the microcrack candidate region map;

[0069] S53. Based on the feature enhancement weight parameter λ in the final fine image segmentation parameter final perform the microcrack region feature enhancement operation on the preliminary microcrack candidate region map I crack to obtain the enhanced preliminary microcrack candidate region map I 增强 :

[0070] I 增强 (x,y) = I crack (x,y) · [1 + λ final · I edge (x,y)];

[0071] wherein, λ final is the weight coefficient, which is used to enhance the edge features of the microcrack region.

[0072] Optionally, the S6 includes the following steps:

[0073] S61. The enhanced preliminary microcrack candidate region map I 增强Input multi-scale dilated convolution sub-module. In the multi-scale dilated convolution sub-module, a set of dilation rates D = {d1, d2, d3, d4} is set, and four groups of dilated convolution branches are constructed in parallel. Each group of branches uses a corresponding dilation rate d i of the dilated convolution kernel Perform dilated convolution operation on the input image to generate a dilated feature map :

[0074]

[0075] where represents the dilated feature response at the position (x, y) under the dilation rate d i ;

[0076] S62. Input all the dilated feature maps into the channel attention sub-module, perform attention weight assignment on the multi-scale channel features extracted from different convolution branches, and obtain a channel-weighted response map

[0077]

[0078] where Pool(·) represents the combined operation of global average pooling and maximum pooling, and MLP(·) is a channel-wise weight learning network represents the response weight of the feature map with the dilation rate d i ;

[0079] S63. Input the channel-weighted response map into the feature fusion decoding sub-module, perform channel dimension concatenation to obtain a fused feature map F fused , and then input the fused feature map F fused into the upsampling decoding network, perform upsampling layer by layer and spatial feature restoration using transposed convolution operation to generate a depth feature map F deep with the same size as the input image

[0080] Optionally, S7 includes the following steps:

[0081] S71. Input the depth feature map F deep into the adaptive threshold module, automatically generate a spatial dynamic threshold map T adapt (x, y) according to the local statistical features, and perform a binary operation in combination with the pixel response value to generate a preliminary segmentation result map I bin ;

[0082] S72. Perform morphological reconstruction operation on the preliminary segmentation result map I bin using the thin linear and connectivity features of microcracks, and adopt a structural element S recPerform morphological closing operation and thinning process to generate a microcrack region map I with continuous structure recon :

[0083]

[0084] Among them, are dilation and erosion operations respectively, Thin(·) represents the thinning operation, and the structural element S rec is set according to the width range of the target crack;

[0085] S73. For the microcrack region map I after structure reconstruction recon Perform connected component filtering and area constraint processing to remove isolated small regions with an area smaller than the set threshold A min to form the final microcrack region segmentation map I final .

[0086] Optionally, the S8 includes the following steps:

[0087] S81. Perform connected component extraction operation on the final microcrack region segmentation map I final to obtain the boundary contour set of all independent microcrack candidate regions and calculate the morphological parameter set of each candidate region respectively Among them, A k is the area of the kth region, is the aspect ratio, is the average response value of the internal depth feature of the region;

[0088] S82. Combine the final fine image segmentation parameters to construct a dynamic microcrack region classification rule, and set the following thresholds:

[0089] Area dynamic threshold where α1 is the area adjustment coefficient;

[0090] Low response intensity threshold T low = α2·T final , high threshold T high = α3·λ final ·T final , where α2 and α3 are response adjustment coefficients respectively;

[0091] S83. According to the performance characteristics of microcrack morphology in actual highway detection, divide the detection results into three categories: suspected microcrack region, obvious microcrack region and non-crack region:

[0092] The classification rule for the non-crack region is to meet any of the following conditions: A k < A thresh , or and

[0093] The classification rules for suspected microcrack regions are that all of the following conditions are met: A k ≥A thresh , and

[0094] The classification rules for obvious microcrack regions are that all of the following conditions are met: A k ≥A thresh , and

[0095] S84. Extract the position information of the classified crack candidate regions and output them in a structured manner. Extract the coordinates of the boundary center points, the circumscribed rectangle bounding boxes, and the region numbers, classification labels, and morphological parameters of each obvious microcrack region to form the final highway pavement microcrack detection result set.

[0096] The beneficial effects of the present invention are as follows:

[0097] (1) By constructing a two-stage bitter fish optimization algorithm model, the present invention introduces a double-layer mechanism of global search and local refinement search in the optimization process of image segmentation parameters, realizing a progressive optimization strategy from coarse-grained to refined. The optimized model not only introduces a crack sensitivity factor in the global search stage to enhance the response ability to weak edges, but also fuses a crack topological structure consistency function in the local refinement stage, effectively improving the structure reduction accuracy of microcrack regions.

