JP5218849B2 - Defect inspection equipment - Google Patents

Description

  The present invention relates to an apparatus for inspecting defects such as cracks.

  A method for automatically analyzing the distribution of cracks on the concrete surface has been proposed. For example, in Non-Patent Document 1, an inspection object is picked up as a visible image with a digital camera, and a cracked portion is specified by performing wavelet conversion and binarization processing. Furthermore, the process which calculates the width | variety of a crack based on the image is also performed.

  Such a method is effective in terms of automatic detection of cracks. However, when the crack width is small, there is a problem that the cracked portion cannot be identified in the visible image. The same problem occurs when cracks are generated not on the surface but inside.

  Therefore, it has been proposed to identify cracks that are difficult to find in the visible image as described above by imaging the inspection object with an infrared camera and performing the same processing as described above (Non-Patent Document 2). This is because, in an infrared camera, a clear temperature difference is generated even in a portion that is difficult to find in a visible image, so that cracks can be identified.

Tsuyoshi Maruya et al. “Practical application of crack investigation of floor slab concrete using wavelet transformation”, Proceedings of Concrete Engineering, Vol. 29, No. 2 (2007) Shigeru Inaba, “Relationship between crack width and measurement distance by thermography method, conference papers on thermography method”, pages 29-32 (April 1992)

  However, the conventional method for identifying cracks using an infrared camera has a problem that the shape and dimensions of the cracks are not sufficient because the resolution of the infrared camera is lower than that of a visible camera.

  In recent years, a technique for increasing the resolution (super-resolution) using a plurality of images whose positions are gradually shifted has been put into practical use. Although it is conceivable to apply these techniques to the infrared image to increase the resolution, the following problems arise.

  That is, in high resolution processing such as super-resolution processing, processing is performed on the basis of the amount of misalignment on the premise that the amount of misregistration of different images is accurately known. In the case of a visible image, a specific part can be determined as a reference point from the image itself, and the amount of deviation of the reference point in a plurality of images can be known. However, an infrared image has a lower contrast than a visible image, and it is difficult to accurately determine a reference point from the image itself, and it is difficult to accurately obtain a reference point shift amount in a plurality of images. For this reason, it has been difficult to perform high resolution processing based on an infrared image.

  Therefore, there is a problem that inspection with a visible image is unavoidable, and it is impossible to detect a defect such as a crack and a size that are difficult to detect with a visible image.

  An object of the present invention is to solve the above-described problems and provide an apparatus for inspecting defects such as cracks that are difficult to capture only with a visible image.

(1) (3) In the defect inspection apparatus according to the present invention, a visible image capturing unit that captures images while moving a target region to be imaged and generates a series of visible images based on visible light intensity, and the visible image capturing unit includes An infrared image capturing unit that captures a target region substantially the same as the target region to be imaged and generates a series of infrared images based on infrared light intensity corresponding to the series of visible images; and the infrared image Infrared image defect region specifying means for performing pattern analysis based on image density difference and specifying a defect region, and infrared image search region setting for setting a search region including the defect region in the infrared image Means, a visible image search area setting means for setting a search area in the series of visible images in correspondence with a search area set in the infrared image, and a high level based on the series of visible images in the search area. Resolution processing Visible image high-resolution means for obtaining a high-resolution search area visible image, and pattern analysis based on the density difference of the image based on the search area visible image to identify a defective part Means.

  Since the site of the defect is specified based on the infrared image, it is possible to find a defect that is difficult to find in the visible image.

(2) (4) A defect inspection apparatus according to the present invention includes a visible image defect size specifying means for specifying a size of a defective part based on a search region visible image.

  Since the site of the defect is specified based on the infrared image, it is possible to find a defect that is difficult to find in the visible image. However, even if a defect can be found by using only the infrared image, the resolution cannot be increased only by the infrared image itself, and it is difficult to specify the dimensions. Therefore, the defect portion found in the infrared image is increased in resolution by the corresponding visible image. Thereby, since only the periphery of the defective part in the visible image is increased in resolution, the defective part can be found also in the visible image, and the defect size can be specified because the resolution is increased.

(7) (9) In the defect inspection apparatus according to the present invention, a visible image capturing unit that captures images while moving a target region to be imaged and generates a series of visible images based on visible light intensity, and the visible image capturing unit includes An infrared image capturing unit that captures a target region substantially the same as the target region to be imaged and generates a series of infrared images based on infrared light intensity corresponding to the series of visible images; and the infrared image Infrared image defect region specifying means for performing pattern analysis based on image density difference and specifying a defect region, and infrared image search region setting for setting a search region including the defect region in the infrared image Means, a visible image search area setting means for setting a search area in the series of visible images in correspondence with a search area set in the infrared image, and the series of visible image search areas, The position of the search area is not Based on the information, a visible image high-resolution processing means for performing high-resolution processing of the visible image in the search area, and a pattern according to the density difference of the image based on a series of visible images in the high-resolution search area Visible image defect site identification means for performing analysis and identifying a defect site is provided.

