Concrete chiseling roughness detection method based on improved bitter fish optimization algorithm

Abstract

The invention discloses a concrete chiseling roughness detection method based on an improved bitter fish optimization algorithm. The method comprises the following steps: S1, acquiring original image data covering a detection area; s2, performing multi-scale grid segmentation on the original image data; s3, image preprocessing is carried out on the multiple pieces of grid segmentation image data; s4, forming a concrete surface roughness feature vector; s5, obtaining an optimal roughness parameter in each grid region; s6, performing global fusion processing on the optimal roughness parameters in each grid region, and integrating multi-scale detection results by adopting a unified fitting model to generate an overall concrete chiseling roughness detection result; and S7, generating a detection report based on the overall concrete chiseling roughness detection result. According to the method, roughness parameters can be finely adjusted, local optimal traps are avoided, and the optimal solution can be more stably converged when the complex heterogeneous concrete chiseled surface is processed.

Description

Technical Field

[0001] The present invention relates to the technical field of concrete roughening roughness detection, and particularly relates to a concrete roughening roughness detection method based on an improved bitter fish optimization algorithm. Background Art

[0002] With the development of building construction and infrastructure construction, the concrete roughening technology, as an important means to improve the bonding strength between new and old concretes, is widely used in the construction processes of bridges, tunnels, subways, and large building structures. The main purpose of concrete roughening is to remove the smooth layer on the concrete surface by mechanical or manual means to form rough textures with a certain depth and uniformity, so as to enhance the interfacial shear strength.

[0003] Currently, the detection of concrete roughening roughness mainly relies on manual visual inspection or traditional contact measurement methods. Manual visual inspection is usually carried out by experienced inspectors to judge the roughness level based on vision. However, this method is highly subjective and is greatly affected by environmental lighting and the experience of inspectors, making it difficult to ensure the consistency and repeatability of the detection results. Although traditional contact measurement methods such as the pin gauge method and the profilometer measurement method can provide quantitative indicators, these methods have many limitations in practical applications, including complex operation, low detection efficiency, and the measurement accuracy being affected by the probe contact area and pressure. In addition, the contact measurement method is difficult to comprehensively cover the microscopic features of the concrete roughening surface, resulting in local deviations in the measurement data and making it difficult to reflect the overall roughness situation.

[0004] In recent years, the development of image processing and computational intelligence technologies has provided new ideas for non-contact concrete surface detection. Some studies have proposed using high-precision image acquisition devices combined with texture analysis methods for roughness detection. However, the existing image analysis methods still have the following deficiencies:

[0005] Insufficient accuracy in image data feature extraction: Traditional image processing methods are easily interfered by light changes and surface noise when dealing with complex textures, resulting in a decrease in feature extraction accuracy.

[0006] Optimization algorithms are prone to falling into local optima: When facing the heterogeneous and multi-scale changing texture features of the concrete surface, the existing roughness optimization algorithms are easily trapped in local optima, leading to a decrease in detection accuracy.

[0007] Lack of fusion of multi-scale texture features: The scale span of the textures on the concrete roughening surface is relatively large. Existing methods often have difficulty taking into account both local and global features simultaneously, resulting in unstable final roughness evaluation results.

[0008] In summary, the existing technologies have problems such as strong subjectivity, low measurement efficiency, insufficient algorithm optimization ability, and inadequate multi-scale feature fusion in the detection of the roughness of concrete roughening, making it difficult to meet the requirements of high-precision, standardized, and intelligent detection. Summary of the Invention

[0009] An object of the present invention is to propose a method for detecting the roughness of concrete roughening based on an improved bitter fish optimization algorithm. The present invention can finely adjust the roughness parameters, avoid local optimal traps, and can converge to the optimal solution more stably when dealing with complex non-homogeneous concrete roughened surfaces.

[0010] A method for detecting the roughness of concrete roughening based on an improved bitter fish optimization algorithm according to an embodiment of the present invention includes the following steps:

[0011] S1. Use an image acquisition device to collect images of the surface after concrete roughening to obtain the original image data covering the detection area;

[0012] S2. Perform multi-scale grid segmentation on the original image data, and divide the original image data into several grid segmentation image data of different scales according to preset scale parameters;

[0013] S3. Perform image preprocessing on multiple grid segmentation image data respectively to generate optimized preprocessed grid segmentation image data;

[0014] S4. Extract the texture features of the concrete roughened surface from the optimized preprocessed grid segmentation image data to form a concrete surface roughness feature vector;

[0015] S5. Input the concrete surface roughness feature vector into the improved bitter fish optimization algorithm, and optimize the roughness features in each grid area according to the dynamic weight adjustment mechanism and the adaptive iteration strategy to obtain the optimal roughness parameters in each grid area;

[0016] S6. Perform global fusion processing on the optimal roughness parameters in each grid area, and use a unified fitting model to integrate the multi-scale detection results to generate an overall detection result of the roughness of concrete roughening;

[0017] S7. Generate a detection report based on the overall detection result of the roughness of concrete roughening, including roughness numerical evaluation, distribution diagram, and abnormal area prompt.

