CN117036356A - Welded pipe weld quality detection method based on artificial intelligence - Google Patents
Welded pipe weld quality detection method based on artificial intelligence Download PDFInfo
- Publication number
- CN117036356A CN117036356A CN202311300696.5A CN202311300696A CN117036356A CN 117036356 A CN117036356 A CN 117036356A CN 202311300696 A CN202311300696 A CN 202311300696A CN 117036356 A CN117036356 A CN 117036356A
- Authority
- CN
- China
- Prior art keywords
- extremely bright
- point
- extremely
- image
- steel pipe
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 29
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 21
- 238000003466 welding Methods 0.000 claims abstract description 102
- 229910000831 Steel Inorganic materials 0.000 claims abstract description 73
- 239000010959 steel Substances 0.000 claims abstract description 73
- 230000002146 bilateral effect Effects 0.000 claims abstract description 12
- 238000001914 filtration Methods 0.000 claims abstract description 12
- 238000013528 artificial neural network Methods 0.000 claims abstract description 6
- 238000000034 method Methods 0.000 claims description 34
- 238000003708 edge detection Methods 0.000 claims description 6
- 238000004891 communication Methods 0.000 claims description 4
- 230000007547 defect Effects 0.000 abstract description 16
- 238000012545 processing Methods 0.000 abstract description 4
- 230000000694 effects Effects 0.000 description 11
- 230000035772 mutation Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 4
- 238000009499 grossing Methods 0.000 description 4
- 238000005286 illumination Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 3
- 230000005856 abnormality Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000001816 cooling Methods 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 239000002184 metal Substances 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 239000000758 substrate Substances 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000002844 melting Methods 0.000 description 1
- 230000008018 melting Effects 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000007789 sealing Methods 0.000 description 1
- 239000002893 slag Substances 0.000 description 1
- 238000004544 sputter deposition Methods 0.000 description 1
- 231100000331 toxic Toxicity 0.000 description 1
- 230000002588 toxic effect Effects 0.000 description 1
- 230000010415 tropism Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30152—Solder
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
Abstract
The application relates to the field of defect detection processing, in particular to a welded pipe welding seam quality detection method based on artificial intelligence, which comprises the steps of collecting a steel pipe welding image; obtaining a bright point in a steel pipe welding image; for each extremely bright point of the steel pipe welding image, obtaining the noise possibility of an extremely bright region where the extremely bright point is located according to the pixel point sequence of the extremely bright point in each direction; obtaining inter-domain non-correlation of the extremely bright areas where the extremely bright points are located according to the main direction and the auxiliary direction of the extremely bright points of the steel pipe welding image and the distribution condition of the surrounding connected domains of the extremely bright points; obtaining the tendency of each pixel point according to the noise point possibility and inter-domain non-correlation of each pixel point of the steel pipe welding image; and (3) improving a bilateral filtering algorithm according to the trend of each pixel point of the steel pipe welding image to obtain a denoised steel pipe welding image, and realizing the quality detection of the welded pipe welding seam through a neural network. Therefore, the denoised steel pipe welding image is clearer, and the quality detection of the welded pipe welding seam is convenient to realize.
Description
Technical Field
The application relates to the field of defect detection processing, in particular to a welded pipe welding seam quality detection method based on artificial intelligence.
Background
The pipeline transportation has wide application, such as cross-region transportation of natural gas, petroleum and the like, sealing transportation of toxic and inflammable substances and the like. The pipelines are often connected by welding. Welding abnormality occurs during welding for many reasons, for example, welding dislocation occurs during welding due to irregular edges of the pipeline or deformation of the pipeline; due to insufficient welding current, the welder is not skilled in the technique, and the situation of incomplete welding occurs; because the melted welding material can not completely fill the welding seam during welding, defects such as undercut, air holes and the like appear; the surface of the steel is not cleaned before welding, sundries are mixed in a welding melting area of the welding material during welding, and slag inclusion occurs.
