Intelligent identification and maintenance method and system for potholes on road and bridge surfaces based on computer vision

Abstract

The present invention relates to the field of computer vision technology, and in particular to a method and system for intelligent identification and maintenance of potholes on the surface of a road and bridge based on computer vision. The present invention processes a road and bridge surface image with position information to obtain a grayscale image; divides the grayscale image into an initial pothole area and an initial normal area according to the grayscale distribution characteristics of the pixels in the grayscale image; obtains the filtered median corresponding to each target pixel in the initial pothole area and the filtered median corresponding to each target pixel in the initial normal area, respectively, to obtain a smoothed image corresponding to median filtering of the grayscale image; obtains the pothole area in the smoothed image according to the grayscale value difference of each pixel in the smoothed image, and maintains the road and bridge surface corresponding to the pothole area according to the position information of the road and bridge surface image. The present invention makes the contrast between the normal area and the pothole area obvious, thereby accurately performing pothole segmentation on the road and bridge surface image containing the pothole area.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and in particular to a method and system for intelligently identifying and maintaining potholes on the surface of a road or bridge based on computer vision. Background Art

[0002] Potholes are common local damages on asphalt pavements. They affect the stability and safety of driving and may cause damage to the vehicle suspension system and tires. If not handled in time, they will shorten the service life of the pavement and accelerate pavement aging and damage. At the same time, potholes on the road and bridge surface may also cause inconvenience to pedestrians. Therefore, it is necessary to identify potholes on the road and bridge surface in a timely and accurate manner and then maintain them. To identify potholes on the road and bridge surface, it is necessary to segment the road and bridge surface image. However, due to the characteristics of asphalt itself, many spots similar to salt and pepper noise will appear in the road and bridge surface image, affecting the pothole segmentation effect.

[0003] In order to solve the above problems, the traditional method uses median filtering to effectively remove the salt and pepper noise generated by image segmentation, thereby better smoothing the spots similar to salt and pepper noise in the image. However, the median filter selects the median point in the window to replace the grayscale value of the target pixel point, and cannot distinguish the normal area from the pit area well, resulting in difficulty in pit segmentation.

[0004] Therefore, how to distinguish the normal area and the pothole area in the road and bridge surface image so as to accurately segment the pothole becomes an urgent problem to be solved. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a method and system for intelligent identification and maintenance of potholes on the surface of a road or bridge based on computer vision, so as to solve the problem of how to distinguish between normal areas and pothole areas in road or bridge surface images, thereby accurately performing pothole segmentation.

[0006] In a first aspect, an embodiment of the present invention provides a method for intelligently identifying and maintaining potholes on a road or bridge surface based on computer vision, the method comprising the following steps:

[0007] Collecting a road bridge surface image with position information, and processing the road bridge surface image to obtain a grayscale image;

[0008] Recording any pixel in the grayscale image as a target pixel, obtaining the abnormality of each target pixel in the grayscale image according to the grayscale distribution characteristics of the pixel in the grayscale image, and dividing the grayscale image into an initial pit area and an initial normal area according to the abnormality of each target pixel in the grayscale image;

[0009] Taking each of the target pixels as the center, respectively establish a filter window of a preset size in the grayscale image, and obtain a filter median corresponding to each target pixel in the initial pit area according to the grayscale distribution characteristics of the pixel points in the filter window corresponding to each target pixel in the initial pit area; and obtain a filter median corresponding to each target pixel in the initial normal area according to the grayscale distribution characteristics of the pixel points in the filter window corresponding to each target pixel in the initial normal area;

[0010] Obtaining a smoothed image corresponding to median filtering of the grayscale image according to a filtered median value corresponding to each target pixel point in the initial pit area and a filtered median value corresponding to each target pixel point in the initial normal area;

[0011] According to the gray value difference of each pixel point in the smoothed image, a pothole area in the smoothed image is obtained, and the road and bridge surface corresponding to the pothole area is maintained according to the position information of the road and bridge surface image.

[0012] Preferably, obtaining the abnormality degree of each target pixel in the grayscale image according to the grayscale distribution characteristics of the pixel in the grayscale image includes:

[0013] Taking the target pixel point as the center, establishing a first target window of a first preset size, and acquiring the complexity of the target pixel point according to the grayscale distribution characteristics of the pixel points in the first target window;

[0014] Taking the target pixel point as the center, establishing a second target window of a second preset size, and acquiring the local similarity of the target pixel point according to the grayscale distribution characteristics of the pixel points in the second target window, wherein the second target window contains the first target window;

[0015] The abnormality degree of the target pixel is obtained according to the complexity and local similarity of the target pixel.

[0016] Preferably, acquiring the complexity of the target pixel point according to the grayscale distribution characteristics of the pixel point in the first target window includes:

[0017] Obtaining a maximum grayscale value and a minimum grayscale value of pixels in the first target window, and calculating an absolute value of a difference between the maximum grayscale value and the minimum grayscale value, which is recorded as a first absolute value of the difference;

[0018] Obtaining the absolute value of the difference between the grayscale value of each pixel point in the first target window and the target pixel point respectively, obtaining a difference absolute value sequence, and obtaining the standard deviation of the difference absolute value sequence;

[0019] The grayscale values ​​of all pixels in the first target window are combined into a first grayscale value sequence to obtain information entropy of the first grayscale value sequence;

[0020] The addition result of the first difference absolute value, the standard deviation and the information entropy is obtained, recorded as a first addition result, and the first addition result is normalized to obtain the complexity of the target pixel in the first target window.

[0021] Preferably, acquiring the local similarity of the target pixel points according to the grayscale distribution characteristics of the pixel points in the second target window includes:

[0022] On the diagonal line of the second target window, pixel points separated from the target pixel point by a preset length are obtained respectively, recorded as neighborhood pixel points, and with each neighborhood pixel point as the center, a first preset size of neighborhood small window is established in the second target window respectively, and the complexity of the target pixel point in the second target window and the complexity of each neighborhood pixel point in the corresponding neighborhood small window are obtained respectively;

[0023] Obtain the absolute value of the difference between the complexity of each neighborhood pixel in the corresponding neighborhood small window and the complexity of the target pixel in the first target window, and obtain a difference absolute value sequence, recorded as a second difference absolute value sequence; obtain the ratio of each difference absolute value in the second difference absolute value sequence to the complexity of the target pixel in the second target window, and obtain the corresponding mean of the ratios; obtain the addition result of the mean of the ratios and a preset value, recorded as a second addition result; normalize the inverse of the second addition result to obtain the local similarity of the target pixel.

