JP6936684B2 - Crack detection device, crack detection method, and computer program - Google Patents

Description

The techniques disclosed herein relate to crack detectors.

A technique for detecting cracks in an image of an object (for example, a wall of a building or a civil engineering structure) by performing image processing on an image of the object in order to diagnose the occurrence of cracks in the image. Is known (see, for example, Patent Document 1). According to the crack detection using such image processing, it is possible to efficiently and accurately diagnose the crack occurrence state in the object as compared with the case of visual inspection by a field worker, for example.

Japanese Unexamined Patent Publication No. 2012-98045

In the above-mentioned conventional technique, there is room for improvement in terms of crack detection accuracy. That is, in the above-mentioned conventional technique, a false detection that is not a crack but has a concentration close to the crack concentration (for example, dirt on the surface of an object) is mistakenly detected as a crack, or a crack that actually exists is detected. Not a few detection omissions that cannot be detected occur. Therefore, the above-mentioned conventional technique has a problem of suppressing the occurrence of such erroneous detection and detection omission and further improving the crack detection accuracy.

The present specification discloses a technique capable of solving the problem of improving the accuracy of crack detection using image processing.

The techniques disclosed herein can be realized, for example, in the following forms.

(1) The crack detection device disclosed in the present specification is a crack detection device that detects cracks in an object, and is a crack candidate region with a high probability of being an image region representing cracks in an image representing the object. A predetermined condition including that the ratio of the length L (however, L ≧ W) to the width W of the crack candidate region among the crack candidate regions is equal to or greater than the first threshold value. A confirmation determination unit for setting the crack candidate region to be filled in the crack region, which is an image region representing the crack, is provided. In general, a crack generated in an object contains a component of a line segment (that is, a component that is elongated and extends linearly). In this crack detection device, the confirmation determination unit determines a crack candidate region that satisfies a predetermined condition including that the ratio of the length L to the width W of the crack candidate region is equal to or greater than the first threshold value among the crack candidate regions. Set in the cracked area. Therefore, according to this crack detection device, a crack candidate region containing a line segment component having a certain length or longer can be set as the crack region, so that the concentration is close to the crack concentration, although it is not a crack. It is possible to suppress the occurrence of erroneous detection in which a portion having the above (for example, dirt on the surface of an object) is erroneously detected as a crack, and it is possible to improve the accuracy of crack detection using image processing.

(2) In the crack detection device, the width W of the crack candidate region is the width Wr of the first minimum rectangle which is the smallest rectangle including the crack candidate region, and the length L of the crack candidate region is , The length Lr of the first minimum rectangle may be used. According to this crack detection device, the ratio of the length L to the width W of the crack candidate region can be calculated quickly and accurately, and the crack detection process using image processing can be speeded up and the accuracy can be calculated. Can be further improved.

(3) In the crack detection device, the predetermined condition is set by extending both ends in the length direction of the crack candidate region with respect to the width of the first minimum rectangle along the direction of the both ends. The configuration may include that the ratio of the widths of the second minimum rectangle, which is the minimum rectangle including the expanded crack candidate region, is equal to or less than the second threshold value. According to this crack detection device, since a crack candidate region having a shape close to a straight line can be set as a crack region, it is possible to effectively suppress the occurrence of erroneous detection, and image processing is used. The accuracy of crack detection can be further improved.

(4) In the crack detection device, the predetermined condition may include a ratio of the area of the crack candidate region to the area of the polygon including the crack candidate region to be equal to or higher than the third threshold value. .. According to this crack detection device, since a crack candidate region having a shape close to a straight line can be set as a crack region, it is possible to effectively suppress the occurrence of erroneous detection, and image processing is used. The accuracy of crack detection can be further improved.

(5) The crack detection device may further include a display control unit that displays an image showing the crack region on the display. According to this crack detection device, the user can grasp the occurrence status of cracks in the object.

(6) In the crack detection device, the object may be a wall-shaped structure. According to this crack detection device, crack detection is performed on a wall-shaped structure that is an object in which crack generation is likely to be a problem and in which crack detection needs to be performed in a relatively wide range. Can be executed efficiently and accurately.

The techniques disclosed in the present specification can be realized in various forms, for example, a crack detection device, a crack detection method, a computer program for realizing the functions of those devices or methods, and the like. It can be realized in the form of a non-temporary recording medium or the like on which a computer program is recorded. Further, the technique disclosed in the present specification is also useful as an elemental technique of an automatic inspection system using a robot.

It is explanatory drawing which shows schematic structure of the crack detection apparatus 100 in this embodiment. It is explanatory drawing which shows an example of the screen (processing screen S1) which is displayed on the display unit 120 at the time of execution of a crack detection process. It is a flowchart which shows the flow of the crack detection process. It is explanatory drawing which shows an example of the processing screen S1 including the result of the crack detection processing. It is a flowchart which shows the flow of the crack candidate detection processing. It is explanatory drawing which shows an example of the 1st binary image Ib (1). It is a flowchart which shows the flow of the 1st half line filter processing. It is explanatory drawing which shows the outline of the 1st half line filter processing. It is explanatory drawing which shows an example of the result of the 1st half line filter processing. It is explanatory drawing which shows an example of the result of the extended processing. It is a flowchart which shows the flow of the 2nd half line filter processing. It is explanatory drawing which shows the outline of the 2nd half line filter processing. It is explanatory drawing which shows an example of the setting result of the crack candidate region CC. It is a flowchart which shows the flow of the crack candidate confirmation processing. It is explanatory drawing which shows the outline of the crack candidate confirmation processing. It is explanatory drawing which shows the condition for setting the crack candidate region CC in the crack region CR in the modification. It is explanatory drawing which shows the condition for setting the crack candidate region CC in the crack region CR in the modification. It is explanatory drawing which shows the condition for setting the crack candidate region CC in the crack region CR in the modification.

