Crack detection method, crack detection program, and crack detection system

Abstract

To provide a crack detection method, a crack detection program and a crack detection system which detect a crack occurrence state of concrete in a shorter time with higher accuracy in comparison to a conventional technique.SOLUTION: A crack detection method comprises: a crack occurrence region extraction step S11 of extracting a crack occurrence region by analyzing an analysis target region image BRG; a crack binarized image creation step S12 of analyzing the crack occurrence region in detail, binarizing the crack occurrence region in an analysis target region image BRG and creating a crack binarized image H2G discriminating between a crack portion and a non-crack portion; a crack element information creation step S13 of creating crack element information HYJ by dividing the crack portion in the crack binarized image into a plurality of crack elements that do not have a branch point; and a crack related information creation step S14 of creating and outputting crack related information HKJ including a crack damage feature amount on the basis of the crack element information HYJ.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a crack detection method, a crack detection program, and a crack detection system for detecting crack occurrence conditions based on image data of the surface of a concrete structure such as a bridge or tunnel. [Background technology]

[0002] Conventionally, as disclosed in Patent Document 1, for example, there has been an image processing method that includes the steps of receiving an image of the surface of a concrete structure, analyzing the received image to detect cracks on the surface of the structure, and creating crack information that indicates a crack image corresponding to the detected crack and the characteristic quantities of the detected crack. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2020-38227 Summary of the Invention [Problem to be solved by the invention]

[0004] The image processing method in Patent Document 1 is configured to perform image analysis of a wide area of ​​the surface of a structure, but if this is done uniformly and without omission, the processing load is large for high-resolution images, and it takes time to detect cracks and calculate feature values. Also, if the image quality is reduced, the volume of image data can be reduced and the analysis time can be shortened, but the accuracy of the detected crack feature values ​​decreases.

[0005] The present invention has been made in consideration of the above-mentioned background technology, and aims to provide a crack detection method, a crack detection program, and a crack detection system that can detect crack occurrence in concrete in a shorter time and with higher accuracy than conventional methods. [Means for solving the problem]

[0006] The present invention is a method for detecting crack occurrence status based on an analysis target area image obtained by capturing an analysis target area on the surface of a concrete structure, the method being executed by a computer system to which the analysis target area image is input, This crack detection method includes a crack occurrence area extraction step of analyzing the analysis area image to extract a crack occurrence area in which a crack and its surrounding area are imaged; a crack binary image creation step of performing a binarization process on the crack occurrence area in the analysis area image to distinguish it into cracked and non-cracked areas, estimating the actual crack layout, and creating a crack binary image that reflects the estimation result; a crack element information creation step of dividing the crack area in the crack binary image at characteristic points in the longitudinal direction of the crack area, thereby dividing it into multiple crack elements that do not have branch points, and creating crack element information that is information about each of the crack elements; and a crack-related information creation step of creating and outputting crack-related information that indicates the crack occurrence status of the concrete structure, the information including crack damage feature values ​​calculated based on the crack element information.

[0007] The crack occurrence area extraction step is configured to include an analysis area image division step for dividing the analysis area image into small images; a crack small image extraction step for using multiple sample images of the surface of an arbitrary concrete structure, labeled as whether they are images of cracks, as training data, and acquiring or creating a trained model created by machine learning based on the training data, and using the trained model to extract multiple crack small images that are recognized as images of cracks from the multiple small images; and a crack occurrence area identification step for designating the area where the crack small images are combined as the crack occurrence area.

[0008] The crack binarized image creation step performs binarization processing on the crack occurrence region in the analysis target region image. and identify the cracked area.Distinguish between crack candidate areas and non-crack candidate areas The areas other than the crack occurrence area in the analysis target area image are uniformly regarded as the non-crack candidate areas. and a crack correction step of performing a process of identifying the cracked portion and the non-cracked portion by estimating the layout of the actual cracked portion by analyzing the cracked portion binary image and correcting the cracked portion based on the estimation result. Make it into a configuration.

