Intelligent inspection method for bridge piers

Abstract

The invention belongs to the technical field of image processing, and particularly relates to a bridge pier intelligent inspection method, which comprises the following steps of: performing Gaussian filtering on an inspection image through Gaussian filters with different scales, and subtracting an original inspection image from obtained background images with different scales to obtain intermediate images with different scales; calculating the enhancement degree of the gradient features of the pixel points in the intermediate images under different scales compared with the gradient features of the pixel points in the original inspection image; calculating the retention degree of the multiple LBP features of the pixel points in the intermediate image under different scales compared with the original inspection image; and according to the enhancement degree and the retention degree, calculating the weights of the gray values of the pixel points in the intermediate image under different scales, and carrying out weighted summation on the gray values of the pixel points in the intermediate image under different scales so as to obtain a uniform light image. According to the invention, the finally obtained uniform light image achieves better balance in detail and texture aspects.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing. More specifically, the present invention relates to an intelligent inspection method for bridge piers. Background Art

[0002] With the rapid development of China's transportation infrastructure, the number of bridges is increasing continuously, and the safe operation and maintenance of bridges have become crucial tasks; as the key load-bearing structure of bridges, the health status of bridge piers is directly related to the overall safety and service life of bridges.

[0003] Traditional inspection of bridge piers mainly relies on manual labor. Workers need to use equipment such as boats, scaffolding, or bridge inspection vehicles to approach the bridge piers for close visual inspection and simple measurements. The inspection efficiency is low. Especially for large bridges and bridge groups distributed in vast areas, the inspection work is time-consuming and laborious, and there are certain safety risks in manual inspection. Especially in adverse weather conditions or complex water area environments, the personal safety of workers is difficult to be fully guaranteed.

[0004] In order to overcome the deficiencies of traditional inspection methods and improve the efficiency, safety, and accuracy of bridge pier inspection, unmanned aerial vehicle (UAV) technology has been introduced into the field of bridge inspection; through the images and videos collected by UAVs, the surface conditions of bridge piers can be clearly observed, and structural damages and defects can be discovered in a timely manner; at the same time, UAV inspection can realize real-time data transmission and remote monitoring, and workers can analyze and evaluate the inspection data in the control center, greatly improving the convenience and timeliness of the inspection work.

[0005] However, due to the influence of environmental factors, lighting conditions, shooting angles, etc., there is uneven lighting phenomenon in the inspection videos of bridge piers collected by UAVs, which affects the subsequent accurate observation, analysis, and evaluation of the surface conditions of bridge piers; the MASK equalization algorithm performs low-pass filtering on the inspection images through a Gaussian filter, so as to obtain and remove the background image simulating the brightness distribution from the inspection images and obtain an image with uniform illumination.

[0006] However, the scale of the Gaussian filter used in the MASK equalization algorithm is single, and it is unable to balance the equalization effect and detail information, which affects the inspection effect. Summary of the Invention

[0007] To solve the technical problem that the scale of the Gaussian filter used in the above MASK equalization algorithm is single and cannot balance the equalization effect and detail information, the present invention provides an intelligent inspection method for bridge piers, including: collecting inspection videos of bridge piers through a camera carried on a drone, and performing equalization processing on each inspection image in the inspection video through the MASK equalization algorithm, including: setting Gaussian filters with different scales; performing Gaussian filtering on the inspection images through Gaussian filters with different scales to obtain background images simulating the brightness distribution at different scales; subtracting the original inspection images from the background images at different scales to obtain intermediate images at different scales; calculating the enhancement degree of the gradient features of pixel points in the intermediate images at different scales compared with the gradient features in the original inspection images; calculating the retention degree of the multiple LBP features of pixel points in the intermediate images at different scales compared with the multiple LBP features in the original inspection images; calculating the weights of the gray values of pixel points in the intermediate images at different scales according to the enhancement degree and the retention degree, and performing weighted summation on the gray values of pixel points in the intermediate images at different scales according to the weights of the gray values of pixel points in the intermediate images at different scales as the gray values of pixel points in the equalized image, so as to obtain the equalized image.

