JP2020154397A - Crack extracting system and crack extraction method - Google Patents

Abstract

To provide a crack extracting system and a crack extraction method for solving a problem in a conventional technology and allowing even a person not expertise in the field to easily perform image analysis process and automatic extract cracks on a concrete surface.SOLUTION: The crack extracting system of the present invention extracts cracks generated on the surface of a concrete construction by using images. The system includes background brightness calculation means, brightness calculation means, peculiar pixel extraction means, shape parameter calculation means, and crack extraction means. In this case, the shape parameter calculation means extracts a region of the peculiar pixel and calculates the shape parameter of the region of the peculiar pixel. Finally, the crack extraction means extracts the region of the peculiar pixel having a shape parameter being lower than a threshold as a crack.SELECTED DRAWING: Figure 2

Description

The present invention is a technique for extracting cracks generated on the surface of a concrete structure, and more specifically, a crack extraction system for extracting cracks by arithmetically processing the brightness of an image of a concrete structure, and a crack extraction system. It relates to an extraction method.

It has been pointed out that the construction infrastructure (hereinafter referred to as "construction infrastructure"), which has been intensively developed during the period of high economic growth, has already deteriorated considerably. In 2014, the "Proposal for Full-scale Implementation of Measures for Road Aging (Social Infrastructure Development Council)" was compiled, and "In the near future, human life and society such as the collapse of bridges," with the example of the Sasago Tunnel in 2012. It will cause a fatal situation related to the equipment, "he warned, strongly advocating the importance of maintenance of construction infrastructure.

Against this background, the national government promulgated a ministerial ordinance to partially revise the Road Law Enforcement Regulations, and regularly showed specific construction infrastructure inspection methods, points of interest for major deformations, and photographs of judgment cases. The inspection procedure is formulated. This periodic inspection procedure targets bridges with a bridge length of 2.0 m or more, which is said to be about 700,000 bridges, and the first inspection is performed within 2 years after the start of service, and the periodic inspection is performed once every 5 years thereafter. It is supposed to be. In addition, even on the underside of the bridge, distant vision using binoculars or the like is not permitted, and as a general rule, inspection must be performed by close visual inspection, which is confirmed with the naked eye. The national government has notified that not only roads but also railways will undergo initial inspections, general inspections, etc., and railway concrete structures will usually undergo general inspections mainly by visual inspection once every two years. ing.

However, it is not so easy to visually inspect the bridge slab (especially the lower surface). Normally, aerial work platforms or scaffolding are used to approach the bridge, but if the clearance is extremely high, a considerable scale aerial work platform or scaffolding is required. Especially when building scaffolding, if it is a bridge that crosses the river, the scaffolding will be assembled in the river, and if it is an overpass or overpass, the scaffolding will be assembled on the road or railroad track. There are even cases where it cannot be built.

In any case, according to the conventional method, a considerable amount of aerial work platforms and scaffolding are used to approach the bridge, which causes a problem that labor, cost, and work period increase. In addition, inspection work using aerial work platforms and scaffolding is so-called aerial work, so there are work safety problems such as the inspector falling over, and inspectors need some experience to visually inspect in close proximity. However, considering the recent labor shortage, this point can also be pointed out as a big problem.

Therefore, in recent years, inspection work using images has also been performed. There is no need to use aerial work platforms or scaffolding to approach the bridge, and the damaged part can be confirmed from the acquired image, so damage can be detected without restrictions on location and time, and other than the inspector. Since the person can also judge, the damage can be detected more objectively. Furthermore, the inspection using images has the advantage that the labor, cost, and work period can be shortened compared to using paper for recording and data management. For example, Patent Document 1 proposes a method of inspecting using a carriage that moves on the bridge surface and an arm attached to the carriage, and the tip of the arm is arranged on the lower surface of the bridge and connected to the tip of the arm. We are proposing a technology in which the dolly takes a picture of the underside of the bridge with a camera.

By the way, in order to automatically (or semi-automatically) extract cracks from the image acquired in the inspection work and to obtain the width dimension of the extracted cracks, it is common to use an image analysis process. For example, Non-Patent Documents 1 to 3 propose a specific analysis method for automatically extracting cracks on a concrete surface by image analysis and calculating the crack width.

Japanese Unexamined Patent Publication No. 2016-79684 "High-precision automatic crack extraction of concrete structures by image processing" Proceedings of the Japan Society of Civil Engineers F, Vol. 66. No. 3, pp. 459-470, 2010 "Development of crack image analysis technology for concrete using Gabor wavelet transform" Proceedings of the Japan Society of Civil Engineers E2, Vol. 68 ・ No. 3, pp. 178-194, 2012 "Detection by image analysis of concrete surface cracks focusing on statistical and geometric features" Proceedings of the Japan Society of Civil Engineers F3, Vol. 70. No. 2,I_1-I_8,2014

According to the image analysis method shown in Non-Patent Documents 1 to 3, cracks on the concrete surface can be automatically extracted, and the crack width can be calculated. However, the conventional image analysis method is extremely complicated and cannot be easily and freely used by persons other than those concerned or experts. In addition, for railway concrete structures, inspections using a digital camera are also permitted, and although a digital camera is actually used, the conventional method can only measure cracks with a width of about 0.3 mm. Moreover, it was a little complicated work to detect the crack.