A. Embodiment:
A-1. Device configuration:
FIG. 1 is an explanatory diagram schematically showing the configuration of the crack detection device 100 in the present embodiment. The crack detection device 100 is a general-purpose computer such as a personal computer, a tablet terminal, or a smartphone. The crack detection device 100 includes a storage unit 110, a display unit 120, an input unit 130, an interface unit 140, and a control unit 170. Each of these parts is communicably connected to each other via a bus 190.

The display unit 120 is composed of, for example, a liquid crystal display or the like, and displays various images and information. The input unit 130 is composed of, for example, a keyboard, a mouse, a microphone, or the like, and receives user operations and voice instructions. The interface unit 140 is composed of, for example, a LAN interface, a USB interface, or the like, and communicates with another device (for example, an image pickup device 320) by wire or wirelessly.

The storage unit 110 is composed of, for example, a hard disk drive (HDD) or the like, and stores various programs and data. For example, the storage unit 110 stores a crack detection program CP for executing a crack detection process described later. The crack detection program CP is provided in a state of being stored in a computer-readable recording medium (not shown) such as a CD-ROM, a DVD-ROM, or a USB memory, and is installed in the crack detection device 100 to store the crack detection program CP. It is stored in 110.

Further, the image data Id is stored in the storage unit 110. The image data Id is generated by, for example, capturing an object (in this embodiment, the wall of the building 310) with an image pickup device 320 such as a digital still camera. The image data Id is input to the crack detection device 100 via the interface unit 140 and stored in the storage unit 110.

The control unit 170 is composed of, for example, a CPU, a ROM, a RAM, or the like, and controls the operation of the crack detection device 100 by executing a computer program read from the storage unit 110. For example, the control unit 170 functions as a crack detection processing unit 200 that executes a crack detection process described later by reading and executing the crack detection program CP. The crack detection processing unit 200 includes a display control unit 210, a candidate detection processing unit 220, and a candidate confirmation processing unit 230. Further, the candidate detection processing unit 220 includes a binary image acquisition unit 222, a candidate determination unit 224, and a candidate area setting unit 226. Further, the candidate confirmation processing unit 230 includes a rectangle calculation unit 232 and a confirmation determination unit 234. The functions of each of these parts will be described in accordance with the description of the crack detection process described later.

A-2. Crack detection process:
Next, the crack detection process executed by the crack detection device 100 will be described. The crack detection process is a process of detecting cracks in an image obtained by capturing an image of an object by performing image processing on the image. In general, a crack generated in an object contains a component of a line segment (that is, a component that is elongated and extends linearly). In the present embodiment, the crack detection process with high accuracy is realized by utilizing such a feature of the crack. In the present embodiment, the object of the crack detection process is a tiled wall (that is, a wall-shaped structure) of the building 310.

FIG. 2 is an explanatory diagram showing an example of a screen (hereinafter, referred to as “processing screen S1”) displayed on the display unit 120 when the crack detection process is executed. The processing screen S1 is displayed on the display unit 120 by the display control unit 210. The processing screen S1 includes an image selection area R1, a method selection area R2, and an image display area R3. The image selection area R1 is an area for selecting a target image IO to be subjected to the crack detection process. The method selection area R2 is an area for selecting a method (algorithm) used for the crack detection process. The image display area R3 is an area for displaying the target image IO and the like.

In the image selection area R1 of the processing screen S1, an image identifier (file name, thumbnail, etc.) represented by the image data Id (see FIG. 1) stored in the storage unit 110 is displayed as an option. When the user selects a desired image (for example, "image 1") from the image choices displayed in the image selection area R1 via the input unit 130, the selected image is set as the target image IO. It is displayed in the image display area R3. The image selection area R1 includes a number display column R11 for displaying the number of cracks detected by the crack detection process.

Further, in the method selection area R2 of the processing screen S1, options of the method (algorithm) used for each of the processes (crack candidate detection process, crack candidate confirmation process) described later that constitute the crack detection process are displayed. From the plurality of methods displayed in the method selection area R2, the user uses the input unit 130 to perform a desired method for each process (for example, "method A" of the crack candidate detection process and the crack candidate confirmation process. When "Method H") is selected and the "Execute" button B1 is selected, the crack detection process using the selected method is started. As shown in FIG. 2, in the present embodiment, the target image IO is an image showing the tiled wall of the building 310, and each pixel constituting the image is represented by a density of 256 gradations. It is a grayscale image. In addition, the target image IO includes pixels representing cracks (hereinafter referred to as "crack pixels Pc") and parts that are not cracks but have a density close to the density of cracks (for example, dirt on the surface of the object). A representative pixel (hereinafter, referred to as "noise pixel Pn") is included.

FIG. 3 is a flowchart showing the flow of the crack detection process. First, the crack detection processing unit 200 executes preprocessing for the target image IO, if necessary (S120). As preprocessing, for example, a process of correcting the density of the target image IO, an image area representing tiles (hereinafter referred to as “tile area”) is extracted from the target image IO, and an image area (joint area) other than the tile area is extracted. Etc.) is excluded from the target range of crack detection. The process of extracting the tile region can be executed by using a known method. Further, when the target range for crack detection in the target image IO is predetermined, it is not necessary to execute the process of extracting the tile area.

Next, the candidate detection processing unit 220 executes the crack candidate detection processing (S130). The crack candidate detection process is a process of detecting a crack candidate region CC including a crack candidate pixel Pcc, which is a pixel representing a crack, from the target image IO. The candidate detection processing unit 220 corresponds to the candidate detection unit within the scope of claims. The details of the crack candidate detection process will be described later.

Next, the crack detection processing unit 200 determines whether or not at least one crack candidate region CC is detected in the crack candidate detection process (S130) (S140). When it is determined that the crack candidate region CC has been detected (S140: YES), the candidate confirmation processing unit 230 executes the crack candidate confirmation process (S150). The crack candidate confirmation process is a process for confirming whether or not the crack candidate region CC detected from the target image IO is the crack region CR, which is an image region that truly represents a crack. The details of the crack candidate confirmation process will be described later. If it is determined that the crack candidate region CC is not detected in the crack candidate detection process (S140: NO), the crack candidate confirmation process (S150) is skipped.