[0018] Optionally, the S2 includes the following steps:

[0019] S21. Set the grid segmentation scale parameters of the concrete roughened surface, and define different scale levels L = {l 1 , l 2,...,l i}, where l i represents the grid segmentation scale of the i-th layer, and i is the number of set scale levels;

[0020] S22. Divide the original image data I into grid regions of different scale levels according to the grid segmentation scale parameter, which represents the image data after segmentation at the grid segmentation scale l i The size of the grid region is determined by the following relationship:

[0021]

[0022] where W and H are the width and height of the original image data I respectively, and represent the number of horizontal and vertical grid divisions at the grid segmentation scale l i respectively;

[0023] S23. For each grid region at the grid segmentation scale l i set the grid region overlap rate and obtain the final segmentation result of the grid region

[0024]

[0025] where, is the preset overlap ratio;

[0026] S24. Index and number the grid regions at all grid segmentation scales l i to form a multi-scale grid segmentation image data set

[0027] Optionally, the S3 includes the following steps:

[0028] S31. Perform noise filtering on the multi-scale grid segmentation image data G, process the image data of each grid region, and filter out the random noise caused by light changes, equipment noise, and environmental interference to obtain the processed grid segmentation image data;

[0029] S32. Enhance the contrast of the grid segmentation image data after noise filtering, optimize the features of different roughness regions on the concrete surface, and obtain the adjusted grid segmentation image data;

[0030] S33. Perform edge detection on the grid segmentation image data after contrast enhancement, extract the key boundaries of the concrete roughened surface texture features, and obtain the extracted edge grid segmentation image data; ​​​

[0031] S34. Edge thinning is performed on the edge grid segmented image data to remove redundant edge information, and combined with double-threshold filtering to enhance the effective edge features, and finally the optimized preprocessed grid segmented image data is obtained.

[0032] S35. For the optimized preprocessed grid segmented image data at all grid segmentation scales Data storage is performed to form an optimized preprocessed image data set:

[0033]

[0034] Optionally, the S4 includes the following steps:

[0035] S41. According to the optimized preprocessed grid segmented image data set G final For the grid segmented image data at each grid segmentation scale Calculate the gray-level co-occurrence matrix

[0036]

[0037] where x and y are gray levels, P and Q are the number of rows and columns of the grid segmented image data respectively, Δp and Δq are relative displacements, and δ is an indicator function used to statistically calculate the joint probability distribution of adjacent pixel gray values;

[0038] S42. According to the gray-level co-occurrence matrix Calculate the contrast entropy correlation and energy

[0039]

[0040]

[0041]

[0042]

[0043] where μ x , μ y are the means of gray levels x and y respectively, and σ x , σ y are the standard deviations respectively;

[0044] S43. Calculate the local binary pattern feature vector

[0045]

[0046] Among them, (p k , q k ) are the neighboring pixels of the pixel point (p, q), K is the number of neighboring pixels, and s(x) is a binary mapping function;

[0047] S44. Vectorize the calculation results of contrast, entropy, correlation, energy, and local binary pattern at all grid segmentation scales, and construct a set of concrete surface roughness feature vectors:

[0048]

[0049] Optionally, the S5 includes the following steps:

[0050] S51. Input the set of concrete surface roughness feature vectors V into the improved bitter fish optimization algorithm. For different grid regions of the concrete roughened surface, calculate the roughness contrast coefficient of each grid according to the information contrast:

[0051]

[0052] Among them, ∈ 0 is a constant to prevent the denominator from being zero;

[0053] Using the roughness contrast coefficient as the weighting factor to generate the initial roughness estimate of each individual S 0 in the initial population P j (0):

[0054]

[0055] Among them, β and γ are preset calibration coefficients, rand(0,1) is a random number obeying a uniform distribution, and j = 1, 2,..., N represents the number of candidate solutions in the population;

[0056] S52. Set the roughness difference factor to describe the roughness change between grid regions and reflect the difference in local roughness of the concrete surface:

[0057]

[0058] Among them, N adj (l i ) is the set of grids adjacent to the l i scale grid region, is the comprehensive roughness index at the l i scale:

[0059]

[0060] Among them, ω 1, ω 2 , ω 3 , ω 4 , ω 5 is the preset weight coefficient;

[0061] S53. Introduce a dynamic weight adjustment mechanism to incorporate the roughness difference factor into the weight update of the bitter fish and the prey, and set the bitter fish influence factor W S (t) and the prey influence factor W P (t):

[0062]

[0063] W P (t) = 1 - W S (t);

