JP2011163866A - Concrete image extraction method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To remove a background and/or an obstacle to be an error, other than a concrete surface, from an image of a concrete structure including the background and the obstacle. <P>SOLUTION: A pixel having the same color as the concrete structure and having the same texture as a texture of the concrete structure is extracted by a structure extraction portion 131 as a concrete candidate region, an edge of the obstacle which exists in the concrete candidate region as a foreground is extracted from the concrete candidate region using a Canny edge extraction method, each texture amount S<SB>k</SB>of 16 neighborhood pixels to a central pixel is calculated using a brightness value of each pixel in the concrete candidate region, the central pixel is extracted as a concrete region when a neighborhood pixel having the maximum amount of texture is in a position of k=9 and a variance value of the amount of 16 textures is 0.0001 or more, and the concrete region is enlarged until the region reaches an edge of the obstacle using the Dilation region expanding method of morphology. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a method for diagnosing a concrete structure using an image. When detecting cracks generated on the surface of concrete structures such as buildings and utility poles by image processing, if an object other than the concrete surface is captured in the image, this is the main cause of false detection of cracks. The present invention relates to a method of eliminating only these erroneous detection factors and extracting only pixels constituting a concrete surface from a captured image.

  At present, anxiety about concrete structures such as aging highways constructed in the post-war high growth period and poorly constructed condominiums has become a social problem. One of the most important items in inspecting the durability of concrete structures is the inspection of cracks on the concrete surface. Cracks are caused by breakage of concrete due to stress applied to a concrete structure by an earthquake or the like. When cracks occur on the concrete surface, rainwater or the like enters from the cracked part, and when it reaches the reinforcing bar inside the concrete, corrosion starts, and the strength of the reinforcing bar decreases, causing a decrease in the durability of the concrete structure. In order to prevent such a decrease in the durability of the concrete structure, it is important to detect cracks on the concrete surface at an early stage and appropriately perform repair and reinforcement.

  When determining the degree of cracking that requires repair, the “width” of the crack appearing on the concrete surface is often used as an index. “Depth”, which is directly related to the arrival of the reinforcing bar, is also an important indicator, but since the depth can be predicted to some extent from the crack width, it is common to use the crack width on the surface to determine whether repairs are necessary or not. It is the target. The allowable crack width for the durability of concrete structures varies slightly depending on the proposed standards in each country because the geology and environmental characteristics differ depending on the geographical location. In general, the width of 0.3 to 0.4 mm is defined as the allowable crack width in dry air, and a value smaller than that is defined in the wet air. Further, cracks whose width has reached 0.2 mm or more are subjected to repair with a filler or coating, or reinforcement with a steel anchor or the like (see Non-Patent Document 1).

  In the current crack diagnosis, visual inspection is the mainstream, but even a skilled worker is difficult to maintain a concentration that does not overlook sub-millimeter width cracks in long-time work. Therefore, it is desired to automate the crack inspection by machines, but from the viewpoint of easy acquisition of equipment and ease of on-site work, a digital camera is used to remotely photograph the wall surface of a concrete structure, and the image is processed by image processing software. The approach of trying to detect cracks is considered promising.

  On the other hand, several algorithms for detecting cracks from an image of a concrete wall by image processing have already been provided (see Non-Patent Document 2), and such existing algorithms generally use feature quantities such as edges. Cracks are detected. An edge is a contour line generated by a sharp change in the density of adjacent pixels, and cracks appear as streaky shadow lines in the image, so edges are regarded as unique features indicating cracks. be able to. Therefore, by using these existing algorithms, the automatic detection of cracks functions effectively for an image in which cracks to be detected only on the wall surface of the concrete structure.

"Concrete crack investigation, guidelines for repair and reinforcement", edited by Japan Concrete Institute, 2003, p274-279 Soichi Oka and three others, "Concrete crack extraction from millimeter-wave near-field images by neural network", Basic and Boundary Park Proceedings, 2009 IEICE Fundamental and Boundary Society Conference, A-4-3, 2009 Year, p66 Supervised by Mikio Takagi, 1 outside, “Image Processing Handbook”, The University of Tokyo Press, p28-29, p516-523, p550-565

  However, in actual sites, there are almost complicated situations where the background and the obstacles that exist behind the concrete structure, such as windows and sashes for buildings and cables and scaffolding nails for utility poles, are complicated. Therefore, it is difficult to photograph only the concrete that is subject to crack inspection. Therefore, when using concrete images taken with such complicated backgrounds and obstacles, the contour edges of those backgrounds and obstacles are mistakenly detected as cracks. There was a problem of not functioning well.

