JP2007024746A - Method and device for defect detection/discrimination inside conduit - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and a device for defect/discriminating inside a conduit greatly reducing the quantity of processing-requiring image data, and performing detection/discrimination of an internal defect of a sewerage pipe or the like. <P>SOLUTION: The defect/discriminating method inside the conduit comprises steps for: (a) acquiring a pipe-inside image by an imaging device loaded on a pipe-inside running device and having a fish-eye lens mounted thereon; (b) performing edge extraction by paying attention to a pipe joint, and then cutting out an image of a domain over the front and rear positions in the pipe axial direction of the pipe joint after masking an image out of a concerned domain by a mask prepared beforehand; and (c) dividing the cut out image, performing luminance conversion of each divided part, operating and calculating correlation between an average image and the individual image of each part, and determining that an image spot part showing less correlation than a fixed threshold has a defect. The device therefor is also provided. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a method for automatically inspecting the presence or absence of defects inside a pipe tub such as a sewer pipe, and an apparatus therefor.

  It is important to detect the presence and extent of defects in sewer pipes and tunnels at an early stage and to take the necessary measures appropriately to ensure and maintain the safety of the infrastructure. For example, when conducting an inspection to detect abnormalities such as precipitates, spring water, floats, cracks, etc. that occur on the inner wall of the tunnel, the inspector enters the tunnel, illuminates the inner wall with a light, and visually identifies the abnormal part. It depends on the method of detecting. In addition, a light and video are mounted on a vehicle traveling on a rail, and the situation of the inner wall surface is photographed and inspected with this image.

On the other hand, in order to acquire an image of a wide space such as the inside of a tunnel, as shown in FIG. 9, a large number of cameras are mounted on a carriage traveling on a rail and imaged, and the images are combined to crack the inner wall of the tunnel. Has been proposed (see, for example, Patent Document 1). In FIG. 9, 110 is a crack detection device, 120 is an inner wall surface, 121 is a tunnel, 130 is an imaging device, 131 is a camera, 132 is an illumination device, 133 is an image storage medium, 140 is a moving device, and 141 is a carriage. In the case of this prior art, as shown in FIGS. 10 and 11, the acquired image data is divided into small sections, subjected to density change processing, and further subjected to defect line feature processing. In this way, defect recognition is performed. 10 and 11, 134 is an image, 135 is a continuous image, 170 is a distribution map, 171 is a wide distribution map, RO is image data, and P1 to Pi are files. This method for detecting cracks in a tunnel is based on the premise that all data is analyzed, and it takes a lot of time to analyze continuous images.
JP 2001-141660 A

  As described above, when the invention disclosed in Patent Document 1 is used, all images captured by many cameras (four cameras in the process shown in FIGS. 1 to 3 in Patent Document 1) are recorded. There is a problem that it takes a lot of image processing time from the place where it needs to be processed, and lacks the real-time property of defect detection / determination.

  The present invention solves the above-mentioned problems in the prior art, significantly reduces the amount of image data required to be processed, and can detect and discriminate internal defects such as sewer pipes in real time. An object of the present invention is to provide a device.

In order to solve the above-mentioned problems, the invention according to claim 1 comprises: a. A step of acquiring an image of the inside of the tube with one imaging device mounted on the in-pipe traveling device that automatically travels inside the tube wirelessly or wiredly and equipped with a fisheye lens;
b. After extracting the edge focusing on the pipe joint where the luminance change is large, the image of the area extending in the tube axis direction before and after the pipe joint is cut out after masking the image outside the region of interest with a mask prepared in advance. Step,
c. The extracted image is divided, luminance image conversion of each divided part is performed, the correlation between the average image and the individual image part of each part is calculated and calculated, and a defect is found in the image part showing the correlation below a certain threshold A method for detecting and discriminating defects inside a tube with a step of determining in real time.

  According to a second aspect of the present invention, there is provided an in-pipe travel device that automatically travels inside a tube wirelessly or in a wired manner, and a single imaging device that is mounted on the in-pipe travel device and is movable in the direction of the tube axis. And after extracting the edge of the pipe seam where the luminance change is large in the original image acquired by the imaging device, the image of the area extending in the tube axis direction of the pipe seam is extracted with a mask prepared in advance. An image processing means that performs after masking an image outside the area and divides the clipped image and performs luminance conversion of each divided part, and calculates and calculates a correlation between the individual image of each part and the average image, and a fixed threshold value An apparatus for detecting and discriminating defects inside a tube with an arithmetic means for detecting and discriminating an image portion showing the following correlation as a defect.