[0008] The present invention constructs a brightness distribution model of the inspection image by setting Gaussian filters with different scales, provides richer background information for subsequent equalization processing, calculates the enhancement degree of the gradient features and the retention degree of the multiple LBP features compared with the original inspection image for the intermediate images at different scales, combines the enhancement degree and the retention degree to obtain the weights of the gray values in the intermediate images at different scales, and then reasonably fuses the information of the intermediate images at different scales through the weights, so that the finally obtained equalized image achieves a better balance in terms of details and textures, gives full play to the advantages of the drone in collecting inspection videos of bridge piers, realizes the intelligent inspection of bridge piers, improves the bridge maintenance level, and ensures the safe operation of the bridge.

[0009] Preferably, the setting of Gaussian filters with different scales includes: taking each odd number within the range as a standard deviation for constructing Gaussian filters with different scales, and are the upper limit and the lower limit of the range respectively, and .

[0010] The present invention can more accurately construct a brightness distribution model of the inspection image by setting Gaussian filters with different scales, provides richer background information for subsequent equalization processing, and then avoids the problem of poor equalization effect caused by improper selection of the filter scale.

[0011] Preferably, calculating the enhancement degree of the gradient feature of the pixel point in the intermediate image at different scales compared with the gradient feature in the original inspection image includes: ; In the formula, , is the number of all scale types, is the enhancement degree of the gradient feature of the pixel point in the intermediate image at the -th scale compared with the gradient feature in the original inspection image, , are respectively the gradient magnitude and gradient direction of the pixel point in the intermediate image at the -th scale, , are respectively the gradient magnitude and gradient direction of the pixel point in the original inspection image, represents taking the absolute value, represents the Sigmoid function.

[0012] The present invention provides a basis for determining the weight of the image gray value in the subsequent process by screening out the scales with better effects in enhancing image details, so that the final equalized illumination image can better retain the detail information of the image.

[0013] Preferably, calculating the retention degree of the multiple LBP feature of the pixel point in the intermediate image at different scales compared with the multiple LBP feature in the original inspection image includes: ; In the formula, , is the number of all scale types, is the retention degree of the multiple LBP feature of the pixel point in the intermediate image at the -th scale compared with the multiple LBP feature in the original inspection image, is the multiple LBP feature of the pixel point in the intermediate image at the -th scale, is the multiple LBP feature of the pixel point in the original inspection image, represents the Hamming distance.

[0014] Calculating the retention degree of the multiple LBP feature of the pixel point in the intermediate image at different scales compared with the multiple LBP feature in the original inspection image by the present invention helps to evaluate the performance of the intermediate images at different scales in maintaining the texture details of the image, so as to balance the relationship between detail enhancement and texture retention in the subsequent process, and avoid texture distortion caused by excessive enhancement or detail loss caused by excessive texture retention.

[0015] Preferably, calculating the weight of the gray value of the pixel point in the intermediate image at different scales according to the enhancement degree and the retention degree includes: ; In the formula, , is the number of types of all scales, is the weight of the gray value of the pixel point in the intermediate image at the -th scale, is the retention degree of the multiple LBP features in the intermediate image of the pixel point at the -th scale compared with the multiple LBP features in the original inspection image, is the minimum limit of the retention degree, is the enhancement degree of the gradient feature in the intermediate image of the pixel point at the -th scale compared with the gradient feature in the original inspection image, represents taking the maximum value, represents the normalization function.

[0016] By calculating the weight of the gray value of the pixel point in the intermediate image at different scales, the present invention can reasonably fuse the information of the intermediate images at different scales, so that the final equalized image achieves a better balance in terms of details and textures, and avoids information loss that may be caused by simply averaging and fusing different-scale images.

[0017] Preferably, the method for obtaining the multiple LBP features of the pixel point includes: constructing a neighborhood centered on the pixel point with a size of 5×5, then the obtained neighborhood includes 1 central pixel point and 24 neighborhood pixel points; calculating the absolute value of the difference between the gray value of each neighborhood pixel point and the central pixel point as the gray difference between each neighborhood pixel point and the central pixel point ; By comparing the gray difference between the -th neighborhood pixel point and the central pixel point with 4 sub-regions, obtaining the marking value of the neighborhood position corresponding to the -th neighborhood pixel point; arranging the marking values of the neighborhood positions corresponding to all neighborhood pixel points in sequence and splicing them into a string as the multiple LBP features of the pixel point.

[0018] Preferably, the method for obtaining the 4 sub-regions includes: setting 3 boundary values within the range , which are respectively , , , and ; Dividing the range into 4 sub-regions by the 3 boundary values, which are respectively sub-region , sub-region , sub-region , sub-region .