  In order to overcome these problems, it is necessary to remove the background and obstacles from the image before attempting to detect cracks. It is difficult to automatically remove various backgrounds and obstacles by applying a pattern recognition technique such as 3.

  For the above reasons, there is a demand for a technique for detecting cracks after deleting a background or an obstacle other than a concrete surface from an image of a concrete structure including a background or an obstacle.

  The present invention has been made in view of the above problems, and removes backgrounds and obstacles other than the concrete surface that cause errors from the image of the concrete structure prior to detecting cracks on the concrete surface using the edge. The task is to do.

  According to the first aspect of the present invention, a first step of capturing a concrete structure to be inspected with a digital camera by a computer and a storage unit storing a captured image including the captured concrete structure. The third step of reading the captured image from the storage means and determining a pixel of the same color as the color of the concrete structure and a pixel of a color other than the color of the concrete structure; A fourth step of reading the captured image from the storage means and determining pixels of the same texture as the texture of the concrete structure and pixels of a texture other than the texture of the concrete structure; and the color of the concrete structure It is judged as the same color and the same texture as that of the concrete structure. A fifth step of extracting the detected pixels as concrete candidate regions, a sixth step of extracting foreground edges existing in the concrete candidate regions from the concrete candidate regions using a predetermined edge extraction method, and the concrete Using the luminance value of each pixel in the candidate area, the texture amount of a plurality of neighboring pixels for an arbitrary pixel is calculated, the neighboring pixel having the maximum texture amount is at a predetermined position, and the plurality of texture amounts When the variance value is equal to or larger than a predetermined specified value, the seventh step of extracting the one arbitrary pixel as a concrete region and the predetermined region enlargement method are used to reach the concrete region to the foreground edge. And an eighth step of expanding the concrete area to a ninth step. And butterflies.

  According to the present invention, a concrete structure to be inspected is imaged with a digital camera, a pixel having the same color as the color of the concrete structure and a pixel having a color other than the color of the concrete structure are determined, and the concrete structure The pixel of the same texture as the texture of the texture and the pixel of the texture other than the texture of the concrete structure are determined, and determined as the same color as the color of the concrete structure, and also determined as the same texture as the texture of the concrete structure Extract a pixel as a concrete candidate area, extract the foreground edge in the concrete candidate area from the concrete candidate area using a predetermined edge extraction method, and use the brightness value of each pixel in the concrete candidate area Calculate the texture amount of multiple neighboring pixels for one pixel When a neighboring pixel having a texture amount is at a predetermined position and a variance value of a plurality of texture amounts is equal to or larger than a predetermined specified value, an arbitrary pixel is extracted as a concrete region, and a predetermined region expansion method is used. Thus, the concrete area is expanded until it reaches the foreground edge, and the expanded concrete area is extracted, so that it is possible to remove backgrounds and obstacles other than the concrete surface from the captured image of the concrete structure. Thereby, it becomes possible to reliably reduce the risk of overlooking cracks in the current visual inspection.

According to a second aspect of the present invention, the fourth step selects and selects n ² pixels that form n vertically and horizontally n (n is a positive odd number) square from the captured image. In addition, the luminance of each pixel is normalized so that the luminance of the pixel having the highest luminance among the n ² pixels is 1, and the standard deviation of the normalized luminance of the n ² pixels is set as the square. It is determined for each pixel of the captured image that the texture value of the center pixel is set, and the determination is made based on whether the texture value of each pixel is included in a predetermined range preset as the texture value of the concrete structure It is characterized by performing.