The object of the present invention is to solve the problems of the prior art, that is, even a person other than a person concerned or an expert can easily and freely perform image analysis processing and crack the concrete surface. It is to provide a crack extraction system capable of automatic extraction, and a crack extraction method.

The present invention focuses on the fact that singular pixels that are candidates for cracks are extracted from the calculated background brightness and actual brightness of each pixel, and cracks are extracted based on the shape parameter of the singular pixel region composed of singular pixels. It is an invention made based on an unprecedented idea.

The crack extraction system of the present invention is a system for extracting cracks generated on the surface of a concrete structure using an image, and is a background brightness calculation means and a brightness difference calculation means, a peculiar pixel extraction means, a shape parameter calculation means, and a crack. It is equipped with an extraction means. Of these, the background brightness calculation means is a means for calculating the background brightness for each pixel of the image, and the brightness difference calculation means calculates the brightness difference (difference between the actual brightness given to the pixels and the background brightness) for each pixel. The singular pixel extraction means is a means for extracting singular pixels (pixels whose actual brightness is darker than the background brightness and whose brightness difference exceeds a predetermined brightness difference threshold). The shape parameter calculating means is a means for extracting a singular pixel region (a region where singular pixels are continuous) and calculating the shape parameter of the singular pixel region, and the crack extracting means has a shape parameter below a predetermined shape threshold value. This is a means for extracting a singular pixel region as a crack. The background luminance calculation means selects a plurality of peripheral pixels around the pixel of interest (pixels for which the background luminance is calculated), and the background of the pixel of interest is based on the representative values of the actual luminance of the pixel of interest and the peripheral pixels. Calculate the brightness. Further, the shape parameter calculation means calculates the shape parameter based on the product of the total pixel area S of the singular pixel region and the reciprocal of the square of the representative length L of the quadrangle including the singular pixel region.

The crack extraction system of the present invention may further include a pixel of interest setting means. This pixel of interest setting means is a means for setting a grid for pixels of interest for an image and setting pixels at intersection positions of the grid for pixels of interest as pixels of interest. In this case, the background luminance calculation means calculates the background luminance of the non-focused pixel based on the background luminance of the focused pixel surrounding the non-focused pixel (the pixel surrounded by the focused pixel).

The crack extraction system of the present invention may also calculate the median value obtained from the actual brightness of the pixel of interest and the peripheral pixels as the background brightness of the pixel of interest.

The crack extraction system of the present invention may also use a luminance difference threshold set according to the resolution of the image.

The crack extraction system of the present invention may further include a crack angle calculating means, a maximum luminance difference calculating means, a widthwise luminance difference calculating means, and a crack width calculating means. This crack angle calculating means obtains the direction (crack direction) of the crack pixel region (unique pixel region extracted as a crack), and is the smallest of the sandwich angles between the two orthogonal axes set on the image and the crack direction. It is a means for calculating the interstitial angle as an angle as a crack angle θ, and the maximum brightness difference calculation means selects a crack representative pixel from the crack pixels (unique pixels constituting the crack pixel region) and also selects the crack representative pixel. This is a means for calculating the maximum brightness difference α, which is the difference between the background brightness and the minimum brightness (set in advance). The width direction brightness difference calculation means extracts a crack pixel that passes through a crack representative pixel and is on an orthogonal axis that is substantially orthogonal (including orthogonal) to the crack direction and is a continuous crack pixel from the crack representative pixel. At the same time, it is a means for calculating the width direction brightness difference β, which is the sum of the brightness differences of the orthogonal pixels. Then, the crack width calculating means calculates the crack width ω by the following equation based on the maximum luminance difference α, the luminance difference β in the width direction, the crack angle θ, the image resolution γ, and the coefficient κ.
ω = κ × β / α × γ / cos θ

The crack extraction system of the present invention may further include a crack representative pixel extraction means. The crack representative pixel extraction means is a means for setting a crack pixel grid for an image and extracting crack pixels intersecting with the crack pixel grid as crack representative pixels.

In the crack extraction system of the present invention, the crack angle calculating means sets a rectangular region centered on a crack representative pixel, and calculates the crack angle θ after linearly approximating the crack pixels distributed in this rectangular region. You can also do it.

The crack extraction method of the present invention is a method of extracting cracks generated on the surface of a concrete structure using an image, and is a photographing step, a background brightness calculation step, a brightness difference calculation step, a peculiar pixel extraction step, and a shape parameter calculation. It is a method including a process and a crack extraction process. Of these, in the photographing step, the concrete structure is photographed to acquire an image, in the background brightness calculation step, the background brightness is calculated for each pixel of the image, and in the brightness difference calculation step, the brightness difference is calculated for each pixel. In the singular pixel extraction step, singular pixels are extracted, and in the shape parameter calculation step, the singular pixel region is extracted and the shape parameter of this singular pixel region is calculated. Then, in the crack extraction step, a peculiar pixel region in which the shape parameter is lower than the shape threshold value is extracted as a crack. In the background brightness calculation step, a plurality of peripheral pixels are selected, and the background brightness of the pixel of interest is calculated based on the representative values of the actual brightness of the pixel of interest and the peripheral pixels. In the shape parameter calculation step, the pixels in the singular pixel region are calculated. The shape parameter is calculated based on the product of the total area S and the reciprocal of the square of the representative length L of the quadrangle containing the singular pixel region.

