Bar arrangement inspection system

Abstract

To provide a bar arrangement inspection system that improves the accuracy of measurement of dimensions or the like associated with bar arrangement on an object surface by extracting only the bar arrangement on an object surface with a marker arranged thereon from overcrowded bar arrangement.SOLUTION: An edge image generation unit 48 generates an edge detection image T for each front position conversion image. An overcrowded bar arrangement separation unit 50 generates an overcrowded bar arrangement separation image V' by overlaying the edge detection image T on a bar arrangement area extraction image V. A displacement amount estimation mask image generation unit 44 generates a displacement amount estimation mask image U, and a parallax mask image generation unit 46 generates a parallax mask image W. A parallax mask bar arrangement separation unit 51 overlays the edge detection image T on the parallax mask image W, and separates a bar arrangement area 86 included in the parallax mask image W. An object bar arrangement area extraction unit 52 overlays the displacement amount estimation mask image U on the parallax mask bar arrangement separation image W', and extracts a bar arrangement area 91 of an object bar arrangement 2.SELECTED DRAWING: Figure 17

Description

[Technical Field]

[0001] The present invention relates to a reinforcement inspection system for inspecting the state of reinforcement construction. [Background technology]

[0002] Conventionally, a bar arrangement inspection system of this type is disclosed in Patent Document 1, for example.

[0003] In this reinforcement inspection system, a marker placed on the target reinforcement layer is photographed from different angles using a monocular camera. The angled images are then transformed into images facing the marker installation surface, with the marker at the center. Reinforcement areas are extracted from the transformed images, and multiple reinforcement area images extracted from the transformed images are superimposed using the marker as a reference to generate a superimposed image.

[0004] Since parallax occurs in the superimposed image except on the surface where the marker is installed, the areas where parallax occurs are deleted and an image (mask image) is generated in which only the reinforcement layer where the marker is installed remains.For the generated mask image, the center line of each rebar in the reinforcement area is determined, and the number of rebars is calculated from the number of center lines, the rebar pitch from the spacing between the center lines, and the rebar diameter from the width of the reinforcement area perpendicular to the center line. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent No. 6801055 Summary of the Invention [Problem to be solved by the invention]

[0006] However, in the conventional reinforcement inspection system disclosed in Patent Document 1, when reinforcement is densely arranged, even if an attempt is made to extract the target reinforcement by extracting the reinforcement area from an image taken with a monocular camera and deleting the parts where parallax occurs from the superimposed image centered on the marker, it can be difficult to extract only the reinforcement on the target surface.

[0007] In other words, in images of densely arranged reinforcement, there are more cases where the reinforcement on the front side where the markers are placed and the reinforcement on the back side are reflected overlapping in the captured image. As a result, the boundary between the overlapping front and back surfaces becomes unclear. For this reason, even in areas that are recognized as the front layer due to parallax, there is a high possibility that the reinforcement areas on surfaces other than the target surface will remain in all images captured from different angles, making it difficult to extract only the reinforcement on the target surface.

[0008] The present invention aims to improve the measurement accuracy of various dimensions of the reinforcement on the target surface by extracting only the reinforcement on the target surface where the marker is placed from the overcrowded reinforcement. [Means for solving the problem]

[0009] For this purpose, the present invention provides: a front-facing transformed image generating unit that generates a plurality of front-facing transformed images by converting a plurality of original images of a marker arranged on a predetermined layer of reinforcement composed of vertical and horizontal reinforcement bars, which are photographed from different angles by a monocular camera, into front-facing images of the marker; a reinforcement area extraction image generation unit that generates a reinforcement area extraction image for each of the front-facing transformed images by extracting a reinforcement area from the front-facing transformed image; an edge image generation unit that generates an edge-detected image in which the edges of the vertical and horizontal stripes are detected for each facing transformed image; a parallax mask image generating unit that generates a parallax mask image by overlapping the reinforcement region extraction images so that the markers match, and leaving the overlapping image portions of the reinforcement region extraction images; a displacement amount estimation mask image generating unit that estimates a layer with a small amount of reinforcement displacement during front-to-back transformation among a plurality of front-to-back transformed images as a predetermined layer, and generates a displacement amount estimation mask image in which the reinforcement area of ​​the predetermined layer is extracted; an overcrowded reinforcement separation unit that generates an overcrowded reinforcement separation image by superimposing the paired edge detection image and reinforcement area extraction image for each pair of edge detection image and reinforcement area extraction image generated from the same frontal transformation image, in which the boundary in the reinforcement area extraction image between the reinforcement area of ​​the target reinforcement in a predetermined layer and the reinforcement area of ​​the non-target reinforcement in a non-target layer different from the predetermined layer is separated by an edge in the edge detection image; a parallax mask reinforcement separation unit that generates a parallax mask reinforcement separated image in which the boundaries in the parallax mask image are separated by edges by superimposing the edge detection image on the parallax mask image; a target reinforcement area extraction unit that overlays the displacement estimation mask image on the parallax mask reinforcement separation image, generates a parallax mask reinforcement extraction image by removing from the parallax mask reinforcement separation image reinforcement areas that do not overlap with reinforcement areas of a predetermined layer in the displacement estimation mask image, and extracts reinforcement areas of the parallax mask reinforcement separation image that overlap with reinforcement areas of a predetermined layer in the displacement estimation mask image as reinforcement areas of the target reinforcement; a reinforcing bar line calculation unit that calculates the center line of the reinforcing bar area in the target reinforcing bar extracted by the target reinforcing bar area extraction unit as a reinforcing bar line; a rebar spacing calculation unit that calculates the rebar spacing of the target reinforcing bar arrangement from the distance between the reinforcing bar lines; a reinforcing bar diameter calculation unit that superimposes the center line on an overcrowded reinforcement separation image, calculates the reinforcing bar diameter of the target reinforcement for each overcrowded reinforcement separation image from the reinforcing bar area of ​​the overcrowded reinforcement separation image where the center line is superimposed, and sets the reinforcing bar diameter with the highest calculation frequency among the calculated reinforcing bar diameters as the reinforcing bar diameter of the target reinforcement; The reinforcement inspection system was constructed with these components.

[0010] According to this configuration, for each normal facing transformed image, the reinforcement area extraction image generation unit generates a reinforcement area extraction image that extracts the reinforcement area consisting of vertical and horizontal reinforcement, and the edge image generation unit generates an edge-detected image from each normal facing transformed image that detects the edges of the vertical and horizontal reinforcement. The dense reinforcement separation unit overlays the edge-detected image on the reinforcement area extraction image that is paired with the edge-detected image to generate an over-dense reinforcement separation image for each pair of edge-detected image and reinforcement area extraction image. In this over-dense reinforcement separation image, the boundary between the reinforcement area of ​​the target reinforcement and the reinforcement area of ​​the non-target reinforcement in the reinforcement area extraction image is separated and clarified by the edges in the edge-detected image.

[0011] Furthermore, the disparity mask image generation unit superimposes each reinforcement area extraction image, leaving the overlapping image portions, and thereby removing reinforcement areas estimated to be in non-target layers from each reinforcement area extraction image due to the parallax caused by differences in shooting angles, thereby generating a disparity mask image representing an estimated target reinforcement area estimated to be the reinforcement area of ​​the target reinforcement. Furthermore, the displacement estimation mask image generation unit estimates a layer with a small amount of reinforcement displacement during frontal transformation to be a specified layer where markers are placed, and generates a displacement estimation mask image extracting the reinforcement area of ​​that layer.

[0012] The parallax mask reinforcement separation unit overlays the edge detection image on the parallax mask image generated by the parallax mask image generation unit to generate a parallax mask reinforcement separation image in which the boundary between the reinforcement area of ​​the target reinforcement and the reinforcement area of ​​the non-target reinforcement in the parallax mask image is separated by the edge in the edge detection image. Furthermore, the target reinforcement area extraction unit overlays the displacement amount estimation mask image generated by the displacement amount estimation mask image generation unit on this parallax mask reinforcement separation image, thereby extracting the reinforcement area of ​​the target reinforcement where the marker is placed, which is included in the parallax mask reinforcement separation image, from the parallax mask image.

[0013] The center lines of the extracted reinforcing bar regions in the target reinforcement are calculated as reinforcing bar lines by the reinforcing bar line calculation unit. The reinforcing bar spacing calculation unit calculates the reinforcing bar spacing of the target reinforcement from the distance between the reinforcing bar lines. The center lines calculated by the reinforcing bar line calculation unit are superimposed on the overcrowded reinforcement separation image. The reinforcing bar diameter calculation unit calculates the reinforcing bar diameter of the target reinforcement for each overcrowded reinforcement separation image from the reinforcing bar regions of the overcrowded reinforcement separation image where the center lines are superimposed. The reinforcing bar diameter with the highest calculation frequency is determined to be the reinforcing bar diameter of the target reinforcement.

[0014] The present invention also provides a rib edge removal unit that superimposes a displacement amount estimation mask image on the edge detection image and removes edges in overlapping areas of each reinforcement area in the edge detection image and the displacement amount estimation mask image as rib edges corresponding to the ribs of the reinforcing bars from each edge detection image; The overcrowded reinforcement separation unit is characterized in that it separates the boundaries in each pair of reinforcement area extraction images by the edges in each edge detection image from which the rib edges have been removed, by superimposing each edge detection image from which the rib edges have been removed on each pair of reinforcement area extraction images.

[0015] According to this configuration, edges in overlapping areas of the reinforcement areas in the edge-detection image and the displacement estimation mask image are removed from each edge-detection image as rib edges by the rib edge removal unit. That is, edges of the edge-detection image that overlap the reinforcement areas in the displacement estimation mask image are regarded as edges corresponding to the ribs of the reinforcing bars and are removed by the rib edge removal unit. Therefore, the overcrowded reinforcement separation unit generates an overcrowded reinforcement separation image in which the boundaries between the reinforcement areas of the target reinforcement and the reinforcement areas of the non-target reinforcement are accurately and clearly separated by the edges in the edge-detection image from which the rib edges have been removed. Therefore, the reinforcing bar areas of the target reinforcement are accurately detected by superimposing the center lines calculated by the reinforcing bar line calculation unit on the overcrowded reinforcement separation image, and the reinforcing bar diameters are accurately calculated by the reinforcing bar diameter calculation unit.

[0016] The present invention also provides the parallax mask reinforcement separation unit overlays the remaining edge detection images, for the number of remaining edge detection images not used in generating the parallax mask reinforcement separation image, on the parallax mask reinforcement extraction image, thereby separating the boundaries remaining in the parallax mask reinforcement extraction image for each of the remaining edge detection images using the edges in each of the remaining edge detection images, and updating the parallax mask reinforcement separation image; The target reinforcement area extraction unit overlays the displacement estimation mask image on the updated parallax mask reinforcement separation image every time the parallax mask reinforcement separation image is updated, removes reinforcement areas that do not overlap with the reinforcement areas of a predetermined layer in the displacement estimation mask image from the updated parallax mask reinforcement separation image, updates the parallax mask reinforcement extraction image, and finally extracts the reinforcement area in the updated parallax mask reinforcement extraction image as the reinforcement area of ​​the target reinforcement. It is characterized by:

[0017] According to this configuration, the parallax mask reinforcement separation unit initially generates a parallax mask reinforcement separation image by overlaying an edge detection image on a parallax mask image. However, once the target reinforcement area extraction unit generates a parallax mask reinforcement extraction image, the remaining edge detection images are overlaid on the parallax mask reinforcement extraction image instead of the parallax mask image, and the parallax mask reinforcement separation image is updated for the number of remaining edge detection images not used to generate the parallax mask reinforcement separation image. Each time the parallax mask reinforcement separation image is updated, the target reinforcement area extraction unit overlays a displacement estimation mask image on the updated parallax mask reinforcement separation image and removes reinforcement areas in the updated parallax mask reinforcement separation image that do not overlap with reinforcement areas of a specified layer in the displacement estimation mask image, thereby updating the parallax mask reinforcement extraction image. Finally, the target reinforcement area is extracted based on the updated parallax mask reinforcement extraction image.

[0018] Therefore, even if there are edge detection images in which the edge corresponding to the boundary between the reinforcement area of ​​the target reinforcement and the reinforcement area of ​​the non-target reinforcement is not extracted among the multiple edge detection images generated from each frontal transformation image, if the edge is extracted in the remaining edge detection images, the reinforcement area of ​​the target reinforcement will be reliably extracted by the processes of the parallax mask reinforcement separation unit and the target reinforcement area extraction unit using the remaining edge detection images.In addition, even if the separation and extraction process of the reinforcement area of ​​the target reinforcement is performed by the parallax mask reinforcement separation unit and the target reinforcement area extraction unit the number of edge detection images, and the process of distinguishing between the target reinforcement and the non-target reinforcement is repeated, the reinforcement area of ​​the target reinforcement will be reliably extracted.