  Since the part of the defect is specified based on the infrared image, it is possible to see the defect that is difficult to find in the visible image.

(8) (10) The defect inspection apparatus according to the present invention further includes a visible image defect size specifying means for specifying the size of the defective portion based on a series of visible images in the search area with high resolution. .

  Therefore, the dimension of the defective part can be specified.

(13) (15) In the defect inspection apparatus according to the present invention, a visible image capturing unit that captures an image while moving a target region to be imaged and generates a series of visible images based on visible light intensity, and the visible image capturing unit includes An infrared image capturing unit that captures a target region substantially the same as the target region to be imaged and generates a series of infrared images based on infrared light intensity corresponding to the series of visible images; and the series of red Infrared image high-resolution processing means for performing high-resolution processing of the infrared image based on the positional deviation information of the visible image corresponding to the series of infrared images using the outside image, and the high resolution Infrared image defect site specifying means for performing pattern analysis based on the density difference of the images based on the series of infrared images and specifying the defect site is provided.

  Since the site of the defect is specified based on the infrared image, it is possible to find a defect that is difficult to find in the visible image. However, even if it is possible to find a defect with only the visible image, it is impossible to increase the resolution with only the visible image itself, and it is difficult to specify the dimensions. Therefore, the resolution of the infrared image is increased based on the position information based on the visible image, and the defect size is specified.

  The “infrared image defect site specifying means” corresponds to step S3 in FIG. 4A in the embodiment.

  The “infrared image search area setting means” corresponds to step S4 of FIG. 4a in the embodiment.

  The “visible image search area setting unit” corresponds to step S6 in FIG. 4A in the embodiment.

  The “visible image high-resolution means” corresponds to step S7 in FIG. 4A in the embodiment.

  “Visible image defect site specifying means” corresponds to step S8 in FIG. 4A in the embodiment.

  The “visible image defect size specifying means” corresponds to step S9 in FIG. 4a in the embodiment.

  The “infrared image high-resolution processing means” corresponds to step S13 in FIG. 21 in the embodiment.

  The “program” is a concept that includes not only a program that can be directly executed by the CPU, but also a source format program, a compressed program, an encrypted program, and the like.

It is a functional block diagram of the defect inspection apparatus by 1st embodiment. It is a figure which shows the whole structure of a defect inspection apparatus. It is a hardware configuration of a defect inspection apparatus. It is a flowchart of a defect inspection program. It is a flowchart of a defect inspection program. It is a figure which shows a series of visible images. It is a figure which shows a series of infrared images. It is a figure which shows typically the process which aligns the magnitude | size and shape of a visible image and an infrared image. It is a figure which shows the infrared image which specified the crack. It is a figure which shows an infrared image search area | region. It is a figure which shows a visible image selection area | region. It is a figure which shows a series of selection area visible images in each visible image selection area. It is a figure for demonstrating a super-resolution process. It is a figure for demonstrating a super-resolution process. It is a figure for demonstrating a super-resolution process. It is a figure for demonstrating a super-resolution process. It is a figure for demonstrating a super-resolution process. It is a figure which shows the search area | region visible image made high resolution. It is a figure which shows the width | variety (alpha) in a crack image. It is a functional block diagram of the defect inspection apparatus by 2nd embodiment. It is a functional block diagram of the defect inspection apparatus by 3rd embodiment. It is a figure which shows the flowchart of a defect inspection program. It is a figure which shows a spatial surface function. It is the figure which considered RGB as a triangle.

1. First embodiment
(1) Functional Block Diagram FIG. 1 shows an overall configuration diagram of a defect inspection apparatus according to an embodiment of the present invention. The visible image capturing unit 2 captures an image while moving a target region to be captured, and generates a series of visible images based on the visible light intensity. The infrared image capturing unit 10 captures a target region that is substantially the same as the target region captured by the visible image capturing unit. Thereby, the infrared image capturing unit 10 generates a series of infrared images based on the infrared light intensity corresponding to the series of visible images.

  The infrared image defect site | part identification means 12 performs a pattern analysis by the density difference of an image based on any one (2 or more sheets) of the said infrared image, and specifies a defect site | part. The infrared image search area setting means 14 sets a search area including the defective part in the infrared image. The visible image search area setting means 4 sets a search area for a series of visible images in correspondence with the search area set for the infrared image.