[0024] Preferably, obtaining the abnormality degree of the target pixel point according to the complexity and local similarity of the target pixel point includes:

[0025] The sum of the inverse of the complexity of the target pixel and the local similarity of the target pixel is obtained, recorded as a third addition result, and the third addition result is normalized to obtain the abnormality degree of the target pixel.

[0026] Preferably, dividing the grayscale image into an initial pit area and an initial normal area according to the abnormality degree of each target pixel in the grayscale image comprises:

[0027] An abnormality degree threshold is set, and all target pixel points in the grayscale image whose abnormality degree is greater than the abnormality degree threshold are obtained to form an initial pit area in the grayscale image. All target pixel points in the grayscale image whose abnormality degree is less than or equal to the abnormality degree threshold are obtained to form an initial normal area in the grayscale image.

[0028] Preferably, obtaining the filtered median corresponding to each target pixel point in the initial pit area according to the grayscale distribution characteristics of the pixel points in the filtering window corresponding to each target pixel point in the initial pit area comprises:

[0029] For any target pixel point in the initial pit area, record the filtering window corresponding to the any target pixel point as a first filtering window, obtain the gray value of each pixel point in the first filtering window, sort the gray values ​​of all pixels in the first filtering window, and obtain the sorted median of the gray values ​​in the first filtering window, which is recorded as a first sorted median;

[0030] Obtaining the probability of occurrence of the grayscale value of each pixel in the first filtering window, and taking the grayscale value with the highest probability of occurrence in the first filtering window as the recommended median value of the first filtering window, recorded as the first recommended median value;

[0031] A first difference threshold is set, and the absolute value of the difference between the first sorted median and the first recommended median is obtained, recorded as the second difference absolute value, and the second difference absolute value is normalized to obtain a first normalized value. If the first normalized value is less than the first difference threshold, the first sorted median is used as the filtered median corresponding to any target pixel point; if the first normalized value is greater than or equal to the first difference threshold, the first recommended median is used as the filtered median corresponding to any target pixel point.

[0032] Preferably, obtaining the filtered median corresponding to each target pixel in the initial normal area according to the grayscale distribution characteristics of the pixel in the filtering window corresponding to each target pixel in the initial normal area includes:

[0033] For any target pixel in the initial normal area, record the filtering window corresponding to the any target pixel as a second filtering window, obtain the gray value of each pixel in the second filtering window, sort the gray values ​​of all pixels in the second filtering window, and obtain the sorted median of the gray values ​​in the second filtering window, which is recorded as the second sorted median;

[0034] In the grayscale image, obtaining a filtered median value of a pixel point belonging to the initial pit region that is closest to any pixel point, and recording the filtered median value as a second recommended median value;

[0035] Setting a second difference threshold, obtaining an absolute value of a difference between the second recommended median and the second sorted median, recording the difference as a third absolute value of the difference, normalizing the third absolute value of the difference to obtain a second normalized value, and if the second normalized value is greater than or equal to the second difference threshold, using the second sorted median as the filtered median corresponding to any target pixel;

[0036] If the second normalized value is less than the second difference threshold, the probability of occurrence of the grayscale value of each pixel in the second filtering window is obtained, and the grayscale value with the highest probability of occurrence in the second filtering window is used as the recommended median of the second filtering window, recorded as the third recommended median, and the absolute value of the difference between the second recommended median and the third recommended median is obtained, recorded as the fourth absolute value of the difference, and the fourth absolute value of the difference is normalized to obtain a third normalized value. When the third normalized value is greater than or equal to the second difference threshold, the third recommended median is used as the filtering median corresponding to any target pixel;

[0037] When the third normalized value is less than the second difference threshold, the grayscale value with the second highest probability of appearing in the second filtering window is taken as the new third recommended median, and the method of obtaining the third normalized value is repeated until the third normalized value is greater than or equal to the second difference threshold, and the new third recommended median is taken as the filtering median corresponding to any target pixel point.

[0038] Preferably, the step of obtaining the pit area in the smoothed image according to the gray value difference of each pixel in the smoothed image comprises:

[0039] The grayscale value of each pixel in the smoothed image is obtained, and a first grayscale value threshold and a second grayscale value threshold are set respectively, wherein the first grayscale value threshold is less than the second grayscale value threshold, and all pixel points in the smoothed image whose grayscale values ​​are less than the first grayscale threshold or greater than the second grayscale threshold are obtained to form a pit area in the smoothed image.

[0040] In a second aspect, an embodiment of the present invention further provides a system for intelligent identification and maintenance of potholes on the surface of roads and bridges based on computer vision, comprising a memory and a processor, wherein the processor executes a computer program stored in the memory to implement the method for intelligent identification and maintenance of potholes on the surface of roads and bridges based on computer vision as described in the first aspect.

[0041] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0042] The present invention collects a surface image of a road or bridge with position information, processes the surface image of the road or bridge to obtain a grayscale image; records any pixel point in the grayscale image as a target pixel point, obtains the abnormality of each target pixel point in the grayscale image according to the grayscale distribution characteristics of the pixel points in the grayscale image, and divides the grayscale image into an initial pit area and an initial normal area according to the abnormality of each target pixel point in the grayscale image; takes each of the target pixel points as the center, establishes a filter window of a preset size in the grayscale image, obtains the initial pit area according to the grayscale distribution characteristics of the pixel points in the filter window corresponding to each target pixel point in the initial pit area, and obtains the initial normal area according to the grayscale distribution characteristics of the pixel points in the filter window corresponding to each target pixel point in the initial pit area. The filtering median corresponding to each target pixel in the pothole area; according to the grayscale distribution characteristics of the pixel points in the filtering window corresponding to each target pixel in the initial normal area, the filtering median corresponding to each target pixel in the initial normal area is obtained; according to the filtering median corresponding to each target pixel in the initial pothole area and the filtering median corresponding to each target pixel in the initial normal area, a smoothed image corresponding to the median filtering of the grayscale image is obtained; according to the grayscale value difference of each pixel in the smoothed image, the pothole area in the smoothed image is obtained, and the road bridge surface corresponding to the pothole area is maintained according to the position information of the road bridge surface image. The present invention first divides the grayscale image into an initial pothole area and an initial normal area according to the abnormality of each target pixel in the grayscale image, adaptively obtains the filtering median corresponding to each pixel in different areas, performs median filtering on the grayscale image, and obtains the corresponding smoothed image, which can not only effectively remove the salt and pepper noise generated by image segmentation, thereby better smoothing the spots similar to the salt and pepper noise in the image, but also makes the contrast between the normal area and the pothole area more obvious, so as to accurately perform pothole segmentation on the road bridge surface image containing the pothole area. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0044] Figure 1 This is a method flow chart of a method for intelligently identifying and maintaining potholes on a road or bridge surface based on computer vision provided in the first embodiment of the present invention;