[0019] Preferably, obtaining the marking value of the neighborhood position corresponding to the th neighborhood pixel point includes: If the gray-scale difference is within the sub-region , then the marking value of the neighborhood position corresponding to the th neighborhood pixel point is 00; if the gray-scale difference is within the sub-region , then the marking value of the neighborhood position corresponding to the th neighborhood pixel point is 01; if the gray-scale difference is within the sub-region , then the marking value of the neighborhood position corresponding to the th neighborhood pixel point is 10; if the gray-scale difference is within the sub-region , then the marking value of the neighborhood position corresponding to the th neighborhood pixel point is 11.

[0020] Preferably, the normalization function includes: , is the retention degree of the multiple LBP features in the intermediate image at the th scale of the pixel point compared to the multiple LBP features in the original inspection image, is the minimum limit of the retention degree, is the enhancement degree of the gradient features in the intermediate image at the th scale of the pixel point compared to the gradient features in the original inspection image, represents taking the maximum value, is the number of all scale types.

[0021] Preferably, the method further includes: performing stretching display processing on the uniform illumination image, and the stretching display processing method includes 2% linear stretching and contrast parameter stretching.

[0022] The beneficial effects of the present invention are as follows: By setting Gaussian filters of different scales, the present invention constructs a brightness distribution model of the inspection image, provides richer background information for subsequent uniform illumination processing, calculates the enhancement degree of the gradient features and the retention degree of the multiple LBP features in the intermediate images at different scales compared to the original inspection image, combines the enhancement degree and the retention degree to obtain the weights of the gray values in the intermediate images at different scales, and then reasonably fuses the information of the intermediate images at different scales through the weights, so that the finally obtained uniform illumination image achieves a better balance in terms of details and textures, gives full play to the advantages of the UAV in collecting bridge pier inspection videos, realizes the intelligent inspection of bridge piers, improves the bridge maintenance level, and ensures the safe operation of the bridge. Description of the Drawings

[0023] Figure 1 is a flowchart schematically showing a method for intelligent inspection of bridge piers in the present invention; Figure 2 is a schematic diagram schematically showing an inspection image with uneven illumination phenomenon; Figure 3 is schematically showing that through the MASK equalization algorithm for Figure 2 when performing equalization processing, a schematic diagram of the equalized image obtained; Figure 4 is schematically showing that through the method of the present invention for Figure 2 when performing equalization processing, a schematic diagram of the equalized image obtained. Specific embodiments

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0025] Next, the specific embodiments of the present invention will be described in detail in conjunction with the accompanying drawings.

[0026] The embodiments of the present invention disclose a method for intelligent inspection of bridge piers. Referring to Figure 1 , it includes steps S1 to S3: S1. Collect inspection videos of bridge piers through a camera mounted on a drone, including multiple inspection images.

[0027] Specifically, according to the pier height, inspection range, and accuracy requirements, select an appropriate model of industrial drone; use Geographic Information System (GIS) and satellite positioning technology (such as GPS or Beidou), combined with the precise coordinate data of the bridge pier, to pre-plan the inspection flight path of the drone to ensure that the drone can perform circular shooting along multiple key points evenly distributed around the pier; and then control the drone to collect inspection videos of the bridge pier according to the inspection flight path through the mounted camera, and the obtained inspection videos are composed of multiple inspection images.

[0028] Furthermore, during the flight of the drone, through a wireless communication module (such as 4G network or 5G network), the collected inspection videos are transmitted back to the ground control station in real time. The ground control station can be deployed in a monitoring room or an operation and maintenance vehicle near the bridge, and is equipped with professional image receiving and processing equipment to ensure the stable reception of the videos.

[0029] It should be noted that drones can quickly reach the areas around bridge piers, covering a large-scale inspection area. Compared with the traditional manual inspection method, the inspection time is greatly shortened and the work efficiency is improved. At the same time, drones can replace humans to enter dangerous areas for close-up shooting, reducing the safety risks of personnel.

[0030] S2. Perform Gaussian filtering on the inspection image through Gaussian filters of different scales to obtain background images simulating the brightness distribution at different scales. Subtract the original inspection image from the background images at different scales to obtain intermediate images at different scales.