Next, the display control unit 210 causes the display unit 120 to display the result of the crack detection process (S160). FIG. 4 is an explanatory diagram showing an example of the processing screen S1 including the result of the crack detection processing. In the example of the processing screen S1 shown in FIG. 4, an image showing the detected crack region CR on the target image IO displayed in the image display region R3 (that is, an image showing at least a part of the crack candidate pixels Pcc). Is displayed. Further, in the number display field R11 of the image selection area R1 of the processing screen S1, the number of cracked areas CR detected from the target image IO (for example, one) is shown. The user can grasp the occurrence state of cracks in the object (for example, the wall of the building 310) represented by the target image IO through the processing screen S1 showing the result of the crack detection process.

A-3. Crack candidate detection process:
Next, the crack candidate detection process (S130 in FIG. 3) will be described in detail. As described above, the crack candidate detection process is a process of detecting a crack candidate region CC including a crack candidate pixel Pcc, which is a pixel representing a crack, from the target image IO. FIG. 5 is a flowchart showing the flow of the crack candidate detection process.

First, the binary image acquisition unit 222 of the candidate detection processing unit 220 executes binarization processing on the target image (original image) IO, and binarizes the image (hereinafter referred to as "first binary image"). Acquire Ib (1) (S210). FIG. 6 is an explanatory diagram showing an example of the first binary image Ib (1). FIG. 6 shows an enlarged configuration of a part (a part corresponding to the X1 part of FIG. 2) of the first binary image Ib (1). In the first binary image Ib (1), the object has two gradations of a density value representing black (0 in this embodiment) and a density value representing white (255 in this embodiment). It is an image to be expressed. That is, the first binary image Ib (1) is an image composed of black pixels Pk and white pixels Pw. In FIG. 6, the black pixel Pk is hatched.

The binarization process for the target image IO uses an arbitrary known method (for example, a method of determining a threshold value with reference to an image histogram according to a discriminant analysis method and binarizing the image using the threshold value). Although it can be executed, in the present embodiment, the following processing is executed as the binarization processing. That is, the pixel of interest is sequentially selected from the pixels constituting the target image (original image) IO which is a grayscale image, and the density value of the pixel of interest is set to the density value of a plurality of peripheral pixels located around the pixel of interest. Compare with the average value. The range of peripheral pixels can be arbitrarily set, and for example, eight pixels surrounding the pixel of interest can be set. When the density value of the pixel of interest is smaller than the average value of the density values of a plurality of peripheral pixels (that is, when it matches the magnitude relationship of the density value (0) representing black with respect to the density value (255) representing white). Converts the density value of the pixel of interest to a density value representing black (that is, 0). Further, when the density value of the attention pixel is larger than the average value of the density values of a plurality of peripheral pixels (that is, when the density value representing black color (0) does not match the density value (255) representing white color). ) Converts the density value of the pixel of interest to a density value representing white (that is, 255). As a result, the first binary image Ib (1) composed of the white pixel Pw and the black pixel Pk is generated. The binary image Ib obtained by binarizing the target image IO does not necessarily have to be displayed as an image.

In the target image IO, which is a grayscale image, the density value of the crack pixel Pc representing the crack is close to the density value of black. Therefore, in the first binary image Ib (1) generated by the binarization process of the target image IO, the cracked pixel Pc becomes the black pixel Pk. However, not all the black pixels Pk in the first binary image Ib (1) are cracked pixels Pc. That is, the noise pixel Pn included in the target image IO, which is not a crack but represents a portion having a density close to the crack density (for example, dirt on the surface of the object), is also the first binary image Ib (1). ) May result in black pixels Pk. In the example shown in FIG. 6, in the first binary image Ib (1), in addition to the black pixel Pk which is the crack pixel Pc, the black pixel Pk which is the noise pixel Pn is included.

Next, the candidate determination unit 224 of the candidate detection processing unit 220 executes the first half-line filter processing on the first binary image Ib (1) (S220). FIG. 7 is a flowchart showing the flow of the first half-line filter processing. Further, FIG. 8 is an explanatory diagram showing an outline of the first half-line filter processing. In FIG. 8, a region composed of the cracked pixel Pc and a region composed of the noise pixel Pn are shown by a thick broken line (the same applies to FIGS. 9, 10 and 13).

First, the candidate determination unit 224 selects one of the pixels constituting the first binary image Ib (1) as the pixel of interest SP (S221). In the present embodiment, the pixel of interest SP is selected from the black pixels Pk constituting the first binary image Ib (1). However, the pixel of interest SP may be selected from all the pixels constituting the first binary image Ib (1). FIG. 8 illustrates three notable pixels SP (SP (1), SP (2), SP (3)). In reality, a plurality of pixels are not selected as the pixel of interest SP at the same time, and each pixel is sequentially selected as the pixel of interest SP one by one.

Next, the candidate determination unit 224 has a plurality of virtual line segments having a predetermined length L1 and different predetermined inclinations on the first binary image Ib (1) with the pixel of interest SP as an end point. SL is set (S222). FIG. 8 exemplifies eight virtual line segment SLs having different inclinations of 45 degrees for each attention pixel SP, and the length L1 of each virtual line segment SL is set to five pixels, but it is virtual. The number and length of the line segment SL can be set arbitrarily.

For each virtual line segment SL, the candidate determination unit 224 states that "the density value of pixels having a predetermined ratio RA1 or more among the pixels located on the virtual line segment SL matches the density value of the black pixel Pk (that is, predetermined). (S223), it is determined whether or not the condition (hereinafter, referred to as “first condition”) of “a pixel having a ratio of RA1 or more corresponds to a black pixel Pk) is satisfied (S223). The predetermined ratio RA1 at this time is set to, for example, one value in the range of 80% to 100%. For example, for the two virtual line segment SLs (virtual line segment SL (1) and SL (2)) set for the pixel of interest SP (1) shown in FIG. 8, the pixels located on the virtual line segment SL. Since all of the above correspond to the black pixel Pk, it is determined that the first condition is satisfied. On the other hand, regarding the other virtual line segment SL (3) set for the attention pixel SP (1) shown in FIG. 8, among the pixels located on the virtual line segment SL, the pixels corresponding to the black pixel Pk Since the ratio is low (less than RA1), it is determined that the first condition is not satisfied.