[0045] Figure 2 A grayscale image of a road or bridge surface provided in the first embodiment of the present invention;

[0046] Figure 3It is a schematic diagram of a second target window provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0047] Embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, but should not be understood as limiting the present disclosure.

[0048] It should be noted that the terms "first", "second", etc. in the specification of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0049] In order to illustrate the technical solution of the present invention, specific embodiments are provided below for illustration.

[0050] See also Figure 1 , is a method flow chart of a method for intelligently identifying and maintaining potholes on a road and bridge surface based on computer vision provided in Embodiment 1 of the present invention, such as Figure 1 As shown, the method may include:

[0051] Step S101, collecting a road or bridge surface image with position information, and processing the road or bridge surface image to obtain a grayscale image.

[0052] The vehicle-mounted road surface shooting system is a road and bridge surface image acquisition system that can capture road and bridge surface images in real time. It uses the vehicle-mounted camera to capture road and bridge surface images in real time during driving, receives satellite signals through the GPS receiver, obtains the vehicle's real-time geographic location information, and associates the geographic location information with the captured road and bridge surface images to obtain road and bridge surface images with location information.

[0053] The vehicle-mounted road surface photography system is used to collect the road and bridge surface image with location information, and OpenCV is used to read and convert it into a grayscale image to obtain a grayscale image of the road and bridge surface, such as Figure 2 As shown, Numpy is used to extract the grayscale value of each pixel in the grayscale image of the road and bridge surface, wherein OpenCV and Numpy belong to the prior art and will not be described in detail here.

[0054] Step S102, record any pixel point in the grayscale image as a target pixel point, obtain the abnormality degree of each target pixel point in the grayscale image according to the grayscale distribution characteristics of the pixel points in the grayscale image, and divide the grayscale image into an initial pit area and an initial normal area according to the abnormality degree of each target pixel point in the grayscale image.

[0055] In order to remove spots similar to salt and pepper noise that appear when performing pothole segmentation on road and bridge surface images, it is necessary to perform median filtering on the grayscale image. However, the median filter selects the median point in the window to replace the grayscale value of the target pixel, and cannot distinguish the normal area from the pothole area well, resulting in difficulty in pothole segmentation. Therefore, it is necessary to first divide the grayscale image into an initial pothole area and an initial normal area according to the grayscale distribution characteristics of the pixels in the grayscale image, and then perform median filtering on the initial pothole area and the initial normal area respectively to obtain a pothole area and a normal area with obvious contrast.

[0056] The pothole part (the pothole part includes the inner area of ​​the pothole and the edge area of ​​the pothole) is a block-shaped depression formed when the road surface is affected by some external influence. Its complexity is lower than that of the background part, that is, the normal road surface part, that is, the grayscale range, grayscale change, and grayscale non-uniformity of the pothole part are smaller than those of the normal road surface part. Moreover, potholes generally appear in block areas, that is, pixels with similar complexity are concentrated in the pothole part, which is specifically manifested as high local similarity. Therefore, the degree of abnormality of each pixel in the grayscale image can be obtained through the complexity and local similarity of each pixel, and then the grayscale image can be distinguished through the degree of abnormality of each pixel in the grayscale image to obtain the initial pothole area and the initial normal area.

[0057] Any pixel in the grayscale image is recorded as the target pixel. The method for obtaining the abnormality degree of the target pixel in the grayscale image through the complexity and local similarity of each pixel in the grayscale image is as follows:

[0058] (1) Establishing a first target window of a first preset size with the target pixel as the center, and acquiring the complexity of the target pixel according to the grayscale distribution characteristics of the pixel in the first target window.

[0059] Specifically, the maximum grayscale value and the minimum grayscale value of the pixel points in the first target window are obtained, and the absolute value of the difference between the maximum grayscale value and the minimum grayscale value is calculated, which is recorded as the first absolute value of the difference;

[0060] Obtaining the absolute value of the difference between the grayscale value of each pixel point in the first target window and the target pixel point respectively, obtaining a difference absolute value sequence, and obtaining the standard deviation of the difference absolute value sequence;

[0061] The grayscale values ​​of all pixels in the first target window are combined into a first grayscale value sequence to obtain information entropy of the first grayscale value sequence;

[0062] The addition result of the first difference absolute value, the standard deviation and the information entropy is obtained, recorded as a first addition result, and the first addition result is normalized to obtain the complexity of the target pixel in the first target window.

[0063] In one embodiment, taking the target pixel point A as an example, with the target pixel point A as the center, an n×n first target window is established. In this embodiment, the value of n is 3, which is not limited here and can be set according to the specific implementation scenario. First, the maximum grayscale value and the minimum grayscale value of the pixels in the first target window are obtained; then the absolute value of the difference between the grayscale value of each pixel point in the first target window and the grayscale value of the target pixel point A is obtained respectively, forming a difference absolute value sequence, and obtaining the standard deviation of the difference absolute value sequence; then the grayscale values ​​of all pixels in the first target window are obtained, and the grayscale values ​​of all pixels in the first target window are formed into a first grayscale value sequence, and the information entropy of the first grayscale value sequence is obtained. The information entropy belongs to the prior art and will not be repeated here. Finally, according to the maximum grayscale value and the minimum grayscale value of the pixels in the first target window, the standard deviation of the difference absolute value sequence, and the information entropy of the first grayscale value sequence, the complexity of the target pixel point A in the first target window is calculated:

[0064]

[0065] Where d is the complexity of the target pixel A in the first target window; is the maximum grayscale value of the pixels in the first target window; is the minimum grayscale value of the pixels in the first target window; is the standard deviation of the difference absolute value sequence; H is the information entropy of the first gray value sequence; norm() is the normalization function; | | is the absolute value symbol.