[0098] (2) In the process of microcrack feature extraction, the present invention introduces a deep feature encoding network that combines multi-scale dilated convolution and channel attention mechanism, which can capture crack texture information of different sizes, directions, and densities more comprehensively. By setting a set of dilation rates to construct multiple parallel dilated convolution branches and combining the response weight allocation mechanism between channels, the enhanced expression of the features of key crack regions is realized, especially showing significant advantages in pavement images with weak contrast and complex background textures. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0100] Figure 1 is a flowchart of a highway pavement microcrack detection system based on an improved swarm algorithm proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0101] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0102] Reference Figure 1 , a highway pavement micro-crack detection system based on an improved swarm algorithm, comprising:

[0103] An image acquisition module, configured to acquire a highway pavement image and complete standardization processing, and output a standardized highway pavement image;

[0104] An image preprocessing module, configured to perform noise suppression and edge enhancement operations on the standardized highway pavement image to generate a preprocessed highway pavement image;

[0105] A parameter optimization module, which performs global search and local refinement search on the preprocessed highway pavement image based on a two-stage bitter fish optimization algorithm model, and outputs the final fine image segmentation parameters;

[0106] An image segmentation module, which takes the preprocessed highway pavement image and the final fine image segmentation parameters as inputs, performs preliminary micro-crack region image segmentation, and generates a preliminary micro-crack candidate region map;

[0107] A feature extraction module, which inputs the preliminary micro-crack candidate region map into a multi-scale dilated convolution module, generates a multi-scale feature response map, and performs channel attention mechanism and feature splicing processing to form a depth feature map;

[0108] A region reconstruction module, which performs adaptive threshold binarization and morphological reconstruction on the depth feature map, and outputs the final highway pavement micro-crack region segmentation result;

[0109] A detection and output module, which performs candidate region detection and positioning on the highway pavement micro-crack region segmentation result, judges and filters the highway pavement micro-crack region by using region contour analysis, and outputs the final highway pavement micro-crack detection result including the position and region information of the micro-cracks.

[0110] A highway pavement micro-crack detection method based on an improved swarm algorithm, which is applied to a highway pavement micro-crack detection system based on an improved swarm algorithm, and comprises the following steps:

[0111] S1. Acquire a highway pavement image and perform standard processing to form a standardized highway pavement image;

[0112] S2. Preprocess the standardized highway pavement image to generate a preprocessed highway pavement image;

[0113] S3. Use the two-stage bitter fish optimization algorithm model constructed based on the preprocessed road pavement image. Initialize the optimization population in the global search stage, and evaluate the image segmentation parameters according to the fitness function to obtain the coarse-grained image segmentation parameters;

[0114] S4. Use the coarse-grained image segmentation parameters to enter the local refinement search stage of the two-stage bitter fish optimization algorithm model, and obtain the final fine image segmentation parameters through local perturbation and regulation;

[0115] S5. Take the preprocessed road pavement image and the final fine image segmentation parameters as inputs, perform image segmentation operations, and generate a preliminary microcrack candidate region map;

[0116] S6. Input the preliminary microcrack candidate region map into the multi-scale dilated convolution module to generate a multi-scale feature response map, and perform channel attention mechanism and feature splicing processing to form a depth feature map;

[0117] S7. Perform adaptive threshold binary processing and morphological reconstruction on the depth feature map to eliminate residual noise and repair the crack continuity, and output the final road pavement microcrack region segmentation result;

[0118] S8. Perform candidate region detection and localization on the road pavement microcrack region segmentation result, use region contour analysis to judge and screen the road pavement microcrack region, and output the final road pavement microcrack detection result including the microcrack position and region information.

[0119] In this embodiment, S1 includes the following steps:

[0120] S11. Set the road pavement image acquisition frequency, and obtain the road pavement image through the road pavement image acquisition device. The road pavement image acquisition resolution is R w ×R h , where R w represents the width of the road pavement image, and R h represents the height of the road pavement image;

[0121] S12. Perform size normalization processing on the acquired original road pavement image, and uniformly scale different-sized road pavement images to the standard size R s =W s ×H s , where R s represents the standard road pavement image size, and W s and H s are the unified width and height of the standard road pavement image respectively;

[0122] S13. Perform a color space conversion operation on the size-normalized road surface image to convert the RGB road surface image into a grayscale space road surface image;

[0123] S14. Perform road surface image enhancement processing on the grayscale road surface image to improve the resolvability of microcracks in low-contrast regions, and obtain the enhanced road surface image I enh , and use the enhanced road surface image as the standardized road surface image.

[0124] In this embodiment, S2 includes the following steps:

[0125] S21. Based on the standardized road surface image I enh , use a Gaussian filter to perform preliminary noise suppression processing on the road surface image, suppress the random noise generated during the environmental acquisition process, and generate the Gaussian smoothed road surface image I gauss ;

[0126] S22. Use the gradient direction of the road surface image and the brightness difference between adjacent pixels to regulate the filtering result of the Gaussian smoothed road surface image, and generate an edge-preserving road surface image;

[0127] S23. Perform multi-channel filtering processing on the edge-preserving road surface image, and respectively construct guiding filter channels in different directions {I θ}, where θ ∈ {0°, 45°, 90°, 135°}, and each directional filtering channel enhances the microcrack linear structure through directional response, and generates a set of direction-enhanced road surface images {I dir,θ};

[0128] S24. Perform response fusion processing on the multi-channel direction-enhanced road surface image set {I dir,θ}, and use a weighted superposition strategy to calculate the fused road surface image I fused ;

[0129] S25. Perform edge detection processing on the fused road surface image, and use the Sobel operator to calculate the gradient response values G x (x, y) and G y (x, y) of the road surface image in the horizontal and vertical directions, and obtain the edge response map I edge of the road surface image;

[0130] S26. Perform pixel-level fusion processing on the edge response map I edge and the Gaussian smoothed road surface image I gauss , and generate a preprocessed road surface image I pre that includes noise suppression and edge enhancement features:

[0131]

[0132] Among them, θ1 is the noise modulation factor, which is used to enhance the noise suppression effect of the Gaussian-smoothed highway pavement image I gauss ; θ2 is the attenuation constant, which is used to regulate the attenuation rate of the influence of the exponential function on the Gaussian-smoothed highway pavement image I in different brightness regions gauss ; θ3 is the edge enhancement scaling factor, which is used to amplify the contribution of the edge response map I edge so as to retain the edge details of the microcracks during the fusion process; θ4 is the edge sensitivity adjustment factor, which adjusts the non-linear mapping of the edge response value through the hyperbolic tangent function to distinguish the low-response and high-response regions; μ is the edge threshold parameter, which sets the reference threshold for the edge response. Only when the edge response map I edge (x, y) exceeds the edge threshold parameter, the edge information is enhanced. exp(·) and tanh(·) represent the exponential function and the hyperbolic tangent function respectively.

