JP6473844B1 - Crack detection device, crack detection method, and crack detection program - Google Patents

Abstract

A crack detection device capable of accurately determining cracks in drainage pavement is provided.
A crack detection device includes a determination unit that determines whether or not a pavement road surface is drainage pavement based on a captured image of the pavement road surface, and an image processing unit that performs image processing based on a determination result of the determination unit. And a detection unit for detecting cracks on the paved road surface based on the image processing result of the image processing unit.
[Selection] Figure 1

Description

  The present disclosure relates to a paved road surface and relates to a detection method for cracks in the paved road surface.

  In order to prevent accidents, it is necessary to inspect the pavement surface for cracks.

  As a conventional inspection method, it has been performed by visual inspection, but such inspection has a problem that it takes time and cost.

  In recent years, inspection efficiency has been improved by performing inspection using a measurement vehicle equipped with a sensor such as a laser range finder or a video camera.

  Patent Document 1 proposes an on-vehicle paved road surface patrol device. According to the paved road surface patrol device, information necessary for road surface inspection can be collected safely and easily.

  Specifically, the cracks on the road surface are automatically detected by analyzing an image taken of asphalt pavement. According to this method, it is possible to detect a crack with high accuracy on a uniform basis as compared with the detection of a crack by visual observation by the user.

  Patent Document 2 discloses a method for determining the presence or absence of cracks based on, for example, a change in luminance during image analysis.

  On the other hand, various materials are used for the paved road surface, and in recent years, not the dense-graded paved road surface but the drainage paved road surface (also referred to as a drainable paved road surface) is increasing.

Patent 6294529 JP 2018-87484 A

  The drainage pavement has a structure in which a gap is provided between aggregates in order to improve drainage performance. Therefore, as in the conventional method, when the presence or absence of a crack is determined based on a change in luminance by image analysis, the gap between aggregates may be determined as a crack.

  The present disclosure has been made to solve the above-described problems, and an object thereof is to provide a crack detection device, a crack detection method, and a crack detection program capable of accurately determining cracks in drainage pavement. And

  The crack detection device according to a certain aspect, a determination unit that determines whether or not the pavement road surface is drainage pavement based on a photographed image of the pavement road surface, an image processing unit that performs image processing based on a determination result of the determination unit, And a detection unit that detects cracks on the paved road surface based on the image processing result of the image processing unit.

  Preferably, the determination unit includes a binarization processing unit that binarizes the shadow region of the captured image and the other region, a totaling unit that tabulates the area of the shadow region, and an area value totalized by the totaling unit. An area determination unit that determines whether or not the value is equal to or greater than a predetermined value.

  Preferably, the area determination unit determines that the pavement road surface is a drainage pavement surface when determining that the area value calculated by the totalization unit is equal to or greater than a predetermined value.

Preferably, the image processing unit performs a color tone correction process on the entire captured image.
Preferably, when it is determined that the paved road surface is a drainage paved road surface as a determination result of the determining unit, the image processing unit performs a color tone correction process on a shadow area of a captured image having a predetermined area or less.

  Preferably, the detection unit detects a crack based on a change in luminance of the captured image corrected by the image processing unit.

Preferably, the detection unit detects cracks on the paved road surface by machine learning.
Preferably, the captured image is divided into patch units, and a data generation unit that generates patch images and machine learning teacher data in which binary images indicating deformed portions and non-deformed portions included in the patch images are associated with each other Is further provided.

  A crack detection method according to a certain aspect, the step of determining whether or not the pavement road surface is drainable pavement based on the photographed image of the pavement road surface, the step of executing image processing based on the determination result, and the image processing result And detecting a crack on the paved road surface.

  A program for causing a computer according to a certain aspect to function as a crack detection device, the program having a determination unit for determining whether or not the paved road surface is drainable pavement based on a photographed image of the paved road surface, and a determination An image processing unit that performs image processing based on the determination result of the unit, and a detection unit that detects cracks on the paved road surface based on the image processing result of the image processing unit.

  The crack detection device, the crack detection method, and the crack detection program of the present disclosure can accurately determine a crack.