  According to the present invention, since the inside of the sewer pipe or the like is picked up by a single camera equipped with a fisheye lens mounted on an in-pipe traveling apparatus that automatically travels, only one camera is required and the imaging region is changed. Therefore, a moving mechanism and / or a swinging mechanism are not required.

  In the case of sewer pipes, the defects are mostly present in pipe joints extending in the circumferential direction of the pipe. Thus, after extracting the edge of the pipe joint where the luminance change of the captured image is large, the image is cut out by masking the part other than the predetermined area in the pipe axis direction of the pipe joint. Combined with imaging using only one camera, the amount of image data required to be processed can be greatly reduced. Therefore, compared with the prior art, the image data processing time can be reduced to about 1/10, and a defect detection / discrimination system having real-time characteristics can be constructed.

  Furthermore, since an image of only a predetermined region in the tube axis direction of the tube seam portion is cut out and other regions are masked, in the core portion and the peripheral portion of the original image, which is unavoidable in the case of imaging with a fisheye lens. Distorted portions can be removed, and the problem of image distortion in imaging with a fisheye lens can also be solved. Thus, a simple in-pipe defect detection / discrimination system can be constructed.

  In the present invention, as the in-pipe traveling apparatus equipped with one camera equipped with a fisheye lens, the in-pipe traveling apparatus shown in FIGS. 7 and 8 proposed by the applicant in Japanese Patent Application No. 2005-018384 is preferably used. Can do. As shown in FIGS. 7 and 8, the in-pipe traveling device is pivotally supported by the frame 2 and rotating shafts 18 to 21 provided on the front right side, the front left side, the rear right side, and the rear left side of the frame 2. The first to fourth arms 3 to 6 that are rotatable in the horizontal plane and the first to fourth arms that are pivotally supported by one ends of the first to fourth arms 3 to 6 and are driven to rotate in the horizontal plane. The driving wheel 11 to 14 and the first and second springs 15 and 16 for applying a restoring force around the rotating shafts 18 to 21 to the first to fourth arms 3 to 6 are provided.

  The in-pipe travel device having the above-described configuration can freely travel on a curved pipe line, a pipe with a changing inner diameter, a pipe with a branch, or the like without using a complicated steering mechanism or an external control mechanism. In particular, it is useful for inspection equipment that inspects the interior of water and sewage pipes and gas pipes buried in the ground.

  The in-pipe imaging system according to the present invention includes a single camera mounted on the above-described in-pipe traveling device and equipped with a fisheye lens. The in-pipe travel device (robot) is equipped with one camera (imaging device) equipped with a fisheye lens, moves autonomously within a sewer such as a sewer pipe, and images the inside of the tube with a camera equipped with a fisheye lens. The image data is stored in a storage device such as a hard disk (not shown) provided in the in-pipe travel device (robot) body or transmitted to an external computer.

  Thus, the defect detection / discrimination device inside the tube of the present invention is an in-pipe travel device that automatically travels inside the tube wirelessly or by wire, and is mounted in the in-pipe travel device and is movable in the tube axis direction within the tube. After extracting an edge of a pipe joint portion having a large luminance change in the original image acquired by the image pickup apparatus and the original image acquired by the image pickup apparatus, an image of an area extending in the tube axis direction of the pipe joint portion is obtained. Image processing means for performing clipping after masking an image outside the region of interest with a mask prepared in advance, dividing the clipped image, and performing luminance conversion of each divided portion, and an individual image and an average image of each portion It comprises calculation means for calculating and calculating a correlation, and detecting and discriminating an image portion showing a correlation below a certain threshold as a defective portion.

  In the case of a sewer pipe or the like targeted by the present invention, most of the defects are present in the pipe joint extending in the circumferential direction of the pipe. Thus, an image inside a pipe such as a sewer pipe is obtained with a camera equipped with a fisheye lens, and edge extraction is performed by paying particular attention to the pipe joint where the luminance change is large. In other words, the pipe joint portion can be extracted by calculating the edge strength of the acquired image. Thereafter, an image of a predetermined region extending in the tube axis direction of the pipe joint portion, for example, 10 cm in the upstream direction of the pipe joint portion, 10 cm in the downstream direction, and a total of 20 cm is cut out. This is performed after masking with a mask prepared in advance. Increasing the predetermined region X across the pipe axis direction of the pipe joint portion makes it possible to detect and discriminate rough defects in a wide range, and reducing it makes it possible to detect and discriminate detailed defects in a narrow range. Thus, the predetermined region X extending in the tube axis direction of the pipe joint is appropriately selected as necessary. The masking can be performed on software in a computer.