[0066] It should be noted that, the smaller the difference between the maximum grayscale value and the minimum grayscale value of the pixel in the first target window, the smaller the range of the grayscale value in the first target window, and the smaller the complexity of the target pixel A in the first target window; the smaller the standard deviation of the difference absolute value sequence, the smaller the data fluctuation of the difference absolute value sequence, that is, the smaller the fluctuation of the grayscale value difference in the first target window, and the smaller the complexity of the target pixel A in the first target window; the smaller the information entropy of the first grayscale value sequence, the simpler the distribution of the grayscale value in the first target window, that is, the lower the grayscale non-uniformity, and the smaller the complexity of the target pixel A in the first target window.

[0067] (2) establishing a second target window of a second preset size with the target pixel as the center, and obtaining the local similarity of the target pixel according to the grayscale distribution characteristics of the pixel in the second target window, wherein the second target window contains the first target window.

[0068] Specifically, on the diagonal line of the second target window, pixel points separated from the target pixel point by a preset length are obtained respectively, recorded as neighborhood pixel points, and with each neighborhood pixel point as the center, a first preset size of neighborhood small window is established in the second target window respectively, and the complexity of the target pixel point in the second target window and the complexity of each neighborhood pixel point in the corresponding neighborhood small window are obtained respectively;

[0069] Obtain the absolute value of the difference between the complexity of each neighborhood pixel in the corresponding neighborhood small window and the complexity of the target pixel in the first target window, and obtain a difference absolute value sequence, recorded as a second difference absolute value sequence; obtain the ratio of each difference absolute value in the second difference absolute value sequence to the complexity of the target pixel in the second target window, and obtain the corresponding mean of the ratios; obtain the addition result of the mean of the ratios and a preset value, recorded as a second addition result; normalize the inverse of the second addition result to obtain the local similarity of the target pixel.

[0070] In one implementation, taking the target pixel point A as an example, a second target window of (2n+1)×(2n+1) is established with the target pixel point A as the center. In this implementation, the value of n is 3, that is, the size of the second target window is 7×7. There is no limitation here. It can be set according to the specific implementation scenario. Among the pixels located on the two diagonals of the second target window, the pixels spaced from the target pixel point A are respectively obtained. Pixels (i.e., 1 pixel apart) are recorded as neighborhood pixels, and the pixels that are 1 pixel apart from the target pixel A are recorded as neighborhood pixels, and 4 neighborhood pixels are obtained. With each neighborhood pixel as the center, 3×3 neighborhood small windows are established in the second target window, such as Figure 3As shown, it is a schematic diagram of the second target window, the target pixel A is the central pixel of the second target window, pixel 1, pixel 2, pixel 3, pixel 4 are the neighborhood pixels separated by 1 pixel from the target pixel A, and the light gray areas are 3×3 neighborhood small windows established with each neighborhood pixel as the center. According to the above method for obtaining the complexity of the target pixel A in the first target window, the complexity of each neighborhood pixel in its neighborhood small window and the complexity of the target pixel A in the second target window are obtained respectively. According to the complexity of each neighborhood pixel in its neighborhood small window, the complexity of the target pixel A in the first target window, and the complexity of the target pixel A in the second target window, the local similarity of the target pixel A is calculated:

[0071]

[0072] in, is the local similarity of the target pixel A; d is the complexity of the target pixel A in the first target window; is the complexity of the bth neighborhood pixel in its neighborhood small window; b is the serial number of the neighborhood pixel in the second target window; k is the number of neighborhood pixels in the second target window; D is the complexity of the target pixel A in the second target window; c is the preset value; norm() is the normalization function; | | is the absolute value symbol.

[0073] It should be noted that the preset value c is a constant 2, which is used to prevent the formula calculation from being wrong when the denominator is taken as 0; the smaller the difference between the complexity of each neighborhood pixel point in its neighborhood small window and the complexity of the target pixel point A in the first target window, the more similar the distribution of grayscale values ​​in the neighborhood small window in the second target window is to that in the first target window, and the greater the local similarity of the target pixel point A.

[0074] (3) Obtaining the abnormality degree of the target pixel point based on the complexity and local similarity of the target pixel point.

[0075] Specifically, the sum of the inverse of the complexity of the target pixel and the local similarity of the target pixel is obtained, recorded as a third addition result, and the third addition result is normalized to obtain the abnormality degree of the target pixel.

[0076] In one implementation, taking target pixel point A as an example, the abnormality degree of target pixel point A is calculated:

[0077]

[0078] in, is the abnormality degree of the target pixel A; d is the complexity of the target pixel A in the first target window; is the local similarity of the target pixel A; norm() is the normalization function.

[0079] It should be noted that the lower the complexity of the target pixel point A in the first target window, the more the target pixel point A conforms to the pixel point in the pit area, and the higher the abnormality of the target pixel point A; the greater the local similarity of the target pixel point A, the more the target pixel point A conforms to the pixel point in the pit area, and the higher the abnormality of the target pixel point A.

[0080] According to the method for obtaining the abnormality of the target pixel A, the abnormality of each target pixel in the grayscale image is obtained. After the abnormality of each target pixel in the grayscale image is obtained, the grayscale image can be distinguished by the abnormality of each target pixel in the grayscale image to obtain the initial pit area and the initial normal area.

[0081] Specifically, an abnormality degree threshold is set, and all target pixel points in the grayscale image whose abnormality degree is greater than the abnormality degree threshold are obtained to form an initial pit area in the grayscale image; and all target pixel points in the grayscale image whose abnormality degree is less than or equal to the abnormality degree threshold are obtained to form an initial normal area in the grayscale image.

[0082] In one embodiment, the abnormality threshold is set to 0.75, which is not limited here. It can be set according to the specific implementation scenario, and all target pixel points with an abnormality greater than 0.75 in the grayscale image are obtained, and all target pixel points with an abnormality greater than 0.75 in the grayscale image are composed of an initial pit area in the grayscale image. All target pixel points with an abnormality less than or equal to 0.75 in the grayscale image are obtained, and all target pixel points with an abnormality less than or equal to 0.75 in the grayscale image are composed of an initial normal area in the grayscale image.