[0009] In the crack candidate binary image creation step, it is preferable that the crack occurrence region in the analysis target region image is subjected to a process for increasing the resolution and then subjected to a binarization process.

[0010] In the crack correction step, a connection process is performed to connect the portions of the crack candidate image where the crack candidate portion is interrupted and where it is estimated that the crack is actually continuous without interruption, thereby forming a connection between the crack portion and the non-crack portion. Identify. The connecting process includes an auxiliary thinning step of auxiliary thinning the crack candidate portion in the crack candidate binary image, an endpoint extraction step of extracting endpoints of multiple thin lines that make up the crack candidate portion thinned in the auxiliary thinning step, a connected point extraction step of extracting, for each thin line, an endpoint of another thin line that is closest to the thin line itself and whose distance between the endpoints is equal to or less than a specified value, and setting the own endpoint and the other endpoint as a pair of connected points, a connected region binary image creation step of setting an area of ​​a predetermined size sandwiched between the pair of connected points as a connected region, performing binarization processing on the connected region in the analysis target region image to distinguish it into connected portions and non-connected portions, and treating areas other than the connected region as the non-connected portion, to create a connected region binary image, and a crack connecting step of connecting cracks from the connected region binary image and the crack candidate binary image. Make it into a configuration.

[0011] The crack element information created in the crack element information creation step preferably includes at least one of information on the length and information on the maximum width of the crack element. Also, the crack damage feature calculated in the crack-related information creation step preferably includes at least one of the density of the cracks in the analysis target area, the maximum width of the cracks in the analysis target area, the shape of the cracks in the analysis target area, and the minimum distance between the crack elements spaced apart from each other in the analysis target area.

[0012] The present invention also provides a crack detection program comprising programs for executing each step, for causing a computer system to execute the above-described crack detection method.

[0013] The present invention also provides a crack detection system that detects crack occurrence status based on an analysis target area image obtained by capturing an analysis target area on the surface of a concrete structure, comprising: The crack detection system includes a crack occurrence area extraction unit that analyzes the analysis target area image to extract a crack occurrence area in which a crack and its surrounding area are imaged; a crack binary image creation unit that performs a binarization process on the crack occurrence area in the analysis target area image to distinguish it into cracked and non-cracked areas, estimates the actual crack layout, and creates a crack binary image that reflects the estimated results; a crack element information creation unit that divides the crack areas in the crack binary image at characteristic points in the longitudinal direction of the crack areas, thereby dividing them into multiple crack elements that do not have branch points, and creates crack element information that is information about each of the crack elements; and a crack-related information creation unit that creates and outputs crack-related information that indicates the crack occurrence status of the concrete structure, including crack damage feature amounts calculated based on the crack element information.

[0014] The crack occurrence area extraction unit executes an analysis area image division step of dividing the analysis area image into small images, a crack small image extraction step of using multiple sample images of the surface of an arbitrary concrete structure, labeled as crack images or not, as training data, and acquiring or creating a trained model created by machine learning based on the training data, and using the trained model to extract multiple crack small images that are recognized as crack images from the multiple small images, and a crack occurrence area identification step of identifying the area where the crack small images are combined as the crack occurrence area.

[0015] The crack binarized image creation unit performs binarization processing on the crack occurrence area in the analysis target area image. and identify the cracked area. Distinguish between crack candidate areas and non-crack candidate areas The areas other than the crack occurrence area in the analysis target area image are uniformly regarded as the non-crack candidate areas. and a crack correction step of performing a process of identifying the cracked portion and the non-cracked portion by estimating the layout of the actual cracked portion by analyzing the cracked portion binary image and correcting the cracked portion based on the estimation result. Make it into a configuration.

[0016] In the crack candidate binary image creation step, it is preferable that the crack occurrence region in the analysis target region image is subjected to a process for increasing the resolution and then subjected to a binarization process.

