Method for detecting and evaluating anticorrosion effect of highway based on image processing technology

Abstract

The present application relates to the technical field of image processing, in particular to a highway anticorrosion effect detection and evaluation method based on image processing technology, which comprises the following steps: obtaining two different types of image corresponding to the detected gray image of the highway anticorrosion layer surface, and then segmenting to obtain the detected sub-image; obtaining two difference images corresponding to the detected sub-image according to the row gray difference and column gray difference of the gray difference image corresponding to the detected sub-image; obtaining the roughness of the detected sub-image, obtaining the preferred scale based on the roughness, decomposing the two detected gray images by wavelet transform according to the preferred scale, and then fusing the feature images according to the decomposition results to obtain the fused feature image; obtaining the defect area of the fused feature image, and obtaining the quality detection result of the highway according to the area characteristics of the defect area. The present application can obtain more accurate quality detection result of the highway.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a method for detecting and evaluating the effectiveness of highway corrosion protection based on image processing technology. Background Technology

[0002] With the deepening development and utilization of marine resources, underwater highways, as a key sealing device for underwater facilities, are playing an increasingly prominent role. Underwater highways are mainly used for sealing underwater pipelines or wellheads, effectively preventing pipeline leaks, avoiding environmental pollution, and protecting the safe and stable operation of underwater facilities. Given the important role of underwater highways in marine engineering, their quality inspection has become a crucial link in ensuring facility safety.

[0003] Under current technological conditions, the quality inspection of underwater highways primarily focuses on the corrosion status of their anti-corrosion coatings, as the integrity of these coatings directly affects the highway's corrosion resistance and service life. Therefore, detecting changes in the anti-corrosion coating has become a routine quality inspection item. Existing inspection methods typically rely on images of the highway's anti-corrosion coating surface captured by cameras, analyzing the corroded areas in the images to assess the highway's quality.

[0004] However, this method has certain limitations in practical applications. Due to the complexity of the underwater environment and the limitations of camera imaging, the acquired surface images often contain a significant amount of noise, which can interfere with the accurate extraction of corroded areas. The presence of noise may lead to misjudgment or missed detection of corroded areas, thus affecting the accuracy of the quality inspection results.

[0005] To improve the accuracy of corrosion detection for underwater highway anti-corrosion coatings, a novel detection and evaluation method based on image processing technology needs to be developed. This method should be able to effectively remove image noise and accurately extract corroded areas, thereby achieving a precise assessment of the anti-corrosion effect of underwater highways. This will help to promptly identify and address problems with the anti-corrosion coating, ensuring the safe and stable operation of underwater highways and extending their service life. Summary of the Invention

[0006] To address the technical problem of inaccurate highway quality inspection results, the present invention aims to provide a highway corrosion protection effect detection and evaluation method based on image processing technology. The specific technical solution adopted is as follows:

[0007] Two different types of images of the surface of the highway anti-corrosion layer are obtained to obtain grayscale images to be detected. The grayscale images to be detected are segmented to obtain sub-images to be detected. Based on the row grayscale difference and column grayscale difference of the grayscale difference image corresponding to the sub-image to be detected, two difference images corresponding to the sub-image to be detected are obtained respectively.

[0008] The roughness of the sub-image to be detected is obtained based on the difference in pixel values ​​and texture information between the two difference images corresponding to the sub-image to be detected.

[0009] The optimal scale is obtained by adjusting the preset scale of wavelet transform using the coarsness. Based on the optimal scale, the two grayscale images to be detected are decomposed using wavelet transform respectively, and then the decomposition results are fused to obtain a fused feature image.

[0010] Defect regions are obtained from the fused feature images, and the quality inspection results of the highway are obtained based on the area characteristics of the defect regions.

[0011] Preferably, the step of obtaining two difference images corresponding to the sub-image to be detected based on the row grayscale difference and column grayscale difference of the grayscale difference image corresponding to the sub-image to be detected specifically involves:

[0012] For any sub-image to be detected, calculate the absolute value of the difference between the gray values ​​of corresponding pixels in adjacent rows in the corresponding gray-level difference image to obtain the first difference image of the sub-image to be detected; calculate the absolute value of the difference between the gray values ​​of corresponding pixels in adjacent columns in the corresponding gray-level difference image to obtain the second difference image of the sub-image to be detected.

[0013] The difference image includes a first difference image and a second difference image.

[0014] Preferably, the step of obtaining the coarseness of the sub-image to be detected based on the difference in pixel values ​​and texture information in the two difference images corresponding to the sub-image to be detected specifically involves:

[0015] For any difference image corresponding to the sub-image to be detected, the variance of the pixel values ​​of all pixels is recorded as the coarseness coefficient; the difference image is thresholded to obtain the first feature region, where the pixel values ​​of the pixels in the first feature region are greater than a preset threshold; the overlapping area of ​​the first feature regions of the two difference images is obtained to obtain the second feature region;

[0016] The first coarseness weight and the second coarseness weight are obtained based on the texture information of the second feature region. The coarseness coefficients of the first difference image and the second difference image are weighted and summed using the first coarseness weight and the second coarseness weight to obtain the coarseness degree corresponding to the sub-image to be detected.