According to a third aspect of the present invention, the plurality of neighboring pixels are separated from the one arbitrary pixel having the luminance value i by a neighboring pixel δ _{k = 1} that is one pixel away in the right direction and two pixels that are separated in the right direction. A neighboring pixel δ _{k = 2} , a neighboring pixel δ _{k = 3} that is three pixels away in the right direction, a neighboring pixel δ _{k = 4} that is four pixels away in the right direction, and one pixel in the right direction and one in the upward direction. Neighboring pixels δ _{k = 5} apart, neighboring pixels δ _{k = 6} apart two pixels in the right direction and two pixels upward, neighboring pixels δ three pixels in the right direction and three pixels away in the upward direction _{k = 7} , neighboring pixel δ _{k = 8} , four pixels away in the right direction and four pixels away in the upward direction, neighboring pixel δ _{k = 9} , one pixel away in the upward direction, and neighboring pixels two pixels away in the upward direction a pixel [delta] _{k = 10,} the neighboring pixel [delta] _{k = 11} away three pixels upward, the neighboring pixels [delta] _{k = 12} spaced four pixels upward one pixel in the left direction Is a neighboring pixel [delta] _{k = 13} spaced one pixel upward, away second pixel in the left direction, and neighboring pixels [delta] _{k = 14} apart second pixel upward, away three pixels to the left, upward three The probability that the neighboring pixel δ _{k = 15 that} is separated by pixels, the neighboring pixel δ _{k = 16 that} is separated by four pixels in the left direction and four pixels away in the upward direction, and that the luminance value of each neighboring pixel is j is P _{When δk} (i, j) is used, the seventh step is:

Using the above equation (1) to calculate the texture amount S _k of each neighboring pixel, respectively, a maximum value texture amount S _k is at k = 9, and the variance of the texture amount S _k 0. In the case of 0001 or more, the arbitrary one pixel is extracted as a concrete region.

  The present invention described in claim 4 is characterized in that the first step performs imaging while focusing on the concrete structure to be inspected.

  The fifth aspect of the present invention is characterized in that the fifth step extracts, as the concrete candidate region, a region having a density per unit area equal to or higher than a predetermined specified value among the concrete candidate regions. .

  According to a sixth aspect of the present invention, the concrete structure has a cylindrical shape, and the first step is such that the optical axis of the lens of the digital camera is orthogonal to the rotational axis of the column, And the said concrete structure is imaged so that the up-down direction of an image may correspond to the said rotating shaft, It is characterized by the above-mentioned.

  In the seventh aspect of the present invention, in the fifth step, when the histogram of the number of pixels in the vertical direction of the image in the concrete candidate region is greater than or equal to a predetermined threshold value, All are extracted as the concrete candidate regions.

  The present invention according to claim 8 is characterized in that the sixth step uses a Canny edge extraction method.

  The present invention according to claim 9 is characterized in that the eighth step uses a morphological dilation region expansion method.

  According to the present invention, it is possible to remove backgrounds and obstacles other than a concrete surface from a captured image of a concrete structure. Thereby, it becomes possible to reliably reduce the risk of overlooking cracks in the current visual inspection.

It is a figure which shows the functional block of the image processing algorithm which comprises a crack detection system. It is a figure which shows the functional block of the image processing algorithm which comprises a search area extraction part. It is a figure which shows an example of the captured image which imaged the concrete structure. It is a figure which shows the result of having processed the picked-up image shown in FIG. 3 by commercially available crack detection software. It is a figure explaining the process of a color determination part. It is a figure explaining the process of a texture determination part. It is a figure explaining the process of a structure determination part. It is a figure which shows the extraction result of the edge of an obstruction. It is a figure explaining the process which determines the filling start point of a concrete pixel. It is a figure which compares and shows the texture amount in a concrete surface and a non-concrete surface. It is a figure explaining the process of a concrete pixel extraction part. It is a figure which shows the extraction result of a concrete surface.

  As described above, a technique has already been disclosed in which a concrete structure such as a building or a utility pole is imaged with a digital camera or the like, and a crack generated on the surface of the concrete structure is detected by image processing. However, when an object other than a concrete surface is imaged together in the image, the object is a main factor of crack detection error. Therefore, before detecting cracks, it is necessary to remove image elements other than the concrete surface from the captured image. Hereinafter, in the present embodiment, a method for removing the background existing behind the concrete structure and the obstacle existing as the foreground in front of the concrete structure will be described.

  The concrete image extraction method according to the present embodiment does not extract and remove pixels having features of the background and obstacles that are not the inspection target from the image captured by the digital camera, but the concrete structure that is the inspection target. It is characterized in that pixels having the above characteristics are extracted and pixels that do not match the pixels are removed. Further, at the time of imaging, it is also characterized in that imaging is performed so that pixels on the concrete structure to be inspected and features on the image represent the characteristics of the concrete structure to the maximum. Hereinafter, an image processing algorithm of the concrete image extraction method according to the embodiment will be described.

  FIG. 1 is a diagram showing functional blocks of an image processing algorithm constituting a crack detection system. The crack detection system 1 according to the present embodiment includes an imaging unit 11, an image storage unit 12, a search area extraction unit 13, a crack search unit 14, and a display unit 15.