[0083] At this point, the initial pit area and the initial normal area in the grayscale image are obtained.

[0084] Step S103, with each of the target pixels as the center, respectively establish a filter window of a preset size in the grayscale image, and obtain the filter median corresponding to each target pixel in the initial pit area according to the grayscale distribution characteristics of the pixel points in the filter window corresponding to each target pixel in the initial pit area; obtain the filter median corresponding to each target pixel in the initial normal area according to the grayscale distribution characteristics of the pixel points in the filter window corresponding to each target pixel in the initial normal area.

[0085] After obtaining the initial pit area and the initial normal area in the grayscale image, adaptive median filtering is performed on the initial pit area and the initial normal area respectively, so as to obtain a smooth image corresponding to the median filtering of the grayscale image, so as to enhance the contrast between the pit area and the normal area in the grayscale image, so that the pit segmentation can be accurately performed.

[0086] When performing median filtering on a grayscale image, you first need to establish a filter window. Take each target pixel in the grayscale image as the center and establish a 5×5 filter window. There is no restriction here and it can be set according to the specific implementation scenario.

[0087] When performing median filtering on the initial pit area, there will also be some interference spots in the initial pit area, but the number is relatively small. In order to prevent this type of pixel points from being used as the filtering median, it is necessary to obtain the sorted median of each filtering window, and count the probability of occurrence of the grayscale value of each pixel point in each filtering window. According to the difference between the grayscale value with the highest probability of appearing in each filtering window and the sorted median of the corresponding filtering window, the filtering median corresponding to each target pixel point in the initial pit area is obtained.

[0088] Among them, according to the difference between the gray value with the highest probability of appearing in each filtering window and the sorted median value of the corresponding filtering window, the method for obtaining the filtering median value corresponding to each target pixel point in the initial pit area is:

[0089] For any target pixel point in the initial pit area, record the filtering window corresponding to the any target pixel point as a first filtering window, obtain the gray value of each pixel point in the first filtering window, sort the gray values ​​of all pixels in the first filtering window, and obtain the sorted median of the gray values ​​in the first filtering window, which is recorded as a first sorted median;

[0090] Obtaining the probability of occurrence of the grayscale value of each pixel in the first filtering window, and taking the grayscale value with the highest probability of occurrence in the first filtering window as the recommended median value of the first filtering window, recorded as the first recommended median value;

[0091] A first difference threshold is set, and the absolute value of the difference between the first sorted median and the first recommended median is obtained, recorded as the second difference absolute value, and the second difference absolute value is normalized to obtain a first normalized value. If the first normalized value is less than the first difference threshold, the first sorted median is used as the filtered median corresponding to any target pixel point; if the first normalized value is greater than or equal to the first difference threshold, the first recommended median is used as the filtered median corresponding to any target pixel point.

[0092] In one embodiment, taking the target pixel point I in the initial pit area as an example, a 5×5 first filtering window is established with the target pixel point I as the center, firstly, the grayscale values ​​of all the pixels in the first filtering window are obtained, the grayscale values ​​of all the pixels in the first filtering window are sorted, and the median of the grayscale values ​​of the pixels in the first filtering window is selected as the sorted median of the first filtering window, and the sorted median of the first filtering window is recorded as the first sorted median; then, the probability of occurrence of the grayscale value of each pixel in the first filtering window is obtained, and the grayscale value with the highest probability of occurrence of the sorted median of the first filtering window is taken as the recommended median of the first filtering window, and the recommended median of the first filtering window is recorded as the first recommended median; finally, the first difference threshold is set to 0.6, which is not limited here and can be set according to the specific implementation scenario, and the filtering median corresponding to the target pixel point I is calculated according to the difference between the first sorted median, the first recommended median and the first difference threshold:

[0093]

[0094] Among them, Lzi is the filtered median corresponding to the target pixel I; Zi is the first sorted median; is the first recommended median; g is the first difference threshold; norm() is the normalization function; | | is the absolute value symbol.

[0095] It should be noted that the greater the difference between the first sorted median and the first recommended median, the more likely it is that the first sorted median is an interference pixel, so the first recommended median is used as the filtered median corresponding to the target pixel I; the smaller the difference between the first sorted median and the first recommended median, the less likely it is that the first sorted median is an interference pixel, so the first sorted median is used as the filtered median corresponding to the target pixel I.

[0096] According to the method for obtaining the filtered median corresponding to the target pixel point I, the filtered median corresponding to each target pixel point in the initial pit area is obtained.

[0097] When performing median filtering on the initial normal area, in order to prevent the grayscale value similar to the grayscale value of the target pixel in the initial pit area from being used as the filtering median, the filtering median of the target pixel in the initial normal area should be as different as possible from the filtering median of the target pixel in the initial pit area. Therefore, it is necessary to adjust the filtering median corresponding to each target pixel in the initial normal area by analyzing the difference between the sorted median of each filtering window in the initial normal area and the filtering median corresponding to the target pixel in the initial pit area closest to it. Among them, the method for obtaining the filtering median corresponding to each target pixel in the initial normal area is:

[0098] (1) For any target pixel in the initial normal area, record the filtering window corresponding to the any target pixel as the second filtering window, obtain the gray value of each pixel in the second filtering window, sort the gray values ​​of all pixels in the second filtering window, and obtain the sorted median of the gray values ​​in the second filtering window, which is recorded as the second sorted median;

[0099] In the grayscale image, obtaining a filtered median value of a pixel point belonging to the initial pit region that is closest to any pixel point, and recording the filtered median value as a second recommended median value;

[0100] Setting a second difference threshold, obtaining an absolute value of a difference between the second recommended median and the second sorted median, recording the difference as a third absolute value of the difference, normalizing the third absolute value of the difference to obtain a second normalized value, and if the second normalized value is greater than or equal to the second difference threshold, using the second sorted median as the filtered median corresponding to any target pixel;

[0101] If the second normalized value is less than the second difference threshold, obtain the probability of occurrence of the grayscale value of each pixel in the second filtering window, take the grayscale value with the highest probability of appearing in the second filtering window as the recommended median of the second filtering window, record it as the third recommended median, obtain the absolute value of the difference between the second recommended median and the third recommended median, record it as the fourth difference absolute value, normalize the fourth difference absolute value to obtain a third normalized value, and when the third normalized value is greater than or equal to the second difference threshold, take the third recommended median as the filtering median corresponding to any target pixel.