[0031] It should be noted that the inspection video of the bridge pier collected by the drone can achieve remote monitoring, clearly observe the surface condition of the bridge pier, and promptly discover potential diseases, greatly improving the convenience and timeliness of the inspection work. However, when collecting the inspection video of the bridge pier by the drone, affected by environmental factors, lighting conditions, shooting angles and other factors, there are varying degrees of differences in the brightness, hue and contrast inside the inspection image, that is, there is uneven illumination phenomenon, which affects the subsequent accurate observation, analysis and evaluation of the surface condition of the bridge pier. Therefore, eliminating the uneven illumination phenomenon in the inspection video of the bridge pier and performing uniform illumination processing on the inspection video of the bridge pier have important theoretical and practical application values for obtaining high-quality inspection videos.

[0032] Exemplarily, a schematic diagram of an inspection image with uneven illumination phenomenon is as Figure 2 shown.

[0033] Furthermore, it should be noted that the MASK uniform illumination algorithm is a method proposed for uneven image illumination. It mainly regards the image as the superposition of a background image with uneven brightness and an image with uniform light reception. From the perspective of the frequency domain, the brightness distribution of the image is mainly reflected as low-frequency information, while the high-frequency information includes edges, details and noises, etc. By enhancing the high-frequency information and suppressing the low-frequency information, the detail contrast of the image is enhanced and the abnormal brightness change of the image is suppressed, so as to achieve the purpose of uniform brightness distribution. Therefore, the MASK uniform illumination algorithm performs low-pass filtering on the original image through a Gaussian filter to obtain a background image simulating the brightness distribution, and then subtracts the original image from the background image to obtain an image with uniform light reception.

[0034] It should be noted that the Gaussian function in the Gaussian filter belongs to the normal distribution. Therefore, the larger the standard deviation of the Gaussian function, the larger the scale of the Gaussian filter. More pixel points participate in the operation but they are widely distributed, resulting in a greater degree of blurring and insufficient simulation effect on the local brightness distribution, leading to a poor equalization effect of the finally obtained equalized image. On the contrary, the smaller the standard deviation of the Gaussian function, the smaller the scale of the Gaussian filter. Fewer pixel points participate in the operation but they are concentrated. The obtained background image can better simulate the brightness distribution, but at the same time, it also contains more detailed information, resulting in the loss of local detailed information in the finally obtained equalized image.

[0035] Therefore, by performing Gaussian filtering on the inspection image using Gaussian filters of different scales, background images simulating the brightness distribution at different scales can be obtained. By subtracting the inspection image from the background images at different scales, intermediate images at different scales can be obtained. Furthermore, by synthesizing the intermediate images at different scales, the final equalized image can be obtained.

[0036] 1. Set Gaussian filters of different scales.

[0037] It should be noted that since a Gaussian filter with a single scale has limitations: when the standard deviation is large, the simulation effect on local brightness is poor; when the standard deviation is small, local detail loss is likely to occur. By setting Gaussian filters of different scales with different standard deviations, Gaussian filters of different scales can respectively simulate the brightness distribution in different ranges, from large-scale brightness changes to small-scale local brightness details, and the brightness distribution information of the inspection image can be obtained more comprehensively.

[0038] Specifically, each odd number within the range is used as a standard deviation to construct Gaussian filters of different scales. and are the upper and lower limits of the range respectively, and ; as the standard deviation increases, the smoothing effect of the constructed Gaussian filter becomes more obvious.

[0039] Among them, the specific values of and in the range can be set according to the actual application scenario and requirements, and , where has a value range of [7, 11], has a value range of [15, 19], and in the present invention, , are respectively set to 9 and 15.

[0040] It should be noted that, compared with the Gaussian filtering of a single scale, setting Gaussian filters of different scales can more accurately construct the brightness distribution model of the inspection images, provide richer background information for subsequent equalization processing, and thus avoid the problem of poor equalization effect caused by improper selection of the filter scale.

[0041] 2. Perform Gaussian filtering on the inspection images through Gaussian filters of different scales to obtain background images simulating the brightness distribution at different scales; subtract the original inspection images from the background images at different scales to obtain intermediate images at different scales.

[0042] It should be noted that the original inspection images contain background information with uneven brightness and detail information of the images. By subtracting the original inspection images from the background images at different scales, the uneven brightness components in the original inspection images can be removed, and intermediate images mainly containing the detail information of the images can be obtained. Moreover, the intermediate images at different scales respectively correspond to the extraction of detail information at different scales.