[0133] In this embodiment, S3 includes the following steps:

[0134] S31. Construct a two-stage bitter fish optimization algorithm model including a global search stage and a local refinement search stage, and initialize the bitter fish optimization population The bitter fish optimization individual X i represents a set of preprocessing highway pavement image segmentation parameters, including the binary threshold T for the preliminary segmentation of the microcrack region i , the filter kernel size K i and the feature enhancement weight parameter λ i , where N is the size of the bitter fish optimization population;

[0135] S32. For the characteristics of the highway pavement microcrack region segmentation detection, construct a fitness function F(X i ) based on the bitter fish optimization individual X i :

[0136]

[0137] Among them, is the between-class variance between the microcrack region and the background region in the preprocessing highway pavement image under the action of the preprocessing highway pavement image segmentation parameter X i , which measures the segmentation discrimination of the microcrack region; E edge (X i ) is the edge saliency of the preprocessing highway pavement image under the action of the preprocessing highway pavement image segmentation parameter X i , which is used to evaluate the clarity of the microcrack edge; R conn (X i ) is the preprocessing highway pavement image segmentation parameter X iPreprocess the connectivity index of the microcrack region in the road surface image under the action, which is used to evaluate the continuity of the microcrack region. α1, α2, and α3 are the corresponding weight coefficients respectively;

[0138] S33. In the global search stage of the two-stage bitter fish optimization, according to the fitness function F(X i ) Calculate the current optimal bitter fish optimization individual position And the average position of the bitter fish optimization population And introduce the crack sensitivity factor

[0139]

[0140] Among them, Is the image edge saliency corresponding to the current optimal bitter fish optimization individual, Is the average value of the edge saliencies of all bitter fish optimization individuals in the current bitter fish optimization population, and β is the crack sensitivity adjustment coefficient;

[0141] S34. Use the crack sensitivity factor To update the bitter fish optimization individual position:

[0142]

[0143] Among them, Respectively represent the positions of the bitter fish optimization individual at the t-th and t+1-th iterations, and r2 and r1 are random factors;

[0144] S35. Set the crack structure convergence threshold in the global search stage, and calculate the fitness value of the current optimal bitter fish optimization individual according to the fitness function. When the fitness value of the current optimal bitter fish optimization individual is greater than or equal to the crack structure convergence threshold, terminate the iteration in the global search stage and obtain the coarse-grained image segmentation parameter X global For the preliminary segmentation of microcracks on the road surface, otherwise return to step S33 to continue the global search process.

[0145] In this embodiment, S4 includes the following steps:

[0146] S41. In the local refinement search stage of the two-stage bitter fish optimization algorithm model, construct a local bitter fish optimization individual set Where M << N, and each local bitter fish optimization individual X′ i Is generated by adding a perturbation vector Δ i To the coarse-grained image segmentation parameter:

[0147]

[0148] Among them, Indicates a mean of 0 and a covariance matrix of Σloc The multi-dimensional Gaussian perturbation distribution is used for fine-grained sampling of the local search space;

[0149] S42. Based on the preprocessed highway pavement image I pre Construct a microcrack topology graph G=(V, E), where V represents the set of crack candidate points extracted by edge detection, and E represents the set of crack structure edges connected by the local gradient direction and distance threshold. Combine the topological structure to construct a crack region consistency function C(X′ i , G) for evaluating the matching degree between the current microcrack segmentation result and the true microcrack topology structure:

[0150]

[0151] where R(X′ i ) represents the microcrack segmentation region obtained under the parameter X′ i , θ uv is the direction consistency angle of the crack structure edge (u, v), is an indicator function for judging whether the endpoints of the crack structure edge are simultaneously in the candidate region;

[0152] S43. Construct a composite fitness function based on the crack region consistency function constructed from the topological structure to evaluate the local bitter fish optimization individuals:

[0153] F local (X′ i ) = γ1·F(X′ i ) + γ2·C(X′ i , G);

[0154] where F(X′ i ) is the fitness function of the local bitter fish optimization individual, and γ1 and γ2 are the weight coefficients of the global search stage and the local refinement search stage;

[0155] S44. Apply the bitter fish optimization update strategy to the local bitter fish optimization individual X′ i . Set the update direction of the local search as the vector difference between the current local optimal bitter fish optimization individual X′ best and the current bitter fish optimization individual position X′ i , and introduce a dynamic step factor to finely control the search process. The dynamic step factor is dynamically attenuated according to the composite fitness function F local (X′ i ) and the preset target composite fitness function F target . The step size gradually decreases as the individual fitness improves;

[0156] S45. When the fitness value of the current locally optimal bitter fish optimization individual meets the termination condition and is greater than or equal to the local crack structure convergence threshold, output the final fine image segmentation parameter X final ={T final ,K final ,λ final} and complete the local search of the two-stage bitter fish optimization.