[0017] In the crack correction step, a connection process is performed to connect the portions of the crack candidate image where the crack candidate portion is interrupted and where it is estimated that the crack is actually continuous without interruption, thereby forming a connection between the crack portion and the non-crack portion. Identify.The connecting process includes an auxiliary thinning step of auxiliary thinning the crack candidate portion in the crack candidate binary image, an endpoint extraction step of extracting endpoints of multiple thin lines that make up the crack candidate portion thinned in the auxiliary thinning step, a connected point extraction step of extracting, for each thin line, an endpoint of another thin line that is closest to the thin line itself and whose distance between the endpoints is equal to or less than a specified value, and setting the own endpoint and the other endpoint as a pair of connected points, a connected region binary image creation step of setting an area of ​​a predetermined size sandwiched between the pair of connected points as a connected region, performing binarization processing on the connected region in the analysis target region image to distinguish it into connected portions and non-connected portions, and treating areas other than the connected region as the non-connected portion, to create a connected region binary image, and a crack connecting step of connecting cracks from the connected region binary image and the crack candidate binary image. Make it into a configuration.

[0018] It is preferable that the crack element information created by the crack element information creation unit includes at least one of information on the length and information on the maximum width of the crack element. Furthermore, it is preferable that the crack damage feature calculated by the crack-related information creation unit includes at least one of the density of the cracks in the analysis target area, the maximum width of the cracks in the analysis target area, the shape of the cracks in the analysis target area, and the minimum distance between the crack elements spaced apart from each other in the analysis target area. [Effects of the Invention]

[0019] The crack detection method, crack detection program, and crack detection system of the present invention are configured to first roughly analyze the entire image of a wide area of ​​the surface of a concrete structure (image of the area to be analyzed) to extract the area where cracks have occurred, and then narrow down the analysis to the area where cracks have occurred and its surrounding area for a detailed analysis.This significantly reduces the processing burden compared to analyzing the entire image of the area to be analyzed in detail, and makes it possible to detect the crack occurrence status in a shorter time than conventional methods, and with accuracy equal to or higher than conventional methods. [Brief explanation of the drawings]

[0020] [Figure 1] 1 is a diagram showing a system configuration of an embodiment of a crack detection system of the present invention and an example of an image of an area to be analyzed. [Figure 2] 1 is a flowchart showing the overall flow of an embodiment of a crack detection method of the present invention. [Figure 3] 10A and 10B are diagrams illustrating a specific example of processing performed in an analysis target area image dividing step. [Figure 4] 10A and 10B are diagrams showing a specific example of processing performed in the crack small image extraction step. [Figure 5] Figure 1 shows an example of training data that forms the basis of the trained model used in the crack small image extraction step, with (a) showing a sample image of a crack and (b) showing a sample image of a non-crack. [Figure 6] 10A to 10C are diagrams showing a specific example of processing performed in the crack candidate binary image creation step. [Figure 7] FIG. 10 is a diagram (part 1) showing a specific example of the processing performed in the crack correction step. [Figure 8] FIG. 2 is a diagram (part 2) showing a specific example of the processing performed in the crack correction step. [Figure 9] FIG. 3 is a diagram (part 3) showing a specific example of the processing performed in the crack correction step. [Figure 10] A figure showing a specific example of processing performed in the crack element information creation step. [Figure 11] 10A and 10B are diagrams showing specific examples of crack-related information created in the crack-related information creation step. DETAILED DESCRIPTION OF THE INVENTION

[0021] Hereinafter, an embodiment of the crack detection method, the crack detection program, and the crack detection system of the present invention will be described with reference to the drawings. In this specification, the term "image" includes the concept of "image data." The crack detection system 10 of this embodiment is a computer system that detects the occurrence of cracks H by analyzing an image (analysis target area image BRG) of a certain wide area (analysis target area BR) on the surface of a concrete structure. As shown in Figure 1, the crack detection system 10 includes a crack occurrence area extraction unit 10a, a crack binary image creation unit 10b, a crack element information creation unit 10c, and a crack-related information creation unit 10d.