[0064] Among them, and are respectively the maximum and minimum values of the bitter fish influence factor, T is the maximum number of iterations, and λ is the amplification coefficient of the roughness difference influence;

[0065] S54. In each iteration, for each candidate solution in the population adopt the improved update rule:

[0066]

[0067] Among them, S best (t) is the bitter fish individual with the smallest detection error based on the concrete surface roughness in the current population, P best (t) is the current best prey individual, r 1 , r 2 are uniformly distributed random numbers, and α is the roughness gradient adjustment coefficient;

[0068] S55. Define a fitness function to measure the deviation between the candidate solution and the actual concrete surface roughness measurement value. The fitness function is:

[0069]

[0070] Among them, is the actual value of the concrete roughened surface roughness at the l i th scale obtained from experiments or on-site detections;

[0071] S56. When the fitness function meets the preset convergence threshold ∈ or reaches the maximum number of iterations T, determine the optimal roughness parameter at the l i th scale is the current optimal candidate solution for fitness, and the set of optimal roughness parameters for all grid regions is expressed as:

[0072]

[0073] Among them, each parameter reflects the local difference and overall characteristics of the concrete roughened surface texture.

[0074] Optionally, S6 includes the following steps:

[0075] S61. Calculate the multi-scale roughness weight coefficient according to the set of optimal roughness parameters in each grid region The multi-scale roughness weight coefficient is used to reflect the contribution of grid regions at different scales to the overall roughness:

[0076]

[0077] Among them, is the comprehensive roughness index at the l i scale, is the mean value of the comprehensive roughness of all grid scales, and κ is the local roughness change influence factor, which is used to amplify or reduce the influence of the grid region on the overall roughness;

[0078] The multi-scale roughness weight coefficient makes the regions with roughness changes greater than the threshold obtain higher weights in the fusion process by comparing the differences between the comprehensive roughness indexes of each grid region and the global mean;

[0079] S62. Weighted fusion of the optimal roughness parameters of each grid region according to the multi-scale roughness weight coefficient W li to make the overall roughness evaluation value reflect the true roughness distribution of the concrete surface by introducing an additional local deviation correction term during the fusion process, and generate the fused roughness parameter P : fusion :

[0080]

[0081] Among them, is the mean value of the optimal roughness parameters of all grid regions, and λ 1 is the roughness deviation adjustment coefficient, which is used to correct the influence of local abnormal points on the overall evaluation during the fusion process;

[0082] S63. Calculate the fused roughness error E fusion , which is used to evaluate the rationality of multi-scale fusion:

[0083]

[0084] Among them, the first item is used to measure the deviation between the local optimal roughness parameter and the fusion parameter, and the second item is used to evaluate the deviation degree of the comprehensive roughness index, so that the fusion result can characterize the roughness characteristics of the concrete roughened surface; if the fusion roughness error meets the preset fusion error threshold ∈ f , the fusion result is valid, otherwise the weight coefficient W li is adjusted and the fusion roughness parameter P fusion is recalculated;

[0085] S64. The nonlinear fitting model F fit is used to fit and optimize the fusion roughness parameter P fusion . The fitting model enables the fusion roughness parameter to conform to the actual concrete surface roughness distribution law through an adaptive optimization process:

[0086]

[0087] Among them, P final is the overall concrete roughening roughness detection output finally.

[0088] The beneficial effects of the present invention are as follows:

[0089] (1) The present invention adopts a multi-scale grid segmentation strategy to perform multi-level regional division on the concrete roughened surface, and extracts the texture features of the concrete surface by segmenting the image data of different scales of grids, which can capture the microscopic and macroscopic features of the concrete surface roughness at different scales, and through an adaptive fusion strategy, the optimal roughness parameters of each grid area are weighted and integrated, so that the overall roughness detection result can not only reflect the detail features but also maintain global consistency.

[0090] (2) The present invention improves the bitter fish optimization algorithm. On the basis of the traditional bitter fish optimization algorithm, a roughness comparison coefficient is introduced, and the comprehensive roughness difference of different grid areas is used as the optimization weight to improve the adaptability of the algorithm to complex texture features. In addition, a dynamic weight adjustment mechanism is adopted to dynamically adjust the influence factors of the bitter fish and the prey according to the roughness gradient during the optimization process, ensuring that the algorithm has a strong global search ability in the initial stage and can finely adjust the roughness parameters in the later stage to avoid local optimal traps. When dealing with complex heterogeneous concrete roughened surfaces, it can converge to the optimal solution more stably.

[0091] (3) The present invention adopts a nonlinear fitting model in the multi-scale roughness fusion process, and uses an adaptive weight adjustment mechanism to perform secondary optimization on the fused roughness parameters to reduce the influence of local outliers on the overall detection result. The nonlinear fitting strategy can automatically adjust the parameters according to the change trend of the concrete surface texture, making the final detection result more conform to the true roughness distribution. Description of the Drawings

[0092] The accompanying 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 accompanying drawings:

[0093] Figure 1 It is a flowchart of a method for detecting the roughness of concrete roughening based on an improved bitter fish optimization algorithm proposed by the present invention. Specific embodiments

[0094] Now, the present invention will be further described in detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic way, so they only show the components related to the present invention.