[0157] In this embodiment, S5 includes the following steps:

[0158] S51. Based on the preprocessed road surface image I pre and the final fine image segmentation parameter X final , perform the preliminary image segmentation operation of the road surface microcracks to generate the preliminary segmentation image I seg :

[0159]

[0160] Among them, I seg (x,y) represents the pixel value at the coordinate (x,y) after the preliminary image segmentation, and T final is the optimal binary threshold in the final fine image segmentation parameter, which is used to distinguish the microcrack area from the non-crack area;

[0161] S52. Based on the filter kernel size K in the final fine image segmentation parameter final perform the morphological closing operation on the preliminary segmentation image I seg to generate the preliminary microcrack candidate region map I crack with enhanced connectivity:

[0162]

[0163] Among them, and respectively represent the morphological dilation and erosion operations, and the size of the structure element is determined by the final fine image segmentation parameter K final to smooth the interference and fracture regions in the microcrack candidate region map;

[0164] S53. Based on the feature enhancement weight parameter λ in the final fine image segmentation parameter final perform the microcrack region feature enhancement operation on the preliminary microcrack candidate region map I crack to obtain the enhanced preliminary microcrack candidate region map I 增强 :

[0165] I 增强 (x,y)=I crack (x,y)·[1+λ final ·I edge(x,y);

[0166] where λ final is a weight coefficient used to enhance the edge features of the microcrack region.

[0167] In this embodiment, S6 includes the following steps:

[0168] S61. Input the enhanced preliminary microcrack candidate region map I 增强 into the multi-scale dilated convolution sub-module. In the multi-scale dilated convolution sub-module, a set of dilation rates D = {d1, d2, d3, d4} is set, and four groups of dilated convolution branches are constructed in parallel. Each branch uses a dilated convolution kernel i with the corresponding dilation rate d to perform a dilated convolution operation on the input image to generate a dilated feature map

[0169]

[0170] where represents the dilated feature response at the position (x, y) under the dilation rate d i ;

[0171] S62. Input all the dilated feature maps into the channel attention sub-module to perform attention weight assignment on the multi-scale channel features extracted by different convolution branches, and obtain a channel-weighted response map

[0172]

[0173] where Pool(·) represents the combined operation of global average pooling and max pooling, and MLP(·) is a channel-wise weight learning network, represents the feature map response weight with the dilation rate d i ;

[0174] S63. Input the channel-weighted response map into the feature fusion decoding sub-module for channel dimension concatenation to obtain a fused feature map F fused , and then input the fused feature map F fused into the upsampling decoding network. Use the transposed convolution operation for layer-by-layer upsampling and spatial feature restoration to generate a depth feature map F deep with the same size as the input image.

[0175] In this embodiment, S7 includes the following steps:

[0176] S71. Input the depth feature map F deep into the adaptive threshold module to automatically generate a spatial dynamic threshold map T according to the local statistical featuresadapt (x, y) performs a binarization operation in combination with the pixel response value to generate a preliminary segmentation result map I bin ;

[0177] S72. For the preliminary segmentation result map I bin Perform a morphological reconstruction operation. Utilize the thin linear and connectivity characteristics of microcracks and adopt a structuring element S rec Perform a morphological closing operation and a thinning process to generate a microcrack region map I with continuous structure recon :

[0178]

[0179] Among them, are the dilation and erosion operations respectively. Thin(·) represents the thinning operation. The size of the structuring element S rec is set according to the width range of the target crack;

[0180] S73. For the microcrack region map I after structure reconstruction recon Perform a connected component filtering and area constraint process to remove isolated small regions with an area smaller than the set threshold A min to form the final microcrack region segmentation map I final .

[0181] In this embodiment, S8 includes the following steps:

[0182] S81. For the final microcrack region segmentation map I final Perform a connected component extraction operation to obtain the boundary contour set of all independent microcrack candidate regions and calculate the morphological parameter set of each candidate region respectively Among them, A k is the area of the k-th region, is the aspect ratio, is the average response value of the internal depth feature of the region;

[0183] S82. Combine the final fine image segmentation parameters to construct a dynamic microcrack region classification rule and set the following thresholds:

[0184] Area dynamic threshold where α1 is the area adjustment coefficient;

[0185] Low response intensity threshold T low = α2·T final , high threshold T high = α3·λ final ·T final where α2 and α3 are the response adjustment coefficients respectively;

[0186] S83. According to the performance characteristics of microcracks in actual highway detection, the detection results are divided into three categories: suspected microcrack areas, obvious microcrack areas, and non-crack areas:

[0187] The classification rule for non-crack areas is to meet any of the following conditions: A k <A thresh , or and

[0188] The classification rule for suspected microcrack areas is to meet all of the following conditions: A k ≥A thresh , and

[0189] The classification rule for obvious microcrack areas is to meet all of the following conditions: A k ≥A thresh , and

[0190] S84. Extract the location information and perform structured output on the classified crack candidate areas, and extract the boundary center point coordinates, circumscribed rectangle bounding box, and area number, classification label, and morphological parameters of each obvious microcrack area to form the final highway pavement microcrack detection result set.

[0191] Example 1:

[0192] At 9:30 am on November 5, 2024, the Traffic Survey and Design Institute of Province A and a scientific research team from a certain university carried out microcrack patrol detection on the Changde East section of the G5513 Chang-Zhang Expressway. The vehicle traffic volume on this section has increased rapidly in the past three months, and the risk of early minor structural damage has increased. To provide accurate data support for preventive maintenance, the implementation team decided to apply the system of the present invention.