[0022] The crack detection method of this embodiment is a method executed by a crack detection system 10. Specifically, as shown in Figure 2, it includes a crack occurrence area extraction step S11, a crack binary image creation step S12, a crack element information creation step S13, and a crack-related information creation step S14, and each of steps S11 to S14 is executed by the crack occurrence area extraction unit 10a, the crack binary image creation unit 10b, the crack element information creation unit 10c, and the crack-related information creation unit 10d, respectively. Furthermore, the crack detection program of this embodiment is composed of programs for executing each step, for causing the crack detection system 10 to execute the crack detection method.

[0023] The contents of each of steps S11 to S14 executed by the crack detection system 10 will be explained below in order. <Crack occurrence area extraction step S11> The crack occurrence area extraction step S11 is a step in which the crack occurrence area HR in which the crack H and its surrounding area are imaged is extracted by analyzing the analysis area image BRG. Specifically, the analysis area image division step S111, the crack small image extraction step S112, and the crack occurrence area identification step S113 are executed in sequence.

[0024] In the analysis area image division step S111, the analysis area image BRG is divided into small images SG of a predetermined size, as shown in Figure 3. If necessary, RGB normalization may be performed on each small image SG. RGB normalization adjusts the brightness and saturation of pixels in the small image SG so that they are distributed from maximum to minimum. By performing this process, the next crack small image extraction step S112 can be performed with higher accuracy.

[0025] In the crack small image extraction step S112, as shown in Figure 4, a predetermined trained model is acquired or created, and the trained model is used to extract crack small images HSG that are recognized as crack images from multiple small images SG.

[0026] The predetermined trained model is a trained model created by machine learning based on training data, which are multiple images of the surface of an arbitrary concrete structure, with sample images labeled as cracked or not. For example, a sample image of a crack is shown in Figure 5(a), and a sample image of a non-crack is shown in Figure 5(b). The amount of training data can be further increased by data augmentation.

[0027] In the crack occurrence area identification step S113, a process is performed to designate the crack occurrence area HR as an area obtained by combining the multiple small crack images HSG extracted in the small crack image extraction step S112. The information format indicating the crack occurrence area HR is free, and for example, a mask image may be created that makes it possible to recognize only the image of the crack occurrence area HR when it is overlaid on the analysis target area image BRG, and this mask image may be sent to the next step.

[0028] The resolution of the analysis area image BRG used in the crack occurrence area extraction step S11 can be lower than the resolution of the analysis area image BRG used in the subsequent crack binary image creation step S12. The crack occurrence area extraction step S11 is a step for narrowing down the area in which the crack H and its surrounding area are imaged, i.e., the area where detailed analysis should be performed, and the detailed analysis will be performed later in the crack binary image creation step S12 or later. Therefore, the resolution of the analysis area image BRG used in the crack occurrence area extraction step S11 needs to be sufficient to determine whether the divided small images SG are crack small images HSG.

[0029] <Crack Binary Image Creation Step S12> The crack binary image creation step S12 is a step in which a binarization process is performed on the crack occurrence area in the analysis area image BRG to create a crack binary image H2G that is divided into cracked areas HB and non-cracked areas HHB. This step S12 also estimates the detailed layout of the actual crack H and reflects the estimation results in the crack binary image H2G.

[0030] As shown in Fig. 2, the crack binary image creation step S12 sequentially executes a crack candidate binary image creation step S121 and a crack correction step S122. In the crack candidate binary image creation step S121, a binarization process is performed on the crack occurrence area HR in the analysis target area image BRG, and a crack candidate binary image HK2G is created that is divided into crack candidate areas HKB (white) and non-crack candidate areas HHKB (black). At this time, as shown in Fig. 6, it is preferable to perform a process to increase the resolution of the crack occurrence area HR in the analysis target area image BRG to enable detailed analysis, and then perform binarization processing using the P-tile method. It is also preferable to further perform a predetermined threshold process to remove noise. This allows subsequent analysis to be performed with higher accuracy. Note that The area other than the crack occurrence area HR in the analysis target area image BRG is uniformly designated as the non-crack candidate area HHKB. In the image shown in FIG. 6, the areas other than the crack occurrence area HR are uniformly shown in black.