The crack extraction method of the present invention can also be a method including a crack angle calculation step, a maximum luminance difference calculation step, a luminance difference calculation step in the width direction, and a crack width calculation step. In this crack angle calculation step, the crack direction is obtained, and the sandwich angle, which is the smaller angle between the two orthogonal axes set on the image and the crack direction, is calculated as the crack angle θ, and the maximum luminance difference calculation step is performed. Then, the crack representative pixel is selected from the crack pixels, and the maximum luminance difference α, which is the difference between the background luminance and the minimum luminance of the crack representative pixel, is calculated. In the width direction brightness difference calculation step, orthogonal pixels that pass through the crack representative pixel and are on the orthogonal axis that is substantially orthogonal (including orthogonal) to the crack direction and are continuous crack pixels are extracted from the crack representative pixel. At the same time, the width direction brightness difference β, which is the sum of the brightness differences of the orthogonal pixels, is calculated. Then, in the crack width calculation step, the crack width ω is calculated by the following equation based on the maximum luminance difference α, the luminance difference β in the width direction, the crack angle θ, the image resolution γ, and the coefficient κ.
ω = κ × β / α × γ / cos θ

The crack extraction system and the crack extraction method of the present invention have the following effects.
(1) Inspection work can be performed by acquiring images, labor, cost, and work period can be reduced, and by avoiding work at heights, inspection work can be performed more safely than with conventional methods, and it is more objective. Since cracks are extracted, it can be carried out without limiting the inspector.
(2) Cracks on the concrete surface can be automatically extracted, and image analysis processing for that purpose can be easily and freely performed, and can be processed at high speed.
(3) Since the cracks are extracted based on the shape parameters, erroneous extraction can be reduced as compared with the conventional method, and as a result, the cracks can be extracted with higher accuracy. It is also possible to obtain a crack width smaller than the resolution of the image (one side length of the pixel).

(A) is an image diagram showing the surface of a concrete structure, and (b) is an enlarged image diagram showing cracks extracted as a result of processing the image. A block diagram showing the main configurations of the crack extraction system. A flow chart showing the main processing flow until the crack extraction system extracts cracks. A flow chart showing the main processing flow until the crack extraction system calculates the crack width. A model diagram showing a pixel of interest selected from the pixels constituting the image. (A) is a model diagram showing eight peripheral pixels adjacent to the pixel of interest, and (b) is a model diagram showing 24 peripheral pixels around the pixel of interest. A model diagram showing 9 non-focused pixels surrounded by 4 focused pixels. (A) is a model diagram showing a peculiar pixel region composed of 19 peculiar pixels, and (b) is a model diagram showing an inclusion quadrangle including a peculiar pixel region composed of 19 peculiar pixels. (A) is a model diagram showing a crack representative pixel selected from the crack pixels constituting the crack pixel region, and (b) is a model diagram showing a crack direction obtained based on the crack pixel in the rectangular region. A model diagram showing a sandwich angle formed by the approximate straight line and the X-axis and a sandwich angle formed by the approximate straight line and the Y-axis. A model diagram showing an orthogonal axis substantially orthogonal to an approximate straight line and orthogonal pixels on this orthogonal axis. The graph which explains the sum total of the luminance difference of the orthogonal pixel. The flow chart which shows the main process of the crack extraction method of this invention.

An example of the crack extraction system of the present invention and the embodiment of the crack extraction method will be described with reference to the drawings.

1. 1. Overall Overview The crack extraction system of the present invention extracts cracks (FIG. 1 (b)) generated on the surface of a concrete structure using an image as shown in FIG. 1 (a). Note that FIG. 1A is an image of the surface of the concrete structure, and FIG. 1B is an enlarged image showing the cracks extracted as a result of processing this image. Hereinafter, the outline processing flow from extracting cracks from the image will be described.

First, from the image, the brightness indicating the concrete structure itself (hereinafter referred to as "background brightness") is obtained for each pixel, and then the brightness actually possessed by the pixel (hereinafter referred to as "actual brightness") and the background. The difference from the brightness (hereinafter referred to as "brightness difference") is obtained. The "luminance" here is an index (so-called gray scale) indicating brightness (darkness), and can be given in the range of 0 to 255, for example. When a brightness difference is obtained for each pixel, pixels considered to be cracks (hereinafter referred to as "singular pixels") are extracted based on the brightness difference, and a region in which the singular pixels are continuous (hereinafter referred to as "singular pixel area") is extracted. ) Is extracted.

Some of the extracted singular pixel regions show cracks, some show dirt on the concrete surface, and some show mere noise. That is, it is necessary to exclude dirt, noise, and the like, and appropriately extract a peculiar pixel region showing cracks (hereinafter, referred to as "cracked pixel region"). Therefore, in the present invention, the cracked pixel region is extracted based on an index value (hereinafter, referred to as “shape parameter”) indicating the shape of the singular pixel region. Since cracks show a characteristic shape (that is, a so-called elongated shape such as a linear shape) unlike dirt and noise, a peculiar pixel region of a shape parameter showing a specific value is extracted as a crack pixel region.