[0193] During the test, the intelligent inspection vehicle equipped with the system of the present invention traveled steadily at a speed of 20 kilometers per hour, and synchronously collected dual-channel images through two FLIR black-and-white industrial cameras (model: FLIR BFS-U3-16S2M) installed at the bottom of the vehicle. The single-channel resolution is 2448×2048, and the frame rate is 8 fps. It traveled 12.2 kilometers from 9:30 am to 12:30 pm on the same day, and the total number of collected images was 17,568.

[0194] The system runs in real time on the on-vehicle edge computing terminal. First, the image acquisition module completes size normalization and color space conversion, unifies the image resolution to 1024×512, and converts it into a grayscale image. After image enhancement processing, a standardized image numbered CDX-11057 is sent to the preprocessing module.

[0195] In the image, there is a horizontal crack with a width of only about 0.35 mm at a position slightly to the left of the center of the image. Due to the influence of natural light reflection, the gray level of some areas of the crack is close to the background, and it is difficult for traditional detection methods to completely extract the crack boundary. However, after noise suppression, the system automatically enters the two-stage bitter fish optimization process. In the first stage, 30 individuals are used to perform coarse-grained global search, and the threshold of the crack area is initially obtained as 0.58. Subsequently, in the second stage, a crack topology map is generated based on the preprocessed image to guide the search individuals to perform local refinement adjustment. Finally, the fine segmentation parameters converge to X final ={T final =0.618,K final =3,λ final =0.92}.

[0196] The system performs segmentation and enters the feature extraction module. Four groups of convolution kernels with hole rates of 1, 2, 4, and 6 are used to extract multi-scale features in parallel. After channel attention enhancement and decoding upsampling processing, a depth feature map is generated. Subsequently, after adaptive threshold mapping and morphological closing reconstruction, the micro-crack area is successfully segmented from the image, and the system immediately enters the detection module.

[0197] The crack is recognized as an obvious micro-crack area by the system. The area of the segmented area is 46 pixels², the aspect ratio is 6.2, and the average depth response value is 0.91, exceeding the high threshold of 0.84 set by the detection rule. The detection results generate the following structured data:

[0198] Image number: CDX-11057;

[0199] Detection time: 2024-11-05 10:12:46;

[0200] Crack type: obvious micro-crack;

[0201] Location information: center point of the image coordinates (512, 244);

[0202] Bounding rectangle of the circumscribed rectangle: (478, 237, 546, 251);

[0203] Response value: 0.91;

[0204] System response time: 1.23 seconds;

[0205] After the system confirms the detection results, it automatically uploads the detection data to the cloud platform through the 5G module and sends a warning reminder to the engineering maintenance personnel responsible for this section of the road through the SMS interface. Ten minutes later, the maintenance personnel arrive at the designated location and use a handheld crack observation instrument to recheck the site. It is confirmed that the crack width is 0.36 mm, the length is 8 cm, and it is located at the edge of the right lane, which is the initial state of a typical temperature-type micro-crack. At 14:00 on the same day, this area was spray-marked and included in the key re-inspection plan for this month.

[0206] In addition, during the inspection of the entire section of the road, the system detected a total of 132 candidate micro-crack areas, including 78 obvious micro-cracks and 39 suspected micro-cracks. The maintenance personnel verified 50 crack targets through sampling inspection and confirmed that the accurate recognition rate reached 92%. Among them, there were 4 false alarms and 1 missed detection, both of which were micro-cracks with extremely small-scale reflective interference. Compared with the average time of 9.4 seconds per image for manual review, this system only requires an average of 1.6 seconds, with the recognition accuracy increased by 18% and the efficiency increased by nearly 5.9 times.

[0207] This real-road inspection case fully demonstrates that the method of the present invention has excellent accuracy and stability in the actual complex road surface environment, effectively overcomes the robustness problems of traditional methods under illumination and texture perturbations, and has the potential for all-weather, rapid deployment, and high-reliability engineering applications. It is particularly suitable for the automatic inspection tasks of micro-damage of key highway structures such as highway sections, tunnels, and overpasses.

[0208] The present invention constructs a two-stage bitter fish optimization algorithm model, introduces a double-layer mechanism of global search and local refinement search in the optimization process of image segmentation parameters, and realizes a progressive optimization strategy from coarse-grained to refined. The optimized model not only introduces a crack sensitivity factor in the global search stage to enhance the response ability to weak edges, but also integrates a crack topology structure consistency function in the local refinement stage, effectively improving the structure reduction accuracy of the micro-crack area.

[0209] The present invention introduces a deep feature encoding network that combines multi-scale dilated convolution and channel attention mechanism in the process of micro-crack feature extraction, which can capture crack texture information of different sizes, directions, and densities more comprehensively. By setting a set of dilation rates to construct multiple parallel dilated convolution branches and combining the response weight allocation mechanism between channels, it realizes the enhanced expression of the features of key crack areas, especially showing significant advantages in road surface images with weak contrast and complex background textures.

[0210] The present invention constructs an image preprocessing process that includes a joint enhancement mechanism for image edge response and Gaussian noise suppression, and proposes an image fusion model based on a pixel-level non-linear mapping function, effectively realizing the enhancement of weak edge signals and the suppression of random noise. By introducing a response modulation strategy of hyperbolic tangent and exponential functions, the edge contribution degree and noise weight are adaptively adjusted while enhancing the image, making the microcrack boundary clearer and significantly reducing background interference.