[0095] Reference Figure 1 , a method for detecting the roughness of concrete roughening based on an improved bitter fish optimization algorithm, includes the following steps:

[0096] S1. Use an image acquisition device to acquire images of the surface of the concrete after roughening, and obtain the original image data covering the detection area;

[0097] S2. Perform multi-scale grid segmentation on the original image data, and divide the original image data into several grid segmentation image data of different scales according to the preset scale parameters;

[0098] S3. Perform image preprocessing on multiple grid segmentation image data respectively to generate optimized preprocessed grid segmentation image data;

[0099] S4. Extract the texture features of the concrete roughened surface from the optimized preprocessed grid segmentation image data to form a concrete surface roughness feature vector;

[0100] S5. Input the concrete surface roughness feature vector into the improved bitter fish optimization algorithm, and optimize the roughness features in each grid area according to the dynamic weight adjustment mechanism and the adaptive iteration strategy to obtain the optimal roughness parameters in each grid area;

[0101] S6. Perform global fusion processing on the optimal roughness parameters in each grid area, and use a unified fitting model to integrate the multi-scale detection results to generate an overall detection result of the concrete roughening roughness;

[0102] S7. Generate a detection report based on the overall detection result of the concrete roughening roughness, including roughness numerical evaluation, distribution diagram, and abnormal area prompt.

[0103] In this embodiment, S2 includes the following steps:

[0104] S21. Set the grid segmentation scale parameter for the concrete roughened surface. According to the size range and surface roughness characteristics of the detection area, define different scale levels L = {l 1 , l 2 ,..., l i}, where l i represents the grid segmentation scale of the i-th layer, and i is the set number of scale levels;

[0105] S22. Divide the original image data I into grid regions of different scale levels according to the grid segmentation scale parameter. represents the image data after segmentation at the grid segmentation scale l i . The size of the grid region is determined by the following relationship:

[0106]

[0107] where W and H are the width and height of the original image data I respectively. and respectively represent the number of horizontal and vertical grid divisions at the grid segmentation scale l i ;

[0108] S23. For each grid region i at the grid segmentation scale l , set the grid region overlap rate and obtain the final segmentation result of the grid region

[0109]

[0110] where, is the preset overlap ratio;

[0111] S24. Index and number the grid regions i at all grid segmentation scales l to form a multi-scale grid segmentation image data set

[0112] In this embodiment, S3 includes the following steps:

[0113] S31. Perform noise filtering on the multi-scale grid segmentation image data G, process the image data of each grid region, and filter out the random noise caused by light changes, equipment noise, and environmental interference to obtain the processed grid segmentation image data;

[0114] S32. Enhance the contrast of the grid segmentation image data after noise filtering, optimize the characteristics of different roughness regions on the concrete surface, and obtain the adjusted grid segmentation image data;

[0115] S33. Edge detect the grid-segmented image data after contrast enhancement, extract the key boundaries of the surface texture features of the chiseled concrete surface, and obtain the edge grid-segmented image data after extraction;

[0116] S34. Refine the edges of the edge grid-segmented image data, remove redundant edge information, and combine double-threshold filtering to enhance the effective edge features, finally obtaining the optimized preprocessed grid-segmented image data

[0117] S35. For the optimized preprocessed grid-segmented image data at all grid segmentation scales Perform data storage to form an optimized preprocessed image data set:

[0118]

[0119] In this embodiment, S4 includes the following steps:

[0120] S41. Based on the optimized preprocessed grid-segmented image data set G final For the grid-segmented image data at each grid segmentation scale Calculate the gray-level co-occurrence matrix

[0121]

[0122] where x and y are gray levels, P and Q are the number of rows and columns of the grid-segmented image data respectively, Δp and Δq are relative displacements, and δ is an indicator function used to statistically analyze the joint probability distribution of adjacent pixel gray values;

[0123] S42. Based on the gray-level co-occurrence matrix Calculate the contrast entropy correlation and energy

[0124]

[0125]

[0126]

[0127]

[0128] where μ x , μ y are the means of gray levels x and y respectively, and σ x , σ y are the standard deviations respectively;

[0129] S43. Calculate local binary pattern feature vector

[0130]

[0131] Among them, (p k ,q k ) is the neighborhood pixel of the pixel point (p, q), K is the number of neighborhood pixels, and s(x) is the binary mapping function;

[0132] S44. The contrast, entropy, correlation, energy and local binary pattern calculation results at all grid segmentation scales are vectorized to construct a set of concrete surface roughness feature vectors:

[0133]

[0134] In this implementation, S5 includes the following steps:

[0135] S51. Input the concrete surface roughness feature vector set V into the improved bitter fish optimization algorithm, and calculate the roughness contrast coefficient of each grid based on the information contrast for different grid areas on the concrete roughened surface:

[0136]

[0137] Among them, ∈ 0 To prevent the denominator from being zero;

[0138] Roughness Contrast Coefficient As a weighting factor to generate the initial population P 0 Each individual S j Initial roughness estimate of (0):

[0139]

[0140] Among them, β and γ are preset calibration coefficients, rand(0,1) is a random number that obeys uniform distribution, and j=1,2,…,N represents the number of candidate solutions in the population;

[0141] S52. Setting the roughness difference factor Describes the roughness variation between mesh areas, which is used to reflect the difference in local roughness of the concrete surface:

[0142]

[0143] Among them, N adj (l i ) is the same as the first i The set of grids adjacent to the scale grid area, For the first iComprehensive roughness index at a certain scale:

[0144]

[0145] Among them, ω 1 , ω 2 , ω 3 , ω 4 , ω 5 are preset weight coefficients;

[0146] S53. Introduce a dynamic weight adjustment mechanism to incorporate the roughness difference factor ΔR li into the weight update of the bitter fish and the prey, and set the bitter fish influence factor W S (t) and the prey influence factor W P (t):

[0147]

[0148] W P (t) = 1 - W S (t);

[0149] Among them, and are respectively the maximum and minimum values of the bitter fish influence factor, T is the maximum number of iterations, and λ is the amplification coefficient of the roughness difference influence;

[0150] S54. In each iteration, for each candidate solution in the population adopt an improved update rule:

[0151]

[0152] Among them, S best (t) is the bitter fish individual with the smallest detection error based on the concrete surface roughness in the current population, P best (t) is the current best prey individual, r 1 , r 2 are uniformly distributed random numbers, and α is the roughness gradient adjustment coefficient;

[0153] S55. Define a fitness function to measure the deviation between the candidate solution and the actual measured value of the concrete surface roughness. The fitness function is:

[0154]

[0155] Among them, is the actual value of the roughness of the chiseled concrete surface at the l i scale obtained from experiments or on-site detections;

[0156] S56. When the fitness function satisfies the preset convergence threshold ∈ or reaches the maximum number of iterations T, determine the optimal roughness parameter i at the l-th scale as the current fitness optimal candidate solution, and represent the set of optimal roughness parameters for all grid regions as:

[0157]

[0158] where each parameter reflects the local difference and overall characteristics of the surface texture of the concrete roughening.

[0159] In this embodiment, S6 includes the following steps:

[0160] S61. Calculate the multi-scale roughness weight coefficient W li based on the set of optimal roughness parameters in each grid region. The multi-scale roughness weight coefficient is used to reflect the contribution of grid regions at different scales to the overall roughness:

[0161]

[0162] where is the comprehensive roughness index at the l-th i scale, is the average value of the comprehensive roughness of all grid scales, and κ is the local roughness change influence factor, which is used to amplify or reduce the influence of the grid region on the overall roughness;

[0163] The multi-scale roughness weight coefficient enables the regions with roughness changes greater than the threshold to obtain higher weights in the fusion process by comparing the differences between the comprehensive roughness indices of each grid region and the global mean;

[0164] S62. Weightedly fuse the optimal roughness parameters of each grid region according to the multi-scale roughness weight coefficient . By additionally introducing a local deviation correction term in the fusion process, make the overall roughness evaluation value reflect the true roughness distribution of the concrete surface, and generate the fused roughness parameter P fusion :

[0165]

[0166] where is the average value of the optimal roughness parameters of all grid regions, and λ 1 is the roughness deviation adjustment coefficient, which is used to correct the influence of local abnormal points on the overall evaluation in the fusion process;

[0167] S63. Calculate the fused roughness error E fusion , which is used to evaluate the rationality of the multi-scale fusion:

[0168]

[0169] Among them, the first item is used to measure the deviation between the local optimal roughness parameter and the fusion parameter, and the second item is used to evaluate the deviation degree of the comprehensive roughness index, so that the fusion result characterizes the roughness characteristics of the concrete roughened surface; if the fusion roughness error satisfies the preset fusion error threshold ∈ f , the fusion result is valid, otherwise the weight coefficient is adjusted and the fusion roughness parameter P is recalculated fusion ;

[0170] S64. The non-linear fitting model F fit is used to fit and optimize the fusion roughness parameter P fusion . The fitting model enables the fusion roughness parameter to conform to the actual concrete surface roughness distribution law through an adaptive optimization process:

[0171]

[0172] Among them, P final is the overall concrete roughening roughness detection of the final output.