Further, the candidate determination unit 224 has two inclinations different from each other by 180 degrees (that is, located on one virtual straight line passing through the pixel of interest SP) in the virtual line segment SL satisfying the first condition. It is determined whether or not a set of virtual line segment SLs (hereinafter referred to as "condition-satisfying line segment pair") exists (S224). In other words, this determination is based on a predetermined ratio RA1 or more of the pixels located on the virtual line segment SL for both of the two virtual line segment SLs whose slopes differ by 180 degrees from each other among the plurality of virtual line segment SLs. It is a determination as to whether or not the first condition that the density value of the pixel of the above matches the density value of the black pixel Pk is satisfied.

When the candidate determination unit 224 determines that a condition-satisfying line segment pair exists (S224: YES), the currently selected pixel of interest SP is a pixel representing a crack, which is a highly probable crack candidate pixel (hereinafter referred to as a crack candidate pixel). It is set to Pcc1 (referred to as "primary crack candidate pixel") (S225). On the other hand, when the candidate determination unit 224 determines that the condition satisfaction line segment pair does not exist (S224: NO), the candidate determination unit 224 skips the process (S225) of setting the attention pixel SP to the first crack candidate pixel Pcc1. For example, the two virtual line segments SL (virtual line segments SL (1) and SL (2)) set for the attention pixel SP (1) shown in FIG. 8 have two virtual line segments having slopes that differ by 180 degrees from each other. It is a set of minute SLs, and both satisfy the first condition. Therefore, it is determined that the condition-satisfying line segment pair exists, and the pixel of interest SP (1) at this time is set to the first crack candidate pixel Pcc1. Similarly, the other noteworthy pixel SP (2) shown in FIG. 8 is also determined to have a condition-satisfying line segment pair, and is set to the first crack candidate pixel Pcc1. On the other hand, regarding the other notable pixel SP (3) shown in FIG. 8, it is determined that the condition satisfaction line segment pair does not exist, and the first crack candidate pixel Pcc1 is not set.

The candidate detection processing unit 220 determines whether or not the selection of all the target pixels as the attention pixel SP has been completed (S226), and the selection of the attention pixel SP has not been completed (that is, the unselected target). If it is determined (S226: NO), the process returns to S221 and the above processing is repeated. On the other hand, when the candidate detection processing unit 220 determines that the selection of all the target pixels as the pixel of interest SP is completed (S226: YES), the candidate detection processing unit 220 ends the first half-line processing.

FIG. 9 is an explanatory diagram showing an example of the result of the first half-line filter processing. In FIG. 9, the pixels set in the first crack candidate pixel Pcc1 in the first half-line filter processing are hatched. As mentioned above, cracks generally contain line segment components. Therefore, in the first half-line filter processing, if a certain percentage or more of the pixels located on the line segment twice the length of the virtual line segment SL is the black pixel Pk, the pixel of interest SP at that time. Is a pixel representing a crack, and is determined to be a crack candidate pixel having a high probability of being cracked. In the example shown in FIG. 9, most of the crack pixel Pc that truly represents a crack is set in the primary crack candidate pixel Pcc1. Further, all the noise pixels Pn shown near the upper left of FIG. 9 are not set in the first crack candidate pixel Pcc1. That is, the occurrence of erroneous detection in which these noise pixels Pn are erroneously detected as cracks is prevented. However, since the noise pixel Pn shown near the lower right of FIG. 9 constitutes a relatively elongated group of regions, a part of the noise pixel Pn is erroneously set as the first crack candidate pixel Pcc1. Has been done.

The first half-line filter processing described above is actually realized by image processing using the filter. The same applies to the second half-line filter processing described later.

After the completion of the first half-line filter processing (S220 in FIG. 5), the candidate detection processing unit 220 executes the expansion processing (S230). In the first half-line filtering process described above, a part of the crack pixel Pc that truly represents the crack, specifically, the edge of the region composed of the group of the crack pixel Pc (particularly, the edge in the longitudinal direction of the crack). The crack pixel Pc located in the portion) may not be set as the primary crack candidate pixel Pcc1. Therefore, the candidate detection processing unit 220 expands the region composed of the group of primary crack candidate pixels Pcc1 set by the first half-line filter processing, and among the pixels included in the expanded region, the first The pixel that was not set as the first crack candidate pixel Pcc1 in the half-line filter processing of 1 is corrected to the first crack candidate pixel Pcc1. Such an extended process can be realized, for example, by a morphology operation.

Based on the first binary image Ib (1), the candidate detection processing unit 220 sets the density value of the first crack candidate pixel Pcc1 at the completion of the expansion processing to a density value representing black (that is, 0). By converting and converting the density values of the other pixels to the density values representing white (that is, 255), the converted binary image (hereinafter referred to as "second binary image") Ib (2). ) Is generated. In the second binary image Ib (2), the density value of the first crack candidate pixel Pcc1 (before the expansion process) in the first binary image Ib (1) is converted into a density value representing black. , It can be said that it is a binary image in which at least a part of the density values of the pixels other than the primary crack candidate pixel Pcc1 (before the expansion process) are converted into the density values representing white.

FIG. 10 is an explanatory diagram showing an example of the result of the expansion processing. In FIG. 10, on the second binary image Ib (2), the pixels set in the first crack candidate pixel Pcc1 are hatched. As is clear from comparing FIG. 10 with FIG. 9, the area composed of the pixels set as the first crack candidate pixel Pcc1 before the execution of the expansion process is expanded by the expansion process. The region composed of the first crack candidate pixels Pcc1 at the completion of the expansion process does not always completely coincide with the region composed of the crack pixels Pc, but the two generally coincide with each other.