  The imaging unit 11 has a function of capturing an image of a concrete structure to be inspected with a digital camera while focusing on the concrete structure.

  The image storage unit 12 has a function of storing a captured image having a concrete structure imaged by the imaging unit 11 in a readable manner. Such an image storage unit 12 can be realized by a storage device such as a memory or a hard disk.

  The exploration area extraction unit 13 has a function of reading a captured image from the image storage unit 12 and extracting a pixel (concrete area) of a concrete surface that is a crack exploration range from the captured image showing a concrete structure. Yes.

  The crack search unit 14 has a function of searching for a crack location using an algorithm disclosed in Non-Patent Document 2 or the like using the concrete region extracted by the search region extraction unit 13 as a search range.

  The display unit 15 has a function of displaying the crack location searched by the crack search unit 14 so that the user of the crack detection system 1 can visually recognize the crack.

  Next, the search area extraction unit 13 will be described in detail. FIG. 2 is a diagram showing functional blocks of an image processing algorithm constituting the search area extraction unit. The exploration area extraction unit 13 includes a color determination unit 131a, a texture determination unit 131b, and a structure determination unit 131c. The structure extraction unit 131 extracts pixels having the characteristics of the concrete structure to be inspected. The obstacle edge extracting unit 132a and the concrete pixel extracting unit 132b include a concrete surface extracting unit 132 that extracts only the concrete surface from the pixels extracted by the structure extracting unit 131. Hereinafter, the function of each of the functional units constituting the exploration region extraction unit 13 will be described in detail using a captured image of the utility pole shown in FIG. 3 as an example of a concrete structure.

  FIG. 4 shows the result of processing the captured image shown in FIG. 3 with commercially available crack detection software. As can be seen from the enlarged view of FIG. 4, buildings, trees, belts, cables, scaffolding nails and the like are erroneously detected as cracks. Therefore, it is necessary to remove these obstacles before the crack detection unit 14 performs the crack detection process.

  First, the structure extraction unit 131 will be described. FIG. 5 is a diagram illustrating the processing of the color determination unit. The color determination unit 131a uses a color of each pixel constituting the captured image (for example, a luminance value, a gray value, a saturation value, etc.) and a color pixel having the possibility of a concrete surface, and other colors. Are determined and extracted. Generally, since the color of the concrete surface is gray, for example, the color saturation is expressed using the CIE Lab color system (an example of a color evaluation system) shown in FIG. A pixel close to 0 is determined as a concrete surface. FIG. 5B shows a color determination result. A determination result 1 is assigned to a pixel having a concrete surface color, and a determination result 0 is assigned to a pixel having no concrete surface color.

  Here, since the captured image includes a house of the same color as that of the concrete structure, it is difficult to extract the utility pole that is the concrete structure only from the color determination result. Therefore, in order to accurately extract only the utility pole, it is necessary to add and use texture information described below in addition to the color information.

  FIG. 6 is a diagram for explaining the processing of the texture determination unit. The texture determination unit 131b determines a texture pixel having the possibility of a concrete surface and other texture pixels and extracts them. For example, the texture determined to have a high probability of a concrete surface is determined as follows.

  First, the luminance is normalized with respect to a certain arbitrary pixel (center pixel) and each adjacent pixel (see FIG. 6A) in the vicinity of the periphery 8 adjacent to the center pixel in the vertical and horizontal directions. . Specifically, the luminance of each pixel is normalized so that the luminance of the pixel with the highest luminance among the nine pixels including the central pixel is 1.

  Next, with respect to these nine pixels, the texture is digitized by the standardized standard deviation of luminance, and the standardized standard deviation of luminance of nine pixels is set as the texture value of the central pixel. The digitized result is shown in FIG. As the normalized texture value of the standard deviation of luminance is closer to 0, an image of an object having a smoother surface appears.

  As a result, when the standard deviation value of the concrete surface is analyzed, it falls within 0.05 or more and less than 0.10. Therefore, the determination result 1 is assigned to the pixels having the texture dispersion value in this range, and the determination result 0 is assigned to the other pixels. Assign. That is, a pixel whose determination result is 1 by the texture determination unit 131b has a possibility of being a concrete image.