It is a figure explaining the crack detection apparatus 10 according to Embodiment 1. FIG. FIG. 5 is a diagram for describing color tone correction processing for a captured image based on Embodiment 1. It is a figure explaining the binarization process based on Embodiment 1. FIG. It is a figure explaining the image process with respect to the image data of the drainage pavement road surface in the image process part 23 according to Embodiment 1. FIG. It is a figure explaining the flow of a process of the crack detection apparatus 10 according to Embodiment 1. FIG. It is a figure explaining the crack detection apparatus 11 according to Embodiment 2. FIG. It is a figure explaining the teacher data according to Embodiment 2.

  Embodiments will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals, and description thereof will not be repeated.

(Embodiment 1)
FIG. 1 is a diagram illustrating a crack detection device 10 according to the first embodiment.

  Referring to FIG. 1, crack detection device 10 includes a CPU (Central Processing Unit) 20, a display unit 14, a memory 16, and an operation unit 18.

The CPU 20 controls the entire crack detection device 10.
The CPU 20 implements various functional blocks by executing a program stored in the memory 16.

  Specifically, the CPU 20 includes a data acquisition unit 21, a detection unit 22, an image processing unit 23, and a determination unit 25.

  The memory 16 is a storage device such as a ROM (Read Only Memory), a RAM (Random Access Memory), or a hard disk built in the computer. The memory 16 stores various programs and various data including processing programs of the data acquisition unit 21, the detection unit 22, the image processing unit 23, and the determination unit 25, and inputs / outputs these information to / from the CPU 20.

  The display unit 14 is an image display device such as a liquid crystal monitor. The operation unit 18 includes a keyboard and a mouse.

  The crack detection device 10 detects a crack based on the acquired captured image of the paved road surface.

  In this example, a case where a line sensor camera is mounted on a vehicle and a captured image obtained by capturing a pavement surface is used will be described as an example. As an example of the paved road surface, an asphalt paved road surface such as a road on which a vehicle travels will be described. However, the road surface is not limited to a road and can be applied to any road surface. Further, the image is not limited to the line sensor camera, and a captured image using another camera or the like can also be used.

  The data acquisition unit 21 acquires image data that is an image captured by the line sensor camera. The image data may be stored in the memory 16 in advance, or may be acquired from an external server via a network.

The detection unit 22 detects cracks on the paved road surface included in the captured image.
The image processing unit 23 performs predetermined image processing on the image data of the paved road surface. Specifically, the image processing unit 23 performs a color tone correction process for correcting the color tone of the image data of the paved road surface.

The determination unit 25 determines whether the pavement road surface of the captured image is a drainage pavement surface.
The determination unit 25 includes a binarization processing unit 26, a totalization unit 27, and an area determination unit 28.

  The binarization processing unit 26 executes binarization processing based on the image data of the captured image of the paved road surface.

The tabulation unit 27 tabulates the area of the shadow area based on the binarized data.
The area determination unit 28 determines whether or not the total area of the shadow area is equal to or greater than a predetermined area value, and determines that the area is a drainage pavement surface if the area is equal to or greater than the predetermined area value.

  FIG. 2 is a diagram for explaining color tone correction processing for a captured image based on the first embodiment.

With reference to FIG. 2, the tone correction processing in the image processing unit 23 will be described.
As shown in FIG. 2A, the color tone correction processing unit 30 calculates the average value and standard deviation of the histogram of the luminance values of the entire image area to be subjected to the color tone correction processing.

  Here, as an example, the case where the average value Mw and the standard deviation Sw are calculated is shown.

  Next, as shown in FIG. 2B, the color tone correction processing unit 30 divides the image area into a plurality of small areas, and calculates the average value and standard deviation of the histogram of luminance values for each area. .

  Here, the case where the average value M1 and standard deviation S1 of the histogram of the luminance value of the small area P as an example is calculated is shown. Further, the case where the average value M2 and the standard deviation S2 of the histogram of the luminance values of the small region Q are calculated is shown.