[0019] Furthermore, the present invention is characterized in that, when the ratio of the area of ​​the overlapping portion between the continuous area of ​​the reinforcement area included in the parallax mask reinforcement separation image and the reinforcement area of ​​the target reinforcement included in the displacement estimation mask image to the area of ​​the continuous area is equal to or less than a predetermined threshold, the target reinforcement area extraction unit does not extract the continuous area from the reinforcement area included in the parallax mask reinforcement separation image as an area of ​​non-target reinforcement.

[0020] With this configuration, a continuous area of ​​reinforcement areas included in a parallax mask reinforcement separation image can be properly determined as being a reinforcement area of ​​the target reinforcement or a reinforcement area of ​​the non-target reinforcement. If the continuous area is determined to be a reinforcement area of ​​the non-target reinforcement, the target reinforcement area extraction unit will not extract the continuous area as a reinforcement area of ​​the target reinforcement from the reinforcement area included in the parallax mask reinforcement separation image. Therefore, the reinforcement area of ​​the target reinforcement can be properly extracted from the reinforcement area included in the parallax mask reinforcement separation image.

[0021] The present invention also provides a parallax mask image separating unit that separates the parallax mask image into a vertical line parallax mask image for the vertical lines and a horizontal line parallax mask image for the horizontal lines; a displacement amount estimation mask image separating unit that separates the displacement amount estimation mask image into a vertical line displacement amount estimation mask image for vertical lines and a horizontal line displacement amount estimation mask image for horizontal lines, an edge image generating unit generates a vertical edge-detected image in which edges of vertical stripes are detected and a horizontal edge-detected image in which edges of horizontal stripes are detected as edge-detected images for each of the facing transformed images; The overcrowded reinforcement separation unit generates an overcrowded reinforcement separation image by superimposing the paired vertical edge detection image and horizontal edge detection image, generated from the same frontal transformation image, on the reinforcement area extraction image, to separate the vertical reinforcement boundaries in the reinforcement area extraction image between the vertical reinforcement areas of the target reinforcement and the vertical reinforcement areas of the non-target reinforcement along the edges in the vertical edge detection image, and the horizontal reinforcement boundaries in the reinforcement area extraction image between the horizontal reinforcement areas of the target reinforcement and the horizontal reinforcement areas of the non-target reinforcement along the edges in the horizontal edge detection image, a parallax mask reinforcement separation unit overlaying the vertical edge detection image on the vertical line parallax mask image to generate a parallax mask vertical line separated image in which vertical line boundaries in the vertical line parallax mask image are separated by edges in the vertical edge detection image, and overlaying the horizontal edge detection image on the horizontal line parallax mask image to generate a parallax mask horizontal line separated image in which horizontal line boundaries in the horizontal line parallax mask image are separated by edges in the horizontal edge detection image; the target reinforcement area extraction unit overlays the vertical line displacement estimation mask image on the parallax mask vertical line separation image to generate a parallax mask vertical line extraction image by removing from the parallax mask vertical line separation image vertical line areas that do not overlap with vertical line areas of a predetermined layer in the vertical line displacement estimation mask image, extracts vertical line areas of the parallax mask vertical line separation image that overlap with vertical line areas of the predetermined layer in the vertical line displacement estimation mask image as vertical line areas of the target reinforcement, overlays the horizontal line displacement estimation mask image on the parallax mask horizontal line separation image to generate a parallax mask horizontal line extraction image by removing from the parallax mask horizontal line separation image horizontal line areas that do not overlap with horizontal line areas of the predetermined layer in the horizontal line displacement estimation mask image, and extracts horizontal line areas of the parallax mask horizontal line separation image that overlap with horizontal line areas of the predetermined layer in the horizontal line displacement estimation mask image as horizontal line areas of the target reinforcement, a reinforcing bar line calculation unit that calculates the center line of the vertical reinforcement area in the target reinforcement extracted by the target reinforcement area extraction unit as a vertical reinforcement line, and calculates the center line of the horizontal reinforcement area in the target reinforcement extracted by the target reinforcement area extraction unit as a horizontal reinforcement line; a rebar spacing calculation unit that calculates the vertical rebar spacing from the distance between the vertical rebar lines, and calculates the horizontal rebar spacing from the distance between the horizontal rebar lines; The reinforcing bar diameter calculation unit superimposes the center line of the vertical reinforcement area on the overdense reinforcement separation image, calculates the reinforcing bar diameter of the vertical reinforcement in the target reinforcement arrangement for each overdense reinforcement separation image from the vertical reinforcement area of ​​the overdense reinforcement separation image where the center line of the vertical reinforcement area is superimposed, superimposes the center line of the horizontal reinforcement area on the overdense reinforcement separation image, calculates the reinforcing bar diameter of the horizontal reinforcement in the target reinforcement arrangement for each overdense reinforcement separation image from the horizontal reinforcement area of ​​the overdense reinforcement separation image where the center line of the horizontal reinforcement area is superimposed, and sets the reinforcing bar diameter with the highest calculation frequency among the calculated diameters of the vertical reinforcement and the horizontal reinforcement as the diameters of the vertical reinforcement and the horizontal reinforcement in the target reinforcement arrangement. It is characterized by:

[0022] According to this configuration, the disparity mask image generated by the disparity mask image generation unit is separated by the disparity mask image separation unit into a vertical line disparity mask image representing an estimation target vertical line region that is estimated to be a vertical line region of the target reinforcement within the estimation target reinforcement region, and a horizontal line disparity mask image representing an estimation target horizontal line region that is estimated to be a horizontal line region of the target reinforcement within the estimation target reinforcement region. Furthermore, the displacement estimation mask image generated by the displacement estimation mask image generation unit is separated by the displacement estimation mask image separation unit into a vertical line displacement estimation mask image for the vertical lines and a horizontal line displacement estimation mask image for the horizontal lines. Furthermore, the edge detection image generated for each facing transformed image is separated by the edge image generation unit into a vertical edge detection image in which the edges of the vertical lines are detected and a horizontal edge detection image in which the edges of the horizontal lines are detected.

[0023] Accordingly, each process performed on each image is divided into image processing for the vertical reinforcement and image processing for the horizontal reinforcement. By performing each process separately for the vertical reinforcement and the horizontal reinforcement in this way, each image, which previously had a rough appearance and contained a lot of noise due to the vertical and horizontal reinforcement intersecting, becomes a simple image that is easy to see and less cluttered. This reduces the number of missed detections of straight lines along the outline of the reinforcing bar area, improving the detection rate of those straight lines and further improving the measurement accuracy of the dimensions of the reinforcement area.

[0024] The present invention also provides a rib edge removal unit that superimposes a vertical line displacement amount estimation mask image on the vertical edge detection image, and removes from each vertical edge detection image vertical edges in overlapping areas of each vertical line region in the vertical edge detection image and the vertical line displacement amount estimation mask image as rib edges corresponding to the vertical line ribs, and superimposes a horizontal line displacement amount estimation mask image on the horizontal edge detection image, and removes from each horizontal edge detection image horizontal edges in overlapping areas of each horizontal line region in the horizontal edge detection image and the horizontal line displacement amount estimation mask image as rib edges corresponding to the horizontal line ribs, The overcrowded reinforcement separation unit separates vertical reinforcement boundaries in each paired reinforcement area extraction image by the vertical edges in each vertical edge detection image from which the rib edges have been removed, by superimposing each vertical edge detection image from which the rib edges have been removed on each paired reinforcement area extraction image, and separates horizontal reinforcement boundaries in each paired reinforcement area extraction image by the horizontal edges in each horizontal edge detection image from which the rib edges have been removed, by superimposing each horizontal edge detection image from which the rib edges have been removed on each paired reinforcement area extraction image. It is characterized by:

[0025] According to this configuration, the rib edges detected in each vertical edge detection image and each horizontal edge detection image, which correspond to vertical and horizontal ribs, overlap the vertical and horizontal reinforcement displacement estimation mask images, and are then separated into simple vertical and horizontal edges by the rib edge removal unit. Therefore, each vertical and horizontal rib edge is detected and removed without omission, compared to when they are detected and removed all at once. Therefore, the vertical and horizontal reinforcement boundaries between the target reinforcement area and the non-target reinforcement area are separated without omission by the overcrowded reinforcement separation unit using the edges in each vertical edge detection image from which the rib edges have been removed and the edges in each horizontal edge detection image from which the rib edges have been removed. Therefore, the center lines of the vertical and horizontal reinforcement areas of the target reinforcement are superimposed, and the vertical and horizontal reinforcement areas of the target reinforcement obtained from the overcrowded reinforcement separation image are accurate. As a result, the rebar diameters of the vertical and horizontal reinforcement areas of the target reinforcement obtained based on these vertical and horizontal reinforcement areas can be accurately calculated by the rebar diameter calculation unit.

[0026] The present invention also provides the parallax mask reinforcement separation unit overlays the remaining vertical edge detection images, for the number of remaining vertical edge detection images not used in generating the parallax mask vertical line separation image, on the parallax mask vertical line extraction image, to separate vertical line boundaries remaining in the parallax mask vertical line extraction image for each remaining vertical edge detection image using edges in each remaining vertical edge detection image, thereby updating the parallax mask vertical line separation image; and overlays the remaining horizontal edge detection images, for the number of remaining horizontal edge detection images not used in generating the parallax mask horizontal line separation image, on the parallax mask horizontal line extraction image, to separate horizontal line boundaries remaining in the parallax mask horizontal line extraction image for each remaining horizontal edge detection image using edges in each remaining horizontal edge detection image, thereby updating the parallax mask horizontal line separation image; The target reinforcement arrangement area extraction unit superimposes the vertical line displacement amount estimation mask image on the updated parallax mask vertical line separation image every time the parallax mask vertical line separation image is updated, removes vertical line areas in the vertical line displacement amount estimation mask image that do not overlap with vertical line areas of a predetermined layer in the vertical line displacement amount estimation mask image from the updated parallax mask vertical line separation image to update the parallax mask vertical line extraction image, extracts the vertical line areas in the last updated parallax mask vertical line extraction image as the vertical line areas of the target reinforcement arrangement, superimposes the horizontal line displacement amount estimation mask image on the updated parallax mask horizontal line separation image every time the parallax mask horizontal line separation image is updated, removes horizontal line areas in the horizontal line displacement amount estimation mask image that do not overlap with horizontal line areas of a predetermined layer in the horizontal line displacement amount estimation mask image from the updated parallax mask horizontal line separation image to update the parallax mask horizontal line extraction image, and extracts the horizontal line areas in the last updated parallax mask horizontal line extraction image as the horizontal line areas of the target reinforcement arrangement. It is characterized by:

[0027] According to this configuration, the process performed by the parallax mask reinforcement separation unit to update the parallax mask reinforcement separation image by overlaying an edge detection image on the parallax mask reinforcement extraction image is performed by converting the edge detection image and the parallax mask reinforcement extraction image into images separated into vertical and horizontal edges. Furthermore, the process performed by the target reinforcement area extraction unit to update the parallax mask reinforcement extraction image by overlaying a displacement estimation mask image on the updated parallax mask reinforcement separation image each time the parallax mask reinforcement separation image is updated is also performed by converting the parallax mask reinforcement separation image and the displacement estimation mask image into images separated into vertical and horizontal edges. Therefore, the processes performed by the parallax mask reinforcement separation unit and the target reinforcement area extraction unit are performed in a simple state where the edges are separated into vertical and horizontal directions, so that each edge is accurately captured and performed reliably.

[0028] Further, the present invention is directed to a method for extracting a target reinforcement area, comprising: If the ratio of the area of ​​the overlapping portion between the continuous vertical line region included in the parallax mask vertical line separation image and the vertical line region of the target reinforcement included in the vertical line displacement estimation mask image to the area of ​​the continuous vertical line region is equal to or less than a predetermined threshold, the continuous vertical line region is not extracted from the reinforcement region included in the parallax mask vertical line separation image as a non-target reinforcement region, If the ratio of the area of ​​the overlapping portion between the continuous region of horizontal reinforcement included in the parallax mask horizontal reinforcement separation image and the horizontal reinforcement region of the target reinforcement included in the horizontal reinforcement displacement estimation mask image to the area of ​​the continuous region of horizontal reinforcement is equal to or less than a predetermined threshold, the continuous region of horizontal reinforcement is not extracted from the reinforcement region included in the parallax mask horizontal reinforcement separation image as a region of non-target reinforcement. It is characterized by:

[0029] According to this configuration, the process of determining whether a continuous area of ​​reinforcement areas included in a parallax mask reinforcement separation image is a reinforcement area of ​​the target reinforcement arrangement or a reinforcement area of ​​the non-target reinforcement arrangement is performed individually when the reinforcement area is divided into vertical reinforcement areas and horizontal reinforcement areas. When the reinforcement area is divided into vertical reinforcement areas and horizontal reinforcement areas, the vertical reinforcement areas and horizontal reinforcement areas are clearly distinguished and can be clearly understood. This makes it possible to more accurately extract the reinforcement area of ​​the target reinforcement arrangement from the reinforcement areas included in a parallax mask reinforcement separation image.