[0211] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A highway pavement microcrack detection system based on an improved swarm algorithm, characterized in that, Comprising: An image acquisition module, configured to acquire a highway pavement image and perform standardization processing, and output a standardized highway pavement image; An image preprocessing module, configured to perform noise suppression and edge enhancement operations on the standardized highway pavement image to generate a preprocessed highway pavement image; A parameter optimization module, which performs global search and local refinement search on the preprocessed highway pavement image based on a two-stage bitter fish optimization algorithm model, and outputs final fine image segmentation parameters; An image segmentation module, which takes the preprocessed highway pavement image and the final fine image segmentation parameters as inputs, performs preliminary microcrack region image segmentation, and generates a preliminary microcrack candidate region map; A feature extraction module, which inputs the preliminary microcrack candidate region map into a multi-scale dilated convolution module, generates a multi-scale feature response map, and performs channel attention mechanism and feature splicing processing to form a depth feature map; A region reconstruction module, which performs adaptive threshold binaryzation and morphological reconstruction on the depth feature map, and outputs the final highway pavement microcrack region segmentation result; A detection and output module, which performs candidate region detection and positioning on the highway pavement microcrack region segmentation result, judges and filters the highway pavement microcrack region by using region contour analysis, and outputs the final highway pavement microcrack detection result including the microcrack position and region information.

2. A method for detecting microcracks on a highway pavement based on an improved swarm algorithm, which is applied to the highway pavement microcrack detection system based on the improved swarm algorithm described in claim 1, and is characterized in that, Including the following steps: S1. Acquire a highway pavement image and perform standard processing to form a standardized highway pavement image; S2. Preprocess the standardized highway pavement image to generate a preprocessed highway pavement image; S3. Use a two-stage bitter fish optimization algorithm model constructed based on the preprocessed highway pavement image, initialize the optimization population in the global search stage, and evaluate the image segmentation parameters according to the fitness function to obtain coarse-grained image segmentation parameters; S4. Use the coarse-grained image segmentation parameters to enter the local refinement search stage of the two-stage bitter fish optimization algorithm model, and obtain the final fine image segmentation parameters through local perturbation and regulation; S5. Take the preprocessed highway pavement image and the final fine image segmentation parameters as inputs, perform image segmentation operations, and generate a preliminary microcrack candidate region map; S6. Input the preliminary microcrack candidate region map into a multi-scale dilated convolution module, generate a multi-scale feature response map, and perform channel attention mechanism and feature splicing processing to form a depth feature map; S7. Perform adaptive threshold binaryzation processing and morphological reconstruction on the depth feature map, eliminate residual noise and repair crack continuity, and output the final highway pavement microcrack region segmentation result; S8. Perform candidate region detection and positioning on the highway pavement microcrack region segmentation result, judge and filter the highway pavement microcrack region by using region contour analysis, and output the final highway pavement microcrack detection result including the microcrack position and region information.

3. A method for detecting microcracks on a highway pavement based on an improved swarm algorithm according to claim 1, characterized in that, The said S1 includes the following steps: S11. Set the acquisition frequency of highway pavement images, and obtain highway pavement images through the highway pavement image acquisition device. The acquisition resolution of the highway pavement images is R w ×R h , where R w represents the width of the highway pavement image, and R h represents the height of the highway pavement image; S12. Perform size normalization on the collected original road surface images to uniformly scale road surface images of different sizes to the standard size R s = W s × H s , where R s represents the standard road surface image size, and W s and H s are the unified width and height of the standard road surface image, respectively; S13. Perform a color space conversion operation on the highway pavement image after size normalization, and convert the RGB highway pavement image into a grayscale space highway pavement image; S14. Perform image enhancement processing on the grayscale road surface image to improve the resolvability of microcracks in low-contrast areas, and obtain the enhanced road surface image I. enh , and use the enhanced road surface image as the standardized road surface image.

4. A method for detecting microcracks on a highway pavement based on an improved swarm algorithm according to claim 1, characterized in that The said S2 includes the following steps: S21. Based on the standardized road surface image I enh , a Gaussian filter is used to perform preliminary noise suppression processing on the road surface image, suppressing the random noise generated during the environmental acquisition process, and generating a Gaussian smoothed road surface image I gauss ; S22. Adjust the filtering result of the Gaussian-smoothed highway pavement image by using the luminance difference between the gradient direction of the highway pavement image and adjacent pixels to generate an edge-preserving highway pavement image; S23. Perform multi-channel filtering on the edge-preserved road surface image, and construct guiding filter channels {I θ} in different directions respectively, where θ ∈ {0°, 45°, 90°, 135°}. Each directional filtering channel enhances the microcrack linear structure through directional response and generates a set of direction-enhanced road surface images {I dir,θ}; S24. Perform response fusion processing on the multi-channel direction enhanced highway pavement image set {I dir,θ}, and calculate the fused highway pavement image I fused using a weighted superposition strategy; S25. Perform edge detection processing on the fused road surface image, and use the Sobel operator to calculate the gradient response values G x (x, y) and G y (x, y) to obtain the edge response map I edge ; S26. Perform pixel-level fusion processing on the edge response map I edge and the Gaussian-smoothed road surface image I gauss to generate a preprocessed road surface image I with noise suppression and edge enhancement features pre : Among them, θ1 is the noise modulation factor, which is used to enhance the noise suppression effect of the Gaussian smoothed highway pavement image I gauss ; θ2 is the attenuation constant, which is used to regulate the attenuation rate of the exponential function on the Gaussian smoothed highway pavement image I gauss in different brightness regions; θ3 is the edge enhancement scaling factor, which is used to amplify the contribution of the edge response map I edge so as to retain the edge details of microcracks during the fusion process; θ4 is the edge sensitivity adjustment factor, which adjusts the non-linear mapping of the edge response value through the hyperbolic tangent function to distinguish the low-response and high-response regions; μ is the edge threshold parameter, which sets the reference threshold of the edge response. Only when the edge response map I edge (x, y) exceeds the edge threshold parameter, the edge information is enhanced. exp(·) and tanh(·) represent the exponential function and the hyperbolic tangent function respectively.