  Then, correction processing is executed so that the average value and standard deviation of the histogram of luminance values for each small region are matched with the average value and standard deviation of the histogram of luminance values for the entire image region.

  Specifically, correction coefficients a1 and b1 according to the small area P and correction coefficients a2 and b2 according to the small area Q are calculated.

Correction coefficient a1 = Mw / M1
Correction coefficient b1 = Mw−a1 × M1
Correction coefficient a2 = Mw / M2
Correction coefficient b2 = Mw−a2 × M2
FIG. 2C shows correction coefficients a1 and b1 according to the small area P, correction coefficients a2 and b2 according to the small area Q, correction coefficients a3 and b3 according to the small area R, and correction coefficients a4 and b4 according to the small area S. Is shown.

  FIG. 2D shows a case where the tone correction processing is executed on an arbitrary pixel by the bilinear interpolation method using the calculated correction coefficient.

  Even when a part of the image data is dark, it can be corrected to the image data smoothed by the color tone correction processing.

FIG. 3 is a diagram illustrating the binarization process based on the first embodiment.
Referring to FIG. 3, here, the binarization process for the drainage pavement surface and the result of the binarization process for the dense grain pavement surface are shown.

  FIG. 3A shows the result of binarization processing by the binarization processing unit 26 on a partial region (dotted line region) of the smoothed drainage pavement surface image data. Specifically, the binarization processing unit 26 executes binarization processing on the image data using a discriminant analysis method. Here, a case where binarization is performed into a black region corresponding to the lubricant and a white region corresponding to the gap between the lubricants is shown. And it is further highlighted. That is, a color is added to the portion corresponding to the gap.

  Here, many regions where the area of the void region between the lubricants is large appear. In this example, the binarization processing unit 26 explained the method of binarization processing using the discriminant analysis method. However, the present invention is not limited to this, and binarization processing may be performed according to the luminance value. It is possible to adopt a simple method.

  FIG. 3B shows the result of the binarization processing by the binarization processing unit 26 on a partial region (dotted line region) of the smoothed fine-grained paved road image data. Specifically, the binarization processing unit 26 executes binarization processing on the image data using a discriminant analysis method. Here, a case where binarization is performed into a black region corresponding to the lubricant and a white region corresponding to the gap between the lubricants is shown. And it is further highlighted. That is, a color is added to the portion corresponding to the gap.

Here, many regions where the area of the void region between the lubricants is small appear.
As shown in the figure, it is possible to determine the type of the paved road surface by counting the area of the void area and comparing it with a predetermined value.

  In Embodiment 1, the area of the space | gap area | region between lubricants is totaled and compared with a predetermined value, and if it is more than a predetermined value, it will determine with a drainage pavement surface. The predetermined value can be set to an arbitrary value.

  FIG. 4 is a diagram illustrating image processing for the image data of the drainage pavement surface in the image processing unit 23 according to the first embodiment.

  Referring to FIG. 4, the image processing unit 23 executes a color tone correction process on the entire image area in accordance with the method described with reference to FIG. Thereby, it is corrected to smoothed image data.

  The determination unit 25 executes binarization processing on the image data that has been subjected to color tone correction processing on the entire image region. Then, after the binarization process, the area of the void area is totalized. It is determined whether or not the total area of the void regions is equal to or greater than a predetermined area value. In this example, it determines with it being a drainage pavement surface that the area of the total void | hole area | region is more than a predetermined area value.

  The image processing unit 23 performs a color tone correction process on the void area based on the determination result that the road surface is drainable.

  Specifically, the image processing unit 23 extracts a void area (shadow area) having a predetermined area or less from the image data of the drainage pavement surface. The predetermined area can be appropriately set to a value such as a median value of the void region by using a simulation or the like.

The image processing unit 23 performs the color tone correction process again on the extracted void area.
Specifically, the same color tone correction processing as described in FIG. 2 is executed.

  The average value and standard deviation of the luminance value histogram of the void area are calculated, and correction processing is executed so as to match the average value and standard deviation of the luminance value histogram of the entire image area.