[0030] The present invention is also characterized in that the edge image generation unit calculates the peak of the brightness gradient of each pixel in the direction perpendicular to the reinforcing bar lines of the vertical or horizontal reinforcing bars in the reinforcing bar image, and among the calculated peaks, peaks with prominence values ​​equal to or greater than a predetermined threshold are designated as edge candidates, and further, among these edge candidates, edge candidates with lower prominence values ​​are removed from among the edge candidates whose distance between edge candidates is shorter than the predetermined threshold, thereby detecting the edges of the vertical or horizontal reinforcing bars that represent the outline of the reinforcing bar area.

[0031] The edges detected by the edge image generation unit are calculated as the peaks of the brightness gradient by calculating the brightness gradient of each pixel in the direction perpendicular to the vertical or horizontal reinforcing bar lines for the reinforcing bar image. When edge candidates representing the outline of the reinforcing bar are extracted from the edges calculated as these peaks based on the peak prominence value, extra edges in areas where the brightness gradient is high due to shadows are extracted as noise. This extra noise may cause the distance between edges to be detected as shorter, resulting in the reinforcing bar diameter being calculated as thinner.

[0032] With this configuration, when the distance between edge candidates extracted based on the prominence values ​​is shorter than a predetermined threshold, the edge candidate with the lower prominence value is removed, thereby reducing the number of edge candidates detected as noise. By repeating this edge candidate removal process, edge candidates detected as noise are eliminated, and vertical or horizontal reinforcing bar edges that represent the outline of the reinforcing bar region of the reinforcement are obtained.

[0033] The present invention is also characterized in that the edge image generating section performs line detection on all edge candidates of vertical or horizontal stripes using a probabilistic Hough transform.

[0034] With this configuration, even if the line segments obtained by connecting edges vertically or horizontally are too short to detect straight lines that follow the contours of the reinforcing bar regions, the edge image generation unit connects the short line segments using a probabilistic Hough transform, thereby detecting straight lines that follow the contours of the reinforcing bar regions. This prevents straight lines that follow the contours of the reinforcing bar regions from going undetected. [Effects of the Invention]

[0035] According to the present invention, it is possible to provide a reinforcement inspection system that can extract only the reinforcement on the target surface where the marker is placed from the overcrowded reinforcement, thereby improving the measurement accuracy of each dimension of the reinforcement on the target surface. [Brief explanation of the drawings]

[0036] [Figure 1] 1 is an overall configuration diagram illustrating the configuration of a bar arrangement inspection system according to an embodiment of the present invention; [Figure 2] 1A is a plan view showing a marker used in a bar arrangement inspection system according to an embodiment in an unfolded state, and FIG. 1B is a plan view showing a marker in a folded state. [Figure 3] 1 is a diagram illustrating photography by a monocular camera performed in a bar arrangement inspection using a bar arrangement inspection system according to an embodiment. FIG. [Figure 4]1 is a block diagram showing a functional configuration of a bar arrangement inspection system according to an embodiment; [Figure 5] 10 is a flowchart showing the flow of a bar arrangement inspection process performed by the bar arrangement inspection system according to one embodiment. [Figure 6] FIG. 1(a) is an original image of an overcrowded reinforcement arrangement photographed during a reinforcement inspection using a reinforcement inspection system according to one embodiment, and FIG. 1(b) is a reinforcement area superimposed image in which multiple reinforcement area extraction images obtained from the original image in FIG. 1(a) are superimposed. [Figure 7] FIG. 1(a) shows a displacement estimation image of reinforcement when generating a displacement estimation mask image used in reinforcement inspection using a reinforcement inspection system according to one embodiment, and FIG. 1(b) shows an output image obtained as a result of displacement estimation. [Figure 8] FIG. 10 is a diagram showing an example of a vertical reinforcement displacement amount estimation mask image used in a reinforcement bar arrangement inspection using the reinforcement bar arrangement inspection system according to one embodiment. [Figure 9] FIG. 1(a) is a diagram showing an original image of overcrowded reinforcement taken during reinforcement inspection using a reinforcement inspection system according to one embodiment, FIG. 1(b) is a horizontal edge detection image obtained from the original image of FIG. 1(a), and FIG. 1(c) is a vertical edge detection image obtained from the original image of FIG. 1(a). [Figure 10] (a) is an image illustrating the shadows cast on reinforcing bars during a reinforcement inspection using a reinforcement inspection system according to one embodiment; (b) is a graph showing the distribution curve of pixel brightness in the image of (a); and (c) is a graph showing the results of differentiating the brightness shown in (b). [Figure 11] FIG. 10A is an image illustrating that in a reinforcement inspection using a reinforcement inspection system according to one embodiment, edges detected by an edge image generation unit become short line segments, resulting in undetected straight lines; and FIG. 10B is an image illustrating that short line segments are connected by a probabilistic Hough transform to detect straight lines. [Figure 12] FIG. 10 is a diagram showing an example of an overcrowded reinforcement separation image in which vertical and horizontal edges in a reinforcement area extraction image are superimposed on the reinforcement area extraction image in a reinforcement inspection using a reinforcement inspection system according to one embodiment. [Figure 13](a) is an original image of overcrowded reinforcement taken during reinforcement inspection using a reinforcement inspection system according to one embodiment; (b) is an edge-detected image in which the rib edges captured in the original image of (a) appear as straight lines within the reinforcement area; and (c) is an image in which the straight lines that appear in the image of (b) are superimposed on the original image of (a). [Figure 14] 13(a) is an edge-detected image showing an enlarged version of the image shown in FIG. 13(b), and FIG. 13(b) is an image obtained by superimposing a vertical streak displacement amount estimation mask image on the image shown in FIG. 13(a). [Figure 15] FIG. 1(a) is an original image of an overcrowded reinforcement bar taken during a reinforcement bar inspection using a reinforcement bar inspection system according to one embodiment, and FIG. 1(b) is an original vertical bar disparity mask image showing the vertical bar area to be estimated, separated from the original image of FIG. 1(a) by a disparity mask image separation unit. [Figure 16] FIG. 10 is a diagram showing an example of a vertical line disparity mask image used in a reinforcement bar arrangement inspection using the reinforcement bar arrangement inspection system according to one embodiment. [Figure 17] FIG. 10 is a diagram illustrating an example of each process performed by a rib edge removal unit, an overcrowded reinforcement separation unit, a parallax mask reinforcement separation unit, and a target reinforcement area extraction unit during reinforcement inspection using a reinforcement inspection system according to one embodiment. [Figure 18] 1A is a diagram showing a parallax mask vertical line extraction image generated by superimposing a vertical line displacement estimation mask image on a parallax mask vertical line separation image in a reinforcement inspection using a reinforcement inspection system according to one embodiment, and FIG. 1B is a diagram illustrating the process of extracting the reinforcement area of ​​the target reinforcement from the parallax mask vertical line extraction image in FIG. [Figure 19] FIG. 1(a) is a diagram illustrating the process of calculating a center line from the reinforcing bar area of ​​the target reinforcement extracted during reinforcement inspection using a reinforcement inspection system according to one embodiment, and FIG. 1(b) is a diagram illustrating the process of calculating the reinforcing bar spacing from the center line calculated in FIG. 1(a). [Figure 20] (a) is a diagram explaining the process of calculating the rebar diameter from the rebar area of ​​the target rebar extracted during a rebar inspection using a rebar inspection system according to one embodiment, and (b) is a graph showing the distribution of the width and frequency of the rebar area used in the rebar diameter calculation process of (a). [Figure 21]FIG. 10 is a diagram showing an analysis result image showing the measurement results of reinforcement obtained in a reinforcement inspection using the reinforcement inspection system according to one embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0037] Next, an embodiment of a reinforcement bar arrangement inspection system according to the present invention will be described.

[0038] [Configuration of the reinforcement inspection system] Fig. 1 is an overall configuration diagram illustrating the configuration of a bar arrangement inspection system 1 according to one embodiment. Fig. 2(a) is a plan view illustrating the schematic configuration of a marker 10 used in the bar arrangement inspection system 1 according to this embodiment in an unfolded state, and Fig. 2(b) is a plan view illustrating the schematic configuration of the marker 10 in a folded state.

[0039] In this embodiment, a reinforcement inspection system 1 will be described that determines whether reinforcement 2 in an overcrowded reinforcement arrangement is correct, that is, whether the reinforcement 2 has been correctly constructed as designed. As shown in Fig. 1, reinforcement 2, in which a plurality of reinforcing bars 5 are arranged, is formed by stacking multiple layers in the front-to-back direction. The reinforcing bars 5 are composed of vertical bars 5a extending vertically and horizontal bars 5b extending horizontally, and are deformed reinforcing bars such as horizontal reinforcing bars, diagonal reinforcing bars, and threaded reinforcing bars.

[0040] The reinforcement inspection system 1 is composed of a marker 10 placed on a specified layer of reinforcement 2, in this embodiment the frontmost surface layer, a tablet terminal 20 equipped with a monocular camera 30 and an image display device 21, and a personal computer as a control unit 40.

[0041] As shown in FIG. 2(a), the marker 10 is composed of a foldable cross-shaped marker body 12 and four marks 11 provided on the marker body 12. The marker body 12 is formed into a cross shape by a strip-shaped first body 12a and a second body 12b. Marks 11 are provided on both ends of the first body 12a and the second body 12b. The distance L between the marks 11 on the first body and the distance L between the marks 11 on the second body can be set to, for example, 500 mm.

[0042] The first body 12a and the second body 12b are configured to be movable between an unfolded state and a folded state by a shaft 13 provided in the center. In the unfolded state, the first body 12a and the second body 12b are locked together in a cross shape, as shown in FIG. 2(a). In the unfolded state, the four marks 11 form a square. In the folded state, the first body 12a and the second body 12b are folded so that they overlap, as shown in FIG. 2(b).

[0043] This allows the marker 10 to be unfolded when in use and folded when not in use. Therefore, the marker 10 can be inserted in a folded state into a layer deeper than the surface layer of the reinforcement 2, and then unfolded and placed. Furthermore, when not in use, the marker 10 can be folded compactly. This makes the marker 10 easy to carry.

[0044] The mark 11 can be, for example, a circle with a diameter of 5 cm. The mark 11 can be a color that does not exist at a construction site, such as purple or green. When installing the marker 10 on the reinforcement 2, the first body 12a and the second body 12b should not overlap the vertical reinforcement 5a and the horizontal reinforcement 5b that make up the reinforcement 2.

[0045] As shown in Fig. 1, tablet terminal 20 includes monocular camera 30 located on the rear surface and image display device 21 located on the front surface. A typical camera installed in a smartphone or tablet terminal can be used as monocular camera 30. Image display device 21 displays a color image.

[0046] The personal computer serving as the control unit 40 is a dedicated personal computer equipped with a GPU (Graphics Processing Unit). The tablet terminal 20 and the personal computer serving as the control unit 40 are connected via a wireless LAN, enabling the exchange of information between them. The tablet terminal 20 and the control unit 40 may also be connected via a wired connection.

[0047] [Monocular camera shooting method] 3 is a diagram illustrating a method for capturing images using monocular camera 30 in this embodiment. The method for capturing images using monocular camera 30 in this embodiment will be described below with reference to FIG.

[0048] As shown in Figure 3, the worker places the marker 10 on the surface layer of the reinforcement 2, positions the marker 10 approximately in the center, and photographs the reinforcement 2 from multiple different directions using the monocular camera 30. For example, the worker photographs the reinforcement 2 including the marker 10 from two positions G1 and G2 on the left side of the marker 10, a position G3 in front of the marker 10, and two positions G4 and G5 on the right side.

[0049] The worker takes the image so that the horizontal and vertical ranges of the image are, for example, a distance 2L+α, which is greater than the distance 2L. The worker also takes the image so that the distance from the reinforcement 2 to the monocular camera 30 is approximately 0.5 m to 2 m. When photographing the reinforcement 2 including the marker 10, the worker takes the image from a position where the four marks 11 can be identified.

[0050] [Functional configuration of the reinforcement inspection system] 4 is a block diagram showing the functional configuration of the bar arrangement inspection system 1 according to this embodiment. As shown in FIG. 4, the bar arrangement inspection system 1 mainly includes a monocular camera 30, a control unit 40, and an image display device 21.

[0051] (monocular camera) The monocular camera 30 photographs the markers 10 placed on the surface layer of the reinforcement 2 from a plurality of different angles, and generates a plurality of images. The images photographed by the monocular camera 30 are transmitted to the control unit 40.