[0043] S3. According to the weights of the gray values of the pixel points in the intermediate images at different scales, perform weighted summation on the gray values of the pixel points in the intermediate images at different scales as the gray value of the pixel points in the equalized image, so as to obtain the equalized image.

[0044] 1. Obtain the detail features of the pixel points in the intermediate images at different scales and the original inspection images.

[0045] Specifically, the detail features of the pixel points include the gradient features and multiple LBP features of the pixel points; among them, the gradient features of the pixel points are calculated through the Sobel operator, and the gradient features include the gradient direction and the gradient amplitude; the multiple LBP features of the pixel points are calculated through the multiple LBP operator constructed by multiple boundary values set within the range inside.

[0046] Among them, the LBP operator is an operator used to describe the local features in the image. Through the multiple LBP operator constructed by multiple boundary values set within the range inside, the multiple LBP features of the pixel points are calculated. The specific steps include: (1) Construct a neighborhood with the pixel point as the central pixel point and a size of 5×5, then the obtained neighborhood includes 1 central pixel point and 24 neighborhood pixel points.

[0047] (2) Calculate the absolute value of the difference between the gray value of each neighborhood pixel point and the gray value of the central pixel point as the gray difference between each neighborhood pixel point and the central pixel point .

[0048] (3) Set 3 boundary values within the range respectively as , , , and ; The range is divided into 4 sub-regions by 3 boundary values which are sub-region , sub-region , sub-region , and sub-region .

[0049] Among them, the specific values of the boundary values , , can be set according to the actual application scenario and requirements, and the value range of the boundary value is [4, 7], the value range of the boundary value is [9, 12], the value range of the boundary value is [18, 25]. In the present invention, the boundary values , , are respectively set to 5, 10, and 20.

[0050] (4) By comparing the gray-scale difference between the th neighborhood pixel point and the central pixel point with the 4 sub-regions, the marking value of the neighborhood position corresponding to the th neighborhood pixel point is obtained: If the gray-scale difference is within the sub-region , the marking value of the neighborhood position corresponding to the th neighborhood pixel point is 00; If the gray-scale difference is within the sub-region , the marking value of the neighborhood position corresponding to the th neighborhood pixel point is 01; If the gray-scale difference is within the sub-region , the marking value of the neighborhood position corresponding to the th neighborhood pixel point is 10; If the gray-scale difference is within the sub-region , the marking value of the neighborhood position corresponding to the th neighborhood pixel point is 11.

[0051] (5) Arrange the marker values at the neighborhood positions corresponding to all neighborhood pixels in sequence and concatenate them into a string as the multi-LBP feature of the pixel. Since the marker values at the neighborhood positions corresponding to each neighborhood pixel consist of two numerical values, 0 and 1, and each marker value includes two numerical values, the multi-LBP feature of the pixel is essentially a binary sequence consisting of two numerical values, 0 and 1, and its length is equal to 24×2 = 48.

[0052] It should be noted that the LBP feature is an effective method for describing the local texture features of an image, and the multi-LBP feature can describe the texture details of the image from multiple perspectives.

[0053] 2. Calculate the weights of the gray values of the pixels in the intermediate images at different scales according to the change situation of the detail features of the pixels in the intermediate images at different scales compared with the detail features in the original inspection images.

[0054] Specifically, the detail features of the pixel include the gradient feature and the multi-LBP feature of the pixel. Therefore, first, calculate the enhancement degree of the gradient feature of the pixel in the intermediate image at different scales compared with the gradient feature in the original inspection image according to the gradient features of the pixel in the intermediate images at different scales and the original inspection image. Second, calculate the retention degree of the multi-LBP feature of the pixel in the intermediate image at different scales compared with the multi-LBP feature in the original inspection image according to the multi-LBP features of the pixel in the intermediate images at different scales and the original inspection image. Finally, calculate the weights of the gray values of the pixels in the intermediate images at different scales according to the enhancement degree of the gradient feature of the pixel in the intermediate image at different scales compared with the gradient feature in the original inspection image and the retention degree of the multi-LBP feature of the pixel in the intermediate image at different scales compared with the multi-LBP feature in the original inspection image.