[0102] In one embodiment, taking the target pixel point J in the initial normal area as an example, a 5×5 second filtering window is established with the target pixel point J as the center, firstly, the grayscale values ​​of all the pixels in the second filtering window are obtained, the grayscale values ​​of all the pixels in the second filtering window are sorted, and the median of the grayscale values ​​of the pixels in the second filtering window is selected as the sorted median of the second filtering window, and the sorted median of the second filtering window is recorded as the second sorted median; then, the pixel point in the grayscale image that is closest to the target pixel point J and belongs to the initial pit area is selected, and the pixel point in the grayscale image that is closest to the target pixel point J and belongs to the initial pit area is obtained. The corresponding filtered median, the filtered median corresponding to the pixel point belonging to the initial pit area that is closest to the target pixel point J in the grayscale image is recorded as the second recommended median; the probability of occurrence of the grayscale value of each pixel point in the second filtering window is obtained, and the grayscale value with the largest probability of appearing in the second filtering window is taken as the recommended median of the second filtering window, and the recommended median of the second filtering window is recorded as the third recommended median; finally, the second difference threshold is set to 0.6, which is not limited here and can be set according to the specific implementation scenario. According to the second sorted median, the second recommended median, the third recommended median and the second difference threshold, the filtered median corresponding to the target pixel point J is calculated:

[0103]

[0104] Among them, Lzj is the filtered median corresponding to the target pixel point J; Zj is the second sorted median; is the third recommended median value; is the second recommended median; f is the second difference threshold; norm() is the normalization function; | | is the absolute value symbol.

[0105] It should be noted that, the greater the difference between the second recommended median and the second sorted median, the less similar the filtered median corresponding to the target pixel in the initial normal area is to the filtered median corresponding to the target pixel in the initial pit area, so the second sorted median is used as the filtered median corresponding to the target pixel J; the smaller the difference between the second recommended median and the second sorted median, the more similar the filtered median corresponding to the target pixel in the initial normal area is to the filtered median corresponding to the target pixel in the initial pit area, so the third recommended median with a large difference from the second recommended median is used as the filtered median corresponding to the target pixel J.

[0106] (2) When the third normalized value is less than the second difference threshold, the grayscale value with the second highest probability of appearing in the second filtering window is used as the new third recommended median, and the method for obtaining the third normalized value is repeated until the third normalized value is greater than or equal to the second difference threshold, and the new third recommended median is used as the filtering median corresponding to any target pixel.

[0107] In one embodiment, when the third normalized value is less than the second difference threshold value of 0.6, the grayscale value with the second highest probability of appearing in the second filtering window is used as the new third recommended median, and the new third normalized value is obtained according to the method for obtaining the third normalized value. If the new third normalized value is greater than or equal to the second difference threshold value of 0.6, the new third recommended median (i.e., the grayscale value with the second highest probability of appearing in the second filtering window) is used as the filtering median corresponding to the target pixel point J; if the new third normalized value is less than the second difference threshold value of 0.6, the grayscale value with the third highest probability of appearing in the second filtering window is used as the new third recommended median, and the new third normalized value is continued to be obtained according to the method for obtaining the third normalized value, and so on, until the new third normalized value is greater than or equal to the second difference threshold value of 0.6, and the new third recommended median corresponding to the new third normalized value is used as the filtering median corresponding to the target pixel point J.

[0108] According to the method for obtaining the filtered median corresponding to the target pixel point J, the filtered median corresponding to each target pixel point in the initial normal area is obtained.

[0109] At this point, the filtered median corresponding to each target pixel point in the initial pit area and the filtered median corresponding to each target pixel point in the initial normal area are obtained.

[0110] Step S104, obtaining a smoothed image corresponding to median filtering of the grayscale image according to the filtered median corresponding to each target pixel in the initial pit area and the filtered median corresponding to each target pixel in the initial normal area.

[0111] According to the filtered median corresponding to each target pixel in the initial pit area and the filtered median corresponding to each target pixel in the initial normal area, the filtered median corresponding to each target pixel in the grayscale image is obtained, and the filtered median corresponding to each target pixel in the grayscale image is used to replace the grayscale value of the corresponding target pixel in the grayscale image, so as to obtain a smooth image with obvious contrast between the pit area and the normal area.

[0112] At this point, the corresponding smoothed image after median filtering of the grayscale image is obtained.

[0113] Step S105, obtaining a pothole area in the smoothed image according to the gray value difference of each pixel point in the smoothed image, and maintaining the road bridge surface corresponding to the pothole area according to the position information of the road bridge surface image.

[0114] In the smoothed image corresponding to the median filtering of the grayscale image, the contrast between the pit area and the normal area is more obvious, and the pit area in the smoothed image can be obtained according to the gray value difference of each pixel in the smoothed image.

[0115] Specifically, the grayscale value of each pixel in the smoothed image is obtained, and a first grayscale value threshold and a second grayscale value threshold are set respectively, the first grayscale value threshold is less than the second grayscale value threshold, and all pixel points in the smoothed image with grayscale values ​​less than the first grayscale threshold or with grayscale values ​​greater than the second grayscale threshold are obtained to form a pit area in the smoothed image.

[0116] In one embodiment, the gray value of each pixel in the smoothed image, as well as the maximum gray value and the minimum gray value of the pixel in the smoothed image are obtained. Since the deep pothole area is generally exposed to the underlying cement roadbed, the gray value in the smoothed image is higher, while the shallow pothole is generally a small pit on the surface of the asphalt road surface, and the gray value in the smoothed image is lower, the first gray value threshold is set to The minimum gray value of the pixel in the smoothed image is set to the second gray value threshold. The maximum grayscale value of the pixel points in the smoothed image is multipled, the pixel points in the smoothed image whose grayscale values ​​are less than the first grayscale value threshold are recorded as the pixel points of the shallow pit area, and the pixel points in the smoothed image whose grayscale values ​​are greater than the second grayscale value threshold are recorded as the pixel points of the deep pit area, and all the pixel points of the shallow pit area and all the pixel points of the deep pit area in the smoothed image are obtained to form the pit area in the smoothed image.