[0052] (Configuration of the control unit and its processing flow) The control unit 40 mainly includes, as functional units, a memory unit 41, a facing transformation image generation unit 42, a reinforcement area extraction image generation unit 43, a displacement estimation mask image generation unit 44, a displacement estimation mask image separation unit 45, a parallax mask image generation unit 46, a parallax mask image separation unit 47, an edge image generation unit 48, a rib edge removal unit 49, an overcrowded reinforcement separation unit 50, a parallax mask reinforcement separation unit 51, a target reinforcement area extraction unit 52, a rebar line calculation unit 53, a rebar spacing calculation unit 54, a rebar diameter calculation unit 55, and an analysis result image generation unit 56. Each of these functional units is realized by the calculation processing of the GPU.

[0053] The configuration of the control unit 40 will be described below with reference to the flow chart shown in FIG. 5, along with the flow of the bar arrangement inspection process performed by the control unit 40.

[0054] The memory unit 41 stores design data 41a and a trained model 41b. The design data 41a is data on the reinforcement drawing of a building, such as the diameter of the reinforcing bars 5, the number of reinforcing bars 5, and the reinforcing bar pitch. The trained model 41b is used to predict reinforcing bars for unknown data, and is a model trained by prior learning using images of reinforcing bars in various environments. Reinforcing bars in various environments include reinforcing bars on sunny days, reinforcing bars on cloudy days, reinforcing bars on rainy days, and reinforcing bars both indoors and outdoors. The trained model 41b is updated using images captured by the monocular camera 30.

[0055] The facing-up transformed image generation unit 42 generates a plurality of facing-up transformed images by converting each of a plurality of original images captured from different angles by the monocular camera 30 into facing-up images in which the marker 10 is viewed from the front (see FIG. 5, step S101). The facing-up transformed images are generated by performing projective transformation on each of the images captured from different angles by the monocular camera 30 into an image in which the marker 10 is viewed from above so that the marks 11 on the marker 10 form the vertices of a square. In each facing-up transformed image, the marker 10 is square and has the same size. The marker 10 is positioned in the center in each facing-up transformed image. The facing-up transformed image generation unit 42 crops the captured image so that the horizontal and vertical lengths of each facing-up transformed image are, for example, 2L.

[0056] The reinforcement area extraction image generation unit 43 generates a reinforcement area extraction image for each normal facing transformed image by extracting the reinforcement area of ​​the reinforcement 2 from the normal facing transformed image (see FIG. 5, step S102). This reinforcement area extraction process is performed by deep learning based on the trained model 41b.

[0057] The parallax mask image generator 46 overlays multiple reinforcement area extraction images so that the markers 10 are aligned, generating a parallax mask image that retains the overlapping image portions of the reinforcement area extraction images (see FIG. 5, step S103). In this process of generating a parallax mask image, if an original image A of densely packed reinforcement is shown in FIG. 6(a), for example, the parallax mask image generator 46 generates a reinforcement area overlapped image B shown in FIG. 6(b), in which multiple reinforcement area extraction images are overlapped. Then, by performing a logical AND operation between these reinforcement area extraction images, the mask image is generated, eliminating the portions where the reinforcement areas do not overlap due to the parallax caused by differences in the shooting angles. This overlapping portion of the reinforcement areas includes the reinforcement area of ​​the target reinforcement 2 in the reinforcement layer where the marker 10 is placed, but a small amount of excess reinforcement area in a reinforcement layer other than the one where the marker 10 is placed remains.

[0058] In this embodiment, the parallax mask image separating unit 47 separates the parallax mask image generated by the parallax mask image generating unit 46 into a vertical line parallax mask image for the vertical lines 5a and a horizontal line parallax mask image for the horizontal lines 5b (see FIG. 5, step S104).

[0059] The displacement estimation mask image generation unit 44 estimates a layer where the displacement of the reinforcement 2 is small during the normal facing transformation between multiple normal facing transformed images as a predetermined layer on which the marker 10 is placed, and generates a displacement estimation mask image by extracting the reinforcement area of ​​that layer (see FIG. 5, step S105). The displacement of the reinforcement 2 is detected by optical flow, and is detected as the optical movement of each point between temporally consecutive image frames, which can be observed by projecting the movement of the reinforcement 2 during the normal facing transformation onto the image.

[0060] FIG. 7(a) shows a displacement estimation image C of the reinforcement 2, and the length and direction of the arrow segments in the image indicate the magnitude and direction of the displacement of the reinforcement 2. FIG. 7(b) shows an output image D obtained as a result of displacement estimation performed by the displacement estimation mask image generation unit 44. Dark pixels in the image indicate small displacements. The displacement estimation mask image generation unit 44 further performs deep learning on the output image D to extract the reinforcement area of ​​a specified front layer from the output image D and generate a displacement estimation mask image.

[0061] In this embodiment, the displacement amount estimation mask image separating unit 45 separates the displacement amount estimation mask image generated by the displacement amount estimation mask image generating unit 44 into a vertical line displacement amount estimation mask image for the vertical lines 5a and a horizontal line displacement amount estimation mask image for the horizontal lines 5b (see FIG. 5, step S106). FIG. 8 shows a vertical line displacement amount estimation mask image E obtained by separating the vertical line region 71 of a predetermined layer on the front side from the output image D.

[0062] The edge image generating unit 48 generates an edge-detected image for each facing transformed image, in which the edges (ends) of the vertical streaks 5a and the horizontal streaks 5b are detected (see FIG. 5, step S107). In this embodiment, the edge image generating unit 48 generates, as edge-detected images for each facing transformed image, a vertical edge-detected image for which the edges of the vertical streaks 5a are detected, and a horizontal edge-detected image for which the edges of the horizontal streaks 5b are detected.

[0063] For example, when an original image F of dense reinforcement is shown in FIG. 9(a), the edge image generating unit 48 generates a horizontal edge detection image G shown in FIG. 9(b) and a vertical edge detection image H shown in FIG. 9(c) for each of the facing transformed images.

[0064] The edges detected by the edge image generation unit 48 may be affected by the shadow cast on the reinforcing bar 5 and may include an edge corresponding to the shadow 70, as in image I of the reinforcing bar 5 shown in Fig. 10(a). Image I is shown as a graph, with the horizontal axis representing the number of pixels in the width direction of the reinforcing bar 5 and the vertical axis representing the number of pixels in the length direction of the reinforcing bar 5.

[0065] The graph shown in Figure 10(b) shows a brightness distribution curve 72 of pixels on a straight line 71 in image I. The horizontal axis of this graph represents the same number of pixels as the horizontal axis shown in Figure 10(a), and the vertical axis represents pixel brightness. The graph shown in Figure 10(c) is a graph showing the results of differentiating the brightness shown in Figure 10(b), and shows a brightness gradient distribution curve 73. The horizontal axis of this graph represents the same number of pixels as the horizontal axis of the graph shown in Figure 10(b), and the vertical axis represents the differential value of brightness. The six maximum points in the brightness gradient distribution curve 73 shown in Figure 10(c) represent peaks p1 to p6 of the brightness gradient.

[0066] The edge image generating unit 48 calculates the peak of the brightness gradient of each pixel in the direction perpendicular to the reinforcing bar line of the vertical reinforcing bars 5a or the horizontal reinforcing bars 5b for the reinforcing bar image, and extracts the peaks p1 to p6 shown in Figure 10(c) as edge candidates by calculating the peak of the brightness gradient of each pixel on the straight line 71 in the direction perpendicular to the reinforcing bar line of the vertical reinforcing bars 5a for the reinforcing bar image I, and extracts the peaks p1 to p6 with prominence values ​​equal to or greater than a predetermined threshold value from the calculated peaks as edge candidates.

[0067] These peaks p1 to p6 correspond to the edges of each point on the straight line 71 in image I. When edge candidates representing the outline of the rebar are extracted from the edges calculated as peaks based on the peak prominence values ​​in this way, extra edges in areas where the brightness gradient is high due to shadows are extracted as noise. In this example, image I of the rebar 5 shown in FIG. 10(a) includes peaks p1, p3, p4, and p5 of edges corresponding to the shadow 70 as noise among the peaks p1 to p6 of the brightness gradient of each edge. This extra noise may cause the distance between edges to be detected as shorter, resulting in the calculated diameter of the rebar being thinner.

[0068] For this reason, the edge image generation unit 48 further detects the edges of the vertical reinforcement 5a or the horizontal reinforcement 5b that represent the outline of the reinforcing bar region of the reinforcement arrangement 2 by removing edge candidates with lower prominence values ​​among edge candidates whose distance between edge candidates is shorter than the predetermined threshold among those whose prominence values ​​are equal to or greater than a predetermined threshold. In this embodiment, the predetermined threshold is set to 30 pixels (px), and the distance between adjacent peaks, i.e., between edge candidates, is sequentially detected, and when the distance between the edge candidates is shorter than 30px, the peaks with lower prominence values ​​are sequentially removed. By this removal process, the edges of peaks p1, p3, p4, and p5 are removed as noise from the edge-detected image.

[0069] In this way, when the distance between edge candidates extracted based on the prominence values ​​is shorter than a predetermined threshold, the edge candidate with the lower prominence value is removed, thereby reducing the number of edge candidates detected as noise. By repeating this edge candidate removal process, edge candidates detected as noise are eliminated, and the edges of the vertical reinforcement 5a or horizontal reinforcement 5b that represent the outline of the reinforcing bar region of the reinforcement arrangement 2 are obtained. Therefore, the reinforcing bar diameter calculated by the reinforcing bar diameter calculation unit 55 (described later) is correctly calculated without being affected by noise.

[0070] Furthermore, in this embodiment, the edge image generation unit 48 performs line detection using a probabilistic Hough transform on all edge candidates of the vertical reinforcement 5a or the horizontal reinforcement 5b. With this line detection, even if the line segments obtained by connecting the edges vertically or horizontally are too short to detect a line that follows the outline of the reinforcing bar region of the reinforcement 2, the edge image generation unit 48 connects the short line segments using the probabilistic Hough transform, thereby detecting a line that follows the outline of the reinforcing bar region of the reinforcement 2. This prevents a line that follows the outline of the reinforcing bar region of the reinforcement 2 from going undetected.

[0071] For example, edges detected by the edge image generation unit 48 may be scattered as shown in edge detection image J in FIG. 11(a), resulting in undetected straight lines along the contours of the reinforcing bar region of the reinforcement bar 2. Note that, for ease of understanding, the edge detection image J shows the reinforcing bars 5 behind the edges. In such cases, the edge image generation unit 48 connects short line segments 75 using a probabilistic Hough transform to detect straight lines 76 along the contours of the reinforcing bar region of the reinforcement bar 2, as shown in edge detection image K in FIG. 11(b). During the probabilistic Hough transform, the "minimum length of the straight lines 76" and the "maximum gap between two line segments 75 considered as one straight line 76" are appropriately selected to perform an optimal probabilistic Hough transform. Therefore, in this embodiment, it is possible to prevent undetected straight lines along the contours of the reinforcing bar region of the reinforcement bar 2.

[0072] The overcrowded reinforcement separation unit 50 generates an overcrowded reinforcement separation image by superimposing the paired edge-detected image and reinforcement area extraction image for each pair of edge-detected images and reinforcement area extraction images generated from the same front-facing transformed image. The overcrowded reinforcement separation unit 50 generates an overcrowded reinforcement separation image by superimposing the paired edge-detected image and reinforcement area extraction image, and the boundary between the reinforcement area of ​​the target reinforcement in a specific layer and the reinforcement area of ​​the non-target reinforcement in a non-target layer, using the edges in the edge-detected image. For example, the overcrowded reinforcement separation unit 50 generates an overcrowded reinforcement separation image by superimposing the paired vertical edge-detected image H and horizontal edge-detected image G shown in Figures 9(b) and 9(c), and the paired reinforcement area extraction image. The boundary between the reinforcement area of ​​the target reinforcement and the reinforcement area of ​​the non-target reinforcement in the paired reinforcement area extraction image is separated using the edges in the vertical edge-detected image H and horizontal edge-detected image G. An example of such an overcrowded reinforcement separation image is shown in Figure 12.

[0073] The overcrowded reinforcement separation image L shown in Fig. 12 is an overcrowded reinforcement separation image in which the vertical and horizontal edges of one front-to-back transformed image are superimposed on the paired reinforcement area extraction image. In this embodiment, the overcrowded reinforcement separation unit 50 superimposes the vertical edges of the vertical edge detection image H on the paired reinforcement area extraction image, and generates an overcrowded reinforcement separation image in which the horizontal edges of the horizontal edge detection image G are superimposed on the same reinforcement area extraction image.

[0074] In this embodiment, the control unit 40 includes a rib edge removal unit 49. The rib edge removal unit 49 overlays a displacement amount estimation mask image on the edge detection image, and removes edges in overlapping areas of each reinforcement area in the edge detection image and the displacement amount estimation mask image as rib edges corresponding to the ribs of the reinforcing bars 5 from each edge detection image (see FIG. 5, step S108).