[0173] Example 1:

[0174] On October 15, 2024, at the construction site of a large bridge, the construction team carried out roughness detection on the surface of the concrete pier after roughening treatment. The goal was to evaluate the roughening quality and ensure that the bonding force between the new and old concrete reached the design requirements. The construction team started the detection operation at 8:00 am, the on-site temperature was 22°C, and the humidity was 55%. The pier surface was treated with a standardized roughening machine, and the roughening depth was set within the range of 5 mm to 8 mm.

[0175] To ensure the scientific nature of the detection, the construction party selected 50 roughened concrete samples from different construction areas. Before the formal detection, the construction party used the traditional manual visual inspection method for preliminary evaluation and found that there were significant differences in the scores given by different inspectors. In the example, for the concrete sample numbered A15, Inspector 1 gave a roughness grade of 3, while Inspector 2 gave a score of 4, and the score difference reached 25%, making it difficult for the construction party to make an accurate quality assessment decision.

[0176] The construction team used a Basler acA2500-60uc industrial camera (resolution 5000×4000 pixels) to collect high-precision images of the roughened surfaces of 50 samples. Five images at different angles were taken for each sample to reduce the influence of light and angle. During the image acquisition process, due to insufficient light conditions, the image brightness of the sample numbered B07 was low. The system automatically adjusted the exposure parameters, and finally clear texture images were obtained.

[0177] The collected image data was input into the detection system. First, noise filtering and edge enhancement processing were performed to remove image artifacts caused by uneven illumination. After processing, a small amount of over-enhancement occurred at the edge of the roughened area of the sample numbered C12. The system detected the anomaly and automatically adjusted the filtering parameters to restore the normal edge detection effect. Subsequently, the system used a multi-scale grid segmentation algorithm to divide each image into 10*10 grid regions to capture texture features at different scales.

[0178] After completing the grid segmentation, the system began to extract key texture features such as gray-level co-occurrence matrix and local binary pattern for each grid region and generate a roughness feature vector. During this process, due to the relatively shallow roughening depth of the surface of the sample numbered D22, there was an obvious deviation between the feature vector and the standard roughened sample. The system automatically marked this sample as "low roughness risk" and prompted the construction party to conduct a review.

[0179] Subsequently, the system used an improved bitter fish optimization algorithm to perform optimization calculations on the feature vector. The optimization process was iterated 30 times, and finally the optimal roughness parameters for each grid region were obtained. In the embodiment, the initially calculated roughness parameter of the sample numbered E35 was 5.12μm, but after optimization, it was finally adjusted to 5.28μm, improving the accuracy of the data.

[0180] After obtaining the optimal roughness parameters for all grid regions, the system used a multi-scale weighted fusion model to calculate the overall roughness evaluation value. In the embodiment, for the sample numbered F19, the initial roughness parameter range of each grid region was between 4.7μm and 5.3μm. After fusion, the final roughness evaluation value was determined to be 5.05μm.

[0181] The system generated a detection report, which included the following content:

[0182] 1. Detection time: October 15, 2024, 10:45;

[0183] 2. Detection location: The construction site of a certain bridge, pier number: B03;

[0184] 3. Detection method: Concrete roughening roughness detection based on the improved bitter fish optimization algorithm;

[0185] 4. Test Results:

[0186] Total number of samples: 50;

[0187] Average roughness value: 5.08μm;

[0188] Highest roughness value: 5.51μm (Sample No.: G41);

[0189] Lowest roughness value: 4.72μm (Sample No.: D22, marked as low roughness risk);

[0190] Roughness qualification rate: 96% (48 samples are qualified, 2 samples need to be rechecked);

[0191] The experimental results show that the method of the present invention not only has higher detection accuracy, but also can maintain a stable detection effect under different lighting, temperature and humidity conditions. At the same time, the method of the present invention can automatically mark abnormal samples, reduce the cost of manual intervention, and improve the intelligent level of construction quality management.

[0192] Finally, the construction team adjusted the construction process according to the test report, carried out secondary chiseling treatment on the pier area where the D22 sample is located, and completed the re-inspection at 12:30. The test results reached the qualified standard. Thus, the roughness detection of the concrete chiseling was successfully completed. The construction party confirmed that the method of the present invention has excellent detection performance and plans to comprehensively promote and apply it in the subsequent bridge construction.

[0193] The present invention adopts a multi-scale grid segmentation strategy to conduct multi-level regional division on the concrete chiseling surface, and extracts the texture features of the concrete surface by segmenting the image data of different scales of grids. It can capture the microscopic and macroscopic features of the concrete surface roughness at different scales, and through an adaptive fusion strategy, weighted integration of the optimal roughness parameters of each grid region is carried out, so that the overall roughness detection result can not only reflect the detailed features, but also maintain global consistency.