[0017] Preferably, obtaining the first coarseness weight and the second coarseness weight based on the texture information of the second feature region specifically involves:

[0018] The second feature region is processed using vertical and horizontal filters respectively to obtain the vertical response results and the horizontal response results.

[0019] For any response result in any direction, the mean and standard deviation of the phase response value are obtained, and a rough weight is obtained based on the mean and standard deviation of the phase response value. The mean is positively correlated with the rough weight, and the standard deviation is negatively correlated with the rough weight. The rough weight corresponding to the response result in the horizontal direction is taken as the first rough weight, and the rough weight corresponding to the response result in the vertical direction is taken as the second rough weight.

[0020] Preferably, the method for obtaining the grayscale difference image corresponding to the sub-image to be detected is as follows:

[0021] For any sub-image to be detected, obtain the center pixel of the sub-image. If the number of center pixels is even, the average pixel value of all center pixels is used as the feature pixel value. If the number of center pixels is one, the pixel value of the center pixel is used as the feature pixel value. Calculate the difference between the pixel value of each pixel in the sub-image to be detected and the feature pixel value to obtain the grayscale difference image.

[0022] Preferably, the preferred scale is obtained by adjusting the preset scale of the wavelet transform using the degree of roughness as follows:

[0023] For any sub-image to be detected, obtain the sum between the preset value and the roughness of the sub-image to be detected, calculate the ratio between the preset scale of the wavelet transform and the sum, and round the ratio to obtain the preferred scale corresponding to the sub-image to be detected.

[0024] Preferably, the acquisition of the grayscale images to be detected corresponding to the two different types of images of the highway anti-corrosion layer surface specifically involves:

[0025] RGB and infrared images of the surface of the highway anti-corrosion layer are acquired respectively. The RGB image is converted to grayscale to obtain a first grayscale image, and the infrared image is converted to grayscale to obtain a second grayscale image. The first grayscale image and the second grayscale image are the grayscale images to be detected.

[0026] Preferably, the step of fusing the decomposition results to obtain the fused feature image specifically involves:

[0027] The decomposition results include the high-frequency coefficients and low-frequency coefficients of the grayscale image to be detected;

[0028] The maximum value of the high-frequency coefficients of the first grayscale image and the second grayscale image is obtained as the reconstructed high-frequency coefficients. The low-frequency coefficient information of the first grayscale image and the second grayscale image is weighted and summed using preset weights to obtain the reconstructed low-frequency coefficients.

[0029] The fused feature image is obtained by using the reconstructed high-frequency coefficients and the reconstructed low-frequency coefficients to perform wavelet decomposition inverse transform.

[0030] Preferably, obtaining the highway quality inspection result based on the area characteristics of the defective area specifically involves:

[0031] The area of ​​each defect region is obtained, and the ratio between the sum of the areas of all defect regions and the total area of ​​the fused feature image is used as the quality coefficient. When the quality coefficient is less than a preset first threshold, the quality inspection result of the corresponding highway anti-corrosion layer is qualified; when the quality coefficient is greater than or equal to the first threshold and less than a preset second threshold, the quality inspection result of the corresponding highway anti-corrosion layer is level one damage; when the quality coefficient is greater than or equal to the second threshold, the quality inspection result of the corresponding highway anti-corrosion layer is level two damage; the first threshold is less than the second threshold.

[0032] The embodiments of the present invention have at least the following beneficial effects:

[0033] This invention first acquires two different types of grayscale images to be detected, enabling subsequent feature fusion based on the decomposition results of the two different image types. This results in a more accurate feature image. The grayscale images to be detected are then uniformly segmented to obtain sub-images to be detected. This allows for analysis of local pixel changes in the sub-images and the acquisition of two difference images for each sub-image, analyzing grayscale differences in two directions separately. Combined with texture information, the roughness of the sub-images to be detected is obtained, leading to more accurate image feature extraction. Then, the roughness is used to adjust the preset scale of the wavelet transform to obtain an optimal scale. This allows for adaptive acquisition of the scale parameter in the wavelet transform during the decomposition of the grayscale images to be detected, resulting in better decomposition results and a fused feature image with improved edge and detail texture. Based on this fused feature image, defect areas are extracted, leading to more accurate quality inspection results for underwater highway anti-corrosion layers. Attached Figure Description

[0034] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart of the highway corrosion protection effect detection and evaluation method based on image processing technology of the present invention. Detailed Implementation

[0036] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the highway corrosion protection effect detection and evaluation method based on image processing technology proposed by the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0038] The following description, in conjunction with the accompanying drawings, details the specific scheme of the highway corrosion protection effect detection and evaluation method based on image processing technology provided by this invention.

[0039] The main objective of this invention is to capture RGB and infrared images of the surface of an underwater highway anti-corrosion layer, optimize the preset scale of wavelet transformation based on the local roughness in the images, and adaptively acquire grayscale images corresponding to two different types of images at different scales for decomposition and fusion. Finally, based on the fused image, regional analysis is performed to obtain the quality inspection results of the highway anti-corrosion layer.