[0055] The specific calculation process is as follows: (1) Calculate the enhancement degree of the gradient feature of the pixel in the intermediate image at different scales compared with the gradient feature in the original inspection image according to the gradient features of the pixel in the intermediate images at different scales and the original inspection image. The calculation formula is: ; In the formula, ,, is the number of types of all scales, is the enhancement degree of the gradient feature of the pixel in the intermediate image at the -th scale compared with the gradient feature in the original inspection image, , are respectively the pixels at the The gradient magnitude and gradient direction in the intermediate image at a certain scale and are respectively the gradient magnitude and gradient direction of the pixel point in the original inspection image represents taking the absolute value represents the Sigmoid function

[0056] Among them, the larger the gradient magnitude of the pixel point, the stronger the gradient feature of the pixel point. Therefore, by calculating the gradient magnitude of the pixel point in the intermediate image at the th scale and the gradient magnitude of the pixel point in the original inspection image The difference The larger the difference the stronger the gradient feature of the pixel point in the intermediate image at the th scale compared to that in the original inspection image; on this basis, through the difference in the gradient direction between the two for correction. When the difference in the gradient direction between the two is smaller, the enhancement degree of the corrected gradient feature is greater is used to normalize the difference

[0057] It should be noted that the gradient feature is an important indicator for measuring edges and other detailed information in the image. In the intermediate image, since the influence of partial brightness non-uniformity has been removed, the gradient feature may be enhanced. By comparing the gradient features of pixel points in the intermediate image and the original image, the enhancement effect of the intermediate image on detailed information at different scales can be quantified, which helps to evaluate the contribution of intermediate images at different scales in highlighting details

[0058] (2) According to the multiple LBP features of the pixel point in the intermediate images at different scales and the original inspection image, calculate the retention degree of the multiple LBP features of the pixel point in the intermediate images at different scales compared to those in the original inspection image. The calculation formula is: ; In the formula , is the number of all scale types is the retention degree of the multiple LBP features of the pixel point in the intermediate image at the th scale compared to those in the original inspection image is the multiple LBP feature of the pixel point in the intermediate image at the th scale The multiple LBP features of a pixel in the original inspection image denotes the Hamming distance.

[0059] Among them, the Hamming distance between the multiple LBP features of the pixel in the intermediate image at the th scale and the multiple LBP features of the pixel in the original inspection image The smaller it is, the greater the similarity between the multiple LBP features of the pixel in the intermediate image at the th scale and the multiple LBP features of the pixel in the original inspection image. Therefore, the local detail features of the pixel in the intermediate image at the th scale are better preserved. The 48 in the denominator is used to normalize the Hamming distance for normalization.

[0060] It should be noted that for objects with obvious texture features such as the surface of a bridge pier, texture information is an important factor in evaluating image quality. Therefore, by comparing the multiple LBP features of pixels in the intermediate image and the original image, the ability of the intermediate image at different scales to preserve image texture information can be determined.

[0061] (3) Calculate the weight of the gray value of the pixel in the intermediate image at different scales according to the enhancement degree of the gradient feature of the pixel in the intermediate image at different scales compared with the gradient feature in the original inspection image and the preservation degree of the multiple LBP feature of the pixel in the intermediate image at different scales compared with the multiple LBP feature in the original inspection image. The calculation formula is: ; In the formula, , is the number of all scale types, is the weight of the gray value of the pixel in the intermediate image at the th scale, is the preservation degree of the multiple LBP feature of the pixel in the intermediate image at the th scale compared with the multiple LBP feature in the original inspection image, is the minimum of the preservation degree, is the enhancement degree of the gradient feature of the pixel in the intermediate image at the th scale compared with the gradient feature in the original inspection image, denotes taking the maximum value, represents the normalization function. In this embodiment, , is the number of all scale types.

[0062] Among them, the minimum of the preservation degree The specific value can be set according to the actual application scenario and requirements, and The value range of is set to 0.78 in the present invention.

[0063] Among them, when the preservation degree of the multiple LBP features of the pixel point in the intermediate image at the th scale compared to the multiple LBP features in the original inspection image is greater than the minimum preservation degree it indicates that the pixel point has maintained the same local detail features in the intermediate image at the th scale as in the original inspection image. Therefore, , and as the preservation degree increases, the weight of the gray value of the pixel point in the intermediate image at the th scale is greater. On this basis, the weight of the gray value of the pixel point in the intermediate image at the th scale increases with the enhancement degree of the gradient features of the pixel point in the intermediate image at the th scale compared to the gradient features in the original inspection image; when the preservation degree of the multiple LBP features of the pixel point in the intermediate image at the th scale compared to the multiple LBP features in the original inspection image is less than or equal to the minimum preservation degree at this time, it indicates that the local detail features of the pixel point in the intermediate image at the th scale have changed greatly. Therefore, the weight of the gray value of the pixel point in the intermediate image at the th scale is maintained at 0.