5. A method for detecting microcracks on a highway pavement based on an improved swarm algorithm according to claim 1, characterized in that, The said S3 includes the following steps: S31. Construct a two-stage bitter fish optimization algorithm model including a global search stage and a local refinement search stage, and initialize the bitter fish optimization population Bitter fish optimization individual X i Represents a set of preprocessing parameters for road pavement image segmentation, including the binary threshold T for the preliminary segmentation of the microcrack region i , the filter kernel size K i and the feature enhancement weight parameter λ i , where N is the size of the bitter fish optimization population; S32. For the detection characteristics of micro-crack area segmentation on highway pavement, based on the bitter fish optimization of individual X i Construct the fitness function F(X i ): Among them, is the between-class variance between the microcrack region and the background region in the preprocessed highway pavement image under the action of the preprocessed highway pavement image segmentation parameter X i , which measures the segmentation discrimination of the microcrack region, and E edge (X i ) is the edge saliency of the preprocessed highway pavement image under the action of the preprocessed highway pavement image segmentation parameter X i , which is used to evaluate the clarity of the microcrack edge, and R conn (X i ) is the connectivity index of the microcrack region in the preprocessed highway pavement image under the action of the preprocessed highway pavement image segmentation parameter X i , which is used to evaluate the continuity of the microcrack region. α1, α2, and α3 are the corresponding weight coefficients respectively; S33. In the global search stage of two-stage bitter fish optimization, according to the fitness function F(X i ), calculate the current optimal bitter fish optimization individual position and the average position of the bitter fish optimization population and introduce the crack sensitivity factor Among them, is the image edge saliency corresponding to the current optimal bitter fish optimization individual, is the average value of the edge saliencies of all bitter fish optimization individuals in the current bitter fish optimization population, and β is the crack sensitivity adjustment coefficient; S34. Update the optimized individual position of bitter fish by using the crack sensitivity factor Update the optimized individual position of bitter fish as follows: Among them, respectively represent the positions of the bitter fish optimized individuals at the t-th and (t + 1)-th iterations, where r2 and r1 are random factors; S35. Set the crack structure convergence threshold for the global search stage, and calculate the fitness value of the current optimal bitter fish optimization individual according to the fitness function. When the fitness value of the current optimal bitter fish optimization individual is greater than or equal to the crack structure convergence threshold, terminate the iteration of the global search stage and obtain the coarse-grained image segmentation parameter X for the preliminary segmentation of microcracks on the highway pavement global , otherwise, return to step S33 to continue the global search process.

6. The method for detecting microcracks on a highway pavement based on an improved swarm algorithm according to claim 5, characterized in that, The said S4 includes the following steps: S41. In the local refinement search stage of the two-stage bitter fish optimization algorithm model, construct a local bitter fish optimization individual set where M << N, and each local bitter fish optimization individual X′ i is generated by adding a perturbation vector Δ to the coarse-grained image segmentation parameters i as follows: Among them, represents a multi-dimensional Gaussian perturbation distribution with a mean of 0 and a covariance matrix of Σ loc for fine-grained sampling of the local search space; S42. Based on the preprocessed road surface image I pre Construct a microcrack topology graph G = (V, E), where V represents the set of crack candidate points extracted by edge detection, and E represents the set of crack structure edges connected by local gradient direction and distance threshold. Combine the topological structure to construct a crack region consistency function C(X′ i , G), which is used to evaluate the matching degree between the current microcrack segmentation result and the true microcrack topology structure: where, R(X′ i ) represents the microcrack segmentation region obtained under the parameter X′ i , θ uv is the direction consistency angle of the crack structure edge (u, v), and II is an indicator function used to determine whether the endpoints of the crack structure edge are both in the candidate region; S43. Evaluate the locally optimized bitter fish individuals by constructing a composite fitness function based on the crack region consistency function of the topological structure: F local (X′ i ) = γ1·F(X′ i ) + γ2·C(X′ i , G); Among them, F(X′ i ) is the fitness function for optimizing individuals by applying local bitter fish, and γ1 and γ2 are the weight coefficients for the global search stage and the local refinement search stage; S44. Optimize the individual X′ for the local bitter fish i Apply the bitter fish optimization update strategy, and set the update direction of the local search to the current local optimal bitter fish optimization individual X′ best The vector difference with the current position of the bitter fish optimization individual X′ i is introduced, and a dynamic step factor is introduced to finely control the search process. The dynamic step factor is based on the composite fitness function F of the local bitter fish optimization individual local (X′ i ) and the preset target composite fitness function F target are dynamically attenuated and adjusted, and the step size gradually decreases as the individual fitness improves; S45. When the fitness value of the current locally optimal bitter fish optimization individual meets the termination condition and is greater than or equal to the local crack structure convergence threshold, output the final fine image segmentation parameter X final ={T final ,K final ,λ final} and complete the local search of the two-stage bitter fish optimization.