[0031] In the crack correction step S122, a process (connection process) is performed to connect the portions of the crack candidate portion HKB in the crack candidate binary image HK2G where the crack candidate portion HKB is interrupted to the portions where it is estimated that the crack H is actually continuous without interruption. As a result, the crack candidate portion HKB and the non-crack candidate portion HHKB are corrected to the crack portion HB and the non-crack portion HHB, and a crack binary image H2G that specifies the layout of the actual crack H is created.

[0032] The concatenation process is The procedure is as follows: First, as shown in Fig. 7, the crack candidate portion HKB in the crack candidate binary image HK2G is subjected to auxiliary thinning (auxiliary thinning step S122-1), and endpoints TT of the multiple thin lines SS that make up the thinned crack candidate portion are extracted (endpoint extraction step S122-2). Then, for each thin line SS, the endpoint TT of another thin line SS that is closest to the own endpoint TT and whose distance between the endpoints TT is equal to or less than a specified value is extracted, and the own endpoint TT and the other endpoint TT are set as a pair of connection target points RT (connection target point extraction step S122-3).

[0033] 8, the area of ​​a predetermined size between the pair of connection target points RT is designated as a connected area RR, and the connected area RR in the analysis area image BRG is subjected to a binarization process to distinguish it into a connected portion RB (white) and a non-connected portion HRB (black), and the area outside the connected area RR is regarded as a non-connected portion HRB and displayed in black, thereby creating a connected area binarized image RR2G (connected area binarized image creation step 122-4). It is preferable to use the P-tile method for the binarization process.

[0034] Once the connected region binary image RR2G has been created, the connected region binary image RR2G is combined with the crack candidate binary image HK2G to create a crack binary image H2G (crack linking step S122-5), as shown in Fig. 9. In other words, the crack portion HB represented in the crack binary image H2G is the crack candidate portion HKB in the crack candidate binary image HK2G to which the connected region RB in the connected region binary image RR2G has been added.

[0035] In this way, by performing the linking process (steps S122-1 to S122-5), a binarized crack image H2G that shows the actual layout of the crack H in detail can be created.

[0036] <Crack element information creation step S13> The crack element information creation step S13 is a step in which the crack portion HB in the crack binary image H2G is divided into multiple crack elements HY that do not have branch points by dividing the crack portion HB at characteristic points in the length direction of the crack portion HB, and crack element information HYJ, which is information about each crack element HY, is created. Characteristic points in the length direction of the crack portion HB include branch points where a single crack branches off midway along the length direction, and bend points where a crack bends at an unnatural angle midway along the length direction.

[0037] When identifying the feature points of the crack portion HB, it is preferable to perform a thinning process on the crack portion HB as shown in Figure 10, which allows the feature points to be identified with high accuracy. For example, in the example of the crack portion HB shown in Figure 10, there is a feature point where the crack intersects in an approximately X-shape, and the crack portion HB is divided into four at this point. Then, in order to identify each divided crack element HY, identification labels HY1, HY2, ... are assigned (labeling).

[0038] The crack element information HYJ is information such as the identification label assigned to each crack element HY and the coordinates indicating its position within the analysis target area image BRG. In addition, here, the length, maximum width, angle relative to the reference axis, etc. of each crack element HY are calculated and included in the crack element information HYJ.

[0039] <Crack-related information creation step S14> The crack-related information creation step S14 is a step of creating and outputting crack-related information HKJ, which is information indicating the crack occurrence status of the concrete structure and includes crack damage feature values ​​calculated based on the crack element information HYJ. The specific items of the crack-related information HKJ will be set individually for each concrete structure to be inspected, but generally, the crack-related information HKJ can be broadly divided into information for each crack element HY and information for the entire analysis target region BR.