  The visible image resolution increasing processing unit 5 increases the resolution of the search area set in the series of visible images. The visible image defect site specifying means 6 performs pattern analysis based on the density difference of the image based on the high-resolution series of visible image search areas, and specifies the defect site. The visible image defect size specifying means 8 specifies the size of the defect site based on a series of visible image search areas.

(2) Overall Configuration FIG. 2 shows the overall configuration of the defect inspection apparatus according to one embodiment of the present invention. A visible camera 16 that is a visible image capturing unit and an infrared camera 18 that is an infrared image capturing unit are provided. The optical axis 20 of the visible camera 16 and the optical axis 22 of the infrared camera 18 are set in parallel, and the inspection object 24 is imaged. As described above, since the optical axes 20 and 22 of the visible camera 16 and the infrared camera 18 are set in parallel, the visible camera 16 and the infrared camera 18 have substantially the same region of the inspection object 24 regardless of the distance. I will take an image.

  The visible camera 16, the infrared camera 18, and the computer 26 are mounted on a car or the like (not shown) and moved in a direction perpendicular to the paper surface. The moving direction of the visible camera 16 and the infrared camera 18 is preferably a direction perpendicular to the cracking direction expected to occur in the inspection object 24.

  The visible image by the visible camera 16 and the infrared image by the infrared camera 18 are input to the computer 26, and a crack location is specified and a dimension is determined by a defect inspection program.

(3) Hardware configuration FIG. 3 shows the hardware configuration of the defect inspection apparatus. A display 30, a visible camera 16, an infrared camera 18, a hard disk 34, and a memory 36 are connected to the CPU 32. An operating system (such as WINDOWS (trademark)) 38 and a defect inspection program 40 are recorded on the hard disk 34. The operating system 38 and the defect inspection program 40 are obtained by installing the operating system 38 and the defect inspection program 40 recorded on the CD-ROM 44 via the CD-ROM drive 42. The defect inspection program 40 performs its function in cooperation with the operating system 38.

(4) Flowchart of Defect Inspection Program FIGS. 4a and 4b show flowcharts of the defect inspection program 40 recorded on the hard disk 34. FIG.

  The CPU 32 records the visible image from the visible camera 16 on the hard disk 34 (step S1). Since the visible camera 16 captures an image while moving with respect to the inspection object 24, a series of images with slightly different positions are recorded on the hard disk 34. In this embodiment, images of 2048 pixels vertically and 2048 pixels horizontally are continuously recorded on the hard disk 34. 5A, 5B and 5C show examples of a series of visible images recorded in this way. Since the visible camera 16 is moving, it can be seen that the images are gradually shifted.

  The CPU 32 records the infrared image from the infrared camera 18 on the hard disk 34 (step S1). Since the infrared camera 18 also performs imaging while moving with respect to the inspection object 24 in the same manner as the visible camera 16, a series of images with slightly different positions are recorded on the hard disk 34. Note that the infrared image has a feature that not only the pixel density is coarser than the visible image but also the contrast of the image is not clear.

  6A, 6B and 6C show examples of a series of infrared images recorded in this way. As is clear from the figure, cracks 31 that are difficult to see appear in the visible images (FIGS. 5A, 5B, and 5C). This is because the infrared camera 18 generates an image based on the temperature difference, so that even a fine crack appears as an image as long as it causes a temperature difference from the surroundings. In addition, since the concrete does not exist and there exists air (or water) in the crack part which arose in concrete etc., a difference arises in temperature from the difference in specific heat etc. with the surrounding concrete.

  Next, the CPU 32 performs correction in consideration of a difference in lens aberration between the visible camera 16 and the infrared camera 18, a difference in distance between the lens optical axes, and the like so that the visible image and the infrared image are in the same position. To do. For example, the lens aberration is corrected for each of the visible image and the infrared image shown in FIG. 7A, and the sizes and shapes of the visible image and the infrared image are made uniform as shown in FIG. 7B. Further, as shown in a thick frame in FIG. 7C, in consideration of the deviation of the lens optical axis (see optical axes 20 and 22 in FIG. 2), a region that can be said to be the same region in the visible image and the infrared image is cut out. . Such processing is performed on a series of images of a visible image and an infrared image.

  The CPU 32 records the visible image and the infrared image thus extracted as the same area on the hard disk 34. Here, it is assumed that a series of visible images and a series of infrared images recorded in this way are as shown in FIGS. 5 and 6 are used to show the recorded image at the time of step S1, but in this specification, they are also used to show the recorded image at the time of step S2 for convenience.

