KR102295541B1 - Image judgment apparatus and weld inspection system - Google Patents

Abstract

The present invention relates to an image reading device and to a welding unit inspection system. According to one embodiment of the present invention, the image reading device includes: an image processing unit for processing the original image photographed for the welding unit of a structure to generate a processed composite image in a form in which the boundary is clear; and an image reading unit for inspecting whether there is a defect in the welding unit by using the composite image as an inspection image.

Description

Image judgment apparatus and weld inspection system

The present invention relates to an image reading apparatus and a welding part inspection system, and in more detail, an image reading apparatus and a welding part inspection for determining whether a defect has occurred in a welding part of a structure through an image photographed for a welding part welded to a structure It's about the system.

In the welding of structures, it is required to determine whether the welding state is defective. For example, the weld soundness of a SAW (Submerged Arc Welding) welded steel pipe is a major quality indicator that determines the quality level of the steel pipe. Therefore, in the case of SAW welded steel pipe, it is essential to inspect for defects in welds, and radiographic inspection is mainly used as the inspection method.

In the case of such a radiograph, in the case of the process of taking an image from a structure, the part related to the operation of the facility can be automated. However, in the process of judging whether welding is defective through a radiographic image taken, an approved non-destructive inspector directly reads and identifies the result, resulting in a decrease in productivity and human error.

Patent Document 1: Republic of Korea Patent Publication No. 10-2118809

An object of the present invention is to provide an image reading apparatus and a welding part inspection system for determining whether a defect has occurred in a welding part of a structure through an image taken for a welding part on which welding is performed on a structure.

In addition, an object of the present invention is to provide an image reading apparatus and a welding inspection system capable of processing an image taken of the structure of the present invention and determining whether a welding part is defective through the processed image, thereby improving the accuracy of the judgment.

According to an aspect of the present invention, there is provided an image processing unit comprising: an image processing unit for processing an original image photographed for a welded portion of a structure to generate a processed composite image in a form in which boundaries are clear; and an image reading unit configured to inspect whether there is a defect in the welding part by using the synthesized image as an inspection image.

In addition, the image processing unit may include: a boundary extractor configured to extract boundaries between different regions by deleting the original image as a basic image and a resolution-down image from the basic image; a nonlinearity adjustment unit for adjusting the nonlinearity of the boundary extraction image extracted by the boundary extraction unit; and an image synthesizing unit configured to generate a synthesized image by synthesizing the basic image with the boundary extraction image whose nonlinearity is adjusted by the nonlinearity adjusting unit.

In addition, the boundary extractor may downscale the base image by 1/2, upscale the downscaled image to correspond to the base image, and then low-pass filter it to generate a resolution down image.

In addition, the boundary extraction unit obtains a resolution-down image by using each of the images downscaled by a scale of ^{(1/2) x from the original image as a basic image, and deletes each resolution-down image from each basic image.} A boundary extraction image may be acquired.

Also, the image synthesizing unit may generate a synthesized image by performing synthesizing with each boundary extraction image while upscaling a basic image having the lowest resolution among the plurality of basic images in stages.

According to another aspect of the present invention, there is provided an image input unit provided to receive an original image to be read for examination; an image reading device for reading the original image and determining whether welding is defective; and an output unit for outputting a reading result, wherein the image reading apparatus includes: an image processing unit for processing the original image to generate a processed composite image in a form in which boundaries are clear; and an image reading unit for inspecting whether there is a defect in the welding part by using the composite image as the inspection image.

According to an embodiment of the present invention, there may be provided an image reading apparatus and a welding part inspection system for determining whether a defect has occurred in a welding part of a structure through an image photographed for a welding part on which welding is performed on a structure.

In addition, according to an embodiment of the present invention, an image reading apparatus and a welding part inspection system capable of processing an image taken of a structure and determining whether a welding part is defective through the processed image, thereby improving the accuracy of the judgment may be provided.