[0075] In this embodiment, the rib edge remover 49 overlays a vertical edge-detection image with a vertical line displacement estimation mask image, and removes from each vertical edge-detection image vertical edges in overlapping areas of each vertical line region in the vertical edge-detection image and the vertical line displacement estimation mask image as rib edges corresponding to the ribs of the vertical line 5a. Also, it overlays a horizontal line displacement estimation mask image with a horizontal edge-detection image, and removes from each horizontal edge-detection image edges in overlapping areas of each horizontal line region in the horizontal edge-detection image and the horizontal line displacement estimation mask image as rib edges corresponding to the ribs of the horizontal line 5b.

[0076] When the original image captured by the monocular camera 30 is the original image M shown in FIG. 13(a), the edge-detected image N shown in FIG. 13(b) is generated by the edge image generation unit 48. Note that, for ease of understanding, the reinforcement 2 is shown in the background of the edge in the edge-detected image N. In this edge-detected image N, a straight line 81 corresponding to the rib 80 of the vertical reinforcement 5a captured in the original image M is detected in a vertical reinforcement region 82. Therefore, the vertical reinforcement region 82 is vertically divided by the straight line 81 into two regions with the straight line 81 as the boundary.

[0077] 13(c), the straight line 81 appears within the region of the vertical reinforcing bars 5a in the edge-superimposed photographed image O, which is obtained by superimposing the straight line 81 on the original image M. For this reason, the reinforcing bar diameter calculated by the reinforcing bar diameter calculation unit 55, which will be described later, is, for example, the bisected reinforcing bar diameter D of 16 mm, which is calculated to be thinner than the actual reinforcing bar diameter D of 35 mm.

[0078] To resolve this issue, the rib edge removal unit 49 generates the displacement estimation mask superimposed image P shown in FIG. 14(b) by superimposing, for example, the vertical line displacement estimation mask image E shown in FIG. 8 on the edge-detected image N, which is an enlarged version of the image shown in FIG. 13(b) shown in FIG. 14(a). Here, the explanation will be given using the edge-detected image N in which vertical and horizontal edges are superimposed, instead of the vertical edge-detected image. The rib edge removal unit 49 removes the edge in the area where the vertical line region 82 in the edge-detected image N and the vertical line region 71 in the vertical line displacement estimation mask image E overlap, i.e., the straight line 81, as a rib edge corresponding to the rib 80 of the vertical line 5a, from the edge-detected image N.

[0079] In this embodiment, the overcrowded reinforcement separation unit 50 overlays each edge detection image from which the rib edges have been removed on each paired reinforcement area extraction image, and separates the boundary between the reinforcement area of ​​the target reinforcement 2 and the reinforcement area of ​​the non-target reinforcement 2 in each paired reinforcement area extraction image using the edges in each edge detection image from which the rib edges have been removed, thereby generating an overcrowded reinforcement separation image (see Figure 5, step S109).

[0080] More specifically, in this embodiment, the overcrowded reinforcement separation unit 50 separates the boundary between the vertical reinforcement area of ​​the target reinforcement 2 and the vertical reinforcement area of ​​the non-target reinforcement 2 using edges obtained by removing rib edges from the vertical edges in the vertical edge detection image H. Also, it separates the boundary between the horizontal reinforcement area of ​​the target reinforcement 2 and the horizontal reinforcement area of ​​the non-target reinforcement 2 using edges obtained by removing rib edges from the horizontal edges in the horizontal edge detection image G.

[0081] The parallax mask reinforcement separation unit 51 generates a parallax mask reinforcement separation image in which the boundary between the reinforcement area of ​​the target reinforcement 2 and the reinforcement area of ​​the non-target reinforcement 2 in the parallax mask image is separated by the edge in the edge detection image after rib edge removal by superimposing the edge detection image after rib edge removal on the parallax mask image (see Figure 5, step S110).

[0082] In this embodiment, the parallax mask reinforcement separation unit 51 generates a parallax mask vertical line separation image in which the boundary between the vertical line region of the target reinforcement 2 and the vertical line region of the non-target reinforcement 2 in the vertical line parallax mask image is separated by the vertical edge in the vertical line edge detection image in which the parallax mask image is removed. Also, the parallax mask reinforcement separation unit 51 generates a parallax mask horizontal line separation image in which the boundary between the horizontal line region of the target reinforcement 2 and the horizontal line region of the non-target reinforcement 2 in the horizontal line parallax mask image is separated by the horizontal edge in the horizontal line edge detection image in which the parallax mask image is removed.

[0083] 15(b) shows a vertical line parallax mask original image R separated by the parallax mask image separator 47 when the original image captured by the monocular camera 30 is the original image Q shown in Fig. 15(a). Marks 90 are attached to the vertical lines 5a of a predetermined layer on the front surface in the original image Q, and the vertical line parallax mask original image R highlights the vertical lines 5a of the target reinforcement 2 on the front surface that are attached with the marks 90, and the vertical lines 5a of the non-target reinforcement 2 on the back surface of the first layer that were not separated by parallax.

[0084] The vertical line disparity mask image separated by the disparity mask image separation unit 47 is generated as a vertical line disparity mask image S corresponding to the original vertical line disparity mask image R, for example, as shown in Fig. 16. This vertical line disparity mask image S shows an estimated target vertical line region including a vertical line region 91 of the vertical lines 5a in the target reinforcement 2 on the front side and a vertical line region 92 of the vertical lines 5a in the non-target reinforcement 2 on the back side. Traces 90a of the mark 90 appear in the vertical line region 91.

[0085] The parallax mask reinforcement separation unit 51 generates a parallax mask vertical line separation image in which the boundary between the vertical line area 91 of the target reinforcement 2 and the vertical line area 92 of the non-target reinforcement 2 in the vertical line parallax mask image S is separated by the vertical edge in the vertical line edge detection image after rib edge removal, by superimposing the vertical line edge detection image after rib edge removal on the vertical line parallax mask image S.

[0086] The target reinforcement area extraction unit 52 generates a parallax mask reinforcement extraction image by superimposing the displacement estimation mask image on the parallax mask reinforcement separation image and removing reinforcement areas that do not overlap with reinforcement areas of a predetermined layer in the displacement estimation mask image from the parallax mask reinforcement separation image.Then, the target reinforcement area extraction unit 52 extracts reinforcement areas of the parallax mask reinforcement separation image that overlap with reinforcement areas of a predetermined layer in the displacement estimation mask image as reinforcement areas of the target reinforcement 2 (see Figure 5, step S111).

[0087] In this embodiment, the target reinforcement area extraction unit 52 overlays the vertical line displacement estimation mask image on the parallax mask vertical line separation image, and generates a parallax mask vertical line extraction image by removing from the parallax mask vertical line separation image any vertical line areas that do not overlap with the vertical line areas of a predetermined layer in the vertical line displacement estimation mask image, and extracts the vertical line areas of the parallax mask vertical line separation image that overlap with the vertical line areas of a predetermined layer in the vertical line displacement estimation mask image as the vertical line areas of the target reinforcement 2. Furthermore, the target reinforcement area extraction unit 52 overlays the horizontal line displacement estimation mask image on the parallax mask horizontal line separation image, and generates a parallax mask horizontal line extraction image by removing from the parallax mask horizontal line separation image any horizontal line areas that do not overlap with the horizontal line areas of a predetermined layer in the horizontal line displacement estimation mask image, and extracts the horizontal line areas of the parallax mask horizontal line separation image that overlap with the horizontal line areas of a predetermined layer in the horizontal line displacement estimation mask image as the horizontal line areas of the target reinforcement 2.

[0088] FIG. 17 is a diagram for explaining the processes performed on the vertical reinforcement 5a by the rib edge removal unit 49, the overcrowded reinforcement separation unit 50, the parallax mask reinforcement separation unit 51, and the target reinforcement area extraction unit 52.

[0089] As described above, the rib edge removal unit 49 overlays the vertical edge detection image T with the vertical reinforcement displacement estimation mask image U, thereby removing the straight lines 81 in the vertical edge detection image T that overlap the vertical reinforcement region 87 in the vertical reinforcement displacement estimation mask image U as rib edges, to generate a vertical edge detection image T' after rib edge removal. The overcrowded reinforcement separation unit 50 overlays this vertical edge detection image T' on the paired reinforcement region extraction image V, to generate an overcrowded reinforcement separated image V' in which the vertical reinforcement region 89 in the reinforcement region extraction image V is separated by the edges 88 in the vertical edge detection image T'. In the vertical reinforcement region 89, the vertical reinforcement boundaries between the vertical reinforcement region 91 of the target reinforcement 2 and the vertical reinforcement region 92 of the non-target reinforcement 2 are separated by these edges 88.

[0090] As described above, the parallax mask reinforcement separation unit 51 overlays the vertical edge detection image T' after rib edge removal on the vertical line parallax mask image W to generate a parallax mask vertical line separated image W' in which the vertical line region 86 in the vertical line parallax mask image W is separated by an edge 88 in the vertical line edge detection image T'. In the vertical line region 86, the vertical line boundary between the vertical line region 91 of the target reinforcement 2 and the vertical line region 92 of the non-target reinforcement 2 is separated by this edge 88. The target reinforcement region extraction unit 52 overlays the vertical line displacement estimation mask image U on the parallax mask vertical line separated image W' to remove the vertical line region 92 that overlaps the vertical line region 87 of the vertical line displacement estimation mask image U to generate a parallax mask vertical line extracted image W". The target reinforcement region extraction unit 52 extracts the vertical line region remaining in the parallax mask vertical line extracted image W" as the vertical line region 91 of the target reinforcement 2.

[0091] In this embodiment, the parallax mask reinforcement separation unit 51 overlays the remaining vertical edge detection images T' from which the rib edges have been removed onto the parallax mask vertical line extraction image W" for the number of remaining vertical edge detection images T not used in generating the parallax mask vertical line separation image W', thereby separating the vertical line boundaries between the vertical line region 91 of the target reinforcement 2 and the vertical line region 92 of the non-target reinforcement 2, which remain in the parallax mask vertical line extraction image W", for each of the remaining vertical edge detection images T' from which the rib edges have been removed, using the vertical edges in each of the remaining vertical edge detection images T' from which the rib edges have been removed, for each of the remaining vertical edge detection images T', and updates the parallax mask vertical line separation image W' (see FIG. 5, step S112).

[0092] Every time the parallax mask vertical line separation image W' is updated, the target reinforcement area extraction unit 52 overlays the vertical line displacement amount estimation mask image U on the updated parallax mask vertical line separation image W', and removes from the updated parallax mask vertical line separation image W' the vertical line areas that do not overlap with the vertical line areas 87 of the vertical line displacement amount estimation mask image U, thereby updating the parallax mask vertical line extraction image W" (see FIG. 5, step S113).

[0093] Then, the control unit 40 determines whether the update process of the parallax mask vertical line separated image W' by the parallax mask reinforcement separation unit 51 and the update process of the parallax mask vertical line extracted image W" by the target reinforcement area extraction unit 52 have been performed for the remaining number of vertical edge detected images T that have not been used to generate the parallax mask vertical line separated image W' (see FIG. 5, step S114).

[0094] If the update processes have not been performed for the remaining number of vertical edge detection images T, the processes of steps S112 to S114 are repeated. Once the update processes have been performed for the remaining number of vertical edge detection images T, the target reinforcement area extraction unit 52 extracts the vertical reinforcement area 91 in the last updated parallax mask vertical reinforcement extraction image W″ as the vertical reinforcement area of ​​the target reinforcement 2.

[0095] The process of extracting the horizontal reinforcement area in the target reinforcement arrangement 2 is also performed in the same manner as the processes performed on the horizontal reinforcement 5b and the vertical reinforcement 5a by the rib edge removal unit 49, the overcrowded reinforcement separation unit 50, the parallax mask reinforcement separation unit 51, and the target reinforcement arrangement area extraction unit 52.

[0096] If the ratio of the area of ​​the overlapping portion between the continuous area of ​​the reinforcement area included in the parallax mask reinforcement separation image and the reinforcement area of ​​the target reinforcement included in the displacement estimation mask image to the area of ​​the continuous area is equal to or less than a predetermined threshold, the target reinforcement area extraction unit 52 does not extract the continuous area from the reinforcement area included in the parallax mask reinforcement separation image as that of non-target reinforcement 2.

[0097] In this embodiment, if the ratio of the area of ​​the overlapping portion between the continuous vertical reinforcement regions 91, 92 included in the parallax mask vertical reinforcement separation image W' and the vertical reinforcement region 87 of the target reinforcement 2 included in the vertical reinforcement displacement estimation mask image U to the area of ​​the continuous vertical reinforcement regions 91, 92 is equal to or smaller than a predetermined threshold, the target reinforcement region extraction unit 52 does not extract the continuous vertical reinforcement regions 91, 92 from the vertical reinforcement regions 91, 92 included in the parallax mask vertical reinforcement separation image W' as belonging to the non-target reinforcement 2. Also, if the ratio of the area of ​​the overlapping portion between the continuous horizontal reinforcement regions included in the parallax mask horizontal reinforcement separation image and the horizontal reinforcement region of the target reinforcement included in the horizontal reinforcement displacement estimation mask image is equal to or smaller than a predetermined threshold, the target reinforcement region extraction unit 52 does not extract the continuous horizontal reinforcement regions from the horizontal reinforcement regions included in the parallax mask horizontal reinforcement separation image as belonging to the non-target reinforcement 2.