  Next, the CPU 32 specifies a cracked part using one image of the series of infrared images (step S3). For example, a crack site is specified using the infrared image of FIG. 6A. The CPU 32 removes noise by binarizing the infrared image of FIG. 6A and searches for a continuous linear shape, thereby obtaining an image as shown in FIG. 8A and specifying a cracked portion. For specific processing of cracked parts, it is possible to use techniques such as Tsuyoshi Maruya, “Practical application of crack investigation of floor slab concrete using wavelet transformation”, Concrete Engineering Vol. 29, No. 2 (2007). it can.

  Next, the CPU 32 sets the infrared image search areas 42, 44, 46, and 48 including the cracked part as shown in FIG. 9 in the image specifying the cracked part as shown in FIG. 8A (step S4). For example, the lateral width of the search area may be determined, and the rectangular area may be set so that the crack area is included according to this.

  Next, the CPU 32 takes out the visible image of FIG. 5A corresponding to the infrared image of FIG. 6A used for setting the infrared image search region of FIG. Then, in the visible image of FIG. 5A, visible image search areas 52, 54, 56, and 58 corresponding to the infrared image search areas 42, 44, 46, and 38 are set (see FIG. 10) (step S5). Since the positional relationship between the infrared image and the visible image corresponds as described above, the visible image search regions 52, 54, 56, and the like are located at the position of the visible image corresponding to the infrared image search region in the infrared image. 58 may be set. Thereby, as shown in FIG. 10, it is possible to set a visible image search region 52 including a crack region that is difficult to find with only a visible image. Each image in the visible image search areas 52, 54, 56, and 58 in FIG. 10 (FIG. 5A) is recorded in the hard disk 34 as a search area visible image.

  Subsequently, the CPU 32 reads the next visible image (FIG. 5B) recorded on the hard disk 34, and sets the visible image search areas 52, 54, 56, and 58 at the same position as the infrared image search area set in FIG. To do. Each image in the visible image search areas 52, 54, 56, and 58 (not shown) in FIG. 5B is recorded in the hard disk 34 as a search area visible image.

  In the same manner as described above, the visible image search areas 52, 54, 56, and 58 are set for a series of visible images, and the images in the areas are recorded on the hard disk 34 as the search area visible images.

  Since the visible image search areas 52, 54, 56, and 58 are fixed in a series of visible images and the imaging position of the visible image is moving, the visible image search areas 52 are shown in FIGS. 11A, 11B, and 11C. , 54, 56, and 58, the cracks are sequentially moved.

  Next, the resolution is increased for each of the areas 52, 54, 56, and 58 based on the series of search area visible images shown in FIG. 11 (step S7). In this embodiment, high resolution is achieved by super-resolution processing.

  The principle of super-resolution processing is shown in FIG. FIG. 12A shows the original luminance (original) curve α. On the other hand, when this is digitized, as shown in FIG. 12B, it becomes a discrete value averaged with the pixel width reflecting the original luminance curve α (digital image 1). When this imaging position is shifted by ½ pixel, a digital image 2 as shown in FIG. 12C is obtained. Therefore, by taking the average of the value of the digital image 1 and the value of the digital image 2 at each position, a high-resolution digital image as shown in FIG. 12D can be obtained. By performing such processing two-dimensionally, it is possible to increase the resolution of a digital image.

  In order to perform the super-resolution processing as described above, it is necessary to accurately grasp the positional deviation amounts of a plurality of images in units smaller than pixels. In this embodiment, the amount of displacement is calculated based on the series of search region visible images themselves shown in FIGS. 11A, 11B, and 11C. For example, the amount of image misregistration can be determined by SSD (residual sum of squares) two-dimensional parabolic fitting. Hereinafter, the SSD two-dimensional parabolic fitting will be described.

  For example, the positional deviation amount between the search area visible image of FIG. 11A in the area 52 and the search area visible image of FIG. 11B in the area 54 is obtained as follows.

  The luminance values at certain coordinates (x, y) of the comparison target image 1 (the image of the region 52 in FIG. 11A) and the image 2 (the image of the region 54 in FIG. 11B) are I1 (x, y) and I2 (x, y), respectively. ), When the displacement in units of pixels of the image 2 with respect to the image 1 is (dx, dy), the SSD is expressed by the following equation.

  Here, W represents a region of interest (subset) for calculating the SSD. When measuring the shift amount in pixel units, it is only necessary to obtain (dx, dy) that minimizes SSD (dx, dy). On the other hand, in order to obtain the movement amount in sub-pixel units (in units finer than the pixels), parabolic fitting that assumes that the SSD value can be approximated to a quadratic function with respect to the movement amount is used as follows. .

  By moving (dx, dy), a movement amount (dx, dy) in pixel units that minimizes SSD (dx, dy) is obtained. The movement amount at this time is (dx, dy) is (i, j) (integer). Further, SSD (i-1,, j), SSD (i ,, j-1), (i ,, j + 1) are obtained.