[0040] The information for each crack element HY includes, for example, the identification label of each crack element HY, its position coordinates (coordinates of the start and end points), and the crack damage feature values ​​of each crack element HY, as shown in the table in Figure 11(a), and the crack damage feature value items include the length, maximum width, angle with respect to the reference axis, etc. of each crack element HY calculated in the crack element information creation step S13. Also, in Figure 11(a), among multiple crack elements HY, those that are continuous via a specific feature point are grouped together, so that the features of each group can also be recognized.

[0041] The information for the entire analysis area BR is, for example, as shown in the table in Figure 11(b), crack damage feature values ​​calculated based on the crack element information HYJ and other information (for example, the date and time when the analysis area image BRG was captured), and the items of the crack damage feature values ​​include the density of crack sections HB in the analysis area BR, the maximum width of the crack sections HB, the shape of the crack sections HB, and the minimum distance between crack elements HY that are spaced apart from each other.

[0042] In addition, Figure 11(a) shows all items of the crack element information HYJ created in the crack element information creation step S13 in order, but the crack-related information HKJ may be limited to only the information on the characteristic crack element HY. Also, in this embodiment, the crack damage feature amount in Figure 11(a) is calculated in the crack element information creation step S13 (crack element information creation unit 10c), but it is also possible to calculate it in the crack-related information creation step S14 (crack-related information creation unit 10d), as with the crack damage feature amount in Figure 11(b).

[0043] As described above, the crack detection method, crack detection program, and crack detection system 10 of this embodiment are configured to first roughly analyze the entire image of a wide area of ​​the surface of a concrete structure (analysis area image BRG) to extract the crack occurrence area HR, and then narrow down the analysis to the crack occurrence area HR and its surrounding area for detailed analysis.This significantly reduces the processing burden compared to analyzing the entire area of ​​the analysis area image BRG in detail, and makes it possible to detect crack occurrence conditions in a shorter time than conventional methods, and with accuracy equal to or higher than conventional methods.

[0044] The crack detection method, crack detection program, and crack detection system of the present invention are not limited to the above-described embodiments.

[0045] In the crack binary image creation step, the actual crack layout is estimated by analyzing the image of the crack occurrence area in the analysis target area image, and this estimation result is reflected in the crack binary image, but the method used to estimate the actual crack layout is not particularly limited. For example, in the above description of the crack binary image creation step S12, a preferred estimation method is described as "performing a process to increase the resolution of the crack occurrence area HR in the analysis target area image BRG, and then performing a process to binarize it using the P-tile method." I explained that , may be estimated using a different method as long as it can be estimated with the desired accuracy.

[0046] The crack-related information HKJ output by the crack-related information creation unit 10d may include, in addition to the text data shown in Figures 11(a) and (b), an analysis target area image BRG, an enlarged image of a portion of the analysis target area image BRG, a graph of crack damage features, etc. [Explanation of symbols]

[0047] 10 Crack detection system 10a Crack occurrence area extraction section 10b Crack binarization image creation section 10c Crack element information creation section 10d Crack-related Information Creation Department BR Analysis Area BRG Analysis target area image H crack H2G Crack binarized image HB Cracked area HHB Non-cracked area HKB crack candidate area HHKB Non-crack candidate part HR crack initiation area HRJ crack occurrence area information HKJ crack related information HK2G Crack candidate binarized image HRB unconnected part HSG small crack image HY crack element HYJ Crack Element Information RB connection part RR connection area RT Pair of connected points S11 Crack occurrence area extraction step S111: Image division step for analysis area S112 Crack small image extraction step S113 Crack occurrence area identification step S12 Crack binarization image creation step S121 Crack candidate binary image creation step S122 Crack correction step S122-1 Auxiliary thinning step (connection processing) S122-2 End point extraction step (connection processing) S122-3 Linking point extraction step (linking process) S122-4 Connected area binary image creation step (connection processing) S122-5 Crack connection step (connection treatment) S13 Crack element information creation step S14 Crack-related information creation step SG small image SS thin wire TT endpoint