These defects may cause a decrease in the sealability or internal bearing capacity of the steel pipe, thereby affecting the safety in use and the service life of the steel pipe. Therefore, the quality detection of the welding seam is required to be carried out on the condition after the welding is finished, so as to ensure that the welding seam of the steel pipe welding put into use meets the quality requirement. The traditional detection method can finish detection of welding abnormality in an image detection mode, but the acquired picture is affected by illumination, shadow, reflection and the like, so that defect identification is not facilitated.
In summary, the application provides a welded pipe weld quality detection method based on artificial intelligence, which is characterized in that noise points in a steel pipe welding image are analyzed by collecting the steel pipe welding image, and the denoised steel pipe welding image is obtained by improving a bilateral filtering algorithm, so that the welded pipe weld quality detection is completed.
Disclosure of Invention
In order to solve the technical problems, the application provides an artificial intelligence-based welded pipe weld quality detection method to solve the existing problems.
The application discloses an artificial intelligence-based welded pipe weld quality detection method, which adopts the following technical scheme:
one embodiment of the application provides a welded pipe weld quality detection method based on artificial intelligence, which comprises the following steps:
collecting a steel pipe welding image;
obtaining a bright point in a steel pipe welding image; for each extremely bright point of the steel pipe welding image, acquiring a pixel point sequence of the extremely bright point in each direction; obtaining edge points of the extremely bright areas in all directions according to pixel point sequences of the extremely bright points in all directions; obtaining the noise possibility of the extremely bright area where the extremely bright point is located according to the distribution of the extremely bright area edge points of the extremely bright point in all directions;
dividing the steel pipe welding image into an upper part image and a lower part image; for the upper part image, acquiring a binary image of the upper part image by adopting an edge detection algorithm; calculating the main direction of the binary image; for each extremely bright point of the binary image, taking the direction which passes through the extremely bright point and is perpendicular to the main direction as the auxiliary direction of the extremely bright point, and obtaining the edge points of each direction of the extremely bright point according to the main direction and the auxiliary direction of the extremely bright point; obtaining inter-domain non-correlation of the extremely bright region where the extremely bright point is located according to the distribution condition of the communication region where the edge point of each direction of the extremely bright point is located; acquiring inter-domain non-correlation of a very bright area where each very bright point of the lower part image is positioned;
obtaining the trend of the noise point connected domain of the extremely bright area where the extremely bright point is located according to the noise point possibility of the extremely bright area where the extremely bright point is located and the inter-domain non-correlation; the method comprises the steps of obtaining the orientation of each pixel point of a steel pipe welding image; and (3) improving a bilateral filtering algorithm according to the trend of each pixel point of the steel pipe welding image to obtain a denoised steel pipe welding image, and realizing the quality detection of the welded pipe welding seam through a neural network.
Preferably, the obtaining the extremely bright point in the steel pipe welding image includes:
and for each pixel point of the steel pipe welding image, if the gray value of the pixel point is larger than the gray values of all the pixel points in the neighborhood, the pixel point is marked as an extremely bright point.
Preferably, the obtaining the pixel point sequence of the extremely bright point in each direction includes:
the method comprises the steps of taking an extremely bright point as a center, extending towards eight neighborhood directions, and selecting N pixel points in each direction to form a pixel point sequence of each direction of the extremely bright point, wherein N is a preset parameter.
Preferably, the obtaining edge points of the extremely bright area in each direction according to the pixel point sequence of the extremely bright points in each direction includes:
and for the pixel point sequences in all directions of the extremely bright points, calculating the gray value difference absolute value of each adjacent pixel point in the pixel point sequences, and taking the next pixel point in the adjacent pixel points with the gray value difference absolute value larger than a first threshold value as the edge point of the extremely bright area in the direction of the pixel point sequences.