[0040] Example:

[0041] Please see Figure 1 The diagram illustrates a flowchart of a highway corrosion protection effect detection and evaluation method based on image processing technology according to an embodiment of the present invention. The method includes the following steps:

[0042] Step 1: Obtain two grayscale images of the surface of the highway anti-corrosion layer to be detected, and segment the grayscale images to be detected to obtain sub-images to be detected; obtain two difference images corresponding to the sub-images to be detected based on the row grayscale difference and column grayscale difference of the grayscale difference image corresponding to the sub-images to be detected.

[0043] First, RGB and infrared images of the highway anti-corrosion layer are captured using professional underwater camera equipment. Underwater image capture may introduce noise, necessitating denoising. In this embodiment, median filtering is used to denoise the images; this process is well-known and will not be elaborated upon here. The RGB image is then converted to grayscale to obtain a first grayscale image, and the infrared image is converted to grayscale to obtain a second grayscale image; these first and second grayscale images are the grayscale images to be detected. Various methods exist for image grayscale conversion; this embodiment uses the maximum value method. The implementer can choose the appropriate method based on the specific implementation scenario.

[0044] The first grayscale image corresponding to the RGB image of the highway anti-corrosion layer surface has a good effect on the presentation of the texture information and detailed edges of various objects in the image. However, there is environmental noise during image acquisition, such as bubbles and suspended matter, which will have a certain impact on the imaging effect of the RGB image. In contrast, the infrared image of the highway anti-corrosion layer surface has a greater tolerance to environmental noise and can better present the characteristics of the observed subject.

[0045] There are certain differences in the infrared radiation results between the corroded and uncorroded parts of the highway anti-corrosion layer. Therefore, if the highway anti-corrosion layer is corroded, its corrosion characteristics can be observed in the second grayscale image corresponding to the infrared image, and it is less affected by environmental factors. However, the second grayscale image corresponding to the infrared image has poor rendering effect on the texture information and detailed edges of various objects in the image. Therefore, in this embodiment, the feature information in the first grayscale image and the second grayscale image are analyzed separately, and the first grayscale image and the second grayscale image are fused. That is, the parts with better rendering effect are merged to avoid the influence of underwater environmental noise, and at the same time, more accurate and clear detail information can be obtained. In this way, the more accurate corrosion area can be extracted from the fused image.

[0046] In this embodiment, a wavelet transform algorithm is used to perform multi-scale decomposition on two different types of grayscale images of the highway anti-corrosion layer surface, obtaining low-frequency and high-frequency coefficients respectively. Different fusion methods are used for different coefficients, and the fused image is obtained through inverse wavelet decomposition transform. When using traditional wavelet transform to decompose an image, the scale is a fixed value, resulting in poor decomposition results. Therefore, in this embodiment, by analyzing features such as texture information in the grayscale image to be detected, the preset scale in the wavelet decomposition transform algorithm is optimized, enabling the image to be decomposed based on an adaptive scale, thus obtaining better decomposition results.

[0047] Therefore, in order to better analyze the detailed information and texture features in the grayscale image to be detected, it is necessary to segment the grayscale image to be detected. That is, by analyzing the local features in the grayscale image to be detected, the optimal scale corresponding to the wavelet decomposition of the local region in the grayscale image to be detected can be obtained.

[0048] In this embodiment, the grayscale image to be detected is uniformly segmented to obtain sub-images to be detected. That is, each sub-image to be detected is uniformly divided into a preset number of regions of the same size. The image of each region is used as the sub-image to be detected. The value of the preset number needs to be determined by the implementer according to the size of the grayscale image to be detected. For example, if the size of the grayscale image to be detected is 16×16, then uniformly segmenting the grayscale image to be detected can obtain four regions of size 4×4.

[0049] Then, the pixel value changes of the corresponding pixels in the sub-image to be detected are analyzed. That is, two difference images corresponding to the sub-image to be detected are obtained based on the row gray-level difference and column gray-level difference of the gray-level difference image corresponding to the sub-image to be detected. In this embodiment, the first gray-level image corresponding to the RGB image and the second gray-level image corresponding to the infrared image are both uniformly segmented. Then, the gray-level difference image of each sub-image to be detected in the first gray-level image and the second gray-level image also needs to be obtained.

[0050] The specific method for obtaining the grayscale difference image is as follows: For any sub-image to be detected, obtain the center pixel of the sub-image to be detected. If the number of center pixels is even, the average pixel value of all center pixels is used as the feature pixel value. If the number of center pixels is one, the pixel value of the center pixel is used as the feature pixel value. Calculate the difference between the pixel value of each pixel in the sub-image to be detected and the feature pixel value to obtain the grayscale difference image.

[0051] For example, if the size of the sub-image to be detected is 5×5, then the pixel located at the exact center of the sub-image is the center pixel. Since there is only one center pixel, its pixel value can be used as the feature pixel. If the size of the sub-image to be detected is 6×6, then the pixel located at the exact center of the sub-image is the center pixel. Since there are four center pixels, the average of the pixel values ​​of the four center pixels can be used as the feature pixel value.

[0052] Furthermore, the differences in pixel values ​​between each pixel in the sub-image to be detected and the center pixel are analyzed. For example, for the first pixel in the sub-image to be detected, the absolute value of the difference between the pixel value of the first pixel and the feature pixel value is calculated. Then, the absolute value of the difference between each pixel in the sub-image to be detected and the feature pixel is calculated as the pixel value at the corresponding position in the grayscale difference image.