  The image cut out in this way is panorama developed so as to easily detect the defective portion, and is subjected to the defect portion emphasis processing. The defect enhancement processing is performed by applying edge strength calculation → luminance image conversion → image smoothing to the panoramic image. Thereafter, the panoramic image is divided into arbitrary portions, and the correlation between the divided individual images and the average image is calculated and calculated, and it is determined that there is a defect in the image portion showing the correlation below a certain threshold.

  FIG. 1 shows a field of view when a target image is acquired by a camera equipped with a fisheye lens. As shown in FIG. 1, a 360 ° field of view of the fisheye lens can be an imaging target. 2 and 3 show a state in which a camera (imaging device) 32 having a fisheye lens 33 is mounted on the in-pipe traveling device 31 shown in FIGS. 7 and 8. 2 is a side view and FIG. 3 is a front view.

  The in-pipe travel device 31 moves autonomously in the tube and stores image data captured by the camera 32 in a storage device of a hard disk (not shown) disposed in the main body of the in-pipe travel device (robot) 31. Transmit to an external computer.

The image data acquired in this way is processed as follows.
Processing Procedure 1) Image Cutout A pipe joint portion having a particularly large luminance change is extracted from the acquired video image. Extraction of the pipe joint portion is performed by calculating the edge strength of the pipe joint portion. Next, a mask prepared in advance on the outer side of the core and the outer periphery of the annular portion having a width X shown in FIG. The annular portion corresponding to the region X is cut out. In the present invention, from the inventor's knowledge that most of the defects in the sewer pipes and the like are present in the joint portion of the pipe, the mask is applied to the input image outside the core and outer periphery of the annular portion having the width X. Thus, the region of the annular portion corresponding to the predetermined region X in the tube axis direction of the tube joint portion 34 is successfully extracted. When setting the mask, it is necessary and sufficient to analyze the image of the front and rear 20 cm (the region X in FIG. 4A) in the tube axis direction of the pipe joint portion 34, as shown in FIG. 4B. The mask is set so that areas other than the X region centering on the pipe joint 34 are masked.

Processing Procedure 2) Panorama Development In order to extract only information on the inner wall of the pipe in the vicinity of the pipe joint portion 34, as shown in FIG. 5, only the image of the hatched portion in FIG. 5 is extracted from the obtained cutout image. Region r in FIG. 5 is displayed as region r in the panoramic development view of FIG. In the panoramic development view, an X portion in FIG. 5 is displayed as a height X.

Processing procedure 3) Defective part emphasis processing A panoramic developed image is subjected to defective part emphasis processing by a process of edge intensity calculation → luminance image conversion → image smoothing so that the defective part in the tube can be detected more easily. In the present invention, the defect enhancement processing is performed based on the inventor's knowledge that a large luminance change has occurred in the defect.

a. Edge Strength Calculation When calculating the edge strength, a Laplacian operator, Sobel operator, Prewitt operator, or the like can be applied as an operator used for edge extraction.
b. Luminance image conversion The (R, G, B) value of each pixel in a panoramic image obtained as a color image is replaced with luminance information Y. In this embodiment, the conversion is performed by the algorithm Y = 0.3R + 0.6G + 0.1B.
c. Image smoothing Smoothes the luminance image using a Gaussian filter. This has the effect of removing noise and softening.

Processing Procedure 4) Discrimination of Defective Part The panoramic image subjected to the above processing is divided into arbitrary portions as shown in FIG. This division is set as appropriate in relation to the processing time and the position accuracy of the defective part. Here, it can be assumed that the sewer pipe having no defect has the same image even if any divided portion is cut out. Therefore, an average image of the divided part of the sewer pipe inner wall is obtained and used as a template image, and a correlation value with each divided part of the divided panoramic image is calculated by the following equation, for example.

ｔ_ｘ，ｔ_ｙ：取得された画像の幅および高さ
ｘ：原画像の水平軸における画素の座標

t _x , t _y : width and height of the acquired image x: pixel coordinates on the horizontal axis of the original image

The average image I _average is calculated by the following equation.

Ｎ＝９４２−５０，ｔ_ｘ＝５０，ｔ_ｙ＝５０
Ｎ：テンプレート画像作成に用いられた画素の数

N = 942-50, t _x = 50, t _y = 50
N: Number of pixels used to create the template image

  Thus, when there is an image division portion that shows a correlation that is equal to or less than a preset threshold value, it is determined that the portion is a defective portion. The threshold value is appropriately set according to the relationship between the defect detection rate (probability that it can be detected that there is a defect) and the false detection rate (probability that it is determined that the defect is not a defect). By being able to detect and discriminate internal defects such as sewer pipes in real time, the defective part is imaged in more detail on the spot, which helps to accurately determine the treatment to be performed.