In the above description, 9 square pixels each consisting of 3 vertical pixels × 3 horizontal pixels have been described as an example, but the number of pixels used for texture determination is a square of n vertical pixels × n horizontal pixels (n is a positive odd number). N ² pixels may be used. That is, n ² pixels of vertical n × horizontal n are selected, and the standard deviation of luminance (the luminance of each pixel is set so that the luminance of the largest pixel is 1) for the selected n ² pixels. If the result is within 0.05 or more and less than 0.1, for example, the judgment result 1 is applied to the center pixel, and the result does not fall within the above range. Judgment result 0 is applied. Increasing the value of n is expected to result in a smoother image after processing, although the statistical parameter in the standard deviation increases and the determination processing speed decreases.

  The imaging unit 11 described above makes the optical axis of the lens of the digital camera orthogonal to the rotation axis of the cylinder when the concrete structure to be inspected has a cylindrical shape such as a utility pole, and the imaging of the digital camera. It is also possible to image the utility pole so that the vertical direction of the image coincides with the rotation axis. In this case, since the utility pole is imaged perpendicular to the horizontal direction of the image frame, the boundary between the utility pole and the background is also perpendicular to the horizontal direction of the image frame. Thus, the boundary between the pixel where the pixel having the possibility of the concrete surface obtained from the processing result of the texture determination unit 131b is distributed and the other pixel is also made perpendicular to the horizontal direction of the image frame. be able to.

  FIG. 7 is a diagram illustrating the processing of the structure determination unit. The structure determination unit 131c multiplies the output result of the color determination unit 131a and the output result of the texture determination unit 131b (takes an AND) to extract a concrete candidate region to be inspected. FIG. 7A shows the multiplication result, and pixels having the possibility of a concrete surface are extracted as concrete candidate regions. FIG. 7B shows a vertical (vertical) histogram obtained from this image. As shown in FIG. 7C, separation processing (threshold cut) is performed with a specific threshold. A utility pole can be extracted. Note that, for example, half of the peak value can be used as the specific threshold value used for the distributed processing.

  Note that if the image is taken so that the boundary between the concrete structure and the background is perpendicular to the horizontal direction of the image frame, if the histogram of the number of pixels in the vertical direction is greater than or equal to a specific threshold, You may make it determine all the pixels of a direction as a pixel of a concrete structure. In this case, although the imaging method is limited, the determination processing calculation is simplified, so that a reduction in erroneous determination and a reduction in processing cost can be expected.

  In addition, when the concrete structure is not a cylinder, when the optical axis of the digital camera lens is not orthogonal to the rotation axis of the cylinder, or when the image of the digital camera is not taken so that the vertical direction matches the rotation axis, The boundary between the pixel where the pixel having the possibility of the concrete surface obtained from the above processing result is distributed and the other pixel is not always perpendicular to the horizontal direction of the image. In such a case, an area where the density per unit area of the concrete candidate area having the possibility of a concrete surface obtained from the multiplication result may be determined as a concrete candidate area. In this case, although the judgment calculation process is advanced, there is an advantage that there is no restriction on the imaging method.

  Even if other concrete structures are reflected in the background, the focus is focused on the object to be inspected, so the concrete structure in the background is blurred and only the object in front is inspected. Can be extracted. As a result, compared with the pixel of the concrete structure to be inspected in focus, the pixel of the background concrete structure has a smaller difference in luminance from the surrounding pixels, and the standard deviation is also reduced. Therefore, the background concrete structure can be excluded by the texture value calculated in the texture determination unit 131b. In this way, by using the calculation based on the standard deviation of the texture determination unit 131b, the inspection target concrete structure can be moved from the background concrete structure only by adjusting the focus (focus) of the digital camera to the inspection target concrete structure. There is an effect that it can be easily separated.

  The above is a description of the method of extracting and extracting the concrete structure to be inspected from the background image. It is impossible to separate and eliminate obstacles existing as a foreground in a concrete structure from the concrete structure by the above method alone, and can be realized by combining with the methods described below.

  Next, the concrete surface extraction unit 132 will be described. According to the processing result of the structure extraction unit 131 shown in FIG. 7C, only the electric pole area excluding the background area is extracted. However, erroneous detection of cracks in cables, belts, scaffolding nails, etc. in front of the electric pole. Since the obstacle which becomes the factor of the cover is covered, it is still incomplete to perform the crack detection processing in the crack search unit 14 in the subsequent stage. In order to extract only the concrete surface while avoiding these obstacles, the obstacle edge extraction unit 132a first extracts the edge of the obstacle that exists as the foreground. This edge extraction process can be realized by using, for example, the Canny edge extraction method. FIG. 8 shows an edge extraction result.