Claims

1. A method for detecting crack occurrence status based on an analysis target area image obtained by capturing an analysis target area on the surface of a concrete structure, the method being executed by a computer system to which the analysis target area image is input, comprising: a crack occurrence region extraction step of extracting a crack occurrence region in which a crack and its surrounding area are captured by analyzing the analysis target region image; a crack binary image creation step of performing a binarization process on the crack occurrence area in the analysis area image to distinguish it into cracked and non-cracked areas, estimating the actual crack layout, and creating a crack binary image that reflects the estimation result; a crack element information creation step of dividing the crack portion in the binarized crack image into a plurality of crack elements that do not have branch points by dividing the crack portion at characteristic points in the length direction of the crack portion, and creating crack element information that is information about each of the crack elements; a crack-related information creation step of creating and outputting crack-related information that indicates the crack occurrence status of the concrete structure, the crack-related information including the crack damage feature amount calculated based on the crack element information; The crack occurrence region extraction step includes: an analysis target area image dividing step of dividing the analysis target area image into small images; a crack small image extraction step in which a plurality of images of the surface of an arbitrary concrete structure, sample images labeled as crack images or not, are used as training data, a trained model is obtained or created by machine learning based on the training data, and a plurality of crack small images that are recognized as crack images are extracted from the plurality of small images using the trained model; a crack occurrence region specifying step of specifying the region into which the small crack image is synthesized as the crack occurrence region, The crack binarized image creation step includes: a crack candidate binary image creation step of performing a binarization process on the crack occurrence region in the analysis target region image to distinguish the crack occurrence region into a crack candidate portion and a non-crack candidate portion, and creating a crack candidate binary image in which regions other than the crack occurrence region in the analysis target region image are uniformly designated as the non-crack candidate portion; a crack correction step of estimating the layout of an actual crack part by analyzing the crack candidate binarized image, and correcting the crack candidate part based on this estimation result, thereby identifying the crack part and the non-crack part; The crack correction step performs a connection process for connecting portions in the crack candidate binary image where the crack candidate portion is interrupted, where the crack is estimated to be continuous without interruption, thereby identifying the crack portion and the non-crack portion; The connection process includes: an auxiliary thinning step of auxiliary thinning the crack candidate portion in the crack candidate binary image; an end point extraction step of extracting end points of a plurality of thin lines constituting the crack candidate portion thinned in the auxiliary thinning step; a link target point extraction step of extracting, for each of the thin lines, an end point of another thin line that is closest to the end point of the thin line itself, the end point being a distance between the end points being equal to or less than a specified value, and defining the end point of the thin line itself and the other end point as a pair of link target points; a connected region binary image creation step in which a region of a predetermined size sandwiched between the pair of connection target points is defined as a connected region, and the connected region in the analysis target region image is subjected to binarization processing to distinguish between connected portions and non-connected portions, and regions other than the connected region are processed to be considered as non-connected portions, thereby creating a connected region binary image; A crack detection method characterized by sequentially performing a crack linking step of linking cracks from the connected area binary image and the crack candidate binary image.

2. A crack detection method as described in claim 1, wherein in the crack candidate binary image creation step, the crack occurrence area in the analysis target area image is subjected to a high-resolution process and then a binary process is performed.

3. A crack detection method according to claim 1 or 2, wherein the crack element information created in the crack element information creation step includes at least one of information on the length and information on the maximum width of the crack element.

4. The crack damage feature amount calculated in the crack-related information creation step includes: A crack detection method described in any one of claims 1 to 3, which includes at least one of the density of the cracks in the analysis area, the maximum width of the cracks in the analysis area, the shape of the cracks in the analysis area, and the minimum spacing between the crack elements spaced apart from each other in the analysis area.