[0098] For example, the target reinforcement area extraction unit 52 generates the parallax mask vertical line extraction image X shown in Fig. 18(a) by superimposing the vertical line displacement estimation mask image E shown in Fig. 8 on the parallax mask vertical line separation image in which the vertical line boundaries in the vertical line parallax mask image S shown in Fig. 16 are separated by vertical edges after rib edge removal. Then, the target reinforcement area extraction unit 52 extracts the vertical line area 91 of the target reinforcement 2 included in the parallax mask vertical line extraction image X from the overlapping area between the vertical line area 71 in the vertical line displacement estimation mask image E and the vertical line areas 91, 92 in the parallax mask vertical line extraction image X.

[0099] More specifically, the target reinforcement area extraction unit 52 assigns an index to each vertical reinforcement area 91, 92 included in the parallax mask vertical reinforcement extraction image X. Here, each vertical reinforcement area 91, 92 is recognized by the target reinforcement area extraction unit 52 as a continuous area of ​​reinforcement 2 in the estimated target reinforcement area. The target reinforcement area extraction unit 52 calculates the area Sa of the overlapping portion between each vertical reinforcement area 91, 92 and the vertical reinforcement area 71 of the target reinforcement 2 included in the vertical reinforcement displacement estimation mask image E. This area Sa is calculated as the common area of ​​the overlapping portion 93 between each continuous vertical reinforcement area 91, 92 and the vertical reinforcement area 71, as conceptually shown in FIG. 18(b).

[0100] The target reinforcement area extraction unit 52 calculates the area ratio Sa / Sb of this common area Sa to the area Sb of the continuous vertical reinforcement areas 91, 92. If this area ratio Sa / Sb is equal to or less than a predetermined threshold, the target reinforcement area extraction unit 52 does not extract the continuous area (vertical reinforcement areas 91, 92) from the estimated target reinforcement area, assuming that it is the non-target reinforcement area 2 on the back side. If the area ratio Sa / Sb exceeds the predetermined threshold, the target reinforcement area extraction unit 52 extracts the continuous area (vertical reinforcement areas 91, 92) from the estimated target reinforcement area, assuming that it is the target reinforcement area 2 on the front side.

[0101] By extracting the reinforcement area of ​​the target reinforcement 2 from the estimated target reinforcement area in this way, the target reinforcement area extraction unit 52 can properly determine whether the continuous area (91, 92) in the estimated target reinforcement area included in the parallax mask vertical reinforcement extraction image X belongs to the reinforcement area of ​​the target reinforcement 2 on the front side or the reinforcement area of ​​the non-target reinforcement 2 on the back side. If the continuous area (91, 92) is determined to belong to the reinforcement area of ​​the non-target reinforcement 2, the target reinforcement area extraction unit 52 will no longer extract the continuous area (91, 92) as the reinforcement area of ​​the target reinforcement 2 from the estimated target reinforcement area included in the parallax mask vertical reinforcement extraction image X. Therefore, the reinforcement area of ​​the target reinforcement 2 can be properly extracted from the estimated target reinforcement area included in the parallax mask vertical reinforcement extraction image X.

[0102] Furthermore, in this embodiment, the process of determining whether a continuous area of ​​reinforcement areas included in a parallax mask reinforcement separation image is a reinforcement area of ​​the target reinforcement 2 or a reinforcement area of ​​the non-target reinforcement 2 is performed individually when the reinforcement area is divided into vertical reinforcement areas and horizontal reinforcement areas. When the reinforcement area is divided into vertical reinforcement areas and horizontal reinforcement areas, the vertical reinforcement areas and horizontal reinforcement areas are clearly distinguished and can be easily understood. This makes it possible to more accurately extract the reinforcement area of ​​the target reinforcement from the reinforcement areas included in a parallax mask reinforcement separation image.

[0103] The reinforcing bar line calculation unit 53 calculates the center line of the reinforcing bar area in the target reinforcement 2 extracted by the target reinforcement area extraction unit 52 as the reinforcing bar line (see FIG. 5, step S115). In this embodiment, the reinforcing bar line calculation unit 53 calculates the center line of the vertical reinforcement area in the target reinforcement 2 extracted by the target reinforcement area extraction unit 52 as the reinforcing bar line of the vertical reinforcement 5a, and calculates the center line of the horizontal reinforcement area in the target reinforcement 2 extracted by the target reinforcement area extraction unit 52 as the reinforcing bar line of the horizontal reinforcement 5b.

[0104] The center line of the reinforcing bar 5 is calculated from the reinforcing bar area of ​​the target reinforcing bar 2 included in the estimated target reinforcing bar area in the parallax mask image because the reinforcing bar area in the displacement estimation mask image has lower resolution than the reinforcing bar area extracted from the parallax mask image.

[0105] Specifically, the reinforcing bar line calculation unit 53 extracts multiple center points from the reinforcing bar region 5A, for example as shown in Figure 19(a), and uses the Hough method to calculate the center line C of the reinforcing bar region 5A as a straight line, and regards the center line C as the reinforcing bar line.

[0106] The rebar spacing calculation unit 54 calculates the rebar spacing of the target rebar arrangement, i.e., the rebar arrangement pitch, from the distance between the rebar lines (see FIG. 5, step S116). In this embodiment, the rebar spacing calculation unit 54 calculates the vertical rebar spacing of the vertical rebars 5a from the distance between the rebar lines of the vertical rebars 5a, and calculates the horizontal rebar spacing of the horizontal rebars 5b from the distance between the rebar lines of the horizontal rebars 5b.

[0107] For example, as shown in Fig. 19(b), the rebar spacing calculation unit 54 calculates distances D1 to D4 between the center lines C of the vertical rebars 5a and distances E1 to E4 between the center lines C of the horizontal rebars 5b. The rebar spacing calculation unit 54 can calculate the distance between adjacent center lines C based on the number of pixels in the overcrowded reinforcement separation image on which the center lines C are superimposed, as will be described later. At this time, the distance between the marks 11 of the markers 10 is used as a reference to convert from the number of pixels to actual size.

[0108] The rebar spacing calculation unit 54 can also calculate the distance between adjacent center lines C by calculating the distance between the intersections of the horizontal center line C and the vertical center line C that is perpendicular to the horizontal center line C. The distances D1 to D4, E1 to E4 between adjacent center lines C are the rebar spacing of the rebars 5, that is, the reinforcing bar pitch. The rebar spacing calculation unit 54 also calculates the number of center lines C in the vertical and horizontal directions.

[0109] The reinforcing bar diameter calculation unit 55 superimposes the center line C calculated by the reinforcing bar line calculation unit 53 on the overcrowded reinforcement separation image generated by the overcrowded reinforcement separation unit 50, and calculates the reinforcing bar diameter of the target reinforcement 2 for each overcrowded reinforcement separation image from the reinforcing bar area of ​​the target reinforcement 2 where the center line C overlaps. Then, the reinforcing bar diameter with the highest calculation frequency among the calculated reinforcing bar diameters is set as the reinforcing bar diameter of the target reinforcement 2 (see FIG. 5, step S117). The reason for calculating the reinforcing bar diameter from the reinforcing bar area in the overcrowded reinforcement separation image in this way is because the reinforcing bar area of ​​the target reinforcement 2 included in the estimated target reinforcement area extracted in the disparity mask image has lower resolution than the reinforcing bar area in the overcrowded reinforcement separation image.

[0110] In this embodiment, the reinforcing bar diameter calculation unit 55 superimposes the center line C of the vertical reinforcement area of ​​the vertical reinforcement 5a on the over-dense reinforcement separation image, and calculates the reinforcing bar diameter of the vertical reinforcement 5a in the target reinforcement arrangement 2 for each over-dense reinforcement separation image from the vertical reinforcement area of ​​the over-dense reinforcement separation image where the center line C of the vertical reinforcement area is superimposed. Also, the reinforcing bar diameter calculation unit 55 superimposes the center line C of the horizontal reinforcement area of ​​the horizontal reinforcement 5b on the over-dense reinforcement separation image, and calculates the reinforcing bar diameter of the horizontal reinforcement 5b in the target reinforcement arrangement 2 for each over-dense reinforcement separation image from the horizontal reinforcement area of ​​the over-dense reinforcement separation image where the center line C of the horizontal reinforcement area is superimposed. Then, the reinforcing bar diameter with the highest calculation frequency among the calculated reinforcing bar diameters of the vertical reinforcement 5a and the horizontal reinforcement 5b in the target reinforcement arrangement 2 is determined as the reinforcing bar diameter of each of the vertical reinforcement 5a and the horizontal reinforcement 5b.

[0111] Specifically, the center line C of the vertical reinforcement region of the vertical reinforcement 5a is superimposed on one overcrowded reinforcement separation image, and the rebar diameter of each vertical reinforcement region is calculated for the number of overlapping center lines C. For example, if seven vertical reinforcement regions 91 overlapping the vertical reinforcement region 71 are extracted as the target reinforcement 2 from the parallax mask vertical reinforcement extraction image X shown in FIG. 18(a), the width of each of the seven vertical regions 91 is calculated as the rebar diameter of the vertical reinforcement 5a. By performing this calculation for multiple overcrowded reinforcement separation images, the rebar diameter of 7 × multiple images is calculated. For the horizontal reinforcement 5b, the rebar diameter of the horizontal reinforcement 5b is calculated in the same way for multiple overcrowded reinforcement separation images. The mode of the rebar diameters of the vertical reinforcement 5a and the horizontal reinforcement 5b calculated in this way is aggregated as a single true value of the rebar diameter of each of the vertical reinforcement 5a and the horizontal reinforcement 5b.

[0112] Since the center line C calculated by the reinforcing bar line calculation unit 53 is unique, the calculated reinforcing bar intervals between the vertical bars 5a and the horizontal bars 5b are true values.

[0113] To calculate the reinforcing bar diameter from the reinforcing bar region, for example, as shown in Fig. 20(a), widths W1, W2, and W3 of the reinforcing bar region 5A in the direction perpendicular to the center line C are detected, and the peak of the smallest width W1 from the distribution of the widths W1, W2, and W3 and the frequency of pixel appearance shown in Fig. 20(b) is calculated as the diameter of the reinforcing bar 5. The reinforcing bar diameter calculation unit 55 calculates the diameter of the reinforcing bar 5 for all reinforcing bar regions of the target reinforcing bar 2 extracted from the overcrowded reinforcing bar separation image.

[0114] 20(b), it can be seen that there is a peak H1 at width W1, a peak H2 at width W2, and a peak H3 at width W3. If width W1 is W, width W2 is about 1.1W to 1.13W, and width W3 is 10W or more.

[0115] In other words, the width W1 of the reinforcing bar region 5A can be estimated to be the diameter of the main body of the reinforcing bar 5 excluding the node 5c part, the width W2 of the reinforcing bar region 5A can be estimated to be the diameter of the node 5c part of the reinforcing bar 5, and the width W3 of the reinforcing bar region 5A can be estimated to be the length of the reinforcing bar 5 arranged in a direction perpendicular to the reinforcing bar 5.

[0116] The reinforcing bar diameter calculation unit 55 calculates the smallest width W1 of the multiple peaks H1, H2, and H3 as the diameter of the reinforcing bar 5. The reinforcing bar diameter calculation unit 55 can also calculate the width W1 of the reinforcing bar region 5A of the most frequent peak H1 as the diameter of the reinforcing bar 5. The diameter of the reinforcing bar 5 can also be calculated from the distance between the edges sandwiching the main body of the reinforcing bar region 5A in the direction perpendicular to the center line C.

[0117] 21, and displays the generated analysis result image Y on the image display device 21 (see FIG. 5, step S118). In this analysis result image Y, the rebar lines of the vertical reinforcement 5a and the horizontal reinforcement 5b are shown as straight lines, and the rebar diameter D of each of the vertical reinforcement 5a and the horizontal reinforcement 5b, the rebar pitch of each of the vertical reinforcement 5a and the horizontal reinforcement 5b, and the number of each of the vertical reinforcement 5a and the horizontal reinforcement 5b are displayed together with the design data 41a stored in the storage unit 41.

[0118] The analysis result image generating unit 56 compares the calculated rebar diameter D of each of the vertical reinforcement 5a and horizontal reinforcement 5b, each rebar arrangement pitch, and each number with the design data 41a, and determines whether the rebar arrangement 2 in the overcrowded rebar arrangement is correct, that is, whether the rebar arrangement 2 has been correctly constructed as designed. If the rebar diameter D or rebar arrangement pitch is not detected as being as designed, the analysis result image generating unit 56 displays these values ​​on the image display device 21 by highlighting them in red.