Next, the concrete pixel extraction unit 132b extracts concrete pixels. In the concrete candidate region extracted by the structure extraction unit 131, as shown in FIG. 9, each of the 16 neighboring pixels δ _{k = 1 to 16} (center) near one arbitrary pixel (center pixel) A Neighboring pixels δ _{k = 1} one pixel away from the pixel A in the right direction, neighboring pixels δ _{k = 2} two pixels away in the right direction, neighboring pixels δ _{k = 3} three pixels away in the right direction, and four pixels rightward neighboring pixels [delta] _{k = 4} apart _pixels, spaced one pixel in the right direction, neighboring pixel [delta] _{k = 5} away one pixel _upward, away second pixel to the right, near distant second pixel upward pixel [delta] _{k = 6.} Neighboring pixel δ _{k = 7} , three pixels away in the right direction, three pixels away in the upward direction, Neighboring pixel δ _{k = 8} , four pixels away in the right direction, four pixels away in the upward direction, one pixel away in the upward direction neighboring pixels [delta] _{k = 9 was,} the neighboring pixels [delta] _{k = 10} km second pixel _upward, near image away three pixels upward [delta] _{k = 11,} neighboring pixels [delta] _{k = 12} km four pixels _upward, away one pixel in the left direction, neighboring pixel [delta] _{k = 13} km one pixel _upward, away second pixel to the left, upwards second pixel neighboring pixels [delta] _{k = 14} km _away three pixels in the left direction, neighboring pixel [delta] _{k = 15} km three pixels _upward, away four pixels in the left direction, neighboring pixel away four pixels upward [delta] _{k = 16)} using each brightness value of the texture amount S _k called homogeneity which is defined by the following equation (1) is calculated for the center pixel a.

However, k (= 1, 2,..., 16) is a position of a neighboring pixel. P _δ (i, j) is a probability that the luminance value of a neighboring pixel δ _k that is δ (r, θ) away from the central pixel A of the luminance value i is j. r is the distance between the center pixel A and the neighboring pixels [delta] _k, theta is the angle of neighboring pixels [delta] _k with respect to the horizontal direction of the center pixel A viewed image. In general, the luminance values of the i and j pixels are 256 gradations, but it is desirable to compress them to 8 gradations in order to shorten the calculation time. As for the specific method of calculating the texture amount S _k, since that is described in 517, pp ～521 Non-Patent Document 3, where the calculation process is omitted.

Here, the neighboring pixels 16 places when calculating the formula (1) can be obtained 16 texture amount S _k for the central pixel A as a scalar quantity. FIGS. 10 (a) and 10 (b) is a plot of the relation between the position k and the texture amount _{S k} neighboring pixels [delta] _k, 10 (a) is a concrete surface, FIG. 10 (b) Represents an area with a non-concrete surface. As can be seen from FIG. 10 (a), the concrete surface, texture amount S is characterized by a maximum in the vicinity pixel [delta] _9.

Therefore, it is defined that “the center pixel A is a concrete pixel when the texture amount S is maximum when the position k = 9 and the variance value of the texture amount S is 0.0001 or more”. Supplementally, in the amount of unevenness is small non concrete surface, since the value of the texture amount S _k becomes totally flats, texture amount S when k = 9 is to be a accidentally maximum narrowly likely, Texture the amount S ₉ is more than just the maximum is insufficient to define the characteristics of the concrete surface. Therefore, a condition that “the dispersion value of the texture amount S is 0.0001 or more” in the subsequent stage is necessary.

  From the utility pole area extracted by the structure extraction unit 131, a pixel corresponding to the above definition is extracted as a concrete area. As shown in FIG. 11, the concrete pixel extraction unit 132b starts filling the extracted pixel. As a point, the starting point is expanded using a region expansion method called Morphological Dilation. The expansion of the pixel is stopped when it collides with the edge of the obstacle extracted by the obstacle edge extracting unit 132a. Thereby, it is possible to extract only the concrete surface excluding the obstacle in the concrete structure region.

  Note that the different texture calculation performed by the structure extraction unit 131 and the concrete surface extraction unit 132 is due to a difference in calculation cost. In order to process the entire captured image, the structure extraction unit 131 uses standard deviations in the vicinity of 8 with a low calculation cost. On the other hand, since the concrete surface extraction unit 132 only searches for the start point of filling several times, a high-precision co-occurrence matrix of 16 is used although the calculation cost is high.