2. 2. Crack extraction system The crack extraction system of the present invention will be described in detail with reference to the drawings. The crack extraction method of the present invention is a method of extracting cracks using the crack extraction system of the present invention. Therefore, the crack extraction system of the present invention will be described first, and then the crack extraction method of the present invention will be described. I will do it.

FIG. 2 is a block diagram showing a main configuration of the crack extraction system 100. As shown in this figure, the crack extraction system 100 includes a background luminance calculation means 101, a luminance difference calculation means 102, a singular pixel extraction means 103, a shape parameter calculation means 104, and a crack extraction means 105, and further sets a pixel of interest. Means 106, crack angle calculation means 107, crack representative pixel extraction means 108, maximum brightness difference calculation means 109, width direction brightness difference calculation means 110, crack width calculation means 111, output means 112 such as display or printer, brightness difference threshold storage means. 113, the shape threshold storage means 114 can also be included.

Of the crack extraction system 100, the background brightness calculation means 101 and the brightness difference calculation means 102, the singular pixel extraction means 103, the shape parameter calculation means 104, the crack extraction means 105, the pixel of interest setting means 106, the crack angle calculation means 107, and the crack representative pixel. The extraction means 108, the maximum luminance difference calculation means 109, the widthwise luminance difference calculation means 110, and the crack width calculation means 111 can be manufactured as dedicated ones, or a general-purpose computer device can be used. This computer device can be configured by a personal computer (PC), a tablet PC such as an iPad (registered trademark), or a mobile terminal including a smartphone. The computer device includes a processor such as a CPU and a memory such as a ROM and a RAM, and some computer devices further include an input means such as a mouse and a keyboard and a display.

Further, the luminance difference threshold storage means 113 and the shape threshold storage means 114 can be constructed on, for example, a database server, can be placed on a local network (LAN: Local Area Network), or via the Internet (that is, wireless communication or wireless communication). It can also be a cloud server that saves via wired communication).

Subsequently, the main processing of the crack extraction system 100 of the present invention will be described with reference to FIGS. 3 and 4. FIG. 3 is a flow chart showing the main processing flow until the crack extraction system 100 extracts the cracks, and FIG. 4 shows the main processing flow until the crack extraction system 100 calculates the crack width. It is a flow chart. In these flow charts, the central column shows the processing to be performed, the left column shows the input information required for the processing, and the right column shows the output information generated from the processing.

As shown in FIG. 3, first, the “pixel of interest” is set by the pixel of interest setting means 106 (FIG. 2) (Step 101 of FIG. 3). As described above, the crack extraction system 100 of the present invention determines the background brightness for each pixel constituting the image. That is, it is necessary to sequentially set the target pixels for which the background brightness is to be obtained, and the background brightness is obtained for each pixel of interest set here.

The background brightness of the pixel of interest is obtained by a calculation process described later. This process can be performed on all the pixels that make up the image (that is, all the pixels are the pixels of interest), but in this case, the calculation time and the calculation cost increase, so a representative pixel is selected as the pixel of interest. Then, the background brightness should be calculated.

An example of the process of selecting the pixel of interest will be described with reference to FIG. FIG. 5 is a model diagram showing a pixel of interest Pa selected from a large number of pixels P constituting the image. As shown in this figure, grids (hereinafter referred to as "grids for pixels of interest") are set on the image at appropriate intervals (for example, 10 pixels x 10 pixels), and among the grids for pixels of interest, the horizontal axis AGx The pixel at the position where the vertical axis AGy intersects can be set as the pixel of interest Pa. Of course, the interval of the grid for the pixel of interest can be arbitrarily designed, and the pixel of interest Pa can be set by a method other than the method shown in FIG.

When the pixel of interest Pa is set, the background luminance calculation means 101 (FIG. 2) calculates the "background luminance" of the pixel of interest Pa (Step 102 of FIG. 3). In calculating the background brightness, "peripheral pixels" are first selected. The peripheral pixels Pb can be selected in a desired number from the pixels around the pixel of interest Pa. For example, in FIG. 6A, _eight peripheral pixels Pb _{1 to} Pb 8 adjacent to the pixel of interest Pa are selected. In FIG. 6B, 24 peripheral pixels Pb _{1 to} Pb ₂₄ around the pixel of interest Pa are selected. When the peripheral pixel Pb is selected, a representative value consisting of the actual brightness of the peripheral pixel Pb and the actual brightness of the pixel of interest Pa is obtained, and this representative value is used as the background brightness of the pixel of interest Pa. For example, in the case of FIG. 6 (a), bright nine real luminance of the target pixel Pa and the eight peripheral pixels Pb ₁ ~Pb ₈ (dark) arranged in order, paying attention to 5-th real luminance is the median pixel Pa Can be the background brightness of. As the representative value, in addition to the median value, an average value or a mode value can be used.