[0119] [Actions and effects of the reinforcement inspection system] According to the reinforcement inspection system 1 of this embodiment, for each front-facing transformed image, the reinforcement area extraction image generation unit 43 generates a reinforcement area extraction image that extracts the reinforcement area consisting of the vertical reinforcements 5a and horizontal reinforcements 5b, and the edge image generation unit 48 generates edge-detected images G and H that detect the edges of the vertical reinforcements 5a and horizontal reinforcements 5b. The overcrowded reinforcement separation unit 50 overlays the edge-detected images G and H on the reinforcement area extraction image paired with the edge-detected image G and H to generate an overcrowded reinforcement separation image for each pair of the edge-detected images G and H and the reinforcement area extraction image. In this overcrowded reinforcement separation image, the boundary between the reinforcement area of ​​the target reinforcement 2 and the reinforcement area of ​​the non-target reinforcement 2 in the reinforcement area extraction image is separated and clarified by the edges in the edge-detected images G and H.

[0120] Furthermore, the parallax mask image generating unit 46 superimposes each reinforcement area extraction image, leaving the overlapping image portions, and thereby removes reinforcement areas estimated to be in non-target layers from each reinforcement area extraction image due to the parallax caused by differences in shooting angles, generating a parallax mask image S representing an estimated target reinforcement area estimated to be the reinforcement area of ​​the target reinforcement 2. Furthermore, the displacement estimation mask image generating unit 44 presumes that the layer with the smallest displacement of the reinforcement 2 during the facing transformation is the specified layer where the marker 10 is placed, and generates a displacement estimation mask image E in which the reinforcement area of ​​that layer is extracted.

[0121] The parallax mask reinforcement separation unit 51 overlays the edge detection images G and H on the parallax mask image S generated by the parallax mask image generation unit 46, thereby generating a parallax mask reinforcement separation image in which the boundary between the reinforcement area of ​​the target reinforcement 2 and the reinforcement area of ​​the non-target reinforcement 2 in the parallax mask image S is separated by the edges in the edge detection images G and H. Furthermore, the target reinforcement area extraction unit 52 overlays the displacement amount estimation mask image generated by the displacement amount estimation mask image generation unit 44 on this parallax mask reinforcement separation image, thereby extracting from the parallax mask image the reinforcement area of ​​the target reinforcement 2 in which the marker 10 is placed, which is included in the parallax mask reinforcement separation image.

[0122] The center line C of the extracted reinforcing bar area in the target reinforcement 2 is calculated as the reinforcing bar line by the reinforcing bar line calculation unit 53. Furthermore, the reinforcing bar spacing of the target reinforcement 2 is calculated from the distance between the reinforcing bar lines by the reinforcing bar spacing calculation unit 54. Furthermore, the center line C calculated by the reinforcing bar line calculation unit 53 is superimposed on the over-dense reinforcement separation image, and the reinforcing bar diameter of the target reinforcement 2 is calculated for each over-dense reinforcement separation image by the reinforcing bar diameter calculation unit 55 from the reinforcing bar area of ​​the target reinforcement 2 in the over-dense reinforcement separation image on which the center line C is superimposed. Then, the reinforcing bar diameter with the highest calculation frequency is determined to be the reinforcing bar diameter of the target reinforcement 2.

[0123] Therefore, according to the reinforcement inspection system 1 of this embodiment, it is possible to provide a reinforcement inspection system 1 that can extract only the target reinforcement 2 on the front side where the marker 10 is placed from the overcrowded reinforcement, thereby improving the measurement accuracy of each dimension, etc. of the target reinforcement 2 on the front side.

[0124] Furthermore, according to the reinforcement inspection system 1 of this embodiment, as shown in FIG. 14 , a straight line 81 in an overlapping region of each vertical reinforcement region 82, 71 in the edge-detection image N and the displacement estimation mask image E is removed from the edge-detection image N as a rib edge by the rib edge removal unit 49. That is, an edge of the edge-detection image N overlapping the reinforcement region of the displacement estimation mask image E is regarded as an edge corresponding to the rib of the reinforcing bar and is removed by the rib edge removal unit 49. Therefore, the overcrowded reinforcement separation unit 50 generates an overcrowded reinforcement separation image in which the boundary between the reinforcement region of the target reinforcement 2 and the reinforcement region of the non-target reinforcement 2 is accurately and clearly separated by the edges in the edge-detection image N from which the rib edges have been removed. Therefore, the reinforcing bar region of the target reinforcement 2 is accurately detected by superimposing the center line C calculated by the reinforcing bar line calculation unit 53 on the overcrowded reinforcement separation image, and the reinforcing bar diameter can be accurately calculated from the reinforcing bar region of the target reinforcement 2 by the reinforcing bar diameter calculation unit 55.

[0125] Furthermore, according to the reinforcement inspection system 1 of this embodiment, the parallax mask reinforcement separation unit 51 initially generates a parallax mask vertical line separated image W' by superimposing an edge-detected image T' on the parallax mask image W, as shown in steps S111 to S114 of Fig. 5 and Fig. 17. However, once the parallax mask vertical line separated image W" is generated by the target reinforcement area extraction unit 52, the remaining edge-detected images T' are superimposed on the parallax mask vertical line separated image W" instead of the parallax mask image W for each of the remaining edge-detected images T', thereby updating the parallax mask vertical line separated images W' for the number of remaining edge-detected images T' that were not used to generate the parallax mask vertical line separated image W'.

[0126] Every time the parallax mask vertical line separation image W' is updated, the target reinforcement area extraction unit 52 overlays the displacement estimation mask image U on the updated parallax mask vertical line separation image W', removes vertical line areas in the updated parallax mask vertical line separation image W' that do not overlap with the vertical line areas 87 of the displacement estimation mask image U, and updates the parallax mask vertical line extraction image W". Then, based on the last updated parallax mask vertical line extraction image W", the vertical line area of ​​the target reinforcement 2 is extracted.

[0127] Therefore, according to the reinforcement inspection system 1 of this embodiment, even if there is an edge-detected image T' in which the edge 88 corresponding to the boundary between the reinforcement area of ​​the target reinforcement 2 and the reinforcement area of ​​the non-target reinforcement 2 is not extracted among the multiple edge-detected images T' generated from each front-view transformed image, as long as the edge 88 is extracted in the remaining edge-detected images T', the reinforcement area of ​​the target reinforcement 2 can be reliably extracted by the processes of the parallax mask reinforcement separation unit 51 and the target reinforcement area extraction unit 52 using the remaining edge-detected images T'. Furthermore, even if the process of separating and extracting the reinforcement area of ​​the target reinforcement 2 is performed by the parallax mask reinforcement separation unit 51 and the target reinforcement area extraction unit 52 for the number of edge-detected images T, and the process of distinguishing between the target reinforcement 2 and the non-target reinforcement 2 is repeated, the reinforcement area of ​​the target reinforcement 2 can be reliably extracted.

[0128] Furthermore, according to the reinforcement arrangement inspection system 1 of this embodiment, the disparity mask image generated by the disparity mask image generation unit 46 is separated by the disparity mask image separation unit 47 into a vertical line disparity mask image S representing a vertical line area to be estimated that is the vertical line area of ​​the target reinforcement 2 within the reinforcement arrangement area to be estimated, and a horizontal line disparity mask image representing a horizontal line area to be estimated that is the horizontal line area of ​​the target reinforcement 2 within the reinforcement area to be estimated. Furthermore, the displacement estimation mask image generated by the displacement estimation mask image generation unit 44 is separated by the displacement estimation mask image separation unit 45 into a vertical line displacement estimation mask image E for the vertical lines and a horizontal line displacement estimation mask image for the horizontal lines. Furthermore, the edge detection image generated for each front-to-back transformed image is separated by the edge image generation unit 48 into a vertical line edge detection image H in which the edges of the vertical lines are detected, and a horizontal line edge detection image G in which the edges of the horizontal lines are detected.

[0129] Accordingly, each process performed on each image is divided into image processing for the vertical reinforcement 5a and image processing for the horizontal reinforcement 5b. By performing each process separately for the vertical reinforcement 5a and the horizontal reinforcement 5b in this way, each image, which previously had a rough appearance and contained a lot of noise due to the vertical reinforcement 5a and the horizontal reinforcement 5b intersecting, becomes a simple image that is easy to see and less cluttered. This reduces the number of missed detections of straight lines along the outline of the reinforcing bar region, improving the detection rate of those straight lines and, in turn, further improving the measurement accuracy of the dimensions of the reinforcement region.

[0130] Furthermore, according to the reinforcement inspection system 1 of this embodiment, the rib edges 81 detected in each vertical edge detection image N, T (see FIGS. 14 and 17) and each horizontal edge detection image, which correspond to the vertical reinforcement 5a and horizontal reinforcement 5b, overlap the vertical reinforcement displacement estimation mask image E, U (see FIGS. 8 and 17) and the horizontal reinforcement displacement estimation mask image, and are then simplified and removed by the rib edge removal unit 49. Therefore, each vertical and horizontal rib edge is detected and removed without omission, compared to when they are detected and removed all at once. Therefore, the vertical and horizontal reinforcement boundaries between the reinforcement areas of the target reinforcement 2 and the reinforcement areas of the non-target reinforcement 2 are reliably separated without omission by the overcrowded reinforcement separation unit 50 using the edges in each vertical edge detection image T' (see FIG. 17) from which the rib edges have been removed and the edges in each horizontal edge detection image from which the rib edges have been removed.

[0131] Therefore, the center lines C of the vertical and horizontal reinforcement areas of the target reinforcement 2 are superimposed, and the vertical and horizontal reinforcement areas of the target reinforcement 2 obtained from the overcrowded reinforcement separated image V' (see FIG. 17) are accurate. As a result, the reinforcing bar diameters of the vertical and horizontal reinforcement areas of the target reinforcement 2 obtained based on these vertical and horizontal reinforcement areas can be accurately calculated by the reinforcing bar diameter calculation unit 55.

[0132] Furthermore, according to the reinforcement inspection system 1 of this embodiment, the process performed by the parallax mask reinforcement separation unit 51 to update the parallax mask vertical line separation image W' by superimposing the edge detection image T' on the parallax mask vertical line separation image W" (see FIG. 5, step S112) is performed by converting the edge detection image and the parallax mask reinforcement extraction image into images separated into vertical edges and horizontal edges. Furthermore, the process performed by the target reinforcement area extraction unit 52 to update the parallax mask vertical line separation image W" by superimposing the displacement amount estimation mask image U on the updated parallax mask vertical line separation image W' every time it is updated is also performed by converting the parallax mask reinforcement separation image and the displacement amount estimation mask image into images separated into vertical edges and horizontal edges. Therefore, each process performed by the parallax mask reinforcement separation unit 51 and the target reinforcement area extraction unit 52 is performed in a simple state where the edges are separated into vertical and horizontal directions, so each edge is reliably captured and performed reliably. [Explanation of symbols]

[0133] 1. Reinforcement inspection system 2…Reinforcement 5...Reinforcing bars 5a...Vertical stripes 5b...Horizontal stripes 10...Marker 30...Monocular camera 40...Control unit (personal computer) 41...Storage section 42...Direction transformation image generation unit 43…Reinforcement area extraction image generation unit 44...Displacement amount estimation mask image generation unit 45...Displacement amount estimation mask image separation unit 46...Disparity mask image generation unit 47... Parallax mask image separation unit 48...Edge image generation unit 49...Rib edge removal section 50...Excessive reinforcement separation section 51... Parallax mask reinforcement separation section 52...Target reinforcement area extraction section 53...Reinforcing bar calculation section 54...Rebar spacing calculation section 55...Rebar diameter calculation section 56...Analysis result image generation unit