[0117] After obtaining the pothole area in the smoothed image, since the road and bridge surface image carries location information, the location information of the road and bridge surface image is matched with the actual road map to obtain the actual road and bridge section where the pothole area is located. According to the actual road and bridge section where the pothole area is located, the potholes in the actual road and bridge section are maintained.

[0118] Since the on-board road surface photography system can collect road and bridge surface images in real time, according to the above-mentioned method for obtaining pothole areas, each road and bridge surface image is analyzed, the road and bridge surface images containing pothole areas are marked, and the actual road and bridge section of each road and bridge surface image containing pothole areas is obtained. According to the actual road and bridge section of each road and bridge surface image containing pothole areas, each pothole is maintained and repaired, thereby avoiding the continuous expansion of potholes and reducing the impact of potholes on the safety of vehicles and pedestrians.

[0119] In summary, the embodiment of the present invention collects a road bridge surface image with position information, processes the road bridge surface image to obtain a grayscale image; records any pixel point in the grayscale image as a target pixel point, obtains the abnormality degree of each target pixel point in the grayscale image according to the grayscale distribution characteristics of the pixel points in the grayscale image, and divides the grayscale image into an initial pit area and an initial normal area according to the abnormality degree of each target pixel point in the grayscale image; takes each of the target pixel points as the center, establishes a filter window of a preset size in the grayscale image, and obtains the initial pit area according to the grayscale distribution characteristics of the pixel points in the filter window corresponding to each target pixel point in the initial pit area. The filtering median corresponding to each target pixel in the initial pothole area; according to the grayscale distribution characteristics of the pixel points in the filtering window corresponding to each target pixel in the initial normal area, the filtering median corresponding to each target pixel in the initial normal area is obtained; according to the filtering median corresponding to each target pixel in the initial pothole area and the filtering median corresponding to each target pixel in the initial normal area, a smoothed image corresponding to the median filtering of the grayscale image is obtained; according to the grayscale value difference of each pixel in the smoothed image, the pothole area in the smoothed image is obtained, and the road bridge surface corresponding to the pothole area is maintained according to the position information of the road bridge surface image. The embodiment of the present invention first divides the grayscale image into an initial pothole area and an initial normal area according to the abnormality of each target pixel in the grayscale image, adaptively obtains the filtering median corresponding to each pixel in different areas, performs median filtering on the grayscale image, and obtains the corresponding smoothed image, which can not only effectively remove the salt and pepper noise generated by image segmentation, thereby better smoothing the spots similar to the salt and pepper noise in the image, but also makes the contrast between the normal area and the pothole area more obvious, so as to accurately perform pothole segmentation on the road bridge surface image containing the pothole area.

[0120] Based on the same inventive concept as the above method, an embodiment of the present invention also provides a road and bridge surface pothole intelligent identification and maintenance system based on computer vision, including a memory and a processor, and the processor executes a computer program stored in the memory to implement the above-mentioned road and bridge surface pothole intelligent identification and maintenance method based on computer vision.

[0121] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A computer vision-based intelligent recognition and maintenance method for potholes on road and bridge surfaces, characterized in that: The method comprises: Collecting a road bridge surface image with position information, and processing the road bridge surface image to obtain a grayscale image; Recording any pixel in the grayscale image as a target pixel, obtaining the abnormality of each target pixel in the grayscale image according to the grayscale distribution characteristics of the pixel in the grayscale image, and dividing the grayscale image into an initial pit area and an initial normal area according to the abnormality of each target pixel in the grayscale image; Taking each of the target pixels as the center, respectively establish a filter window of a preset size in the grayscale image, and obtain a filter median corresponding to each target pixel in the initial pit area according to the grayscale distribution characteristics of the pixel points in the filter window corresponding to each target pixel in the initial pit area; and obtain a filter median corresponding to each target pixel in the initial normal area according to the grayscale distribution characteristics of the pixel points in the filter window corresponding to each target pixel in the initial normal area; Obtaining a smoothed image corresponding to median filtering of the grayscale image according to a filtered median value corresponding to each target pixel point in the initial pit area and a filtered median value corresponding to each target pixel point in the initial normal area; According to the gray value difference of each pixel point in the smoothed image, a pothole area in the smoothed image is obtained, and the road and bridge surface corresponding to the pothole area is maintained according to the position information of the road and bridge surface image.

2. The computer vision-based intelligent recognition and maintenance method for potholes on road and bridge surfaces according to claim 1 is characterized in that: The step of obtaining the abnormality degree of each target pixel in the grayscale image according to the grayscale distribution characteristics of the pixel in the grayscale image includes: Taking the target pixel point as the center, establishing a first target window of a first preset size, and acquiring the complexity of the target pixel point according to the grayscale distribution characteristics of the pixel points in the first target window; Taking the target pixel point as the center, establishing a second target window of a second preset size, and acquiring the local similarity of the target pixel point according to the grayscale distribution characteristics of the pixel points in the second target window, wherein the second target window contains the first target window; The abnormality degree of the target pixel is obtained according to the complexity and local similarity of the target pixel.

3. The computer vision-based intelligent recognition and maintenance method for potholes on road and bridge surfaces according to claim 2 is characterized in that: The obtaining the complexity of the target pixel point according to the grayscale distribution characteristics of the pixel point in the first target window includes: Obtaining a maximum grayscale value and a minimum grayscale value of pixels in the first target window, and calculating an absolute value of a difference between the maximum grayscale value and the minimum grayscale value, which is recorded as a first absolute value of the difference; Obtaining the absolute value of the difference between the grayscale value of each pixel point in the first target window and the target pixel point respectively, obtaining a difference absolute value sequence, and obtaining the standard deviation of the difference absolute value sequence; The grayscale values ​​of all pixels in the first target window are combined into a first grayscale value sequence to obtain information entropy of the first grayscale value sequence; The addition result of the first difference absolute value, the standard deviation and the information entropy is obtained, recorded as a first addition result, and the first addition result is normalized to obtain the complexity of the target pixel in the first target window.