7. A method for detecting microcracks on a highway pavement based on an improved swarm algorithm according to claim 6, characterized in that The said S5 includes the following steps: S51. Based on the preprocessed road surface image I pre and the final fine image segmentation parameter X final , perform the preliminary image segmentation operation of the road surface microcracks to generate the preliminary segmentation image I seg : Among them, I seg (x, y) represents the pixel value at the coordinate (x, y) after preliminary image segmentation, and T final is the optimal binary threshold in the final fine image segmentation parameters, which is used to distinguish the microcrack region from the non-crack region; S52. Based on the filter kernel size K in the final fine image segmentation parameters final Perform morphological closing on the preliminary segmented image I seg to generate a preliminary microcrack candidate region map I with enhanced connectivity crack : Among them, and respectively represent morphological dilation and erosion operations, and the size of the structural element S Kfinal is determined by the final fine image segmentation parameter K final to smooth the interference and fracture regions in the microcrack candidate region map; S53. Based on the feature enhancement weight parameter λ in the final fine image segmentation parameters final Perform a microcrack region feature enhancement operation on the preliminary microcrack candidate region map I crack to obtain the enhanced preliminary microcrack candidate region map I 增强 : I 增强 (x,y) = I crack (x,y)·[1 + λ final ·I edge (x,y)]; Among them, λ final is a weight coefficient used to enhance the edge features of the microcrack region.

8. A method for detecting microcracks on a highway pavement based on an improved swarm algorithm according to claim 7, characterized in that, The said S6 includes the following steps: S61. Input the enhanced preliminary microcrack candidate region map I 增强 into the multi-scale dilated convolution sub-module. In the multi-scale dilated convolution sub-module, set the dilation rate set D = {d1, d2, d3, d4}, and construct four groups of dilated convolution branches in parallel. Each group of branches uses a dilated convolution kernel with the corresponding dilation rate d i Perform dilated convolution operations on the input image to generate dilated feature maps ​ Among them, represents the hole feature response at the position (x, y) under the porosity d i ; S62. Input all the hole feature maps into the channel attention sub-module, perform attention weight allocation on the multi-scale channel features extracted from different convolutional branches, and obtain the channel weighted response map Among them, Pool(·) represents the combined operation of global average pooling and max pooling, and MLP(·) is the inter-channel weight learning network. represents the dilation rate of d i feature map response weights; S63. The channel weighted response map is input into the feature fusion decoding sub-module for channel dimension splicing to obtain the fused feature map F fused . Then, the fused feature map F fused is input into the upsampling decoding network, and deconvolution operations are used for layer-by-layer upsampling and spatial feature restoration to generate the depth feature map F with the same size as the input image deep .

9. The method for detecting microcracks on a highway pavement based on an improved swarm algorithm according to claim 1, characterized in that, The said S7 includes the following steps: S71. Input the depth feature map F deep into the adaptive threshold module to automatically generate a spatially dynamic threshold map T adapt based on local statistical features, and perform a binarization operation in combination with the pixel response values to generate a preliminary segmentation result map I bin ; S72. For the preliminary segmentation result diagram I bin Perform morphological reconstruction operations. Utilize the thin linear and connectivity characteristics of microcracks, and adopt the structural element S rec Perform morphological closing operations and thinning processing to generate a microcrack region diagram I with continuous structure recon : Among them, are dilation and erosion operations respectively, Thin(·) represents the thinning operation, and the size of the structure element S rec is set according to the width range of the target crack; S73. The microcrack region map I after structural reconstruction recon Perform connected component filtering and area constraint processing to remove isolated small regions with an area smaller than the set threshold A min to form the final microcrack region segmentation map I final .

10. A method for detecting microcracks on a highway pavement based on an improved swarm algorithm according to claim 9, characterized in that, The said S8 includes the following steps: Segmentation of the final microcrack region, Figure I final Perform a connected component extraction operation to obtain a set of boundary contours of all independent microcrack candidate regions And calculate the set of morphological parameters of each candidate region respectively Among them, A k Is the area of the k-th region, Is the aspect ratio, Is the average response value of the internal depth feature of the region; S82. Combine the final fine image segmentation parameters to construct a dynamic micro-crack region classification rule and set the following thresholds: Area dynamic threshold where α1 is the area adjustment coefficient; Low threshold T of response intensity low = α2·T final , high threshold T high = α3·λ final ·T final , where α2 and α3 are response adjustment coefficients respectively; S83. According to the performance characteristics of the micro-crack morphology in actual highway detection, divide the detection results into three categories: suspected micro-crack regions, obvious micro-crack regions, and non-crack regions: The classification rule for non-crack regions is that any of the following conditions is satisfied: A k <A thresh , or and The classification rules for the suspected microcrack area are that all of the following conditions are met: A k ≥A thresh , and The classification rules for obvious microcrack regions are that all of the following conditions are satisfied: A k ≥A thresh , and S84. Extract the position information of the classified crack candidate regions and output them in a structured manner. Extract the boundary center point coordinates, circumscribed rectangle bounding box, and region number, classification label, and morphological parameters of each obvious micro-crack region to form the final highway pavement micro-crack detection result set.

Images