When the background brightness of the pixel of interest Pa is obtained, the background brightness of the “non-focused pixel” is calculated by the background brightness calculation means 101 (FIG. 2) (Step 103 of FIG. 3). Here, the non-focused pixel is a pixel that is not set as the focused pixel Pa, and the background brightness of the non-focused pixel can be calculated based on the background brightness of the focused pixel Pa around it. For example, in FIG. 7, nine non-focused pixels Pn ₁ to non-focused pixels Pn ₉ are surrounded by _four focused pixels Pa ₁ to focused pixels Pa ₄ , and in this case, non-focused pixels Pn ₁ to non-focused pixels Pn ₁ to non-focused pixels. The background brightness of Pn ₉ can be calculated by interpolating the background brightness of the _four pixels of interest Pa ₁ to the pixels of interest Pa ₄ . When obtaining the background luminance for all the pixels constituting the image (that is, all the pixels are the pixels of interest Pa), the process of calculating the background luminance of the non-focused pixel Pn (Step 103 in FIG. 3) is naturally performed. It can be omitted.

When the background brightness of the focused pixel Pa and the non-focused pixel Pn is obtained, the “luminance difference” is calculated by the luminance difference calculating means 102 (FIG. 2) (Step 104 in FIG. 3). As described above, this luminance difference is the difference between the actual luminance and the background luminance in the pixel, and is a value obtained by subtracting the actual luminance from the background luminance (background luminance minus the actual luminance) or an absolute value that does not consider positive or negative. You should ask. The luminance difference is repeatedly calculated for all the target pixels as shown in the flow chart of FIG. 3 (arrows on the right side of Step 101 to Step 104).

When the luminance difference is calculated for all the target pixels, the "singular pixel" is extracted by the singular pixel extracting means 103 (FIG. 2) (Step 105 in FIG. 3). Here, the singular pixel is a pixel whose actual brightness is darker (smaller) than the background brightness, and whose brightness difference exceeds a predetermined brightness difference threshold value. More specifically, the singular pixel extracting means 103 reads the "luminance difference threshold" from the luminance difference threshold storage means 113 as shown in FIG. 2, and the actual brightness is a pixel darker than the background brightness and exceeds the luminance difference threshold. Pixels are extracted as singular pixels. The inventors have found that cracks can be extracted more accurately by setting the luminance difference threshold value according to the magnitude of the image resolution (side length of one pixel). Therefore, the luminance difference threshold value may be appropriately set according to the resolution of the acquired image.

When the singular pixel is extracted, the "shape parameter" is calculated by the shape parameter calculation means 104 (FIG. 2) (Step 109 in FIG. 3). Hereinafter, the flow of processing until the shape parameters are calculated will be described in detail.

First, a region formed by continuously arranged singular pixels is extracted as a "singular pixel region" (Step 106 in FIG. 3). For example, in FIG. 8A, 19 peculiar pixels Ps are continuously arranged, and this region is extracted as a peculiar pixel region Rs. At this time, the positional relationship in which the singular pixels Ps are in direct contact with each other may be referred to as "continuous", and the positional relationship in which the singular pixels Ps are close to each other (for example, every other singular pixel Ps is adjacent to each other) may be "continuous". You may say "to do".

Next, the pixel area of the singular pixel region Rs (hereinafter referred to as “total pixel area S”) is calculated (Step 107 in FIG. 3). The total pixel area S is the total area of the peculiar pixels Ps constituting the peculiar pixel region Rs, that is, the number of peculiar pixels Ps constituting the peculiar pixel region Rs in the area (unit area) of one pixel P. It is a value multiplied by. For example, in the case of FIG. 8A, since there are 19 singular pixels Ps constituting the singular pixel region Rs, the total sum S of the pixel areas can be obtained by “unit area × 19”.

Further, the representative length L of the quadrangle containing the singular pixel region Rs (hereinafter referred to as “conflict quadrangle Qc”) is also obtained (Step 108 in FIG. 3). For example, in FIG. 8 (a), the range indicated by the broken line is cut out as the included quadrangle Qc, and as shown in FIG. 8 (b), the long side La or short side Lb forming the included quadrangle Qc, or the length of the diagonal line ( The square root of the sum of squares of the long side La and the short side Lb) is obtained as the representative length L.

When the total pixel area S and the representative length L are obtained, the shape parameters of the singular pixel region Rs are calculated (Step 109 in FIG. 3). The shape parameter is obtained based on the product of "total pixel area S" and "reciprocal of the square of the representative length L", and can be calculated by, for example, the following equation. Note that m in the following equation is a preset coefficient, and m = 1 can be set, or can be set in the range of m <10 or m> 1.
(Shape parameter) = m × S ÷ L ²

When the shape parameter of the singular pixel region Rs is obtained, the crack is extracted by the crack extracting means 105 (FIG. 2) (Step 110 in FIG. 3). More specifically, the crack extracting means 105 reads out the “shape threshold” from the shape threshold storage means 114 as shown in FIG. 2, and compares the shape threshold with the shape parameter to create a singular pixel region in which the shape parameter is lower than the shape threshold. Rs is defined as a crack pixel region, that is, this is extracted as a crack. The singular pixel region Rs having a relatively large shape parameter has a shape such as a circle, a quadrangle, or a polygon that can secure a considerable length in the vertical and horizontal directions, while the singular pixel region Rs having a relatively small shape parameter has a linear or rod shape. Therefore, the singular pixel region Rs is selected to be the cracked pixel region or other depending on the magnitude of the shape parameter with the shape threshold as the boundary.