  In the vicinity of (i, j), SSD (x, y) can be expressed by a quadratic equation Dp (dx.dy), and its coefficients are SSD (ij), SSD (i-1, j), SSD (i , j-1), SSD (i + 1, j), SSD (i, j + 1). As a result, (dx, dy) that minimizes SSD (dx.dy) is obtained as the vertex of Dp (dx, dy). This can be the position of displacement in translation.

  Next, a method for obtaining the vertex of Dp (dx, dy) that minimizes SSD (dx, dy) will be described.

  Here, Dp (dx, dy) is assumed to be expressed as follows.

  The values of SSD (ij), SSD (i-1, j), SSD (i, j-1), SSD (i + 1, j), SSD (i, j + 1) are obtained and substituted into the following equation The coefficient ak (K = 0, 1, 2, 3, 4) is calculated.

It is. Equation (2) can be modified as follows.

  At this time, if the coefficient ak is obtained, the movement amount in sub-pixel units can be calculated as the vertex (-a1 / a0, -a3 / a2) of the quadratic function Dp (dx, dy).

  As described above, the shift amount between the image of FIG. 11A in the region 52 and the image of FIG. 11B in the region 54 can be calculated in subpixel units (units smaller than the pixels).

  Subsequently, a high-resolution image is obtained based on the amount of deviation calculated as described above. As a technique for increasing the resolution, there are a pixel shifting method, a pixel shifting method considering weights, a method using local iterative calculation, and the like.

  The pixel shifting method is a method of taking the average of the corresponding luminance values of a plurality of measured low-resolution images in order to obtain the luminance value of a certain pixel in the high-resolution image. For example, consider a case where the moving pixel amount of a certain low resolution image indicated by a dotted line is (k + dx, l + dy) with respect to a reference low resolution image indicated by a solid grid as shown in FIG. . At this time, all the luminance values of the pixels in the high-resolution image included in the area A, which is an area where the pixel (i, j) of Iref and the pixel (i + k, j + l) of If are overlapped, are {Iref (i, j) + If (i + k, j + l)} / 2.

  If a plurality of different image data with a movement amount within one pixel is used, the number of divisions can be increased, and thus a higher resolution image can be obtained in principle.

  The pixel shifting method considering the weight is executed as follows. In simple image superposition in the pixel shifting method, a luminance value of a certain pixel of a high resolution image and a pixel of a low resolution image are generated under the same influence from the large or small influence. For example, as shown in FIG. 14, three reference low-resolution images A, B, and C are used, and a pixel S1 having a high-resolution image that divides each pixel of the reference image A into three has A, B, and C. Affected by pixels A2, B1, B2, C1. However, in fact, the one having the value closest to the luminance value of S1 is considered to be C1 having the pixel center closest to the pixel center of S1. Therefore, when obtaining the luminance value of a pixel of a high-resolution image, an average is obtained by applying a weight function so that the luminance value of the pixel of the low-resolution image whose pixel centers are close to each other has a greater influence.

  In the method using the local iterative calculation, the plurality of measured low resolution images fi are considered to be images obtained by sampling from the true high resolution image CReal.

  If the luminance value fi (u, v) of the low resolution image is sampled from nine points near the luminance value CReal (xu, yv) of the true high resolution image as shown in the upper right diagram of FIG. v) is expressed as Equation (6). However, (xu, yv) represents the coordinates of a pixel including the center of fi (u, v) in the coordinate system when a high resolution image is obtained.

hf _i (k, l) is a sampling function of fi for the high-resolution image Creal, and the pixel of the high-resolution image separated by (k, l) from (xu, yv) is the pixel (u, v ) And the area ratio in contact with.

  A high resolution image can be constructed by repeating the following process. The conceptual diagram is shown in FIG.

  Step 1: Set the initial value C0 (x, y) of the luminance value of the high resolution image. Since the selection of the initial value does not affect the convergence result, any image may be used. In this study, an image obtained by subdividing the reference image used for alignment by linear interpolation was used from the viewpoint of convergence speed. . Here, the number of repetitions n = 0.

  Step 2: Create low-resolution image Fi (U, V) by applying sampling function hfi (k, l) of measurement data fi (u, v) for Creal to Cn (x, y) . i represents the observation data number.

  Step 3: If the value of Cn (x, y) is the true luminance value Creal (x, y) of the high-resolution image, the low-resolution image Fi (u, v) generated in Step 2 is actually It should match the measured low resolution image fi (u, v). Therefore, it is assumed that the observation error En is defined as Equation (6), and the process is terminated when En takes the minimum value.