  FIG. 12 is a diagram illustrating the extraction result of the concrete surface. As described above, the crack search unit 14 performs crack detection using only the concrete region extracted by the search region extraction unit 13 as a processing range target using the algorithm disclosed in Non-Patent Document 2. Thereafter, the display unit 15 displays the detection result using a monitor of a personal computer.

  Next, a concrete image extraction process flow of the crack detection system will be described. First, a concrete structure to be inspected is imaged with a digital camera by the imaging unit 11 (S1).

  Next, the image storage unit 12 stores a captured image having the concrete structure imaged in S1 (S2).

  Next, the color determination unit 131a reads a captured image from the image storage unit 12, and determines pixels having the same color as the color of the concrete structure and pixels having a color other than the color of the concrete structure (S3). ).

  Next, the texture determination unit 131b reads a captured image from the image storage unit 12, and determines pixels having the same texture as the texture of the concrete structure and pixels having a texture other than the texture of the concrete structure (S4). ).

  Subsequently, the pixel determined as the same color as the color of the concrete structure in S3 by the structure determination unit 131c and determined as the same texture as the texture of the concrete structure in S4 is extracted as a concrete candidate region (S5). ).

  Next, the obstacle edge extraction unit 132a uses the Canny edge extraction method to extract the edge of an obstacle existing as a foreground in the concrete candidate area from the concrete candidate area extracted in S5 (S6).

Then, the concrete pixel extracting section 132b, a texture amount S _k of 16 neighboring pixels for any center pixel respectively calculated using the luminance value of each pixel of the concrete candidate region, neighboring pixels having the maximum texture amount Is at position k = 9 and the variance value of the 16 texture amounts is 0.0001 or more, the central pixel is extracted as a concrete region (S7).

  Next, the concrete pixel extraction unit 132b expands the concrete region extracted in S7 until reaching the edge of the obstacle extracted in S6 using the morphological dilation region expansion method (S8).

  Next, the concrete area extracted in S8 is extracted by the concrete pixel extraction unit 132b (S9).

  Next, the crack detection unit 14 performs crack detection on the concrete area extracted in S9 (S10).

  Finally, the crack detection result of S10 is displayed on the monitor by the display unit 15 (S11).

  As described above, by using this embodiment, it is possible to detect a crack after deleting a background or an obstacle other than a concrete surface from an image of a concrete structure.

  Finally, the crack detection system 1 can be configured by a computer having an arithmetic processing device such as a CPU and a storage device such as a memory, and the processing of each unit is executed by a program. It should be noted that this program is stored in a storage device and can be recorded on a recording medium or provided through a communication network.

According to the present embodiment, a concrete structure to be inspected is imaged with a digital camera, and is determined as the same color as the color of the concrete structure using the captured image, and is the same as the texture of the concrete structure. Pixels determined as textures are extracted as concrete candidate areas, the edges of obstacles existing as foregrounds in the concrete candidate areas are extracted from the concrete candidate areas using the Canny edge extraction method, and each pixel in the concrete candidate area is extracted. there of the 16 neighboring pixels for any central pixel using the luminance value of the texture amount S _k respectively calculated, in the vicinity of the pixel position k = 9 having the greatest texture amount, the dispersion of 16 texture amount If the value is greater than or equal to 0.0001, the center pixel is extracted as a concrete area and the morph The area of the concrete is expanded until it reaches the edge of the obstacle using the dilation area expansion method of the topology, and the enlarged concrete area is extracted, so the background and obstacles other than the concrete surface are removed from the captured image of the concrete structure It becomes possible to do. Thereby, it becomes possible to reliably reduce the risk of overlooking cracks in the current visual inspection.

DESCRIPTION OF SYMBOLS 1 ... Crack detection system 11 ... Imaging part 12 ... Image storage part 13 ... Exploration area extraction part 131 ... Structure extraction part 131a ... Color determination part 131b ... Texture determination part 131c ... Structure determination part 132 ... Concrete surface extraction part 132a ... Obstacle edge extraction unit 132b ... Concrete pixel extraction unit 14 ... Crack search unit 15 ... Display unit S1 to S11 ... Step

Claims (9)