Preferably, the obtaining the noise probability of the extremely bright area where the extremely bright point is located according to the distribution of the edge points of the extremely bright area where the extremely bright point is located in each direction includes:
the Euclidean distance between the extremely bright point and the extremely bright region edge point in each direction is obtained, the average value of the Euclidean distances of the extremely bright region edge points in all directions of the extremely bright point is calculated, the sum of the squares of the difference values of the Euclidean distances of the extremely bright region edge points in all directions of the extremely bright point and the average value is calculated, and the reciprocal of the product of the sum and the average value is used as the noise possibility of the extremely bright region where the extremely bright point is located.
Preferably, the calculating the main direction of the binary image includes:
and performing straight line fitting on each edge straight line of the binary image to obtain each fitted straight line, taking the included angle between each fitted straight line and the rightward direction of the horizontal direction as the approximate direction of each edge straight line, and taking the approximate direction average value of all the edge straight lines of the binary image as the main direction of the binary image.
Preferably, the obtaining the edge points of the extreme light point in each direction according to the main direction and the auxiliary direction of the extreme light point includes:
and marking the edge points of other connected areas except the connected area where the extreme light point is located, which are firstly contacted by the straight line passing through the main direction and the auxiliary direction of the extreme light point, as the edge points of all directions of the extreme light point.
Preferably, the obtaining the inter-domain non-correlation of the extremely bright area where the extremely bright point is located according to the distribution situation of the connected domain where the extremely bright point is located and the connected domain where the edge point in each direction of the extremely bright point is located includes:
for edge points in all directions of the extremely bright point, calculating Euclidean distances between the edge points and the extremely bright point, calculating Euclidean distance average values of the Euclidean distances of all the edge points of the extremely bright point, and taking the square of the difference between the Euclidean distances of the edge points and the Euclidean distance average value as an index of an exponential function taking a natural constant as a base;
and calculating the product of the square of the difference value of the area of the connected domain where the edge point is positioned and the connected domain where the extremely bright point is positioned and the exponential function, and taking the average value of the product of the edge points in all directions of the extremely bright point as the inter-domain uncorrelation of the extremely bright area where the extremely bright point is positioned.
Preferably, the obtaining the noise connected domain trend of the extremely bright area where the extremely bright point is located according to the noise probability and the inter-domain non-correlation of the extremely bright area where the extremely bright point is located includes:
the noise connected domain trend of the extremely bright region where the extremely bright point is located is the product of the noise probability of the extremely bright region where the extremely bright point is located and the inter-domain uncorrelation.
Preferably, the acquiring the trend of each pixel point of the steel pipe welding image includes:
for each pixel point of the steel pipe welding image, when the pixel point is in an extremely bright area in the binary image, taking the trend of a noise point connected area in the extremely bright area as the trend of the pixel point; when the pixel is not in the extremely bright area in the binary image, the tendency of the pixel is marked as 0.
The application has at least the following beneficial effects:
the method analyzes the shape and the size of the interior of the extremely bright area where the extremely bright point is positioned based on the regular punctiform distribution of the noise points, judges the possibility that the extremely bright point is the noise point, and can eliminate the interference of the area without the punctiform distribution characteristic; meanwhile, the position distribution situation among the extremely bright areas where the extremely bright points are located is calculated, and inter-domain non-correlation among the extremely bright areas where the extremely bright points are located is obtained according to the degree of correlation among the connected domains, so that the characteristic of discrete distribution of the noise points can be effectively integrated into the noise point judgment index;
by combining the two indexes, when the positions of the pixel points are close to an isolated extremely-bright area with smaller area, the modified Gaussian parameters are larger, the smoothing effect of the Gaussian function is stronger, so that the denoising effect of bilateral filtering on the pixel points is better, the denoised steel pipe welding image is clearer, defects in the steel pipe welding image such as cracks, air holes and other lines are more consistent, the edges are more complete, and the steel pipe welding image is easier to be recognized by a machine.
Drawings
In order to more clearly illustrate the embodiments of the application or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the application, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a welded pipe weld quality detection method based on artificial intelligence;
fig. 2 is an image of fine material inclusions in a weld zone.