1 is a view showing a welding inspection system according to an embodiment of the present invention.
2 is a diagram showing the configuration of an image reading apparatus.
3 is a diagram illustrating a state in which linearity is changed according to a change in a value and a p value in the nonlinearity adjusting unit.
4 is a diagram illustrating an image reading apparatus according to another exemplary embodiment.

Hereinafter, an embodiment of the present invention will be described in more detail with reference to the accompanying drawings. Embodiments of the present invention may be modified in various forms, and the scope of the present invention should not be construed as being limited to the following embodiments. This embodiment is provided to more completely explain the present invention to those of ordinary skill in the art. Accordingly, the shapes of elements in the drawings are exaggerated to emphasize a clearer description.

1 is a view showing a welding inspection system according to an embodiment of the present invention.

Referring to FIG. 1 , a welding part inspection system 1 according to an embodiment of the present invention includes an image input unit 2 , an image reading device 3 , and an output unit 4 . The welding part inspection system 1 determines whether a defect has occurred in the welding part of the structure through the image taken for the welding part of the structure. The image may be a welding part radiography (X-ray) image taken for a portion on which welding is performed on a metal structure. The structure on which the inspection is performed may be a SAW (Submerged Arc Welding) welded steel pipe. A SAW welded steel pipe is a steel pipe made by forming a carbon steel sheet or a low alloy steel sheet in a circular shape and welding the seam in the SAW (Submerged Arc Welding) method. In general, steel pipes are provided with a thickness of 8 mm or more and a length of 4000 mm or more. After welding is performed on a steel pipe, it is inevitably required to inspect the integrity of the welded part. At this time, for the non-destructive inspection, an image is taken of the welded portion of the structure, and through the reading of the captured image, it is determined whether the welding is defective.

The image input unit 2 is provided to receive an original image to be read for examination. For example, the image input unit 2 may be an imaging device that performs radiographic imaging of a welded structure. Also, the image input unit 2 may be an input device capable of receiving image data captured by the photographing device in the form of data. For example, the image input unit 2 is provided as an input device provided to receive image data from an external storage medium such as a Universal Serial Bus (USB), a Solid State Drive (SSD), a Hard Disk Drive (HDD), or the like. , it may be provided as a communication device capable of receiving image data in the manner of wireless communication or wired communication.

The image reading device 3 reads the original image input through the image input unit 2 to determine whether welding is defective.

FIG. 2 is a diagram illustrating the configuration of an image reading apparatus, and FIG. 3 is a diagram illustrating a state in which linearity is changed according to changes in a value and p value in the nonlinearity adjusting unit.

2 and 3 , the image reading apparatus 3 includes an image processing unit 10 and an image reading unit 20 .

The image processing unit 10 processes the original image input from the image input unit 2 to generate a synthesized image. The image processing unit 10 includes a boundary extraction unit 110 , a nonlinearity adjustment unit 120 , and an image synthesis unit 130 .

The boundary extraction unit 110 deletes the resolution-down image from the basic image, and each different region in the image (eg, a base material region where welding is not performed in a steel plate used for forming a steel pipe, welding is performed and re-melting after melting) The boundary between the solidified weld area and the void contained within the weld area) is extracted. Specifically, the boundary extractor 110 uses the original image I ₀ ^{of 2 n} Х2 ⁿ input through the image input unit 2 as a basic image, low-pass-filters the basic image, and downscales 2 ^(n-1) Х2 ^(n-1) Obtain image I _(1/2). Thereafter, the downscaled image I _(1/2) is upscaled to correspond to the base image I ₀ , and low-pass filtering is performed to generate _{a resolution down image I′ 0 .} Thereafter, as shown in Equation 1, by using the original image I ₀ as the basic image, the resolution-down image I' _{0 is removed from the basic image I 0} to obtain the boundary extraction image Eo.