  Step 4: When En is not sufficiently small, Cn (x, y) is corrected to Cn + 1 (x, y), and the process returns to Step 2. The pixel (x, vy) of the low resolution image affected by the pixel (x, y) of the high resolution image is represented as (ux, vy), and the correction amount Δei (ux, vy) relating to the low resolution image is expressed as in Equation (7).

At this time, a is an appropriate scaling parameter.

  Since there are a maximum of 4 pixels in the low-resolution image affected by one pixel in the high-resolution image, when the weight for each Δei (ux, vy) is wi (ux, vy), the weighted average is high. Assuming that the total correction amount Δei (x, y) for the pixel (x, y) of the resolution image, Cn + 1 (x, y) is expressed as Equation (8).

  Here, the sampling function hfi of the measurement data fi (u, v) for Creal is used as the weighting function wi. It has also been proposed to use a smoothing filter and a sharpening filter as candidates for wi. At this time, when the formula (8) is rewritten, the formula (9) is obtained.

  The CPU 32 performs the super-resolution processing described above using a series of search area visible images of each area 52, and obtains a search area visible image with a high resolution. Similarly, with respect to a series of search area visible images of the other areas 54, 56, and 58, a high-resolution search area visible image is obtained by super-resolution processing.

  FIG. 17 shows a search area visible image with high resolution of each of the areas 52, 54, 56, and 58 obtained in this way. Due to the higher resolution, the cracks in the region 52, which were difficult to discriminate in the original visible image, are clarified. In the regions 54, 56, and 58, the crack width is clarified by the high resolution.

  The CPU 32 performs a crack identification process on the visible images of the regions 52, 54, 56, and 58 that have been increased in resolution in the same manner as in step S3 (step S8). The CPU 32 displays the cracks thus identified on the display 30.

Further, the width of the crack portion in the image is defined as a direction perpendicular to the extending direction, and the size of the largest width (for example, α portion in FIG. 18) in each crack is calculated as the crack size.

2. Second Embodiment FIG. 19 is an overall configuration diagram of a defect inspection apparatus according to a second embodiment. In this embodiment, unlike the first embodiment, the visible image high-resolution processing means 5 is not provided. The visible image capturing unit 2 captures an image while moving a target region to be captured, and generates a series of visible images based on the visible light intensity. The infrared image capturing unit 10 captures a target region that is substantially the same as the target region captured by the visible image capturing unit. Thereby, the infrared image capturing unit 10 generates a series of infrared images based on the infrared light intensity corresponding to the series of visible images.

  The infrared image defect site | part identification means 12 performs a pattern analysis by the density difference of an image based on any one (2 or more sheets) of the said infrared image, and specifies a defect site | part. The infrared image search area setting means 14 sets a search area including the defective part in the infrared image. The visible image search area setting means 4 sets a search area for a series of visible images in correspondence with the search area set for the infrared image.

  The visible image defect site identification means 6 performs pattern analysis based on the density difference of the images based on a series of visible image search areas, and identifies the defect site. The visible image defect size specifying means 8 specifies the size of the defect site based on a series of visible image search areas.

  The flowchart of the defect inspection program in the second embodiment is almost the same as that in FIGS. 4a and 4b. However, the difference is that there is no high resolution processing in step S7.

In this embodiment, since the visible image high-resolution processing means 5 is not provided, it is difficult to find a defective part as compared with the first embodiment. However, the search area including the part identified as a defect in the infrared image is set in the visible image. Therefore, it is guaranteed that there is a defect in the search area, and it is easy to find the defect based on the visible image.

3. Third Embodiment FIG. 20 shows an overall configuration diagram of a defect inspection apparatus according to a third embodiment. The visible image capturing unit 2 captures an image while moving a target region to be captured, and generates a series of visible images based on the visible light intensity. The infrared image capturing unit 10 captures a target region that is substantially the same as the target region captured by the visible image capturing unit. Thereby, the infrared image capturing unit 10 generates a series of infrared images based on the infrared light intensity corresponding to the series of visible images.

  The infrared image high-resolution processing means 11 performs high-resolution processing of the infrared image using the positional deviation information of the visible image corresponding to the series of infrared images. In other words, the amount of displacement described in the first embodiment is calculated based on the visible image, and super-resolution processing such as a pixel shifting method, a pixel shifting method that considers weights, and a method that uses local iterative calculation is red. This is performed on the outside image.

  The infrared image defect site | part identification means 13 performs a pattern analysis by the density difference of an image based on a series of infrared images, and specifies a defect site | part. The infrared image defect size specifying means 15 specifies the size of the defect site based on a series of infrared images.

  In the first embodiment and the second embodiment, the identification of the defective part and the determination of the dimensions are finally performed based on the visible image. Therefore, it is not possible to determine the size of an internal defect that does not appear in the visible image.