Claims

1. a front-facing transformed image generating unit that generates a plurality of front-facing transformed images by converting a plurality of original images of markers arranged on a predetermined layer of reinforcement composed of vertical and horizontal reinforcements, which are photographed from different angles by a monocular camera, into front-facing images of the markers; a reinforcement area extraction image generation unit that generates a reinforcement area extraction image for each of the front-facing transformed images by extracting a reinforcement area from the front-facing transformed image; an edge image generation unit that generates an edge-detected image in which the edges of the vertical streaks and the horizontal streaks are detected for each of the facing transformed images; a parallax mask image generating unit that overlays the reinforcement region extraction images so that the markers are aligned with each other and generates a parallax mask image that leaves the overlapping image portions of the reinforcement region extraction images; a displacement amount estimation mask image generating unit that estimates a layer in which the amount of displacement of the reinforcement is small during the front-to-rear transformation among the plurality of front-to-rear transformed images as the predetermined layer, and generates a displacement amount estimation mask image in which the reinforcement area of ​​the predetermined layer is extracted; an overcrowded reinforcement separation unit that generates an overcrowded reinforcement separated image by superimposing the edge detection image and the reinforcement area extraction image that form a pair, for each pair of the edge detection image and the reinforcement area extraction image generated from the same frontal transformation image, in which the boundary in the reinforcement area extraction image between the reinforcement area of ​​target reinforcement in the predetermined layer and the reinforcement area of ​​non-target reinforcement in a non-target layer different from the predetermined layer is separated by an edge in the edge detection image; a parallax mask reinforcement separation unit that generates a parallax mask reinforcement separated image in which the boundaries in the parallax mask image are separated by the edges by superimposing the edge detection image on the parallax mask image; a target reinforcement area extraction unit that generates a parallax mask reinforcement extraction image by superimposing the displacement estimation mask image on the parallax mask reinforcement separation image and removing from the parallax mask reinforcement separation image reinforcement areas that do not overlap with the reinforcement areas of the specified layer in the displacement estimation mask image, and extracts the reinforcement areas of the parallax mask reinforcement separation image that overlap with the reinforcement areas of the specified layer in the displacement estimation mask image as the reinforcement areas of the target reinforcement; a reinforcing bar line calculation unit that calculates the center line of the reinforcing bar area in the target reinforcing bar area extracted by the target reinforcing bar area extraction unit as a reinforcing bar line; a rebar spacing calculation unit that calculates the rebar spacing of the target reinforcing bar arrangement from the distance between the reinforcing bar lines; a reinforcing bar diameter calculation unit that superimposes the center line on the overcrowded reinforcement separation image, calculates the reinforcing bar diameter of the target reinforcement for each overcrowded reinforcement separation image from the reinforcing bar area of ​​the overcrowded reinforcement separation image where the center line is superimposed, and sets the reinforcing bar diameter of the target reinforcement that is the most frequently calculated of the calculated reinforcing bar diameters; A reinforcement inspection system equipped with

2. a rib edge removal unit that overlays the displacement amount estimation mask image on the edge detection image and removes the edges in overlapping areas of each reinforcement area in the edge detection image and the displacement amount estimation mask image as rib edges corresponding to ribs of reinforcing bars from each edge detection image, The overcrowded reinforcement separation unit separates the boundaries in the paired reinforcement area extraction images by the edges in the edge-detected images from which the rib edges have been removed, by superimposing each of the edge-detected images from which the rib edges have been removed on each of the paired reinforcement area extraction images. The reinforcing bar inspection system according to claim 1 .

3. the parallax mask reinforcement separation unit overlays the remaining edge detection images, for the number of remaining edge detection images not used in generating the parallax mask reinforcement separation image, on the parallax mask reinforcement extraction image, thereby separating the boundaries remaining in the parallax mask reinforcement extraction image for each of the remaining edge detection images using edges in the remaining edge detection images, and updating the parallax mask reinforcement separation image; The target reinforcement area extraction unit overlays the displacement estimation mask image on the updated parallax mask reinforcement separation image every time the parallax mask reinforcement separation image is updated, removes reinforcement areas in the displacement estimation mask image that do not overlap with the reinforcement area of ​​the specified layer from the updated parallax mask reinforcement separation image, and updates the parallax mask reinforcement extraction image, finally extracting the reinforcement area in the updated parallax mask reinforcement extraction image as the reinforcement area of ​​the target reinforcement.

3. The reinforcing bar arrangement inspection system according to claim 1 or 2.

4. 3. The reinforcement inspection system according to claim 1, wherein the target reinforcement area extraction unit does not extract the continuous area from the reinforcement area included in the parallax mask reinforcement separated image as the non-target reinforcement area when the ratio of the area of ​​the overlapping portion between the continuous area of ​​the reinforcement area included in the parallax mask reinforcement separated image and the reinforcement area of ​​the target reinforcement included in the displacement estimation mask image to the area of ​​the continuous area is equal to or less than a predetermined threshold.

5. a parallax mask image separating unit that separates the parallax mask image into a vertical line parallax mask image for the vertical lines and a horizontal line parallax mask image for the horizontal lines; a displacement amount estimation mask image separating unit that separates the displacement amount estimation mask image into a vertical streak displacement amount estimation mask image for the vertical streak and a horizontal streak displacement amount estimation mask image for the horizontal streak, the edge image generation unit generates, for each of the facing transformed images, a vertical edge-detected image in which edges of the vertical streaks are detected and a horizontal edge-detected image in which edges of the horizontal streaks are detected, as the edge-detected images; The overcrowded reinforcement separation unit generates an overcrowded reinforcement separation image by superimposing the paired vertical edge detection image and horizontal edge detection image generated from the same frontal transformation image and the reinforcement area extraction image on the reinforcement area extraction image, thereby separating the vertical reinforcement boundaries in the reinforcement area extraction image between the vertical reinforcement area of ​​the target reinforcement and the vertical reinforcement area of ​​the non-target reinforcement along the edges in the vertical edge detection image, and separating the horizontal reinforcement boundaries in the reinforcement area extraction image between the horizontal reinforcement area of ​​the target reinforcement and the horizontal reinforcement area of ​​the non-target reinforcement along the edges in the horizontal edge detection image, the parallax mask reinforcement separation unit generates a parallax mask vertical line separated image in which the vertical line boundaries in the vertical line parallax mask image are separated by the edges in the vertical line edge detected image by superimposing the vertical edge detected image on the vertical line parallax mask image, and generates a parallax mask horizontal line separated image in which the horizontal line boundaries in the horizontal line parallax mask image are separated by the edges in the horizontal line edge detected image by superimposing the horizontal edge detected image on the horizontal line parallax mask image, the target reinforcement area extraction unit overlays the vertical line displacement estimation mask image on the parallax mask vertical line separation image to generate a parallax mask vertical line extraction image by removing from the parallax mask vertical line separation image any vertical line line area that does not overlap with the vertical line line area of ​​the specified layer in the vertical line displacement estimation mask image, extracts the vertical line area of ​​the parallax mask vertical line separation image that overlaps with the vertical line area of ​​the specified layer in the vertical line displacement estimation mask image as the vertical line area of ​​the target reinforcement, overlays the horizontal line displacement estimation mask image on the parallax mask horizontal line separation image to generate a parallax mask horizontal line extraction image by removing from the parallax mask horizontal line separation image any horizontal line area that does not overlap with the horizontal line area of ​​the specified layer in the horizontal line displacement estimation mask image, and extracts the horizontal line area of ​​the parallax mask horizontal line separation image that overlaps with the horizontal line area of ​​the specified layer in the horizontal line displacement estimation mask image as the horizontal line area of ​​the target reinforcement, the reinforcing bar line calculation unit calculates the center line of the vertical reinforcement area in the target reinforcement extracted by the target reinforcement area extraction unit as the reinforcing bar line of the vertical reinforcement, and calculates the center line of the horizontal reinforcement area in the target reinforcement extracted by the target reinforcement area extraction unit as the reinforcing bar line of the horizontal reinforcement, The reinforcing bar spacing calculation unit calculates the vertical reinforcing bar spacing of the vertical reinforcing bars from the distance between the reinforcing bar lines of the vertical reinforcing bars, and calculates the horizontal reinforcing bar spacing of the horizontal reinforcing bars from the distance between the reinforcing bar lines of the horizontal reinforcing bars, The reinforcing bar diameter calculation unit superimposes the center line of the vertical reinforcement area on the overcrowded reinforcement separation image, calculates the reinforcing bar diameter of the vertical reinforcement in the target reinforcement arrangement for each of the overcrowded reinforcement separation images from the vertical reinforcement area of ​​the overcrowded reinforcement separation image where the center line of the vertical reinforcement area is superimposed, superimposes the center line of the horizontal reinforcement area on the overcrowded reinforcement separation image, calculates the reinforcing bar diameter of the horizontal reinforcement in the target reinforcement arrangement for each of the overcrowded reinforcement separation images from the horizontal reinforcement area of ​​the overcrowded reinforcement separation image where the center line of the horizontal reinforcement area is superimposed, and sets the reinforcing bar diameter of the vertical reinforcement and the horizontal reinforcement in the target reinforcement arrangement to the reinforcing bar diameter of each of the vertical reinforcement and the horizontal reinforcement in the target reinforcement arrangement that is most frequently calculated among the calculated reinforcing bar diameters of the vertical reinforcement and the horizontal reinforcement. The reinforcing bar inspection system according to claim 1 .

6. a rib edge removal unit that superimposes the vertical line displacement amount estimation mask image on the vertical edge-detected image, and removes from each of the vertical edge-detected images the vertical edges in overlapping areas of each vertical line region in the vertical edge-detected image and the vertical line displacement amount estimation mask image as rib edges corresponding to the vertical line ribs, and that superimposes the horizontal line displacement amount estimation mask image on the horizontal edge-detected image, and removes from each of the horizontal edge-detected images the horizontal edges in overlapping areas of each horizontal line region in the horizontal edge-detected image and the horizontal line displacement amount estimation mask image as rib edges corresponding to the horizontal line ribs, The overcrowded reinforcement separation unit separates the vertical reinforcement boundaries in each pair of reinforcement area extraction images by the vertical edges in each pair of the vertical edge detection images from which the rib edges have been removed, by superimposing each of the vertical edge detection images from which the rib edges have been removed on each pair of the reinforcement area extraction images, and separates the horizontal reinforcement boundaries in each pair of the reinforcement area extraction images by the horizontal edges in each pair of the horizontal edge detection images from which the rib edges have been removed, by superimposing each of the horizontal edge detection images from which the rib edges have been removed on each pair of the reinforcement area extraction images. The reinforcing bar inspection system according to claim 5 .

7. the parallax mask reinforcement separation unit overlays the remaining vertical edge detection images, for the number of remaining vertical edge detection images not used in generating the parallax mask vertical line separation image, on the parallax mask vertical line extraction image, to separate the vertical line boundaries remaining in the parallax mask vertical line extraction image for each of the remaining vertical edge detection images using edges in each of the remaining vertical edge detection images, thereby updating the parallax mask vertical line separation image; and overlays the remaining horizontal edge detection images, for the number of remaining horizontal edge detection images not used in generating the parallax mask horizontal line separation image, on the parallax mask horizontal line extraction image, to separate the horizontal line boundaries remaining in the parallax mask horizontal line extraction image for each of the remaining horizontal edge detection images, thereby updating the parallax mask horizontal line separation image; The target reinforcement area extraction unit overlays the vertical line displacement amount estimation mask image on the updated parallax mask vertical line separation image every time the parallax mask vertical line separation image is updated, removes vertical line areas in the vertical line displacement amount estimation mask image that do not overlap with the vertical line areas of the specified layer in the vertical line displacement amount estimation mask image from the updated parallax mask vertical line separation image to update the parallax mask vertical line extraction image, and finally extracts the vertical line areas in the updated parallax mask vertical line extraction image as the vertical line areas of the target reinforcement, and overlays the horizontal line displacement amount estimation mask image on the updated parallax mask horizontal line separation image every time the parallax mask horizontal line separation image is updated, removes horizontal line areas in the horizontal line displacement amount estimation mask image that do not overlap with the horizontal line areas of the specified layer in the horizontal line displacement amount estimation mask image from the updated parallax mask horizontal line separation image to update the parallax mask horizontal line extraction image, and finally extracts the horizontal line areas in the updated parallax mask horizontal line extraction image as the horizontal line areas of the target reinforcement.

7. The reinforcing bar arrangement inspection system according to claim 5 or 6.

8. The target reinforcement area extraction unit If the ratio of the area of ​​the overlapping portion of the continuous vertical line region included in the parallax mask vertical line separation image and the vertical line region of the target reinforcement included in the vertical line displacement estimation mask image to the area of ​​the continuous vertical line region is equal to or less than a predetermined threshold, the continuous vertical line region is not extracted as the non-target reinforcement region from the reinforcement region included in the parallax mask vertical line separation image, If the ratio of the area of ​​the overlapping portion of the continuous region of horizontal reinforcement areas included in the parallax mask horizontal reinforcement separation image and the horizontal reinforcement area of ​​the target reinforcement included in the horizontal reinforcement displacement estimation mask image to the area of ​​the continuous region of horizontal reinforcement areas is equal to or less than a predetermined threshold, the continuous region of horizontal reinforcement areas is not extracted from the reinforcement area included in the parallax mask horizontal reinforcement separation image as the non-target reinforcement area.

7. The reinforcing bar arrangement inspection system according to claim 5 or 6.

9. The edge image generation unit calculates the peak of the brightness gradient of each pixel in the direction perpendicular to the reinforcing bar line of the vertical or horizontal reinforcing bar for the reinforcing bar image, and among the calculated peaks, determines peaks whose prominence value is equal to or greater than a predetermined threshold as edge candidates. Furthermore, among these edge candidates, edge candidates whose prominence value is lower than the predetermined threshold are removed, thereby detecting the edges of the vertical or horizontal reinforcing bars that represent the outline of the reinforcing bar area of ​​the reinforcing bar.

7. The reinforcing bar arrangement inspection system according to claim 1, claim 2, claim 5 or claim 6.

10. 10. The bar arrangement inspection system according to claim 9, wherein the edge image generating unit performs straight line detection on all edge candidates of the vertical or horizontal bars using a probabilistic Hough transform.

Images