In addition, the boundary extraction unit 110 generates a resolution-down image by using each of the images downscaled by a scale of (1/2) ^x _{, (X is an integer), from the original image I 0 as a basic image, and} A boundary extraction image may be obtained by deleting each resolution-down image from the basic image. Specifically, ^{2 (n-1)} * 2 ^(n-1) image I _(1/2) downscaled by 1/2 from _{image I 0} is used as the base image, and the base image is low-pass filtered in the above-described manner. to generate a resolution down image I' _{(1/2) for image I (1/2)} by up-scale and low-pass filtering corresponding to 1/2 down-scale, base image, and _{then image I (1/2)} A boundary extraction image E _(1/2) is obtained by removing the resolution-down image I' _{(1/2) from .}

This process is ^{iteratively performed until the base image becomes I (1/2^(n-1))} of the 2*2 scale from _{I 0} , so that n base images can be obtained. And, for each basic image, n resolution down images I' ₀ ~ I' ^{(1/2^(n-1))} ^' is generated and each resolution-down image is controlled in each base image to obtain n ^{boundary extraction images Eo ~ E (1/2^(n-1) ).}

The nonlinearity adjustment unit 120 adjusts the nonlinearity of the boundary extraction image extracted by the boundary extraction unit 110 through Equation 2 below.

Here, x is the pixel value of the boundary extraction image E, and M is the maximum pixel value of the boundary extraction image E. a is an amplification coefficient, a factor limiting the pixel value to be within the dynamic range of the existing boundary extraction image, and corresponds to the slope of y(x). p is an index determining the nonlinearity of Equation 2 and has a value between 0 and 1. a has a value greater than 0 and less than or equal to 2. In Equation 2, the shape changes nonlinearly as shown in FIG. 3 according to a and p values. As the p-value increases and approaches 1, the graph form becomes more linear. It can be seen that the smaller the a, the smoother the graph, and the larger the a, the sharper the graph. Accordingly, the contrast of the boundary extraction image may be adjusted by adjusting the a and p values.

The image synthesizing unit 130 generates a synthesized image by combining the basic image with the boundary extraction image with non-linearity adjusted. Accordingly, the composite image becomes a form with clear boundaries (that is, a form with enhanced contrast), and the readability of the presence or absence of supply to the welded portion is improved. The image synthesizer 130 may generate a synthesized image by performing synthesizing with each boundary extraction image while stepwise upscaling a basic image having the lowest resolution among a plurality of basic images. Specifically, the image synthesizing unit 130 converts I ^{(1/2^(n-1)), which has the smallest scale among the basic images, to the boundary-extracted image E (1/2^()} having the same scale among the boundary-extracted images. ^{n-1) )} to obtain a synthesis stage image. Thereafter, the synthesis stage image is upscaled and low-pass filtered to obtain an image on one scale (ie, an image having the same scale as ^{I (1/2^(n-2))).} This is again combined with the boundary extraction image E ^{(1/2^(n-2) )} having the same scale to obtain an image of the synthesis stage of the next stage. The scale-up through upscaling and low-pass filtering and the synthesis of the boundary extraction image having the same scale as the scaled-up image are _{repeated until they become the same scale as the original image I 0} and then synthesized with the boundary extraction image Eo of the highest resolution. carry out Accordingly, all of the plurality of boundary extraction images are reflected, and a synthesized image with clear boundaries is obtained.

The image reading unit 20 inspects whether there is a defect in the welding part through the inspection image. The image reading unit 20 extracts the welding area from the inspection image, determines whether a void is located within the welding region, and determines whether there is a welding defect through the size of the void, the number of voids, etc. when it is determined that there is a void. is provided The synthesized image is provided as an inspection image, and may be used to determine whether or not welding is defective in a welding part.