  The method and apparatus for detecting and discriminating defects inside pipe culverts of the present invention greatly reduces the amount of image processing required when a defective part can be identified, such as a pipe joint where most defects occur in sewer pipes and the like. This is particularly effective in that the real-time property of defect detection / determination can be ensured. However, even when a defect occurrence site such as a tunnel cannot be specified, it is sufficient to acquire an image with only one camera (imaging device) equipped with a fisheye lens.

The top view which shows a visual field when imaging with the camera which attached the fisheye lens based on one Example of this invention The side view which shows the state which mounted the camera equipped with the fisheye lens based on one Example of this invention in the in-pipe travel apparatus (robot). The front view which shows the state which mounted the camera equipped with the fish-eye lens based on one Example of this invention in the in-pipe travel apparatus (robot). The schematic diagram which shows the predetermined area | region X in a pipe-axis direction of the pipe joint part which is an image cut-out area | region, such as a sewer pipe, based on one Example of this invention (a) is a front view (b) seen from the pipe-axis direction Schematic diagram showing image cutout area and mask area Schematic diagram showing an image cutout area X (shaded area), a mask area m, and a certain area r in the image cutout area in a panoramic development view according to an embodiment of the present invention. The schematic diagram which shows the image division | segmentation part in a panoramic development based on one Example of this invention The top view which shows the traveling apparatus (robot) in a pipe | tube based on one Example of this invention. The front view which shows the in-pipe travel apparatus (robot) based on one Example of this invention. Front view showing a conventional crack detection device for the inner wall of the tunnel Explanatory drawing which shows the relative position of the camera in the crack detection apparatus of the inner wall surface of a tunnel shown in FIG. Explanatory drawing which shows the condition of the image processing in the crack detection apparatus of the tunnel inner wall surface shown in FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 In-pipe traveling apparatus 2 Frame 3 1st arm 4 2nd arm 5 3rd arm 6 4th arm 7 1st drive motor 8 2nd drive motor 9 3rd drive motor 10 4th drive motor DESCRIPTION OF SYMBOLS 11 1st driving wheel 12 2nd driving wheel 13 3rd driving wheel 14 4th driving wheel 15 1st spring 16 2nd spring 17 Payload 18 Rotating shaft 19 Rotating shaft 20 Rotating shaft 21 times Driving shaft 22 Main pipe 31 In-pipe travel device (robot)
32 Camera (imaging device)
33 Fisheye Lens 34 Pipe Joint X X Predetermined Area in Pipe Axis Direction Including Pipe Joint r Area 110 Crack Detection Device 120 Internal Wall 121 Tunnel 130 Imaging Device 131 Camera 132 Illumination Device 133 Image Storage Medium 134 Image 135 Continuous Image 140 Moving Device 141 Bogie 170 Distribution map 171 Wide distribution map RO Image data P1 to Pi files

Claims (2)

a. A step of acquiring an image of the inside of the tube with an imaging device mounted on an in-pipe traveling device that automatically travels inside the tube wirelessly or by wire, and equipped with a fisheye lens;
b. After performing edge extraction paying attention to the pipe joint part, cutting out the image of the predetermined area across the pipe axis direction of the pipe joint part after masking the image outside the region of interest with a mask prepared in advance,
c. The extracted image is divided, luminance image conversion of each divided part is performed, the correlation between the average image and the individual image part of each part is calculated and calculated, and a defect is found in the image part showing the correlation below a certain threshold And a method for detecting and discriminating a defect inside the tube, characterized by comprising: determining in real time that there is. An in-pipe travel device that automatically travels inside the tube wirelessly or by wire, an imaging device that is mounted on the in-pipe travel device and is movable in the tube axis direction, and that is equipped with a fisheye lens, and an original image acquired by the imaging device After extracting the edge of the pipe seam, the image of the predetermined area across the pipe axis direction of the pipe seam is cut out after masking the image outside the region of interest with a mask prepared in advance and the cut image Image processing means that performs luminance conversion of each divided part, and calculates and calculates the correlation between the individual image of each part and the average image, and detects and discriminates an image portion that shows a correlation below a certain threshold as a defective part An apparatus for detecting and discriminating defects inside the tube, characterized by comprising:

Images