[0194] The present invention improves the bitter fish optimization algorithm. On the basis of the traditional bitter fish optimization algorithm, a roughness comparison coefficient is introduced, and the comprehensive roughness difference of different grid regions is used as the optimization weight to improve the adaptability of the algorithm to complex texture features. In addition, a dynamic weight adjustment mechanism is adopted to dynamically adjust the influence factors of the bitter fish and the prey according to the roughness gradient during the optimization process, ensuring that the algorithm has strong global search ability in the initial stage and can finely adjust the roughness parameters in the later stage to avoid local optimal traps. When dealing with complex heterogeneous concrete chiseling surfaces, it can converge to the optimal solution more stably.

[0195] In the process of multi-scale roughness fusion, the present invention adopts a non-linear fitting model, and uses an adaptive weight adjustment mechanism to perform secondary optimization on the fused roughness parameters to reduce the influence of local outliers on the overall detection result. The non-linear fitting strategy can automatically adjust the parameters according to the change trend of the concrete surface texture, making the final detection result more conform to the true roughness distribution.

[0196] The above is only a preferred specific embodiment 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 method for detecting the roughness of concrete chiseling based on an improved bitterfish optimization algorithm, characterized in that: The steps include: S1. Use an image acquisition device to acquire images of the surface of the concrete after roughening to obtain original image data covering the detection area; S2. Perform multi-scale grid segmentation on the original image data, and divide the original image data into grid segmented image data of several different scales according to a preset scale parameter; S3. Perform image preprocessing on multiple grid segmentation image data to generate optimized preprocessed grid segmentation image data; S4. Extracting the texture features of the chiseled concrete surface from the optimized preprocessed grid segmentation image data to form a concrete surface roughness feature vector; S5. Input the concrete surface roughness feature vector into the improved bitter fish optimization algorithm, and optimize the parameters of the roughness characteristics in each grid area according to the dynamic weight adjustment mechanism and the adaptive iteration strategy to obtain the optimal roughness parameters in each grid area; S6. Perform global fusion processing on the optimal roughness parameters in each grid area, integrate the multi-scale detection results using a unified fitting model, and generate the overall concrete chiseling roughness detection results; S7. Generate a test report based on the overall concrete chiseling roughness test results, including roughness numerical evaluation, distribution diagram and abnormal area prompts.

2. The concrete chiseling roughness detection method based on the improved bitter fish optimization algorithm according to claim 1 is characterized in that: The S2 comprises the following steps: S21. Set the mesh segmentation scale parameters of the concrete roughened surface, and define different scale levels L = {l1,l2,...,l i }, where l i Indicates the grid segmentation scale of the i-th layer, where i is the number of scale levels set; S22. Divide the original image data I into grid areas of different scale levels according to the grid segmentation scale parameter, Indicates the grid segmentation scale l i The image data after segmentation, grid area The size of is determined by the following relationship: Where W and H are the width and height of the original image data I, respectively. and Respectively represent the grid segmentation scale l i The number of horizontal and vertical grid divisions under ; S23. For each grid segmentation scale l i The grid area below Set the grid area overlap ratio And get the final segmentation result of the grid area in, is the preset overlap ratio; S24. For all grid segmentation scales l i The grid area below Perform index numbering to form a multi-scale grid segmentation image data set 3. The concrete chiseling roughness detection method based on the improved bitter fish optimization algorithm according to claim 1 is characterized in that: The S3 comprises the following steps: S31. Perform noise filtering on the multi-scale grid segmentation image data G, process the image data of each grid area, filter out random noise caused by illumination changes, equipment noise and environmental interference, and obtain processed grid segmentation image data; S32. Performing contrast enhancement on the mesh segmentation image data after noise filtering, optimizing the features of different roughness areas on the concrete surface, and obtaining adjusted mesh segmentation image data; S33. Perform edge detection on the contrast-enhanced mesh segmentation image data to extract key boundaries of the texture features of the chiseled concrete surface to obtain the extracted edge mesh segmentation image data; S34. Refine the edge of the edge grid segmentation image data, remove redundant edge information, and combine double threshold filtering to enhance the effective edge features, and finally obtain the optimized pre-processed grid segmentation image data S35. Optimize preprocessing of mesh segmentation image data at all mesh segmentation scales Store data to form an optimized pre-processed image data set:

4. The method for detecting the roughness of concrete chiseling based on the improved bitterfish optimization algorithm according to claim 1 is characterized in that: The S4 comprises the following steps: S41. Segmenting the image data set G based on the optimized preprocessed grid final For each grid segmentation scale, the grid segmentation image data Calculate the gray-level co-occurrence matrix Where x, y are gray levels, P and Q are the number of rows and columns of grid segmented image data, Δp and Δq are relative displacements, and δ is an indicator function used to count the joint probability distribution of gray values ​​of adjacent pixels; S42. Based on the gray level co-occurrence matrix Calculate the contrast at each grid segmentation scale entropy Relevance and energy Among them, μ x , μ y are the means of gray levels x and y, σ x , σ y are the standard deviations, respectively; S43. Calculate local binary pattern feature vector Among them, (p k ,q k ) is the neighborhood pixel of the pixel point (p, q), K is the number of neighborhood pixels, and s(x) is the binary mapping function; S44. The contrast, entropy, correlation, energy and local binary pattern calculation results at all grid segmentation scales are vectorized to construct a set of concrete surface roughness feature vectors:

5. The method for detecting the roughness of concrete chiseling based on the improved bitterfish optimization algorithm according to claim 1 is characterized in that: The S5 comprises the following steps: S51. Input the concrete surface roughness feature vector set V into the improved bitter fish optimization algorithm, and calculate the roughness contrast coefficient of each grid based on the information contrast for different grid areas on the concrete roughened surface: Among them, ∈0 is a constant to prevent the denominator from being zero; Roughness Contrast Coefficient As a weighting factor, each individual S in the initial population P0 is generated j Initial roughness estimate of (0): Among them, β and γ are preset calibration coefficients, rand(0,1) is a random number that obeys uniform distribution, and j=1,2,…,N represents the number of candidate solutions in the population; S52. Setting the roughness difference factor Describes the roughness variation between mesh regions, which is used to reflect the difference in local roughness of the concrete surface: Among them, N adj (l i ) is the same as the first i The set of grids adjacent to the scale grid area, For the first i Comprehensive roughness index under scale: Among them, ω1, ω2, ω3, ω4, ω5 are preset weight coefficients; S53. Introduce a dynamic weight adjustment mechanism to adjust the roughness difference factor Integrate it into the weight update of bitter fish and prey, and set the bitter fish influence factor W at the tth iteration S (t) and prey influence factor W P (t): W P (t)=1-W S (t); in, and are the maximum and minimum values ​​of the bitter fish influence factor, T is the maximum number of iterations, and λ is the amplification factor of the roughness difference; S54. In each iteration, for each candidate solution in the population Adopt improved update rule: Among them, S best (t) is the bitter fish individual with the smallest detection error based on the concrete surface roughness in the current population, P best (t) is the current best prey individual, r1, r2 are uniformly distributed random numbers, and α is the roughness gradient adjustment coefficient; S55. Define the fitness function to measure candidate solutions The deviation between the measured value of the actual concrete surface roughness and the fitness function for: in, is the first i The actual value of the roughness of the chiseled concrete surface under the scale; S56. When the fitness function When the preset convergence threshold ∈ is met or the maximum number of iterations T is reached, the lth i Optimal roughness parameters at scale is the optimal candidate solution for the current fitness, and the optimal roughness parameter set of all mesh regions is expressed as: Among them, each parameter reflects the local differences and overall characteristics of the texture of the chiseled concrete surface.

6. The method for detecting the roughness of concrete chiseling based on the improved bitterfish optimization algorithm according to claim 1 is characterized in that: The S6 comprises the following steps: S61. Calculate the multi-scale roughness weight coefficient based on the optimal roughness parameter set in each grid area The multi-scale roughness weight coefficient is used to reflect the contribution of grid areas of different scales to the overall roughness: in, For the first i The comprehensive roughness index under the scale, is the comprehensive roughness mean of all mesh scales, κ is the influencing factor of local roughness change, which is used to amplify or reduce the influence of the mesh area on the overall roughness; The multi-scale roughness weight coefficient compares the difference between the comprehensive roughness index of each grid area and the global mean, so that the area with a roughness change greater than the threshold will obtain a higher weight in the fusion process; S62. Based on multi-scale roughness weight coefficient Optimal roughness parameters for each mesh region Weighted fusion is performed, and the local deviation correction term is additionally introduced in the fusion process to make the overall roughness evaluation value reflect the actual roughness distribution of the concrete surface, and generate the fusion roughness parameter P fusion : in, is the mean value of the optimal roughness parameters of all grid areas, λ1 is the roughness deviation adjustment coefficient, which is used to correct the influence of local outliers on the overall evaluation during the fusion process; S63. Calculate the fusion roughness error E fusion , used to evaluate the rationality of multi-scale fusion: Among them, the first item is used to measure the deviation between the local optimal roughness parameter and the fusion parameter, and the second item is used to evaluate the deviation degree of the comprehensive roughness index, so that the fusion result can represent the roughness characteristics of the concrete chiseled surface; if the fusion roughness error meets the preset fusion error threshold ∈ f , then the fusion result is valid, otherwise adjust the weight coefficient And recalculate the fusion roughness parameter P fusion ; S64. Using nonlinear fitting model F fit For the fusion roughness parameter P fusion Perform fitting optimization. The fitting model makes the fusion roughness parameters conform to the actual concrete surface roughness distribution law through the adaptive optimization process: Among them, P final Overall concrete chiseling roughness test for final output.

Images