[0053] Gray-scale difference images can characterize the changes of pixels in the corresponding sub-image to be detected relative to the center of the region. If there are no defects in the sub-image to be detected, the gray-scale difference between each pixel in the sub-image to be detected and the pixel at the center of the region is small. If there are defects in the sub-image to be detected, the gray-scale difference changes in the image can be analyzed by obtaining the relative difference between each pixel in the sub-image to be detected and the pixel at the center of the region.

[0054] For any sub-image to be detected, calculate the absolute value of the difference between the gray values ​​of corresponding pixels in adjacent rows in the corresponding gray-level difference image to obtain the first difference image of the sub-image to be detected; calculate the absolute value of the difference between the gray values ​​of corresponding pixels in adjacent columns in the corresponding gray-level difference image to obtain the second difference image of the sub-image to be detected; the difference image includes the first difference image and the second difference image.

[0055] For example, the absolute difference between the gray values ​​of corresponding pixels in the first and second rows of the grayscale difference image is calculated. Then, the absolute difference between the gray values ​​of corresponding pixels in the second and third rows is calculated, and so on, until the difference between the gray values ​​of corresponding pixels in every adjacent pair of rows is calculated, resulting in the first difference image of the sub-image to be detected. Similarly, the absolute difference between the gray values ​​of corresponding pixels in the first and second columns of the grayscale difference image is calculated. Then, the absolute difference between the gray values ​​of corresponding pixels in the second and third columns is calculated, and so on, until the difference between the gray values ​​of corresponding pixels in every adjacent pair of columns is calculated, resulting in the second difference image of the sub-image to be detected.

[0056] The first difference image reflects the horizontal variation of the grayscale difference between each pixel in the sub-image to be detected relative to the center pixel, while the second difference image reflects the vertical variation of the grayscale difference between each pixel in the sub-image to be detected relative to the center pixel.

[0057] Step 2: Obtain the roughness of the sub-image to be detected based on the difference in pixel values ​​and texture information between the two difference images corresponding to the sub-image to be detected.

[0058] Each grayscale image to be detected corresponds to multiple sub-images to be detected, and each sub-image to be detected corresponds to two difference images, including a first difference image and a second difference image. When there are defective parts in the sub-image to be detected, the grayscale difference between each pixel in the sub-image to be detected and the center pixel may change significantly. Therefore, the feature region with large grayscale difference can be obtained based on the change of pixel values ​​in the difference image.

[0059] Based on this, for any difference image corresponding to the sub-image to be detected, the difference image is thresholded to obtain a first feature region, where the pixel value of the pixel in the first feature region is greater than a preset threshold; the overlapping area of ​​the first feature regions of the two difference images is obtained to obtain a second feature region.

[0060] In this embodiment, the first difference image is used as an example. The segmentation threshold in the first difference image is obtained using the maximum inter-class variance method as a preset threshold. The region formed by pixels with pixel values ​​greater than the preset threshold in the first difference image is taken as the first feature region in the first difference image. The first feature region corresponding to the second difference image can be obtained in the same way. The first feature region in the first difference image represents the part of the gray-level difference between each pixel in the sub-image to be detected and the center pixel that changes significantly in the horizontal direction. The first feature region corresponding to the second difference image represents the part of the gray-level difference between each pixel in the sub-image to be detected and the center pixel that changes significantly in the vertical direction. The overlapping part between the parts with significant changes in both directions is recorded as the second feature region. Thus, the second feature region can represent the part of the sub-image to be detected where the overall gray-level difference between each pixel and the center pixel changes significantly.

[0061] For any difference image corresponding to the sub-image to be detected, the variance of the pixel values ​​of all pixels is recorded as the roughness coefficient. The roughness coefficient corresponding to the first difference image represents the degree of fluctuation of the gray-level difference in the sub-image to be detected in the horizontal direction, and the roughness coefficient corresponding to the second difference image represents the degree of fluctuation of the gray-level difference in the sub-image to be detected in the vertical direction. When there are defective parts in the sub-image to be detected, the larger the value of the roughness coefficient of the corresponding difference image, the greater the degree of fluctuation of the gray-level difference.

[0062] When analyzing the roughness of the corresponding sub-image to be detected based on the roughness coefficients of the first and second difference images, if the roughness distribution in the sub-image to be detected is more biased towards the vertical direction, then the roughness coefficient corresponding to the second difference image needs to be given a larger weight; if the roughness distribution in the sub-image to be detected is more biased towards the horizontal direction, then the roughness coefficient corresponding to the first difference image needs to be given a larger weight.

[0063] In this embodiment, the roughness distribution in the sub-image to be analyzed is characterized by analyzing the texture features of the parts with large overall changes in grayscale differences. That is, the first roughness weight and the second roughness weight are obtained based on the texture information of the second feature region.