When the crack pixel region is extracted (that is, when the crack is extracted), the "crack width ω" is calculated by the crack width calculating means 111 (FIG. 2) (Step 118 in FIG. 4). Hereinafter, the flow of processing until the crack width ω is calculated will be described in detail.

First, the "crack representative pixel" is extracted by the crack representative pixel extracting means 108 (FIG. 2) (Step 111 in FIG. 4). The crack width ω is calculated at a predetermined position (so to speak, fragmentarily) in the crack pixel region. The crack representative pixel is a so-called reference pixel indicating a position for calculating the crack width ω. An example of the process of extracting the crack representative pixel Pm will be described with reference to FIG. As shown in this figure, grids (hereinafter referred to as "grid pixel grids") are set on the image at appropriate intervals (for example, 100 pixels x 100 pixels), and among the crack pixel grids, the horizontal axis MGx When the position where the vertical axis MGy intersects overlaps with the peculiar pixel (hereinafter, referred to as "crack pixel Pc") constituting the crack pixel region Rc, the crack pixel Pc can be set as the crack representative pixel Pm. Of course, the spacing between the grids for crack pixels can be arbitrarily designed, and the crack representative pixels Pm can be extracted by a method other than the method shown in FIG. 9A.

Further, while the crack representative pixel Pm is extracted, the "crack direction" is also calculated (Step 113 in FIG. 4). The crack direction is a direction obtained based on the crack pixel region Rc. More specifically, as shown in FIG. 9B, after setting the rectangular region Qr centered on the crack representative pixel Pm (FIG. 4). Step112), the direction of the approximate straight line Lc obtained based on the cracked pixel Pc in the rectangular region Qr. This approximate straight line Lc is a so-called linear model of the cracked pixel region Rc, and is calculated by various conventional methods such as the least squares method for the cracked pixel Pc in the rectangular region Qr. You can ask. The rectangular area Qr shown in FIG. 9B is a square area of 9 pixels × 9 pixels centered on the cracked pixel Pc, but of course, the rectangular area Qr can be set with an arbitrary size (pixel × pixel). In addition, the rectangular area Qr can be set not only as a square but also as a rectangle, and the rectangular area Qr can be set by a method other than the method shown in FIG. 9B.

When the crack direction is obtained (that is, when the approximate straight line Lc is set), the "crack angle θ" is calculated by the crack angle calculating means 107 (FIG. 2) (Step 114 in FIG. 4). The crack angle θ is a narrowing angle formed by two orthogonal axes such as the X-axis and the Y-axis shown in FIG. 9B and an approximate straight line Lc. However, as shown in FIG. 10, the sandwich angle θx formed by the approximate straight line Lc and the X-axis (first axis) and the sandwich angle θy formed by the approximate straight line Lc and the Y-axis (second axis) are smaller. In the case of FIG. 10, the sandwich angle θx) is calculated as the crack angle θ.

While the crack angle θ is calculated, the “maximum luminance difference α” is also calculated by the maximum luminance difference calculating means 109 (FIG. 2) (Step 115 in FIG. 4). The maximum brightness difference α is the difference between the preset “minimum brightness” and the background brightness of the crack representative pixel Pm, and is a value obtained by subtracting the minimum brightness from the background brightness of the crack representative pixel Pm (background of the crack representative pixel Pm). Luminance-minimum brightness). This minimum brightness is set as the brightness corresponding to the crack, and for example, the brightness of the pixel completely covered with the crack may be set as the minimum brightness.

Along with the crack angle θ and the maximum luminance difference α, the “luminance difference β in the width direction” is also calculated by the luminance difference calculating means 110 (FIG. 2) in the width direction (Step 117 in FIG. 4). The width direction luminance difference β is a value obtained by summing the luminance differences of the crack pixels Pc continuously arranged in the direction (that is, the width direction of the crack) substantially orthogonal (including orthogonal) to the approximate straight line Lc. Hereinafter, the method of calculating the luminance difference β in the width direction will be described in more detail with reference to FIGS. 11 and 12.

First, the orthogonal axis Lv shown in FIG. 11 is set. The orthogonal axis Lv is a straight line that passes through the crack representative pixel Pm and is substantially orthogonal (including orthogonal) to the approximate straight line Lc. Next, the crack pixel Pc on the orthogonal axis Lv and further arranged continuously from the crack representative pixel Pm is extracted as the “orthogonal pixel Pv” (Step 116 in FIG. 4). For example, in the case of FIG. 11, four orthogonal pixels Pv1 to Pv4 are extracted. At this time, as in the extraction process of the singular pixel region, the positional relationship in which the orthogonal pixels Pv are in direct contact with each other may be "continuous", or the orthogonal pixels Pv are close to each other (for example, every other pixel Pv is adjacent to each other). The positional relationship may be referred to as "continuous".