Detailed Description
In order to further describe the technical means and effects adopted by the application to achieve the preset aim, the following detailed description is given below of the artificial intelligence-based welding seam quality detection method for welded pipes according to the application, which is based on the specific implementation, structure, characteristics and effects thereof, with reference to the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
The following specifically describes a specific scheme of the welding line quality detection method of the welded pipe based on artificial intelligence provided by the application with reference to the accompanying drawings.
The embodiment of the application provides an artificial intelligence-based welded pipe weld quality detection method.
Specifically, the following method for detecting the quality of welded pipe weld based on artificial intelligence is provided, referring to fig. 1, the method comprises the following steps:
and S001, collecting a steel pipe welding image.
In this embodiment, the defect edge detection technique is used to extract the defect edge of the welded pipe surface. Firstly, a CCD camera is adopted to collect welding images of the steel pipes, the smooth surfaces of the steel pipes are extremely easy to generate reflection influence, and an image enhancement algorithm is adopted to remove illumination reflection influence. The image enhancement algorithm in this embodiment adopts the Retinex theory, which is a known technique, and is not described in detail in this embodiment.
Therefore, the acquisition of the steel pipe welding image can be completed through the method, and the steel pipe welding image can be conveniently analyzed further.
And step S002, analyzing noise points in the steel pipe welding image, and obtaining the denoised steel pipe welding image by improving a bilateral filtering algorithm.
Because the sputtering of the metal solution in the welding process can form a plurality of tiny metal balls on the surface of the steel pipe, as shown in fig. 2, the tiny areas on the surface of the steel pipe are raised due to the inclusion of tiny substances on the surface in the welding area, and under the illumination condition, a plurality of obvious noise points can be generated by light reflection, which is not beneficial to edge detection, thereby influencing the accuracy of defect identification.
One part of the extremely bright spots in the steel pipe welding image is a reflection area generated by air holes or crack edge pits, and the area is a normal reflection area with larger area; the other part is a reflection area generated by the tiny bulges, and the reflection area is a noise point and can influence the defect identification of the image. Accordingly, whether the bright spot is a noise spot or not is analyzed, so that the denoising effect of bilateral filtering is effectively improved, and the quality detection of the welding seam of the welding pipe is facilitated.
In the whole steel pipe welding image, local areas are extremely bright, the extremely bright areas can be noise points, each pixel point in the steel pipe welding image is taken as a central pixel point, and 3*3 neighborhood is adopted to calculate the gray value change of the central pixel point and the neighborhood pixel points, so that the position of the extremely bright point is judged.
And comparing the gray values of the neighborhood pixels of each pixel of the steel pipe welding image, if the gray value of the central pixel in the neighborhood is larger than the gray value of all pixels in the neighborhood, marking the bright point, and recording the coordinates of the bright point.
For each of the extreme bright spots in the steel pipe welding image, one of the extreme bright spots is taken as an example for analysis.
The extremely bright points are taken as starting points and extend to eight directions of 0 degree, 45 degree, 90 degree, 135 degree, 180 degree, 225 degree, 270 degree and 315 degree respectively, N pixel points are selected from each direction to form a pixel point sequence of each direction, wherein N is a preset parameter, and the embodiment takes a verification value of 50, and can be set by an implementer.
Taking the 0-degree direction as an example, calculating the gray difference value of each adjacent pixel point in the sequence by the pixel point sequence in the direction; if the pixel point is still in the extremely bright area, the change amount of the adjacent gray value is small, if the change amount of the gray value is suddenly increased, the gray value of the adjacent pixel point is changed severely, the position is the edge of the extremely bright area where the extremely bright point is located, and the pixel point is taken as the edge point of the extremely bright area where the extremely bright point is located in the direction. The judgment condition for the sudden increase of the gray value variation is as follows: the absolute value of the gray value difference between two adjacent pixel points is larger than the first threshold, and the first threshold is taken to be an empirical value 10 according to the embodiment, so that an implementer can set the first threshold by himself.