[0064] Specifically, taking the second feature region corresponding to any sub-image to be detected as an example, the second feature region is processed using vertical and horizontal filters respectively to obtain vertical and horizontal response results. For any response result in either direction, the mean and standard deviation of the phase response value are obtained, and a coarse weight is obtained based on the mean and standard deviation of the phase response value. The mean is positively correlated with the coarse weight, and the standard deviation is negatively correlated with the coarse weight. The coarse weight corresponding to the horizontal response result is used as the first coarse weight, and the coarse weight corresponding to the vertical response result is used as the second coarse weight.

[0065] In this embodiment, a Gabor filter is used to process the second feature region in the sub-image to be detected. The Gabor filter extracts the image feature responses of the second feature region in different directions and at different scales. The Gabor filter is a filter based on the Gabor function, which is a sine wave modulated by a Gaussian function and can be used to describe the texture features of an image. The response result of the Gabor filter is a two-dimensional complex matrix, representing the response of the texture features of the input image in different directions and at different scales. Specifically, the amplitude of the response result represents the intensity of the texture features of the input image in different directions and at different scales, and the phase of the response result represents the distribution direction of the texture features of the image.

[0066] To obtain the texture coarseness of the second feature region in the sub-image to be detected in both the horizontal and vertical directions, Gabor filters in the vertical and horizontal directions are used to process the second feature region, respectively, to obtain the vertical and horizontal response results. Specifically, Gabor filters in both directions are convolved with the second feature region in the sub-image to be detected, resulting in vertical and horizontal phase response maps. The mean and standard deviation of all phase response values ​​in each direction's phase response map are then calculated.

[0067] In this embodiment, for any second feature region in a sub-image to be detected, taking the response result in the horizontal direction as an example, the calculation formula for the first coarse weight can be expressed as:

[0068]

[0069] Where α represents the first coarse weight corresponding to the second feature region in the sub-image to be detected. This represents the mean of all phase response values ​​in the horizontal direction response results. Norm() represents the standard deviation of all phase response values ​​in the horizontal response results, and represents the normalization function. This represents the first preset value, which is 0.01 in this embodiment. To prevent the denominator from being 0, the implementer can set it according to the specific implementation scenario.

[0070] The mean of all phase response values ​​in the horizontal direction response results The larger the value of , the larger the phase response value in the horizontal direction, which in turn indicates that the texture feature distribution in the second feature region of the sub-image to be detected is more concentrated in the horizontal direction. This reflects that the bias direction of the roughness distribution in the second feature region is closer to the horizontal direction, and the corresponding value of the first roughness weight is larger.

[0071] The standard deviation of all phase response values ​​in the horizontal direction response results The larger the value of , the greater the fluctuation of the phase response value in the horizontal direction, indicating that the texture feature distribution in the horizontal direction is less uniform, and thus the texture direction is less concentrated. This reflects that the less concentrated the roughness distribution in the second feature region is in the horizontal direction, the smaller the value of the corresponding first roughness weight.

[0072] The second coarseness weight can be obtained using the same formula as the first coarseness weight. The first coarseness weight characterizes the degree to which the coarse texture distribution of the second feature region in the sub-image to be detected is concentrated in the horizontal direction, while the second coarseness weight characterizes the degree to which the coarse texture distribution of the second feature region in the sub-image to be detected is concentrated in the vertical direction. Simultaneously, the second feature region characterizes areas in the sub-image to be detected with large variations in grayscale, which may correspond to defective parts. Therefore, the texture information corresponding to the second feature region is used to weight the overall coarseness distribution of the sub-image to be detected.

[0073] Specifically, the roughness of the sub-image to be detected is obtained by weighted summation of the roughness coefficients of the first and second difference images using the first and second roughness weights. The formula for calculating the roughness of the sub-image to be detected can be expressed as follows:

[0074]

[0075] Where Q represents the roughness of the sub-image to be detected, and α represents the first roughness weight. β represents the coarsening coefficient of the first difference image, and β represents the second coarsening weight. This represents the roughness coefficient of the second difference image.

[0076] The larger the value of the first roughness weight, the more concentrated the rough texture distribution in the sub-image to be detected is in the horizontal direction. The larger the value of the roughness coefficient of the first difference image, the greater the fluctuation of the gray-level difference in the sub-image to be detected in the horizontal direction. The larger the corresponding roughness value, the more likely there are defective parts in the sub-image to be detected.

[0077] The larger the value of the second roughness weight, the more concentrated the rough texture distribution in the sub-image to be detected is in the vertical direction. The larger the value of the roughness coefficient of the second difference image, the greater the fluctuation of the gray-level difference change in the vertical direction of the sub-image to be detected. The larger the value of the corresponding roughness, the more likely there are defective parts in the sub-image to be detected.

[0078] Step 3: Adjust the preset scale of wavelet transform using the coarsness to obtain the optimal scale. Based on the optimal scale, use wavelet transform to decompose the two grayscale images to be detected, and then fuse them based on the decomposition results to obtain the fused feature image.

[0079] When fusing two grayscale images to be detected, namely the first grayscale image and the second grayscale image, the first grayscale image has a better effect on representing the texture information and detail edges of the defective parts of the image, while the second grayscale image has a better effect on representing the main features of the defective parts of the image. Therefore, the wavelet decomposition transform algorithm is first used to analyze the grayscale images to be detected separately to obtain low-frequency and high-frequency coefficients. Different fusion methods are used for different coefficients corresponding to the two grayscale images to be detected. Then, the inverse wavelet decomposition transform is used to obtain the fused image of the two grayscale images to be detected. This can better fuse the parts of the two grayscale images to be detected that have better representation effects, making the subsequent defect analysis based on the fused image more accurate.