The image reading unit 20 may be provided with artificial intelligence to enable learning in the process of performing an examination on a plurality of examination images. The image reading unit 20 may be an artificial intelligence implemented to enable deep learning based on a residual network (ResNet). The image reading unit 20 receives a plurality of synthesized images from the image processing unit 10 for one original image, uses each synthesized image as an inspection image to determine whether welding is defective, and compares each result with each other. have. Specifically, in Equation 2, the non-linearity adjusting unit 120 increases or decreases each of a value and p value stepwise, and the image processing unit 10 converts the resulting synthesized image to the image reading unit 20 . , and the image reading unit 20 may determine whether welding is defective by using each composite image as an inspection image. In addition, the original image provided in the initial learning process of the image reading unit 20 may be provided in a form in which a reading value, which is a result of a confirmed determination of whether there is a welding defect, is attached together. Accordingly, the image reading unit 20 may compare the determination result of the examination image with the reading provided together with the original image. And, when the determination result and the read value do not match, the a-value, p-value, and the combination of a-value and p-value used for synthesizing the provided synthetic image are checked, and in the process of determining the plurality of original images The results can be cumulative. Accordingly, a-value, p-value, and a combination of a-value and p-value, which are confirmed to have a large error value, are derived, and then, a-value, p-value, and a-value and p-value confirmed to have a large error value are used. A feedback may be given to the image processing unit 10 so as to skip generation of the synthesized image. Then, thereafter, the image processing unit 10 delays the generation of a composite image using the a-value, p-value, and a-value and p-value received feedback from the image reading unit 20 to a lower priority, or omits generation of the composite image. can

The output unit 4 outputs the result of reading by the image reading device 3 . Also, the output unit 4 may display and output an inspection image used for reading by the image reading device 3 and a portion determined as a welding defect in the inspection image. The output unit 4 may be provided as a display device, or as a data storage device such as HDD or SSD.

4 is a diagram illustrating an image reading apparatus according to another exemplary embodiment.

Referring to FIG. 4 , the image reading apparatus 3 includes image processing units 11 and 12 and an image reading unit 20 .

The image processing units 11 and 12 include a first image processing unit 11 and a second image processing unit 12 .

The first image processing unit 11 includes the boundary extraction unit 110 , the nonlinearity adjustment unit 120 , and the image synthesis unit 130 described above with reference to FIGS. 1 and 2 . The boundary extractor 110 , the non-linearity adjuster 120 , and the image synthesizer 130 generate a synthesized image, and the image reader 20 inspects the synthesized image provided by the first image processing unit 11 . Since the method of performing the inspection using an image is the same as described above, a repeated description will be omitted. Accordingly, the image reading unit 20 derives regions of a-value, p-value, and a-value and p-value having high reading accuracy in the form of a band, and then a-value, p-value, a-value, and p of the derived region Reading is performed using a composite image made of values as an inspection image, and whether or not welding is defective and accuracy verification of the reading result can be simultaneously performed through whether the read results are identical.

The second image processing unit 12 generates an auxiliary image by processing the original image in a different way from that of the first image processing unit 11 , and provides it to the image reading unit 20 . For example, the second image processing unit 12 converts the original image into one or two of Multi-Scale Contrast Enhancement (MCE), Histogram Equalization (HE), Adaptive Histogram Equalization (AHE), and Contrast Limited Adapted Histogram Equalization (CLAHE). The auxiliary image processed in more than one way may be provided to the image reading unit 20 .

The image reading unit 20 may additionally read whether welding is defective through the auxiliary image. In addition, the reading result provided by the first image processing unit 11 may be verified by comparing the reading result with a reading result obtained by using the composite image provided by the first image processing unit 11 as an inspection image. Specifically, in the process of the image reading unit 20 learning by using the original image provided with the read value, the judgment result through the synthesized image created for each band region of a value, p value, and a value and p value is determined. They may appear different from each other alternately. In this case, the image reading unit 20 determines whether the welding is defective through the auxiliary image, performs the determination through the auxiliary image made by one or more methods, and generates an auxiliary image with high accuracy when checking through the read value. and a-value, p-value, and band regions of a-value and p-value associated with the corresponding auxiliary image generating method and having high accuracy.

According to an embodiment of the present invention, an inspection image for determining whether welding is defective may be automatically performed. Conventionally, whether or not welding is defective through reading an inspection image of a welding part has been manually made by a person with certain qualifications. However, in the case of reading by a human, there are problems of a temporary reading error of a human, a possibility of an increase in the reading error due to an increase in fatigue due to a continuous reading operation, and a possibility of a difference in reading accuracy depending on the reading operator. On the other hand, in the case of the present invention, such a problem can be solved as such a reading operation is performed through artificial intelligence. In addition, the inspection image is processed prior to final reading and provided in a form with enhanced contrast, so that the accuracy of the reading result can be improved.