For each extremely bright point, the Euclidean distance between the extremely bright point and the edge point of the extremely bright area in each direction is obtained and is recorded as the abrupt change distanceThe mutation distances in all directions of the extremely bright point are averaged to obtain a mutation mean value r, so that the possibility that the extremely bright area where the extremely bright point is a noise point is obtained by calculation, and a specific calculation formula is as follows:
；
in the method, in the process of the application,is a very bright spot->Noise probability in extremely bright region, +.>Is a very bright spot->Number of mutation distances,/->Is a very bright spot->Mutation mean of mutation distance in all directions, < >>Is a very bright spot->Mutation distance in the i-th direction.
It should be noted that the number of the substrates,the smaller the difference between each abrupt change distance and the average value, the more similar the abrupt change distance of the extremely bright point in all directions is, namely the extremely bright area where the extremely bright point is located is more likely to be in regular point-shaped distribution. />The smaller the distance from the extremely bright point to the edge in each direction, the smaller the area of the extremely bright region, which is more suitable for the feature of smaller noise point, and the higher the possibility that the extremely bright region where the extremely bright point is located is the noise point.
And traversing all the extremely bright points of the steel pipe welding image, and carrying out noise probability of the extremely bright areas where the extremely bright points are positioned in the mode.
Arc-shaped waves appear after welding and cooling, the reflective areas of the waves are not connected, so that continuous small extremely bright areas appear, the areas can be calculated as noise points, and therefore whether the extremely bright areas are connected or not needs to be judged, and if the connection between one extremely bright area and other extremely bright areas is larger, the possibility that the extremely bright area is the noise point is smaller; the smaller the association, the greater the likelihood of being a noise point.
Because the weld joint in the steel pipe welding image has symmetrical properties up and down and approximately trends on the upper side and the lower side of the weld joint are relatively different, the steel pipe welding image is divided into an upper part image and a lower part image, the upper part image is analyzed, and the lower part image is consistent with the analysis method of the upper part image.
And (3) obtaining a binary image of the upper part image by adopting a Canny operator edge detection algorithm on the upper part image, wherein part of edge lines are irregular lines, so that each edge line in the binary image is subjected to line fitting by adopting a least square method, and an included angle between the fitted line and a horizontal rightward direction is used as an approximate direction of each edge line, wherein the Canny operator edge detection algorithm and the least square method are known techniques, and are not repeated herein.
And averaging all the edge straight line approximate directions in the binary image of the upper part of the image to obtain a main direction. The connected domain surrounded by all edges in the binary image comprises all extremely bright areas of the steel pipe welding image, and as extremely bright points of the extremely bright areas are known, the association degree between the connected domain where the extremely bright point is located and other connected domains is calculated for a certain extremely bright point in the binary image.
To be extremely brightFor example, in the binary image, the extreme bright point +.>The direction perpendicular to the main direction is taken as the extreme bright point +.>The secondary direction of the main direction, along the passing extreme bright point +.>The straight lines of the main direction and the auxiliary direction respectively find other four connected domains closest to the extremely bright area where the extremely bright point is located, the coordinates of the edge points of the connected domains which are found in the four directions and are firstly contacted in the main direction and the direction perpendicular to the main direction are respectively recorded as +.>、/>、/>、/>。
Counting extremely bright spots by taking the number of pixel points contained in the connected domain as the area of the connected domainAnd the areas of the four direction communicating domains are +.>、/>、/>、/>、/>Very bright spot->The Euclidean distance to the nearest edge point of the other four direction communicating domains is +.>、/>、/>、/>、/>。
Calculating an extreme bright spotEuclidean distance mean value of nearest edge points of other four-direction connected domains +.>Then->The specific expression of inter-domain non-correlation between the dot communicating domain and other communicating domains is:
；
in the method, in the process of the application,is a very bright spot->Inter-domain uncorrelation in the very bright region,/->Is a very bright spot->Number of directions in the main direction and perpendicular to the main direction, +.>Is a very bright spot->The area of the connected domain in the i-th direction, +.>Is a very bright spot->The area of the communicating region is->Is an exponential function based on a natural constant e, +.>Is a very bright spot->The nearest edge point of the i-th direction communicating domain, < >>Is a very bright spot->The Euclidean distance mean value of the nearest edge point of the connected domain in other four directions.