Next, the candidate determination unit 224 executes a second half-line filter process on the second binary image Ib (2) (S240 in FIG. 5). FIG. 11 is a flowchart showing the flow of the second half-line filter processing. Further, FIG. 12 is an explanatory diagram showing an outline of the second half-line filter processing. In FIG. 12, the region composed of the first crack candidate pixel Pcc1 is indicated by a thick alternate long and short dash line.

First, the candidate determination unit 224 selects one of the first crack candidate pixels Pcc1 constituting the second binary image Ib (2) as the pixel of interest SP (S241). FIG. 12 illustrates two notable pixels SP (SP (1), SP (2)). In reality, a plurality of pixels are not selected as the pixel of interest SP at the same time, and each pixel is sequentially selected as the pixel of interest SP one by one.

Next, the candidate determination unit 224 has a plurality of virtual line segments having a predetermined length L2 and different predetermined inclinations on the second binary image Ib (2) with the pixel of interest SP as an end point. SL is set (S242). FIG. 12 exemplifies eight virtual line segment SLs having different inclinations of 45 degrees for each attention pixel SP, and the length L2 of each virtual line segment SL is set to 10 pixels, but it is virtual. The number and length of the line segment SL can be set arbitrarily. However, the length L2 of the virtual line segment SL set in the second half-line filter processing is longer than the length L1 of the virtual line segment SL set in the first half-line filter processing (L2> L1). The length L2 is preferably twice or more the length L1.

For each virtual line segment SL, the candidate determination unit 224 states that "the density value of pixels having a predetermined ratio RA2 or more among the pixels located on the virtual line segment SL matches the density value of the black pixel Pk (that is, predetermined). (S243), it is determined whether or not the condition (hereinafter, referred to as “second condition”) of “a pixel having a ratio of RA2 or more corresponds to a black pixel Pk) is satisfied (S243). The predetermined ratio RA2 at this time is lower than the ratio RA1 in the first half-line filter processing, and is set to, for example, one value in the range of 60% to 80%.

The candidate determination unit 224 determines whether or not there is a virtual line segment SL (hereinafter, referred to as “condition satisfaction line segment”) that satisfies the second condition (S244), and determines that the condition satisfaction line segment exists. In the case (S244: YES), the currently selected pixel of interest SP is set to a new crack candidate pixel (hereinafter, referred to as “secondary crack candidate pixel”) Pcc2 (S245). That is, the second crack candidate pixel Pcc2 is also included in the second half-line filter processing among the pixels determined to be the crack candidate pixel (first crack candidate pixel Pcc1) in the first half-line filter processing. It is a pixel determined to be a crack candidate pixel (secondary crack candidate pixel Pcc2), and can be said to be an updated crack candidate pixel. On the other hand, when the candidate determination unit 224 determines that the condition satisfaction line segment does not exist (S244: NO), the candidate determination unit 224 skips the process (S245) of setting the attention pixel SP to the second crack candidate pixel Pcc2. That is, in this case, even if the pixel is determined to be the crack candidate pixel (first crack candidate pixel Pcc1) in the first half-line filter processing, the crack candidate pixel is determined in the second half-line filter processing. It is determined that it is not (secondary crack candidate pixel Pcc2). For example, with respect to the attention pixel SP (1) shown in FIG. 12, all the pixels located on the virtual line segment SL (1) set for the attention pixel SP (1) correspond to the black pixel Pk. Therefore, since the virtual line segment SL (1) corresponds to the condition satisfaction line segment, the attention pixel SP (1) is set to the second crack candidate pixel Pcc2. On the other hand, with respect to the other attention pixel SP (2) shown in FIG. 12, since the condition satisfaction line segment does not exist, the attention pixel SP (2) is not set in the second crack candidate pixel Pcc2.

The candidate detection processing unit 220 determines whether or not the selection of all the target pixels as the attention pixel SP has been completed (S246), and the selection of the attention pixel SP has not been completed (that is, the unselected target). If it is determined (S246: NO), the process returns to S241 and the above processing is repeated. On the other hand, when the candidate detection processing unit 220 determines that the selection of all the target pixels as the pixel of interest SP is completed (S246: YES), the candidate detection processing unit 220 ends the second half-line processing.

After the completion of the second half-line filter processing (S240 in FIG. 5), the candidate area setting unit 226 of the candidate detection processing unit 220 includes the second crack candidate pixel Pcc2 set in the second half-line filter processing. The image region is set as a crack candidate region CC with a high probability of being an image region representing a crack (S250). The crack candidate region CC may be composed of only the secondary crack candidate pixel Pcc2, or may include a small number of pixels that are not the secondary crack candidate pixel Pcc2. For example, if there are a few pixels that are not the secondary crack candidate pixels Pcc2 inside the region surrounded by the group of secondary crack candidate pixels Pcc2, they are not such secondary crack candidate pixels Pcc2. The crack candidate region CC may be set including the pixels.

FIG. 13 is an explanatory diagram showing an example of the setting result of the crack candidate region CC. In FIG. 13, hatching is added to the pixel (crack candidate region CC) set in the secondary crack candidate pixel Pcc2 in the second half-line filter processing. As mentioned above, cracks generally contain line segment components. Therefore, in the second half-line filter processing, if a certain percentage or more of the pixels located on the virtual line segment SL having a length L2 having the attention pixel SP as an end point are black pixels Pk, the attention pixel SP is It is determined that the pixel represents a crack and is a candidate pixel for a crack with a high probability.

In the second half-line filter processing, the length of the virtual line segment SL (L2) is made longer than the length of the virtual line segment SL (L1) in the first half-line filter processing. Therefore, it is possible to suppress the occurrence of erroneous detection in which the noise pixel Pn is erroneously set as the secondary crack candidate pixel Pcc2. For example, in the example shown in FIG. 13, most of the crack pixels Pc that truly represent cracks are set as the secondary crack candidate pixels Pcc2, while the noise pixels shown near the lower right of FIG. Pn is not set in the second crack candidate pixel Pcc2. That is, the occurrence of erroneous detection in which these noise pixels Pn are erroneously detected as cracks is prevented.