The above detailed description is illustrative of the present invention. In addition, the above description shows and describes preferred embodiments of the present invention, and the present invention can be used in various other combinations, modifications, and environments. That is, changes or modifications are possible within the scope of the concept of the invention disclosed herein, the scope equivalent to the written disclosure, and/or within the scope of skill or knowledge in the art. The written embodiment describes the best state for implementing the technical idea of the present invention, and various changes required in the specific application field and use of the present invention are possible. Accordingly, the detailed description of the present invention is not intended to limit the present invention to the disclosed embodiments. Also, the appended claims should be construed as including other embodiments.

2: image input unit 3: image reading device
4: output unit 10: image processing unit
20: image reading unit

Claims (6)

an image processing unit for processing the original image taken for the welded part of the structure to generate a processed composite image in a form in which the boundary is clear; and
An image reading unit for inspecting whether there is a defect in the welding part by using the composite image as an inspection image,
The image processing unit,
a boundary extraction unit for extracting boundaries between different regions by deleting the original image as a basic image and a resolution-down image from the basic image;
a nonlinearity adjustment unit for adjusting the nonlinearity of the boundary extraction image extracted by the boundary extraction unit; and
An image synthesizing unit for generating a synthesized image by synthesizing the boundary extracted image whose nonlinearity is adjusted by the nonlinearity adjusting unit with the basic image,
The boundary extractor downscales the base image by 1/2, upscales the downscaled image to correspond to the base image, and then low-pass filters to generate a resolution down image, but (1/2) ^{x greater than the original image.} A resolution-down image is obtained by using each image downscaled by a scale of (X is an integer) as a basic image, and a plurality of boundary extraction images are obtained by deleting each resolution-down image from each basic image,
The non-linearity adjusting unit adjusts the non-linearity according to the following equation,

(x is the pixel value of the edge-extracted image, M is the maximum pixel value of the boundary-extracted image, a is an amplification coefficient value greater than 0 and less than 2, and p is an index determining the nonlinearity of the equation and a value between 0 and 1 )
The image synthesizing unit generates the synthesized image by sequentially performing synthesizing with each of a plurality of boundary extraction images while stepwise upscaling a basic image having the lowest resolution among the plurality of basic images. delete delete delete delete an image input unit provided to receive an original image to be read for examination;
an image reading device for reading the original image and determining whether welding is defective; and
An output unit for outputting a reading result,
The image reading device,
an image processing unit for processing the original image to generate a processed composite image in a form in which boundaries are clear; and
An image reading unit for inspecting whether there is a defect in the welding part by using the composite image as an inspection image,
The image processing unit,
a boundary extraction unit for extracting boundaries between different regions by deleting the original image as a basic image and a resolution-down image from the basic image;
a nonlinearity adjustment unit for adjusting the nonlinearity of the boundary extraction image extracted by the boundary extraction unit; and
An image synthesizing unit for generating a synthesized image by synthesizing the boundary extracted image whose nonlinearity is adjusted by the nonlinearity adjusting unit with the basic image,
The boundary extractor downscales the base image by 1/2, upscales the downscaled image to correspond to the base image, and then low-pass filters to generate a resolution down image, but (1/2) ^{x greater than the original image.} A resolution-down image is obtained by using each image downscaled by a scale of (X is an integer) as a basic image, and a plurality of boundary extraction images are obtained by deleting each resolution-down image from each basic image,
The non-linearity adjusting unit adjusts the non-linearity according to the following equation,

(x is the pixel value of the edge-extracted image, M is the maximum pixel value of the boundary-extracted image, a is an amplification coefficient value greater than 0 and less than 2, and p is an index determining the nonlinearity of the equation and a value between 0 and 1 )
The image synthesizing unit, while upscaling a basic image having the lowest resolution among a plurality of basic images, sequentially performs synthesizing with each of a plurality of boundary extraction images to generate the synthesized image.

Images