By computer
A first step of imaging a concrete structure to be inspected with a digital camera;
A second step of storing a captured image having the captured concrete structure in a storage means;
A third step of reading the captured image from the storage means and determining a pixel of the same color as the color of the concrete structure and a pixel of a color other than the color of the concrete structure;
A fourth step of reading the captured image from the storage means and determining a pixel of the same texture as the texture of the concrete structure and a pixel of a texture other than the texture of the concrete structure;
A fifth step of extracting, as a concrete candidate region, pixels determined as the same color as the color of the concrete structure and determined as the same texture as the texture of the concrete structure;
A sixth step of extracting a foreground edge existing in the concrete candidate region from the concrete candidate region using a predetermined edge extraction method;
Using the luminance value of each pixel in the concrete candidate region, calculate the texture amount of a plurality of neighboring pixels for an arbitrary pixel, and the neighboring pixel having the maximum texture amount is at a predetermined position, and the plurality of neighboring pixels A seventh step of extracting the arbitrary one pixel as a concrete region when the variance value of the texture amount is equal to or greater than a predetermined specified value;
An eighth step of enlarging the concrete area using a predetermined area enlarging method until the foreground edge is reached;
A ninth step of extracting the expanded concrete area;
A concrete image extraction method comprising: The fourth step includes
N vertical pieces horizontal n pieces from the captured image (n is a positive odd number) and select the n ² pixels constituting the square, the luminance of the large pixel of the brightest of the n ² pixels selected For each pixel of the captured image, the luminance of each pixel is normalized so as to be 1, and the standard deviation of the normalized luminance of the n ² pixels is used as the texture value of the central pixel of the square. ,
The concrete image extraction method according to claim 1, wherein the determination is performed based on whether the texture value of each pixel is included in a predetermined range set in advance as a texture value of the concrete structure. The plurality of neighboring pixels are obtained from the one arbitrary pixel having a luminance value i,
Neighboring pixels δ _{k = 1} one pixel away in the right direction,
Neighboring pixels δ _{k = 2} separated by two pixels in the right direction,
Neighboring pixels δ _{k = 3} three pixels away in the right direction,
Neighboring pixels δ _{k = 4} four pixels away in the right direction,
Neighboring pixels δ _{k = 5} one pixel away in the right direction and one pixel away in the upward direction,
Neighboring pixels δ _{k = 6} , two pixels away in the right direction and two pixels away in the upward direction,
Neighboring pixels δ _{k = 7} that are 3 pixels apart in the right direction and 3 pixels away in the upward direction,
Neighboring pixels δ _{k = 8} that are 4 pixels apart in the right direction and 4 pixels away in the upward direction,
Neighboring pixels δ _{k = 9} one pixel away in the upward direction,
Neighboring pixels δ _{k = 10} two pixels away in the upward direction,
Neighboring pixels δ _{k = 11} three pixels away in the upward direction,
Neighboring pixels δ _{k = 12} , four pixels away in the upward direction,
Neighboring pixels δ _{k = 13} one pixel away in the left direction and one pixel away in the upward direction,
Neighboring pixels δ _{k = 14} two pixels away in the left direction and two pixels away in the upward direction,
Neighboring pixels δ _{k = 15} three pixels away in the left direction and three pixels away in the upward direction,
Neighboring pixels δ _{k = 16} that are 4 pixels apart in the left direction and 4 pixels away in the upward direction,
Because
When the probability that the luminance value of each neighboring pixel is j is P _δk (i, j),
The seventh step includes
Using the above equation (1) to calculate the texture amount S _k of each neighboring pixel, respectively, a maximum value texture amount S _k is at k = 9, and the variance of the texture amount S _k 0. The concrete image extraction method according to claim 1, wherein in the case of 0001 or more, the one arbitrary pixel is extracted as a concrete region. The first step includes
The concrete image extraction method according to claim 1, wherein the concrete structure to be inspected is focused and imaged. The fifth step includes
5. The concrete image extraction method according to claim 1, wherein an area having a density per unit area equal to or greater than a predetermined specified value is extracted as the concrete candidate area among the concrete candidate areas. . The concrete structure has a cylindrical shape,
The first step includes
6. The concrete structure is imaged so that the optical axis of the lens of the digital camera is orthogonal to the rotation axis of the cylinder and the vertical direction of the image coincides with the rotation axis. The concrete image extraction method according to any one of the above. The fifth step includes
The all-vertical pixel in the captured image is extracted as the concrete candidate region when the histogram of the number of pixels in the vertical direction of the image in the concrete candidate region is greater than or equal to a predetermined threshold value. Concrete image extraction method as described in 4. The sixth step includes
The concrete image extraction method according to claim 1, wherein a Canny edge extraction method is used. The eighth step includes
The concrete image extraction method according to claim 1, wherein a morphological dilation area expansion method is used.