5. A crack detection program comprising programs for executing each step for causing a computer system to execute the crack detection method according to any one of claims 1 to 4.

6. A crack detection system that detects crack occurrence status based on an image of an analysis target area obtained by capturing an image of an analysis target area on the surface of a concrete structure, a crack occurrence region extraction unit that extracts a crack occurrence region in which a crack and its surrounding area are captured by analyzing the analysis target region image; a crack binary image creation unit that performs a binarization process on the crack occurrence area in the analysis area image to distinguish it into cracked areas and non-cracked areas, estimates the actual crack layout, and creates a crack binary image that reflects the estimated results; a crack element information creation unit that divides the crack portion in the binarized crack image into a plurality of crack elements that do not have branch points by dividing the crack portion at characteristic points in the length direction of the crack portion, and creates crack element information that is information about each of the crack elements; a crack-related information creation unit that creates and outputs crack-related information that indicates the crack occurrence status of the concrete structure, the crack-related information including the crack damage feature amount calculated based on the crack element information; The crack occurrence region extraction unit an analysis target area image dividing step of dividing the analysis target area image into small images; a crack small image extraction step in which a plurality of images of the surface of an arbitrary concrete structure, sample images labeled as crack images or not, are used as training data, a trained model is obtained or created by machine learning based on the training data, and a plurality of crack small images that are recognized as crack images are extracted from the plurality of small images using the trained model; and a crack occurrence region specifying step of specifying the region into which the small crack image is synthesized as the crack occurrence region, The crack binarized image creation unit a crack candidate binary image creation step of performing a binarization process on the crack occurrence region in the analysis target region image to distinguish the crack occurrence region into a crack candidate portion and a non-crack candidate portion, and creating a crack candidate binary image in which regions other than the crack occurrence region in the analysis target region image are uniformly designated as the non-crack candidate portion; a crack correction step of estimating the layout of the actual crack portion by analyzing the crack candidate binarized image, and correcting the crack candidate portion based on this estimation result, thereby identifying the crack portion and the non-crack portion; The crack correction step performs a connection process for connecting portions in the crack candidate binary image where the crack candidate portion is interrupted, where the crack is estimated to be continuous without interruption, thereby identifying the crack portion and the non-crack portion; The connection process includes: an auxiliary thinning step of auxiliary thinning the crack candidate portion in the crack candidate binary image; an end point extraction step of extracting end points of a plurality of thin lines constituting the crack candidate portion thinned in the auxiliary thinning step; a link target point extraction step of extracting, for each of the thin lines, an end point of another thin line that is closest to the end point of the thin line itself, the end point being a distance between the end points being equal to or less than a specified value, and defining the end point of the thin line itself and the other end point as a pair of link target points; a connected region binary image creation step in which a region of a predetermined size sandwiched between the pair of connection target points is defined as a connected region, and the connected region in the analysis target region image is subjected to binarization processing to distinguish between connected portions and non-connected portions, and regions other than the connected region are processed to be considered as non-connected portions, thereby creating a connected region binary image; A crack detection system characterized by sequentially performing a crack linking step of linking cracks from the connected area binary image and the crack candidate binary image.

7. A crack detection system as described in claim 6, wherein in the crack candidate binary image creation step, the crack occurrence area in the analysis target area image is subjected to a high-resolution process and then a binary process is performed.

8. A crack detection system as described in claim 6 or 7, wherein the crack element information created by the crack element information creation unit includes at least one of information on the length and information on the maximum width of the crack element.

9. The crack damage feature amount calculated by the crack-related information creation unit includes: A crack detection system as described in any one of claims 6 to 8, including at least one of the density of the cracks in the analyzed area, the maximum width of the cracks in the analyzed area, the shape of the cracks in the analyzed area, and the minimum spacing between the crack elements spaced apart from each other in the analyzed area.

Images