It should be noted that the number of the substrates,representation->Point connected domain and->Difference of areas of connected domains where points are located, +.>Is->Point to->Difference in point distance from average distance. The larger the area difference value is, the different the sizes of the two connected domains are, the lower the similarity is, and the higher the uncorrelation is; the greater the distance difference, the farther the two communicating regions are from each other, the less likely the non-correlation is on the same or adjacent weld cooling corrugations. The higher the non-correlation, i.e. +.>The more likely the point is a discrete noise point in the extremely bright region.
Repeating the steps to obtain the inter-domain uncorrelation of the area where each extremely bright point of the lower part image in the steel pipe welding image is located.
The two indexes respectively illustrate the self attribute of the noise point about the shape and the small-range locality characteristic of the noise point about the surrounding distribution condition, and the noise point can be evaluated jointly from two distinct layers. Accordingly, the two indexes are combined to obtain the extremely bright pointThe calculation formula of the tendency of the extremely bright region as the noise point connected region is as follows:
；
in the method, in the process of the application,is a very bright spot->Noise point connected domain tendencies of the located extremely bright region,>is a very bright spot->Noise probability of the extremely bright region, +.>Is a very bright spot->Inter-domain uncorrelation of the very bright areas.
When the light is extremely brightWhen the noise probability and inter-domain uncorrelation of the extremely bright region are large, i.e. the more likely the extremely bright region is a noise regular punctual region, the farther the region is from other surrounding regions, the less similar features are provided, i.e. the more likely the extremely bright region is a noise region, i.e. the noise is continuousThe greater the general domain tropism.
Judging whether each pixel point in the steel pipe welding image is positioned in a certain extremely bright region connected domain in the binary image, and if so, taking the noise point connected domain trend of the extremely bright region connected domain as the noise point trend K of the pixel point; if the pixel point is not in the communication domain of the extremely bright area, the tendency K of the pixel point is 0. And carrying out the same processing on all the pixel points in the gray level graph, so that each pixel point corresponds to one trend K.
For each pixel point of the steel pipe welding image, parameters of Gaussian function used by the pixel point in bilateral filteringThe larger the Gaussian function is, the smaller the peak of the Gaussian function is, the better the smoothing effect on the gray value is, and the modified Gaussian parameter is obtained by combining the trend K of each pixel point>The method comprises the following steps:
；
in the method, in the process of the application,gaussian parameters corrected for each pixel point, < ->Gaussian parameters before correction for each pixel point, < ->Is an exponential function based on a natural constant e, +.>Is the orientation of each pixel point.
When the K value is larger, the position of the pixel point is closer to an isolated extremely bright area with smaller area, and the corrected Gaussian parameterThe larger the value of (c), the stronger the smoothing effect of the gaussian function, and the better the denoising effect of the bilateral filtering.
By means of changed parametersAnd carrying out bilateral filtering calculation, denoising the steel pipe welding image, and obtaining a clearer steel pipe welding image after denoising.
And S003, performing defect edge extraction on the denoised steel pipe welding image by adopting a self-coding neural network.
The obtained steel pipe welding image after denoising is clearer, no noise point interference exists in the image, defects in the image such as cracks, air holes and other lines are more coherent, the edge is more complete, and the steel pipe welding image is easier to identify by a machine.
Inputting the denoised pipeline welding images into a self-coding neural network to extract the defect edges, and obtaining the defect edges in each pipeline welding image. The self-coding neural network is a known technology, and the embodiment is not described in detail.