  Since the infrared image is an image based on a temperature difference, defects that do not appear in the visible image can be clearly captured. However, the position information cannot be obtained accurately only with the image information of the infrared image, and the resolution of the image cannot be increased. In the third embodiment, an infrared image and a visible image with matching position information are used to increase the resolution of the infrared image based on the position information of the visible image. Therefore, even a defect that does not appear in the visible image can be specified up to its size.

FIG. 21 shows a flowchart of the defect inspection program in the third embodiment. Steps S11 and S12 are the same as steps S1 and S2 in FIG. 4a. Step S13 is the above-described high resolution processing. Steps S14 and S15 correspond to steps S8 and S9 in FIG. 4b.

4). Other embodiments
(1) In the above embodiments, cracks are targeted as defects. However, in addition to cracks, it can also be used on concrete surfaces and internal voids generated during construction.

(2) In each of the above embodiments, the determination of the defect size is performed. However, only the defect detection may be performed.

(3) In each of the above embodiments, the width of the crack is measured as the size of the defect. However, the length of cracks and the like may be measured.

(4) In each of the above embodiments, cracks are identified and dimensions are measured based on the super-resolution processing of the infrared image and the visible image. However, if a clear contrast appears in the infrared image, it is possible to determine the amount of positional deviation only with the infrared image and to perform super-resolution processing. For example, if the spider's web is stretched around the crack, a clear contrast can be obtained due to the temperature difference of the spider's web. Because it can.

(5) In each of the above embodiments, the crack is specified by image processing. However, the setting of the infrared image search area based on the infrared image (step S4 in FIG. 4a) may be performed manually as follows. The CPU 32 displays an infrared image on the display 30. The user sets an infrared image search area using a mouse, a keyboard, or the like (not shown).

  In addition, the CPU 32 does not perform the processes of steps S8 and S9 (steps S14 and S15), displays a high-resolution search area visible image on the display 30, and the user visually identifies the defect and determines the size. You may do it.

(6) In each of the above-described embodiments, the positional shift amount of the image by parabolic fitting or the like is calculated based on the gray scale (grayscale image) for the positional shift in the resolution enhancement of the visible image in step S6. However, the positional deviation amount may be obtained by parabolic fitting or the like using RGB color visible images.

Each pixel value of RGB is set to C _R , C _G , and C _B. Further, the reference image and the image for obtaining the positional deviation amount are represented by subscript numerals 0 and 1, respectively. At this time, the RGB three-dimensional residual sum of squares SSD is expressed as follows.

W is a subset region. Hereinafter, this SSD may be minimized as in the case of the gray image.

i) Method Using Spatial Plane Function When the above SSD becomes zero, perfect matching is achieved and there is no deviation amount. The minimum value of SSD is calculated as the amount of deviation. However, if the SSD is simply minimized, it cannot be said that the amount of deviation is naturally obtained. Therefore, the amount of deviation may be calculated using a spatial surface function.

  For example, it is assumed that the coordinates (x, y) of the reference image are compared with the coordinates CR1 (x + i, y + i) of the image for which the positional deviation information is obtained. (CRO (x, y) −CR1 (x + i, y + i)) = 10, (CGO (x, y) −CG1 (x + i, y + i)) = 0, (CB0 (x, y ) −CB1 (x + i, y + i)) = 0 (CRO (x, y) −CR1 (x + i, y + i)) = 6, (CGO (x, y) −CG1 ( x + i, y + i)) = 6 and (CBO (x, y) -CB1 (x + i, y + i)) = 6 are considered to be closer in visual color (RGB is 16 bits) in the case of).

  Therefore, R, G, and B are (x, θ, z) = (1, 0, CR), (1, 2π / 3, CG), and (1, 4π / 3) in the (x, θ, z) plane, respectively. , CB). If so defined, a diagram as shown in FIG. 22 can be drawn. Therefore, a function C (x, θ) of the surface S passing through these three points is considered. There are three variables in the function C (x, θ), and matching can be performed by using these three variables in place of CR, CG, and CB in Equation 10. Since two of the three variables are coefficients representing the slope of the surface S and the other is an intercept, weighting may be performed according to the degree of influence of the three variables on the pattern matching accuracy. . That is, when pattern matching is performed using color information, it is possible to examine which variable can be used to achieve the most accurate pattern matching, and weight the variables based on the result.

ii) Method of considering RGB as a triangle As shown in FIG. 23, a triangle is formed by extending RGB pixel values. The correct amount of movement is obtained by searching for something similar to this triangle.