  By the color tone correction process, the void area between the aggregates can be corrected to smoothed image data. In other words, the cracked area other than the void area between the aggregates is corrected to the image data in which the area is obvious.

Then, crack detection is executed on the image data subjected to the color tone correction processing.
Specifically, the detection unit 22 detects cracks in the paved road surface included in the image data of the paved road surface. For example, the detection unit 22 can determine the presence or absence of a crack based on a change in luminance.

FIG. 5 is a diagram illustrating a processing flow of the crack detection apparatus 10 according to the first embodiment.
Referring to FIG. 5, CPU 20 performs a color tone correction process on the image data of the paved road surface (step S0). Specifically, the color tone correction processing unit 30 executes the color tone correction processing using the bilinear interpolation method on the entire image data of the paved road surface as described in FIG.

  Next, CPUY20 performs a binarization process with respect to the image data of a paved road surface. Specifically, as described with reference to FIG. 3, the binarization processing unit 26 executes binarization processing using a discriminant analysis method on the image data subjected to the color tone correction processing.

Next, the CPU 20 adds up the areas of the void regions (shadow regions) (step S4).
Specifically, the totaling unit 27 totals the areas of the void regions between the lubricants sorted by the binarization process.

  Next, the CPU 20 determines whether or not the area of the void area (shadow area) is greater than or equal to a predetermined value (step S6). The area determination unit 28 determines whether the gap area between the lubricants counted by the counting unit 27 is equal to or greater than a predetermined value.

  In step S6, if the CPU 20 determines that the area of the void area (shadow area) is greater than or equal to a predetermined value (YES in step S6), the CPU 20 determines that the surface is a drainage pavement (step S8). The area determination unit 28 determines the drainage pavement surface when it is determined that the gap area between the lubricants counted by the counting unit 27 is equal to or greater than a predetermined value.

Next, the CPU 20 extracts a void area (shadow area) (step S10).
Specifically, as described with reference to FIG. 4, the image processing unit 23 extracts a void area (a shaded area) for the image data determined to be a drainage pavement surface.

  Next, the CPU 20 performs a color tone correction process on the image data of the void area (shadow area) (step S12). Specifically, the image processing unit 23 performs a color correction process using a bilinear interpolation method on the image data of the void area determined to be a drainage pavement surface. That is, the average value and standard deviation of the brightness value histogram of the void area are calculated, and correction processing is executed so as to match the average value and standard deviation of the brightness value histogram of the entire image area.

  Next, the CPU 20 executes crack detection (step S14). Specifically, the detection unit 22 detects cracks on the paved road surface based on the image data subjected to the color tone correction process. Specifically, as an example, it is possible to determine the presence or absence of a crack based on a change in luminance and extract the region.

Then, the process ends (END).
On the other hand, if the CPU 20 determines in step S6 that the area of the void area (shadow area) is less than the predetermined value (NO in step S6), the process after step S8 is skipped and the process proceeds to step S14. move on. Since the subsequent processing is the same, detailed description thereof will not be repeated.

  In the method according to the first embodiment, a void area is specified by binarization processing, and the area of the void area is totaled to determine whether the surface is a drainage pavement surface. And when it determines with it being a drainage pavement road surface, the said space | gap area | region is extracted and a color tone correction process is performed with respect to the extracted said space | gap area. By this processing, the image data in which the gap area between the aggregates is smoothed, that is, the image data in which the crack area other than the gap area between the aggregates is revealed is corrected.

  Thereby, it can suppress that a space | gap area | region (shadow area | region) is judged to be a crack, and can determine the crack of drainage pavement accurately.

(Embodiment 2)
In the second embodiment, a method for efficiently performing crack detection by machine learning will be described.

FIG. 6 is a diagram illustrating a crack detection device 11 according to the second embodiment.
Referring to FIG. 6, crack detection device 11 is different in that CPU 20 is replaced with CPU 20 #. Since other configurations are the same, detailed description thereof will not be repeated.

CPU 20 # further includes a data generation unit 40 as compared with CPU 20.
The data generation unit 40 creates teacher data used in machine learning.