[0064] It should be noted that the intermediate images at different scales have their own advantages and disadvantages in terms of detail enhancement and texture preservation. By comprehensively considering the enhancement degree of gradient features and the preservation degree of multiple LBP features, a weight can be assigned to the gray values of each pixel point in the intermediate images at different scales. This weight reflects the contribution of the intermediate image at that scale to the pixel point. The scale image with a higher weight performs better in terms of comprehensive detail enhancement and texture preservation.

[0065] 3. According to the weights of the gray values of the pixel points in the intermediate images at different scales, perform weighted summation on the gray values of the pixel points in the intermediate images at different scales, and use it as the gray value of the pixel point in the equalization image, thereby obtaining the equalization image.

[0066] Exemplarily, through the MASK equalization algorithm for​Figure 2 When performing even illumination processing, the schematic diagram of the obtained evenly illuminated image is as follows Figure 3 shown; when performing even illumination processing on Figure 2 by the method of the present invention, the schematic diagram of the obtained evenly illuminated image is as follows Figure 4 shown; by comparing Figure 3 and Figure 4 it can be seen that Figure 3 in the light area and the shadow area, there are still obvious gray - level jumps, and the evenly illuminated image as a whole presents a gray - foggy state, Figure 4 not only is the light - dark transition natural, but also details such as cracks in the evenly illuminated image are highlighted, the clarity of the image increases, and the light - illumination consistency effect is stronger.

[0067] It should be noted that, based on the weights calculated previously, the gray - level values of the intermediate images at different scales are weighted and summed, which can integrate the advantages of each scale, make full use of the information of the intermediate images at different scales in terms of detail enhancement and texture preservation, and generate the final evenly illuminated image. The gray - level value of each pixel point can better reflect the actual condition of the pier surface, eliminate the influence of uneven illumination, thereby obtaining a high - quality evenly illuminated image, enabling the pier inspection video to more accurately display the true situation of the pier surface, facilitating more effective observation, analysis, and evaluation of the surface condition of the pier, and improving the quality and efficiency of the inspection work.

[0068] In addition, since the previous processing process involves subtraction operations, the overall contrast of the obtained evenly illuminated image, especially in the darker areas, will decrease. In order to improve the overall contrast of the image, while maintaining the overall contrast consistency of the image, highlighting the image details, and maintaining the image clarity, it is necessary to perform stretching display processing on the evenly illuminated image to finally obtain an evenly illuminated image with uniform light reception.

[0069] Among them, the commonly used stretching display methods include, but are not limited to, 2% linear stretching and contrast - parameter stretching, and both 2% linear stretching and contrast - parameter stretching are well - known technologies, and will not be elaborated here.

Claims

1. An intelligent inspection method for bridge piers, characterized in that, Including: Collecting inspection videos of bridge piers through cameras mounted on drones, and performing uniform illumination processing on each inspection image in the inspection video through the MASK uniform illumination algorithm, including: Setting Gaussian filters of different scales; performing Gaussian filtering on the inspection images through Gaussian filters of different scales to obtain background images simulating brightness distributions at different scales; subtracting the original inspection images from the background images at different scales to obtain intermediate images at different scales; Calculating the enhancement degree of the gradient features of pixel points in the intermediate images at different scales compared to the gradient features in the original inspection images; calculating the retention degree of the multiple LBP features of pixel points in the intermediate images at different scales compared to the multiple LBP features in the original inspection images; According to the enhancement degree and the retention degree, calculating the weights of the gray values of pixel points in the intermediate images at different scales, and performing weighted summation on the gray values of pixel points in the intermediate images at different scales according to the weights of the gray values of pixel points in the intermediate images at different scales, as the gray values of pixel points in the uniform illumination image, thereby obtaining the uniform illumination image.