Further, in the second half-line filter processing, the second condition (the density value of the pixels having a predetermined ratio RA2 or more among the pixels located on the virtual line segment SL matches the density value of the black pixel Pk). The predetermined ratio RA2 in the above is used in the first half-line filter processing. The density value of the pixels having the predetermined ratio RA1 or more among the pixels located on the virtual line segment SL is the black pixel Pk. It is set lower than the predetermined ratio RA1 in (corresponding to the concentration value). Therefore, it is possible to suppress the occurrence of detection omission in which the crack pixel Pc that truly represents the crack is not set in the secondary crack candidate pixel Pcc2.

By the crack candidate detection process described above, the crack candidate region CC including the crack candidate pixel (secondary crack candidate pixel Pcc2) having a high probability of being a pixel representing the crack is detected from the target image IO.

A-4. Crack candidate confirmation process:
Next, the crack candidate confirmation process (S150 in FIG. 3) will be described in detail. As described above, the crack candidate confirmation process is a process for confirming whether or not the crack candidate region CC detected from the target image IO is the crack region CR, which is an image region that truly represents a crack. FIG. 14 is a flowchart showing the flow of the crack candidate confirmation process. Further, FIG. 15 is an explanatory diagram showing an outline of the crack candidate confirmation process.

First, the candidate confirmation processing unit 230 selects one crack candidate region CC (S310). FIG. 15 shows one selected crack candidate region CC. As described above, in the present embodiment, the crack candidate region CC is an image region composed of the secondary crack candidate pixel Pcc2.

Next, the rectangle calculation unit 232 of the candidate confirmation processing unit 230 calculates the minimum rectangle RE including the selected crack candidate region CC (S320). FIG. 15 shows the smallest rectangle RE that includes the selected crack candidate region CC.

Next, in the confirmation determination unit 234 of the candidate confirmation processing unit 230, the ratio of the length Lr (however, Lr ≧ Wr) to the width Wr of the minimum rectangle RE (hereinafter, referred to as “aspect ratio (Lr / Wr)”) is determined. It is determined whether or not it is equal to or higher than a predetermined threshold value Th1 (S330). The threshold Th1 corresponds to the first threshold in the claims. The width Wr of the minimum rectangle RE corresponds to the width W of the crack candidate region CC, and the length Lr of the minimum rectangle RE corresponds to the length L of the crack candidate region CC.

When the aspect ratio (Lr / Wr) of the minimum rectangle RE is equal to or greater than the threshold Th1 (S330: YES), the confirmation determination unit 234 determines that the selected crack candidate region CC is a crack region CR that truly represents a crack. It is determined that there is (S340). On the other hand, when the aspect ratio (Lr / Wr) of the minimum rectangle RE is less than the threshold Th1 (S330: NO), the confirmation determination unit 234 determines that the selected crack candidate region CC is a crack region that truly represents a crack. It is determined that something that is not a CR and is not a crack is erroneously detected (erroneous detection) (S350). As described above, since the crack generated in the object contains a line segment component, the aspect ratio of the crack region CR that truly represents the crack becomes a value to some extent. Therefore, in the present embodiment, when the aspect ratio of the crack candidate region CC (the minimum rectangular RE including the crack candidate region CC) is relatively large, it is determined that the crack candidate region CC is a true crack region CR, and the crack candidate region CC is determined. When the aspect ratio of (the smallest rectangular RE including) is relatively small, it is determined that the crack candidate region CC is not a true crack region CR.

The candidate confirmation processing unit 230 determines whether or not the selection of all the crack candidate regions CC is completed (S360), and if it is determined that there is an unselected crack candidate region CC (S360: NO), S310. Return to and repeat the above process. On the other hand, when the candidate confirmation processing unit 230 determines that the unselected crack candidate region CC does not exist (S360: YES), the candidate confirmation processing unit 230 ends the crack candidate confirmation processing.

By the crack candidate confirmation process described above, it is confirmed whether or not each crack candidate region CC detected from the target image IO in the crack candidate detection process is a crack region CR that truly represents a crack. As described above, the crack region CR confirmed by the crack candidate confirmation process is shown on the target image IO displayed in the image display area R3 of the processing screen S1 (see FIG. 4).

A-5. Effect of this embodiment:
As described above, the crack detection device 100 of the present embodiment includes a candidate detection processing unit 220. The candidate detection processing unit 220 detects a crack candidate region CC, which is an image region representing a crack, in an image representing an object (target image IO), which is highly probable. Further, the crack detection device 100 of the present embodiment includes a confirmation determination unit 234. The confirmation determination unit 234 determines that the crack candidate region CC satisfies a predetermined condition including the ratio of the length L (however, L ≧ W) to the width W of the crack candidate region CC to the threshold Th1 or more. CC is set to the crack region CR, which is an image region representing cracks.

Here, in general, the crack generated in the object contains a component of a line segment (that is, a component that is elongated and extends linearly). In the crack detection device 100 of the present embodiment, the confirmation determination unit 234 determines the crack candidate region CC that satisfies a predetermined condition including that the ratio of the length L to the width W is the threshold Th1 or more among the crack candidate regions CC. , Set to the cracked area CR. Therefore, according to the crack detection device 100 of the present embodiment, the crack candidate region CC containing the component of the line segment having a certain length or longer can be set in the crack region CR, so that the crack is not a crack. It is possible to suppress the occurrence of erroneous detection in which a portion having a density close to the crack concentration (for example, dirt on the surface of an object) is erroneously detected as a crack, and the accuracy of crack detection using image processing can be improved. Can be improved.