[0080] In this embodiment, discrete wavelet transform and inverse transform are used to process the first grayscale image and the second grayscale image, respectively. At the same time, the Dobesi wavelet basis function is used as the wavelet basis function in the discrete wavelet transform algorithm. The scale is one of the important parameters in the wavelet basis function. Using a fixed scale value results in poor decomposition effect.

[0081] Based on this, the surface of the underwater highway anti-corrosion layer, after being corroded, appears dark gray in the corresponding grayscale image to be detected, and the corroded parts may have depressions, resulting in high roughness. Conversely, the surface of a normal anti-corrosion layer without corrosion appears lighter in color in the corresponding grayscale image to be detected, and there are no depressions, resulting in lower roughness. When performing discrete wavelet transform processing on local regions of the grayscale image to be detected, if the local region contains corrosion defects, the corresponding roughness is high. Using a smaller scale results in a higher frequency of the corresponding wavelet basis function, allowing for more accurate detail information, i.e., high-frequency information. If the local region does not contain corrosion defects, the corresponding roughness is low. Using a larger scale results in a lower frequency of the corresponding wavelet basis function, better representing approximate information in the image, i.e., low-frequency information.

[0082] Based on this, the roughness of the sub-image to be detected in the grayscale image represents the roughness of the texture in the sub-image. The optimal scale is obtained by adjusting the preset scale of the wavelet transform using the roughness. Specifically, for any sub-image to be detected, the sum of the preset value and the roughness of the sub-image is obtained. The ratio between the preset scale of the wavelet transform and the sum is calculated. The ratio is then rounded to obtain the optimal scale corresponding to the sub-image to be detected. In this embodiment, the formula for calculating the optimal scale can be expressed as:

[0083]

[0084] in, Let j represent the preferred scale corresponding to the sub-image to be detected, j represent the preset scale, and Q represent the coarseness of the sub-image to be detected. This is the floor function. The value is a preset value, and in this embodiment it is 1. To prevent the denominator from being 0, the implementer can set it according to the specific implementation scenario.

[0085] It should be noted that the preset scale is the scale parameter in the wavelet basis function and scaling function in wavelet decomposition transform. Different scale parameters correspond to different decomposition results. In this embodiment, the preset scale is set to 5. At the same time, the coarsening degree is used to adjust the preset scale, and the adjusted value is processed by the floor function so that the final scale value is an integer. The implementer can set it according to the specific implementation scenario.

[0086] A higher roughness value indicates a coarser surface texture distribution in the sub-image to be detected, suggesting the potential presence of defects. The corresponding higher function frequency indicates better detail information in the sub-image when decomposed using this preferred scale. Conversely, a lower roughness value indicates a smoother surface texture, suggesting the absence of defects. The corresponding lower function frequency indicates better representation of the main subject in the sub-image when decomposed using this preferred scale.

[0087] Thus, we can obtain the adaptive optimal scale corresponding to each sub-image to be detected in the two grayscale images to be detected. Based on the optimal scale, we use the Dobesi wavelet basis function in the discrete wavelet transform algorithm to decompose each sub-image to be detected in each grayscale image to be detected, and obtain the decomposition result of each grayscale image to be detected.

[0088] It should be noted that for any grayscale image to be detected, each sub-image to be detected has a corresponding region within the grayscale image. That is, when performing discrete wavelet transform on the grayscale image, the corresponding region of each sub-image to be detected is decomposed to obtain the decomposition result of the grayscale image. The decomposition result of the grayscale image includes the high-frequency coefficients and low-frequency coefficients of the grayscale image.

[0089] Traditional discrete wavelet transform algorithms output two sets of coefficients: approximation coefficients and detail coefficients. The approximation coefficients represent the output of the low-pass filter of the wavelet transform, and the detail coefficients represent the output of the high-pass filter of the wavelet transform. In this embodiment, the output approximation coefficients are denoted as low-frequency coefficients, and the output detail coefficients are denoted as high-frequency coefficients.

[0090] The maximum value of the high-frequency coefficients of the first grayscale image and the second grayscale image is obtained as the reconstructed high-frequency coefficients, which can preserve the edge and texture information in the image to the greatest extent. The low-frequency coefficients of the first grayscale image and the second grayscale image are weighted and summed using preset weights to obtain the reconstructed low-frequency coefficients, so that the low-frequency optical characteristics have a greater influence on the fusion result and avoid distortion. The fused feature image is obtained by performing wavelet decomposition inverse transform using the reconstructed high-frequency coefficients and the reconstructed low-frequency coefficients.

[0091] Specifically, in this embodiment, the first weight corresponding to the low-frequency coefficients of the first grayscale image is set to 0.3, and the second weight corresponding to the low-frequency coefficients of the second grayscale image is set to 0.7. The implementer can set these weights according to the specific implementation scenario, and then use the first weight and the second weight to weighted sum the low-frequency coefficient information of the first grayscale image and the second grayscale image to obtain the reconstructed low-frequency coefficients.