2. The intelligent inspection method for bridge piers according to claim 1, wherein The setting of Gaussian filters of different scales includes: Each odd number within the range is used as a standard deviation to construct Gaussian filters of different scales. and are the upper and lower limits of the range respectively, and .

3. The intelligent inspection method for bridge piers according to claim 1, wherein, The calculation of the enhancement degree of the gradient features of pixel points in the intermediate images at different scales compared to the gradient features in the original inspection images includes: ； Wherein, , is the number of types at all scales, is the enhancement degree of the gradient feature of the pixel point in the intermediate image at the -th scale compared with the gradient feature in the original inspection image, , are respectively the gradient magnitude and gradient direction of the pixel point in the intermediate image at the -th scale, , are respectively the gradient magnitude and gradient direction of the pixel point in the original inspection image, represents taking the absolute value, represents the Sigmoid function.

4. The intelligent inspection method for bridge piers according to claim 1, characterized in that, The calculation of the retention degree of the multiple LBP features of pixel points in the intermediate images at different scales compared to the multiple LBP features in the original inspection images includes: ； Wherein, , is the number of types at all scales, is the retention degree of the multiple LBP features of the pixel point in the intermediate image at the th scale compared to the multiple LBP features in the original inspection image, is the multiple LBP features of the pixel point in the intermediate image at the th scale, is the multiple LBP features of the pixel point in the original inspection image, represents the Hamming distance.

5. The intelligent inspection method for bridge piers according to claim 1, characterized in that The calculation of the weights of the gray values of pixel points in the intermediate images at different scales according to the enhancement degree and the retention degree includes: ； In the formula, , is the number of types for all scales, is the weight of the gray value of the pixel point in the intermediate image at the -th scale, is the retention degree of the multiple LBP features of the pixel point in the intermediate image at the -th scale compared to the multiple LBP features in the original inspection image, is the minimum of the retention degree, is the enhancement degree of the gradient feature of the pixel point in the intermediate image at the -th scale compared to the gradient feature in the original inspection image, means taking the maximum value, means the normalization function.

6. The intelligent inspection method for bridge piers according to claim 5, characterized in that The normalization function includes: , is the retention degree of the multiple LBP features of the pixel points in the intermediate image at the -th scale compared to the multiple LBP features in the original inspection image, is the minimum of the retention degree, is the enhancement degree of the gradient features of the pixel points in the intermediate image at the -th scale compared to the gradient features in the original inspection image, denotes taking the maximum value, is the number of all scale types.

7. The intelligent inspection method for bridge piers according to claim 1, characterized in that, The method for obtaining the multiple LBP features of pixel points includes: Constructing a neighborhood centered on the pixel point with a size of 5×5, then the obtained neighborhood includes 1 central pixel point and 24 neighborhood pixel points; Calculate the absolute value of the difference between the grayscale values of each neighborhood pixel and the central pixel, which is used as the grayscale difference between each neighborhood pixel and the central pixel ; By comparing the gray - level difference between the th neighborhood pixel and the central pixel with four sub - regions, the marking value of the neighborhood position corresponding to the th neighborhood pixel is obtained; ​ Arranging the marker values at the neighborhood positions corresponding to all neighborhood pixel points in sequence and splicing them into a string as the multiple LBP features of the pixel point.

8. The intelligent inspection method for bridge piers according to claim 7, characterized in that, The method for obtaining the 4 sub-regions includes: Set 3 demarcation values within the range respectively as , , , and ; Divide the range by 3 boundary values into 4 sub-regions, namely sub-region , sub-region , sub-region , and sub-region .

9. The intelligent inspection method for bridge piers according to claim 7, wherein The obtaining of the marking value of the neighborhood position corresponding to the th neighborhood pixel point includes: If the grayscale difference is within the sub-region , then the marking value of the neighborhood position corresponding to the th neighborhood pixel point is 00; If the grayscale difference is within the sub-region , then the marking value at the neighborhood position corresponding to the th neighborhood pixel is 01; If the grayscale difference is within the sub-region , then the marking value of the neighborhood position corresponding to the th neighborhood pixel point is 10; If the grayscale difference is within the sub-region , then the marking value at the neighborhood position corresponding to the th neighborhood pixel is 11.

10. The intelligent inspection method for bridge piers according to claim 1, wherein, The method further includes: Performing stretching display processing on the uniform illumination image, and the stretching display processing method includes 2% linear stretching and contrast parameter stretching.

Images