Then, the luminance difference of the extracted orthogonal pixels Pv is summed to calculate the luminance difference β in the width direction (Step 117 in FIG. 4). FIG. 12 is a graph showing the brightness difference Dg1 to the brightness difference Dg4 of the four orthogonal pixels Pv1 to the orthogonal pixels Pv4 and the brightness difference Dgm of the crack representative pixel Pm shown in FIG. The luminance difference β in the width direction is the sum of the luminance differences Dg1 to the luminance difference Dg4 and the luminance difference Dgm shown in this figure, that is, corresponds to the area of this graph.

When the crack angle θ, the maximum luminance difference α, and the luminance difference β in the width direction are obtained, the crack width ω is calculated by the crack width calculating means 111 (FIG. 2) (Step 118 in FIG. 4). The crack width ω is obtained based on the coefficient k, the resolution γ (side length of one pixel), the crack angle θ, the maximum luminance difference α, and the luminance difference β in the width direction, and can be calculated by, for example, the following equation. Note that k in the following equation is a preset coefficient, and k = 1 can be set, k <10 can be set, or k> 1 can be set.
ω = κ × β × γ ÷ (α × cos θ)

The crack extracted by the crack extraction system 100, or the calculated crack width, is output by an output means 112 such as a display or a printer.

3. 3. Crack Extraction Method Next, the crack extraction method of the present invention will be described with reference to the drawings. The crack extraction method of the present invention is a method of extracting cracks using the crack extraction system 100 described so far. Therefore, avoiding explanations that overlap with the contents described in the crack extraction system 100, the cracks of the present invention should be avoided. Only the contents specific to the extraction method will be explained. That is, the contents not described here are the same as those described in "2. Crack extraction system".

FIG. 13 is a flow chart showing the main steps of the crack extraction method of the present invention. As shown in this figure, first, the surface of the concrete structure is photographed and an image is acquired (Step 201). Then, the background brightness is obtained for each pixel P of the acquired image (Step202), and the brightness difference is obtained for each pixel P (Step203).

When the brightness difference for each pixel P is obtained, the singular pixel Ps is extracted (Step204) and the singular pixel region Rs is extracted (Step205). Then, a shape parameter is calculated from the total pixel area S and the representative length L (Step 206), and a crack (crack pixel region Rc) is extracted based on this shape parameter (Step 207).

When the crack pixel region Rc is obtained, the crack direction (approximate straight line Lc) is obtained to calculate the crack angle (Step208), the maximum luminance difference α is calculated (Step209), and the orthogonal pixel Pv is extracted and the width direction is obtained. The brightness difference β is calculated (Step 210). Then, the crack width ω is calculated based on the coefficient k, the resolution γ, the crack angle θ, the maximum luminance difference α, and the luminance difference β in the width direction (Step211).

The crack extraction system and the crack extraction method of the present invention can be used for various concrete structures such as concrete dams, tunnel lining concrete, and concrete retaining walls, in addition to concrete floor slabs. According to the invention of the present application, it is possible to grasp the deterioration status of the construction infrastructure in service, and it is possible to take early repair and reinforcement measures according to the deterioration status, which in turn leads to a longer life of the construction infrastructure. Is an invention that can be used not only industrially but also can be expected to make a great contribution to society.

100 Crack extraction system 101 Background brightness calculation means (of crack extraction system) 102 Brightness difference calculation means (of crack extraction system) 103 Singular pixel extraction means (of crack extraction system) 104 Shape parameter calculation means (of crack extraction system) 105 ( Crack extraction means (of crack extraction system) 106 (of crack extraction system) Focused pixel setting means 107 (of crack extraction system) crack angle calculation means 108 (of crack extraction system) crack representative pixel extraction means 109 (of crack extraction system) maximum Brightness difference calculation means 110 Width direction brightness difference calculation means (for crack extraction system) 111 (for crack extraction system) Crack width calculation means 112 (for crack extraction system) Output means 113 (for crack extraction system) Brightness difference threshold storage means 114 Shape threshold storage means (of the crack extraction system) AGx Horizontal axis of the grid for the pixel of interest AGy Vertical axis of the grid for the pixel of interest Lc Approximate straight line Lv Orthogonal axis MGx Horizontal axis of the grid for the cracked pixel MGy Vertical axis of the grid for the cracked pixel Pa Focus Pb Peripheral pixel Pc Cracked pixel Pm Cracked representative pixel Pn Non-focused pixel Ps Singular pixel Pv Orthogonal pixel Rc Cracked pixel area Rs Singular pixel area Qc Containing square Qr Rectangular area

Claims (9)