Further, in the crack detection device 100 of the present embodiment, the confirmation determination unit 234 uses the width Wr of the minimum rectangle RE including the crack candidate region CC as the width W of the crack candidate region CC, and the length of the crack candidate region CC. As L, the length Lr of the minimum rectangle RE is used. Therefore, according to the crack detection device 100 of the present embodiment, the ratio of the length L to the width W of the crack candidate region CC can be calculated quickly and accurately, and the crack detection process using image processing can be performed. The accuracy can be further improved while speeding up.

Further, the crack detection device 100 of the present embodiment further includes a display control unit 210 for displaying an image showing the crack region CR on the display unit 120. Therefore, according to the crack detection device 100 of the present embodiment, it is possible for the user to grasp the occurrence status of cracks in the object.

Further, in the crack detection device 100 of the present embodiment, the object of the crack detection process is a tiled wall of the building 310, that is, a wall-shaped structure. A wall-shaped structure is an object in which the occurrence of cracks is likely to be a problem, and it is necessary to detect cracks in a relatively wide range. According to the crack detection device 100 of the present embodiment, it is possible to efficiently and accurately detect cracks in such an object.

B. Modification example:
The technique disclosed in the present specification is not limited to the above-described embodiment, and can be transformed into various forms without departing from the gist thereof, and for example, the following modifications are also possible.

The configuration of the crack detection device 100 in the above embodiment is merely an example and can be variously modified. For example, the crack detection device 100 may include an imaging unit, and the crack detection processing unit 200 may execute the crack detection processing on an image generated by imaging an object by the imaging unit.

Further, in the above embodiment, the processing screen S1 and the like are displayed on the display unit 120 of the crack detection device 100, but the processing screen S1 and the like are not the display unit 120 provided in the crack detection device 100, but an external display. It may be displayed in. In this case, the crack detection device 100 does not need to include the display unit 120.

Further, in the above embodiment, an image in which the density value of the pixel representing the crack is closer to the density value of black than white is used as the target image IO, but conversely, the density value of the pixel representing the crack is higher than that of black. An image close to the white density value may be used as the target image IO.

Further, the content of the crack detection process in the above embodiment is merely an example and can be variously modified. For example, the processing content of the crack candidate detection process (FIG. 5) in the above embodiment is merely an example, and the crack candidate region CC may be detected by the crack candidate detection process having any other content.

Further, in the above embodiment, the width W of the minimum rectangle RE including the crack candidate region CC is used as the width W of the crack candidate region CC, and the length L of the minimum rectangle RE is used as the length L of the crack candidate region CC. However, the width W and length L of the crack candidate region CC may be defined by other methods.

Further, in the above embodiment, in the crack candidate confirmation process (FIG. 14), the ratio of the length L to the width W of the crack candidate region CC (more specifically, the ratio of the length Lr to the width Wr of the minimum rectangular RE). The crack candidate region CC is set to the crack region CR if the condition that is equal to or higher than the threshold Th1 is satisfied. However, if other conditions are satisfied in addition to the above conditions, the crack candidate region CC is cracked. It may be set in the area CR.

16 and 17 are explanatory views showing an example of the above other conditions. In column A of FIGS. 16 and 17, a crack candidate region CC and a minimum rectangular RE including the crack candidate region CC are shown. Further, in column B of FIGS. 16 and 17, the extended crack candidate region CCe set by extending both ends of the crack candidate region CC in the length direction along the directions of both ends, and the expanded crack candidate region CCe and the expanded crack candidate. The extended minimum rectangle REe, which is the minimum rectangle containing the region CCe, is shown. The above-mentioned "both ends of the crack candidate region CC in the length direction" are, for example, n from each of both ends of the crack candidate region CC when the crack candidate region CC is evenly divided into N pieces in the length direction. Identified as part of. Further, the above-mentioned "direction of both ends" is specified as, for example, the direction of the center line CL at each of the both ends when the center line CL passing through the center in the width direction of the crack candidate region CC is specified.

As the above other condition, a condition that the ratio of the width Wre of the extended minimum rectangle REe to the width Wr of the minimum rectangle RE is equal to or less than a predetermined threshold value Th2 may be used. As shown in FIG. 16, for the crack candidate region CC having a shape extending relatively linearly (that is, the one having a high probability of being a true crack), the width Wre of the extended minimum rectangle REe with respect to the width Wr of the minimum rectangle RE. The ratio is a relatively small value. On the other hand, as shown in FIG. 17, for the crack candidate region CC having a shape extending in a relatively curved shape (that is, the one having a low probability of being a true crack), the width of the extended minimum rectangle REe with respect to the width Wr of the minimum rectangle RE. The Wre ratio is a relatively large value. Therefore, in addition to the condition that the ratio of the length L to the width W of the crack candidate region CC defined in the above embodiment is the threshold Th1 or more, the ratio of the width Wre of the extended minimum rectangle REe to the width Wr of the minimum rectangle RE. If the condition that the threshold value is Th2 or less is satisfied and the crack candidate region CC is set in the crack region CR, it is possible to more reliably suppress the occurrence of false detection and perform image processing. The accuracy of crack detection used can be effectively improved. The threshold Th2 corresponds to the second threshold in the claims.

FIG. 18 is an explanatory diagram showing another example of the above other conditions. In columns A and B of FIG. 18, a crack candidate region CC and a minimum rectangular RE including the crack candidate region CC are shown. As the above other condition, the condition that the ratio of the area of the crack candidate region CC to the area of the minimum rectangle RE including the crack candidate region CC is the threshold Th3 or more may be used. As shown in column A of FIG. 18, for the crack candidate region CC having a shape extending relatively linearly (that is, the one having a high probability of being a true crack), the area of the crack candidate region CC with respect to the area of the minimum rectangular RE. The ratio of is a relatively large value. On the other hand, as shown in column B of FIG. 18, for the crack candidate region CC having a shape extending in a relatively curved shape (that is, the one having a low probability of being a true crack), the crack candidate region CC with respect to the area of the minimum rectangular RE. The area ratio of is relatively small. Therefore, in addition to the condition that the ratio of the length L to the width W of the crack candidate region CC defined in the above embodiment is the threshold Th1 or more, the ratio of the area of the crack candidate region CC to the area of the minimum rectangular RE is the threshold. If the condition of Th3 or higher is satisfied and the crack candidate region CC is set to the crack region CR, it is possible to more reliably suppress the occurrence of false detection, and image processing is used. The accuracy of crack detection can be effectively improved. The threshold Th3 corresponds to a third threshold in the claims.