Claims (9)

A visible image capturing unit that captures an image while moving a target region to be imaged, and generates a series of visible images based on visible light intensity;
An infrared image capturing unit that captures a target region substantially the same as the target region captured by the visible image capturing unit and generates a series of infrared images based on infrared light intensity corresponding to the series of visible images. When,
Infrared image high-resolution processing means for performing high-resolution processing of the infrared image based on positional shift information of the visible image corresponding to the series of infrared images, using the series of infrared images;
Based on the high-resolution series of infrared images, pattern analysis is performed according to the density difference of the image, infrared image defect site identification means for identifying the defect site,
Defect inspection device with In the defect inspection apparatus according to claim 1,
Infrared image defect size specifying means for specifying the size of the defect site based on a series of infrared images in the high-resolution search area,
A defect inspection apparatus further comprising: A defect inspection program for causing a computer to function as a defect inspection apparatus,
A series of visible images based on visible light intensity captured while moving the target area to be imaged, and the series of visible images in which a target area substantially the same as the target area captured by the visible image capturing unit is captured. Means for accessing a recording unit that records a series of infrared images based on infrared light intensity corresponding to the images;
Using the series of infrared images, and based on the positional deviation information of the visible image corresponding to the series of infrared images, an infrared image enhancement processing means for performing the resolution enhancement processing of the infrared image;
Based on the high-resolution series of infrared images, pattern analysis is performed according to the density difference of the image, infrared image defect site specifying means for specifying the defect site,
Defect inspection program for realizing a computer. The defect inspection program according to claim 3, further comprising: an infrared image defect size specifying means for specifying a dimension of the defect portion based on a series of infrared images in the search area having the high resolution.
A defect inspection program realized by a computer. Capture a target area to be imaged, generate a series of visible images based on visible light intensity,
Imaging a target region substantially the same as the target region captured by the visible image capturing unit, and generating a series of infrared images based on infrared light intensity corresponding to the series of visible images;
Using the series of infrared images, based on the positional deviation information of the visible image corresponding to the series of infrared images, performing a high resolution processing of the infrared image,
Based on the high-resolution series of infrared images, pattern analysis is performed according to the density difference of the image, to identify the defect site,
A defect inspection method characterized by that. In the defect inspection method according to claim 5,
Identifying the dimension of the defect site based on a series of infrared images in the high-resolution search region;
A defect inspection method further comprising: A visible image capturing unit that captures an image while moving a target region to be imaged, and generates a series of visible images based on visible light intensity;
An infrared image capturing unit that captures a target region substantially the same as the target region captured by the visible image capturing unit and generates a series of infrared images based on infrared light intensity corresponding to the series of visible images. When,
Based on the series of infrared images, pattern analysis is performed by the density difference of the image, infrared image defect site identification means for identifying the defect site,
In the series of infrared images, an infrared image search area setting means for setting a search area including the defect site;
Using the search area set in the series of infrared images, high-resolution processing of the infrared image in the search area is performed based on positional deviation information of the search area of the visible image corresponding to the series of infrared images. Infrared image high-resolution processing means;
Infrared image defect size specifying means for specifying the size of the defect site based on a series of infrared images in the high-resolution search area,
Defect inspection device with A defect inspection program for causing a computer to function as a defect inspection apparatus,
A series of visible images based on visible light intensity captured while moving the target area to be imaged, and the series of visible images in which a target area substantially the same as the target area captured by the visible image capturing unit is captured. Means for accessing a recording unit that records a series of infrared images based on infrared light intensity corresponding to the images;
Based on the series of infrared images, pattern analysis is performed by the density difference of the image, infrared image defect site identification means for identifying the defect site,
In the series of infrared images, an infrared image search area setting means for setting a search area including the defect site;
Using the search area set in the series of infrared images, the resolution enhancement processing of the infrared image in the search area is performed based on the positional deviation information of the search area of the visible image corresponding to the series of infrared images. Performing infrared image high-resolution processing means;
Infrared image defect size specifying means for specifying the size of the defect site based on a series of infrared images in the high-resolution search area;
Defect inspection program for realizing a computer. Capture a target area to be imaged, generate a series of visible images based on visible light intensity,
Imaging a target region substantially the same as the target region captured by the visible image capturing unit, and generating a series of infrared images based on infrared light intensity corresponding to the series of visible images;
Based on the series of infrared images, pattern analysis is performed according to the density difference of the image, the defect site is identified,
In the series of infrared images, set a search area including the defect site,
Using the search area set in the series of infrared images, the resolution enhancement processing of the infrared image in the search area is performed based on the positional deviation information of the search area of the visible image corresponding to the series of infrared images. Done
Identifying the dimension of the defect site based on a series of infrared images in the high-resolution search region;
A defect inspection method characterized by that.