[0092] After obtaining the reconstructed high-frequency and low-frequency coefficients, these coefficients are used as the decomposition results of the discrete wavelet transform. Then, the inverse discrete wavelet transform is performed on the decomposition results to obtain the reconstructed image, which is denoted as the fused feature image. Discrete wavelet transform and inverse discrete wavelet transform are well-known techniques, and only a brief introduction is given here; their transformation processes will not be elaborated upon in detail.

[0093] Step 4: Obtain the defect region from the fused feature image, and obtain the highway quality inspection result based on the area characteristics of the defect region.

[0094] The obtained fused feature image minimizes underwater environmental noise and effectively highlights the detailed features of defective parts in the image, thus enabling accurate extraction of defective regions. Specifically, the implementer can extract defective regions from the fused feature image using methods such as threshold segmentation, connected component extraction, and region growing. In this embodiment, a region growing algorithm is used to process the fused feature image.

[0095] Specifically, the pixel corresponding to the minimum pixel value in the fused feature image is used as the initial seed point. The condition for region growing is that the absolute value of the difference between pixel values ​​is less than or equal to a pixel threshold. In this embodiment, the pixel threshold is set to 20, but the implementer can set it according to the specific implementation scenario. Further, region growing is performed according to the initial seed point and the region obtained is recorded as the defect region.

[0096] Because the corrosion defects in the underwater highway anti-corrosion layer are darker in color than the normal parts of the anti-corrosion layer surface, the defect areas obtained through regional growth are the corrosion defect areas of the underwater highway anti-corrosion layer. Then, the quality inspection results of the highway are obtained based on the area characteristics of the defect areas.

[0097] Specifically, the area of ​​each defective region is obtained, and the ratio between the sum of the areas of all defective regions and the total area of ​​the fused feature image is used as a quality coefficient. When the quality coefficient is less than a preset first threshold, the corresponding highway anti-corrosion layer's quality inspection result is qualified; when the quality coefficient is greater than or equal to the first threshold but less than a preset second threshold, the corresponding highway anti-corrosion layer's quality inspection result is Level 1 damage; when the quality coefficient is greater than or equal to the second threshold, the corresponding highway anti-corrosion layer's quality inspection result is Level 2 damage; the first threshold is less than the second threshold. Level 2 damage is more severe than Level 1 damage.

[0098] In this embodiment, the first threshold is set to 0.15, and the second threshold is set to 0.3. The implementer can set these values ​​according to the specific implementation scenario. When the quality coefficient is less than 0.15, it indicates that there are few defects on the surface of the underwater highway anti-corrosion layer, meaning only minor damage exists, which does not affect the continued use of the anti-corrosion layer. Therefore, the quality inspection result of the highway anti-corrosion layer is qualified. When the quality coefficient is greater than or equal to 0.15 and less than 0.3, it indicates that there are many defects on the surface of the underwater highway anti-corrosion layer, meaning moderate damage exists, indicating a certain quality problem. Relevant personnel need to take timely measures to avoid safety hazards. Therefore, the quality inspection result of the highway anti-corrosion layer is Level 1 damage. When the quality coefficient is greater than 0.3, it indicates that there are the most defects on the surface of the underwater highway anti-corrosion layer, meaning severe damage exists. This indicates a serious quality problem on the surface of the underwater highway anti-corrosion layer, requiring even more timely measures to avoid safety hazards. Therefore, the quality inspection result of the highway anti-corrosion layer is Level 2 damage.

[0099] In summary, this paper improves the quality inspection method for highway anti-corrosion coatings, avoiding the shortcomings of conventional discrete wavelet transform methods. It employs image processing techniques, first acquiring RGB and infrared images of the underwater highway anti-corrosion coating surface. Then, discrete wavelet transform is used to decompose the corresponding grayscale images of the two images. Different fusion methods are applied to coefficients at different frequencies to avoid the influence of underwater environmental noise and to highlight the details of corrosion defects. Simultaneously, the roughness of the sub-image to be detected is obtained based on the pixel feature distribution of the corresponding sub-image. The scale parameter of the Dobesi wavelet basis function in the discrete wavelet transform is adaptively adjusted according to the roughness, resulting in better performance of discrete wavelet transform and inverse discrete wavelet transform. This leads to a fused feature image with better edge and detail texture. Defect regions are extracted based on this fused feature image, thus making the quality inspection results of the underwater highway anti-corrosion coating more accurate.