FIG. 7 is a diagram for explaining teacher data according to the second embodiment.
As shown in FIG. 7, the data generation unit 40 divides the image data into small regions (in units of patches), and a binary image indicating deformed (cracked) locations and non-modified locations included in each patch. Then, teacher data in association with the corresponding image is generated. Here, as an example, a case where 20 teacher data are generated is shown.

  The binary image indicating the deformed portion and the non-deformed portion may be generated based on the detection result of the detection unit 22, or may be generated by the administrator marking the portion using the operation unit 18. May be.

In addition, the data generation unit 40 adds teacher data by machine learning.
The detection unit 22 can detect a crack using, for example, pattern matching or the like using teacher data generated by machine learning.

With this configuration, it is possible to efficiently perform crack detection.
An application that can be executed by a personal computer may be provided as the program in the present embodiment. At this time, the program according to the present embodiment may be incorporated as a partial function of various applications executed on the personal computer.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present disclosure is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

  10, 11 Crack detection device, 14 display unit, 16 memory, 18 operation unit, 21 data acquisition unit, 22 detection unit, 23 image processing unit, 25 determination unit, 26 binarization processing unit, 27 counting unit, 28 area determination Part, 40 data generation part.

Claims (9)

A determination unit for determining whether or not the paved road surface is drainable pavement based on a photographed image of the paved road surface;
An image processing unit that executes image processing based on a determination result of the determination unit;
A detection unit for detecting cracks in the paved road surface based on the image processing result of the image processing unit ;
The image processing unit
When the pavement road surface is a drainage pavement road surface as a determination result of the determination unit, a color tone correction process is performed on a shadow area of a predetermined area or less of the captured image
The crack detection apparatus which does not perform a color tone correction process on a shadow area having a predetermined area or less in the photographed image when the paved road surface is not a drainage paved road surface as a determination result of the determination unit. The determination unit
A binarization processing unit that binarizes the shadow area of the captured image and the other area;
A totaling unit that totalizes the area of the shadow region;
The crack detection apparatus of Claim 1 including the area determination part which determines whether the area value totaled by the said total part is more than predetermined value.   The crack detection device according to claim 2, wherein the area determination unit determines that the pavement road surface is a drainage pavement surface when the area value calculated by the totalization unit is greater than or equal to a predetermined value. The crack detection apparatus according to claim 2, wherein the determination unit performs a tone correction process using a bilinear interpolation method on the entire captured image. Wherein the detection unit detects the cracks on the basis of a change in luminance of the photographed image corrected by the image processing section, cracks detecting apparatus according to claim 1. Wherein the detection unit detects the cracking of the pavement by machine learning, crack detection apparatus according to claim 1. Data generation that divides the photographed image into patch units and generates the patch image and teacher data for the machine learning in which binary images indicating deformed portions and non-deformed portions included in the patch image are associated with each other The crack detection apparatus according to claim 6 , further comprising a section. Determining whether the paved road surface is drainable pavement based on a photographed image of the paved road surface;
Executing image processing based on the determination result;
Detecting cracks on the paved road surface based on image processing results,
The step of executing the image processing includes:
When the paved road surface is a drainage paved road surface as the determination result, a step of performing color tone correction processing on a shadow area of a predetermined area or less of the captured image;
A crack detection method including a step of not performing a color tone correction process on a shadow area of a predetermined area or less of the photographed image when the paved road surface is not a drainage paved road surface as the determination result . A program for causing a computer to function as a crack detection device,
The program is stored in the computer.
A determination unit for determining whether or not the paved road surface is drainable pavement based on a photographed image of the paved road surface ;
An image processing unit that executes image processing based on a determination result of the determination unit;
Based on the image processing result of the image processing unit, function as a detection unit for detecting cracks in the paved road surface ,
The image processing unit
When the pavement road surface is a drainage pavement road surface as a determination result of the determination unit, a color tone correction process is performed for a shadow area of a predetermined area or less of the captured image,
A crack detection program that does not perform color tone correction processing on a shadow area that is equal to or smaller than a predetermined area of the photographed image when the paved road surface is not a drainage paved road surface as a determination result of the determination unit .