Further, in addition to the condition that the ratio of the length L to the width W of the crack candidate region CC defined in the above embodiment is the threshold Th1 or more, the width Wr of the minimum rectangle RE described with reference to FIGS. 16 and 17 The condition that the ratio of the width Wre of the extended minimum rectangle REe to the above is the threshold Th2 or less, and the condition that the ratio of the area of the crack candidate region CC to the area of the minimum rectangle RE described with reference to FIG. 18 is the threshold Th3 or more. , The crack candidate region CC may be set to the crack region CR. Moreover, other conditions may be weighted.

In the modified example shown in FIG. 18, the condition that the ratio of the area of the crack candidate region CC to the area of the minimum rectangle RE including the crack candidate region CC is equal to or more than the threshold Th3 is adopted. Not limited to the minimum rectangular RE containing the crack candidate area CC, a polygon containing the crack candidate area CC defined according to a predetermined rule can be used. As the other condition, the condition that the ratio of the area of the crack candidate region CC to the area of the polygon including the crack candidate region CC is the threshold Th3 or more may be used.

Further, in the above embodiment, the object of the crack detection treatment is a tiled wall of the building 310, but other wall-like structures (for example, a concrete wall or a civil engineering structure of a building). It may be an object other than a wall-like structure.

100: Crack detection device 110: Storage unit 120: Display unit 130: Input unit 140: Interface unit 170: Control unit 190: Bus 200: Crack detection processing unit 210: Display control unit 220: Candidate detection processing unit 222: Binary image Acquisition unit 224: Candidate determination unit 226: Candidate area setting unit 230: Candidate confirmation processing unit 232: Rectangular calculation unit 234: Confirmation determination unit 310: Building 320: Imaging device CC: Crack candidate area CCe: Extended crack candidate area CL: Center line CP: Crack detection program CR: Crack area IO: Target image Ib (1): First binary image Ib (2): Second binary image Id: Image data Pc: Crack pixel Pcc1: Primary Crack candidate pixel Pcc2: Secondary crack candidate pixel Pk: Black pixel Pn: Noise pixel Pw: White pixel R1: Image selection area R11: Number display field R2: Method selection area R3: Image display area RE: Minimum rectangular REe: Expansion Minimum rectangle S1: Screen during processing SL: Virtual line SP: Pixel of interest

Claims (6)

A crack detection device that detects cracks in an object.
In the image representing the object, a candidate detection unit for detecting a crack candidate region having a high probability of being an image region representing a crack, and a candidate detection unit.
Among the crack candidate regions, the crack candidate region satisfying a predetermined condition including that the ratio of the length L (however, L ≧ W) to the width W of the crack candidate region is equal to or more than the first threshold value is cracked. A confirmation judgment unit set in the cracked area, which is an image area representing
Equipped with a,
The width W of the crack candidate region is the width Wr of the first minimum rectangle which is the smallest rectangle including the crack candidate region, and the length L of the crack candidate region is the length of the first minimum rectangle. Is Lr
The predetermined condition includes an extended crack candidate region set by extending both ends in the length direction of the crack candidate region with respect to the width of the first minimum rectangle along the direction of the both ends. A crack detection device comprising a ratio of widths of a second minimum rectangle, which is the smallest rectangle, to be less than or equal to a second threshold. The crack detection device according to claim 1.
The predetermined condition is a crack detection device, wherein the ratio of the area of the crack candidate region to the area of the polygon including the crack candidate region is equal to or more than a third threshold value. The crack detection device according to claim 1 or 2, further comprising.
A crack detection device including a display control unit that displays an image showing the crack region on a display. The crack detection device according to any one of claims 1 to 3.
The object is a crack detection device, which is a wall-shaped structure. It is a crack detection method that detects cracks in an object.
In the image representing the object, a step of detecting a crack candidate region having a high probability of being an image region representing a crack, and a step of detecting the crack candidate region.
Among the crack candidate regions, the crack candidate region satisfying a predetermined condition including that the ratio of the length L (however, L ≧ W) to the width W of the crack candidate region is equal to or more than the first threshold value is cracked. The process of setting the cracked area, which is the image area representing
Equipped with a,
The width W of the crack candidate region is the width Wr of the first minimum rectangle which is the smallest rectangle including the crack candidate region, and the length L of the crack candidate region is the length of the first minimum rectangle. Is Lr
The predetermined condition includes an extended crack candidate region set by extending both ends in the length direction of the crack candidate region with respect to the width of the first minimum rectangle along the direction of the both ends. A method for detecting cracks, which comprises a ratio of widths of a second minimum rectangle, which is the smallest rectangle, to be less than or equal to a second threshold. A computer program for detecting cracks in an object
In the image representing the object, a candidate detection function for detecting a crack candidate region having a high probability of being an image region representing a crack, and
Among the crack candidate regions, the crack candidate region satisfying a predetermined condition including that the ratio of the length L (however, L ≧ W) to the width W of the crack candidate region is equal to or more than the first threshold value is cracked. Confirmation judgment function to set in the cracked area, which is the image area representing
To the computer ,
The width W of the crack candidate region is the width Wr of the first minimum rectangle which is the smallest rectangle including the crack candidate region, and the length L of the crack candidate region is the length of the first minimum rectangle. Is Lr
The predetermined condition includes an extended crack candidate region set by extending both ends in the length direction of the crack candidate region with respect to the width of the first minimum rectangle along the direction of the both ends. A computer program that includes that the ratio of the widths of the second smallest rectangles, which are the smallest rectangles, is less than or equal to the second threshold.

Images