[0100] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for detecting and evaluating the anticorrosive effect of a road based on image processing technology, characterized in that, The method comprises the following steps: The method comprises the following steps: Obtaining two different types of image corresponding to the detected gray image of the road anticorrosion layer surface, and obtaining the detected sub-image by segmenting the detected gray image; obtaining two difference images corresponding to the detected sub-image according to the row gray difference and column gray difference of the gray difference image corresponding to the detected sub-image; Obtaining the roughness of the detected sub-image according to the difference of pixel value and the texture information of the two difference images corresponding to the detected sub-image; Adjusting the preset scale of wavelet transform by using the roughness to obtain the optimal scale, decomposing the two detected gray images by using wavelet transform according to the optimal scale, and obtaining the fusion feature image by fusing the decomposition results; 2. The method for detecting and evaluating the anticorrosive effect of a road based on image processing technology according to claim 1, characterized in that, Obtaining the defect area of the fusion feature image, and obtaining the quality detection result of the road according to the area characteristics of the defect area. The method comprises the following steps: For any one of the detected sub-image, the absolute value of the difference between the gray values of the pixel points corresponding to the adjacent rows in the corresponding gray difference image is calculated to obtain the first difference image of the detected sub-image; the absolute value of the difference between the gray values of the pixel points corresponding to the adjacent columns in the corresponding gray difference image is calculated to obtain the second difference image of the detected sub-image; 3. The method for detecting and evaluating the anticorrosive effect of a road based on image processing technology according to claim 2, characterized in that, The difference image comprises the first difference image and the second difference image. The method comprises the following steps: For any one of the difference images corresponding to the detected sub-image, the variance of the pixel value of all pixel points is taken as the roughness coefficient; the threshold segmentation is performed on the difference image to obtain a first feature area, and the pixel value of the pixel points in the first feature area is greater than a preset threshold; the overlapping area of the first feature areas of the two difference images is obtained to obtain a second feature area; 4. The method for detecting and evaluating the anticorrosive effect of a road based on image processing technology according to claim 3, characterized in that, The first roughness weight and the second roughness weight are obtained according to the texture information of the second feature area, and the roughness coefficients of the first difference image and the second difference image are weighted and summed by using the first roughness weight and the second roughness weight to obtain the roughness corresponding to the detected sub-image. The method comprises the following steps: The vertical direction filter and the horizontal direction filter are used to process the second feature area respectively to obtain the response results in the vertical direction and the horizontal direction; 5. The method for detecting and evaluating the anticorrosive effect of a road based on image processing technology according to claim 1, characterized in that, For any one of the response results, the mean value and the standard deviation of the phase response value are obtained, the roughness weight is obtained according to the mean value and the standard deviation of the phase response value, the mean value and the roughness weight are positively correlated, and the standard deviation and the roughness weight are negatively correlated; the roughness weight corresponding to the response result in the horizontal direction is taken as the first roughness weight, and the roughness weight corresponding to the response result in the vertical direction is taken as the second roughness weight. The method for obtaining the gray difference image corresponding to the detected sub-image comprises the following steps: For any one of the to-be-detected sub-image, the center pixel point of the to-be-detected sub-image is obtained, if the number of the center pixel point is even, the average value of the pixel value of all the center pixel points is taken as the feature pixel value, if the number of the center pixel point is one, the pixel value of the center pixel point is taken as the feature pixel value; The difference between the pixel value of each pixel point in the to-be-detected sub-image and the feature pixel value is calculated respectively to obtain a gray difference image.

6. The method for detecting and evaluating the anticorrosive effect of a road based on image processing technology according to claim 1, characterized in that, The adjusting the preset scale of the wavelet transform by using the roughness to obtain an optimal scale is specifically: For any one of the to-be-detected sub-image, the sum value between the preset value and the roughness of the to-be-detected sub-image is obtained, the ratio between the preset scale of the wavelet transform and the sum value is calculated, and the optimal scale corresponding to the to-be-detected sub-image is obtained by rounding the ratio.

7. The method for detecting and evaluating the anticorrosive effect of a road based on image processing technology according to claim 1, characterized in that, The to-be-detected gray image corresponding to the two different types of images of the surface of the highway anticorrosive layer is specifically: The RGB image and the infrared image of the surface of the highway anticorrosive layer are obtained respectively, the RGB image is processed to obtain a first gray image, and the infrared image is processed to obtain a second gray image; the first gray image and the second gray image are to-be-detected gray images.

8. The method for detecting and evaluating the anticorrosive effect of a road based on image processing technology according to claim 7, characterized in that, The fusion feature image is obtained according to the decomposition result, and the decomposition result includes the high-frequency coefficient and the low-frequency coefficient of the to-be-detected gray image. The maximum value of the high-frequency coefficient of the first gray image and the second gray image is taken as the reconstructed high-frequency coefficient, and the low-frequency coefficient information of the first gray image and the second gray image is weighted and summed by using a preset weight to obtain a reconstructed low-frequency coefficient. The fusion feature image is reconstructed by using the reconstructed high-frequency coefficient and the reconstructed low-frequency coefficient for wavelet inverse transform. The quality detection result of the highway is obtained according to the area characteristics of the defect area, and the area of each defect area is obtained, the ratio between the sum value of the areas of all the defect areas and the total area of the fusion feature image is taken as a quality coefficient, when the quality coefficient is less than a preset first threshold value, the quality detection result of the corresponding highway anticorrosive layer is qualified; when the quality coefficient is greater than or equal to the first threshold value and less than a preset second threshold value, the quality detection result of the corresponding highway anticorrosive layer is a first-level damage; when the quality coefficient is greater than or equal to the second threshold value, the quality detection result of the corresponding highway anticorrosive layer is a second-level damage; 9. The method for detecting and evaluating the anticorrosive effect of a road based on image processing technology according to claim 1, characterized in that, The first threshold value is less than the second threshold value. ​ ​

Images