In a system for extracting cracks generated on the surface of a concrete structure using an image of the concrete structure.
A background brightness calculation means for calculating the background brightness for each pixel of the image,
Luminance difference calculation means for calculating the luminance difference, which is the difference between the actual luminance given to the pixels and the background luminance, for each pixel.
A singular pixel extraction means for extracting singular pixels whose actual brightness is darker than the background brightness and whose brightness difference exceeds a predetermined brightness difference threshold value.
A shape parameter calculation means for extracting a singular pixel region, which is a region in which the singular pixels are continuous, and calculating a shape parameter of the singular pixel region.
A crack extraction means for extracting the peculiar pixel region in which the shape parameter is below a predetermined shape threshold value as a crack is provided.
The background brightness calculation means selects a plurality of peripheral pixels around the pixel of interest for which the background brightness is to be calculated, and the pixel of interest is based on the representative value of the pixel of interest and the actual brightness of the peripheral pixel. Calculate the background brightness of
The shape parameter calculating means calculates the shape parameter based on the product of the total pixel area S of the singular pixel region and the reciprocal of the square of the representative length L of the quadrangle including the singular pixel region.
A crack extraction system characterized by this. A pixel of interest setting means for setting a grid for the pixel of interest for the image and setting the pixel at an intersection position of the grid for the pixel of interest as the pixel of interest is further provided.
The background brightness calculation means calculates the background brightness of the non-focused pixel surrounded by the non-focused pixel based on the background brightness of the focused pixel surrounding the non-focused pixel.
The crack extraction system according to claim 1, wherein the crack extraction system is characterized in that. The background brightness calculation means calculates the median value obtained from the actual brightness of the pixel of interest and the peripheral pixels as the background brightness of the pixel of interest.
The crack extraction system according to claim 1 or 2, wherein the crack extraction system is characterized in that. The luminance difference threshold is a value set according to the resolution of the image.
The crack extraction system according to any one of claims 1 to 3, wherein the crack extraction system is characterized. The crack direction, which is the direction of the crack pixel region, which is the peculiar pixel region extracted as a crack, is obtained, and the crack angle, which is the smaller angle between the two orthogonal axes set on the image and the crack direction, is determined. The crack angle calculation means calculated as the crack angle θ, and
A crack representative pixel is selected from the crack pixels which are the peculiar pixels constituting the crack pixel region, and the maximum brightness difference α which is the difference between the background brightness of the crack representative pixel and the set minimum brightness is calculated. Maximum brightness difference calculation means and
The orthogonal pixels that pass through the crack representative pixel and are on an orthogonal axis that is orthogonal to or substantially orthogonal to the crack direction and are continuous crack pixels from the crack representative pixel are extracted and the orthogonal pixels are extracted. The width direction brightness difference calculation means for calculating the width direction brightness difference β which is the sum of the brightness differences of
A crack width calculating means for calculating the crack width ω by the following equation based on the maximum luminance difference α, the luminance difference β in the width direction, the crack angle θ, the resolution γ of the image, and the coefficient κ.
ω = κ × β / α × γ / cos θ
The crack extraction system according to any one of claims 1 to 4, further comprising. A crack representative pixel extraction means for setting a crack pixel grid for the image and extracting the crack pixels intersecting with the crack pixel grid as the crack representative pixels.
5. The crack extraction system according to claim 5, further comprising. The crack angle calculating means sets a rectangular region centered on the crack representative pixel, and calculates the crack angle θ after linearly approximating the crack pixels distributed in the rectangular region.
The crack extraction system according to claim 5 or 6, wherein the crack extraction system is characterized. In the method of extracting cracks on the surface of concrete structures using images,
A photographing process of photographing the concrete structure and acquiring the image, and
A background brightness calculation step of calculating the background brightness for each pixel of the image, and
A luminance difference calculation process for calculating the luminance difference, which is the difference between the actual luminance given to the pixels and the background luminance, for each pixel.
A singular pixel extraction step of extracting singular pixels whose actual brightness is darker than the background brightness and whose brightness difference exceeds a predetermined brightness difference threshold value.
A shape parameter calculation step of extracting a singular pixel region, which is a region in which the singular pixels are continuous, and calculating a shape parameter of the singular pixel region.
A crack extraction step of extracting the peculiar pixel region in which the shape parameter is below a predetermined shape threshold value as a crack is provided.
In the background brightness calculation step, a plurality of peripheral pixels around the pixel of interest for which the background brightness is to be calculated are selected, and the pixel of interest and the pixel of interest are based on the representative values of the actual brightness of the pixel of interest and the peripheral pixel. Calculate the background brightness of
In the shape parameter calculation step, the shape parameter is calculated based on the product of the total pixel area S of the singular pixel region and the reciprocal of the square of the representative length L of the quadrangle containing the singular pixel region.
A crack extraction method characterized by this. The crack direction, which is the direction of the crack pixel region, which is the peculiar pixel region extracted as a crack, is obtained, and the crack angle, which is the smaller angle between the two orthogonal axes set on the image and the crack direction, is determined. The crack angle calculation process calculated as the crack angle θ, and
A crack representative pixel is selected from the crack pixels which are the peculiar pixels constituting the crack pixel region, and the maximum brightness difference α which is the difference between the background brightness of the crack representative pixel and the set minimum brightness is calculated. Maximum brightness difference calculation process to be performed and
The orthogonal pixels that pass through the crack representative pixel and are on an orthogonal axis that is orthogonal to or substantially orthogonal to the crack direction and are continuous crack pixels from the crack representative pixel are extracted and the orthogonal pixels are extracted. The width direction brightness difference calculation step of calculating the width direction brightness difference β which is the sum of the brightness differences of
A crack width calculation step of calculating the crack width ω by the following equation based on the maximum luminance difference α, the luminance difference β in the width direction, the crack angle θ, the resolution γ of the image, and the coefficient k.
ω = κ × β / α × γ / cos θ
The crack extraction method according to claim 8, wherein the crack extraction method is provided.

Images