4. The computer vision-based intelligent recognition and maintenance method for potholes on road and bridge surfaces according to claim 2 is characterized in that: The acquiring the local similarity of the target pixel points according to the grayscale distribution characteristics of the pixel points in the second target window includes: On the diagonal line of the second target window, pixel points separated from the target pixel point by a preset length are obtained respectively, recorded as neighborhood pixel points, and with each neighborhood pixel point as the center, a first preset size of neighborhood small window is established in the second target window respectively, and the complexity of the target pixel point in the second target window and the complexity of each neighborhood pixel point in the corresponding neighborhood small window are obtained respectively; Obtain the absolute value of the difference between the complexity of each neighborhood pixel in the corresponding neighborhood small window and the complexity of the target pixel in the first target window, and obtain a difference absolute value sequence, recorded as a second difference absolute value sequence; obtain the ratio of each difference absolute value in the second difference absolute value sequence to the complexity of the target pixel in the second target window, and obtain the corresponding mean of the ratios; obtain the addition result of the mean of the ratios and a preset value, recorded as a second addition result; normalize the inverse of the second addition result to obtain the local similarity of the target pixel.

5. The computer vision-based intelligent recognition and maintenance method for potholes on road and bridge surfaces according to claim 2 is characterized in that: The step of obtaining the abnormality degree of the target pixel point according to the complexity and local similarity of the target pixel point comprises: The sum of the inverse of the complexity of the target pixel and the local similarity of the target pixel is obtained, recorded as a third addition result, and the third addition result is normalized to obtain the abnormality degree of the target pixel.

6. The computer vision-based intelligent recognition and maintenance method for potholes on road and bridge surfaces according to claim 1, characterized in that: The step of dividing the grayscale image into an initial pit region and an initial normal region according to the abnormality degree of each target pixel in the grayscale image comprises: An abnormality degree threshold is set, and all target pixel points in the grayscale image whose abnormality degree is greater than the abnormality degree threshold are obtained to form an initial pit area in the grayscale image. All target pixel points in the grayscale image whose abnormality degree is less than or equal to the abnormality degree threshold are obtained to form an initial normal area in the grayscale image.

7. The computer vision-based intelligent recognition and maintenance method for potholes on road and bridge surfaces according to claim 1, characterized in that: The step of obtaining the filtered median corresponding to each target pixel point in the initial pit groove area according to the grayscale distribution characteristics of the pixel points in the filtering window corresponding to each target pixel point in the initial pit groove area comprises: For any target pixel point in the initial pit area, record the filtering window corresponding to the any target pixel point as a first filtering window, obtain the gray value of each pixel point in the first filtering window, sort the gray values ​​of all pixels in the first filtering window, and obtain the sorted median of the gray values ​​in the first filtering window, which is recorded as a first sorted median; Obtaining the probability of occurrence of the grayscale value of each pixel in the first filtering window, and taking the grayscale value with the highest probability of occurrence in the first filtering window as the recommended median value of the first filtering window, recorded as the first recommended median value; A first difference threshold is set, and the absolute value of the difference between the first sorted median and the first recommended median is obtained, recorded as the second difference absolute value, and the second difference absolute value is normalized to obtain a first normalized value. If the first normalized value is less than the first difference threshold, the first sorted median is used as the filtered median corresponding to any target pixel point; if the first normalized value is greater than or equal to the first difference threshold, the first recommended median is used as the filtered median corresponding to any target pixel point.

8. The computer vision-based intelligent recognition and maintenance method for potholes on road and bridge surfaces according to claim 1, characterized in that: The step of obtaining the filtered median corresponding to each target pixel in the initial normal area according to the grayscale distribution characteristics of the pixel in the filtering window corresponding to each target pixel in the initial normal area comprises: For any target pixel in the initial normal area, record the filtering window corresponding to the any target pixel as a second filtering window, obtain the gray value of each pixel in the second filtering window, sort the gray values ​​of all pixels in the second filtering window, and obtain the sorted median of the gray values ​​in the second filtering window, which is recorded as the second sorted median; In the grayscale image, obtaining a filtered median value of a pixel point belonging to the initial pit region that is closest to any pixel point, and recording the filtered median value as a second recommended median value; Setting a second difference threshold, obtaining an absolute value of a difference between the second recommended median and the second sorted median, recording the difference as a third absolute value of the difference, normalizing the third absolute value of the difference to obtain a second normalized value, and if the second normalized value is greater than or equal to the second difference threshold, using the second sorted median as the filtered median corresponding to any target pixel; If the second normalized value is less than the second difference threshold, the probability of occurrence of the grayscale value of each pixel in the second filtering window is obtained, and the grayscale value with the highest probability of occurrence in the second filtering window is used as the recommended median of the second filtering window, recorded as the third recommended median, and the absolute value of the difference between the second recommended median and the third recommended median is obtained, recorded as the fourth absolute value of the difference, and the fourth absolute value of the difference is normalized to obtain a third normalized value. When the third normalized value is greater than or equal to the second difference threshold, the third recommended median is used as the filtering median corresponding to any target pixel; When the third normalized value is less than the second difference threshold, the grayscale value with the second highest probability of appearing in the second filtering window is taken as the new third recommended median, and the method of obtaining the third normalized value is repeated until the third normalized value is greater than or equal to the second difference threshold, and the new third recommended median is taken as the filtering median corresponding to any target pixel point.

9. The computer vision-based intelligent recognition and maintenance method for potholes on road and bridge surfaces according to claim 8, characterized in that: The step of obtaining the pit region in the smoothed image according to the gray value difference of each pixel in the smoothed image comprises: The grayscale value of each pixel in the smoothed image is obtained, and a first grayscale value threshold and a second grayscale value threshold are set respectively, wherein the first grayscale value threshold is less than the second grayscale value threshold, and all pixel points in the smoothed image whose grayscale values ​​are less than the first grayscale threshold or greater than the second grayscale threshold are obtained to form a pit area in the smoothed image.

10. A computer vision-based intelligent recognition and maintenance system for potholes on road and bridge surfaces, comprising a memory and a processor, characterized in that: The processor executes the computer program stored in the memory to implement the computer vision-based intelligent identification and maintenance method for potholes on the surface of a road or bridge as described in any one of claims 1 to 9.

Images