And counting the number of pixel points in all the defect edges, and judging that the welding seam effect of the steel pipe welding is poor when the number of pixel points is greater than T times of the number of all the pixel points of the steel pipe welding image. In this embodiment, the empirical value of T is 0.01, which can be set by the practitioner.
So far, the quality detection of the welded seam of the welded pipe can be finished by the method.
In summary, the embodiment of the application provides a welded pipe welding seam quality detection method based on artificial intelligence, which is implemented by collecting a steel pipe welding image, analyzing noise points in the steel pipe welding image, and obtaining a denoised steel pipe welding image by improving a bilateral filtering algorithm to finish the welded pipe welding seam quality detection.
According to the embodiment of the application, based on the regular punctiform distribution of the noise points, the shape and the size of the interior of the extremely bright area where the extremely bright points are positioned are analyzed, the possibility that the extremely bright points are the noise points is judged, and the interference of the area without the punctiform distribution characteristics can be eliminated; meanwhile, the position distribution situation among the extremely bright areas where the extremely bright points are located is calculated, and inter-domain non-correlation among the extremely bright areas where the extremely bright points are located is obtained according to the degree of correlation among the connected domains, so that the characteristic of discrete distribution of the noise points can be effectively integrated into the noise point judgment index;
by combining the two indexes, when the positions of the pixel points are close to an isolated extremely-bright area with smaller area, the modified Gaussian parameters are larger, the smoothing effect of the Gaussian function is stronger, so that the denoising effect of bilateral filtering on the pixel points is better, the denoised steel pipe welding image is clearer, defects in the steel pipe welding image such as cracks, air holes and other lines are more consistent, the edges are more complete, and the steel pipe welding image is easier to be recognized by a machine.
It should be noted that: the sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. And the foregoing description has been directed to specific embodiments of this specification. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and the same or similar parts of each embodiment are referred to each other, and each embodiment mainly describes differences from other embodiments.
The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; the technical solutions described in the foregoing embodiments are modified or some of the technical features are replaced equivalently, so that the essence of the corresponding technical solutions does not deviate from the scope of the technical solutions of the embodiments of the present application, and all the technical solutions are included in the protection scope of the present application.
Claims (9)
1. The welded pipe weld quality detection method based on artificial intelligence is characterized by comprising the following steps of:
collecting a steel pipe welding image;
obtaining a bright point in a steel pipe welding image; for each extremely bright point of the steel pipe welding image, acquiring a pixel point sequence of the extremely bright point in each direction; obtaining edge points of the extremely bright areas in all directions according to pixel point sequences of the extremely bright points in all directions; obtaining the noise possibility of the extremely bright area where the extremely bright point is located according to the distribution of the extremely bright area edge points of the extremely bright point in all directions;
dividing the steel pipe welding image into an upper part image and a lower part image; for the upper part image, acquiring a binary image of the upper part image by adopting an edge detection algorithm; calculating the main direction of the binary image; for each extremely bright point of the binary image, taking the direction which passes through the extremely bright point and is perpendicular to the main direction as the auxiliary direction of the extremely bright point, and obtaining the edge points of each direction of the extremely bright point according to the main direction and the auxiliary direction of the extremely bright point; obtaining inter-domain non-correlation of the extremely bright region where the extremely bright point is located according to the distribution condition of the communication region where the edge point of each direction of the extremely bright point is located; acquiring inter-domain non-correlation of a very bright area where each very bright point of the lower part image is positioned;
obtaining the trend of the noise point connected domain of the extremely bright area where the extremely bright point is located according to the noise point possibility of the extremely bright area where the extremely bright point is located and the inter-domain non-correlation; the method comprises the steps of obtaining the orientation of each pixel point of a steel pipe welding image; according to the trend of each pixel point of the steel pipe welding image, a bilateral filtering algorithm is improved to obtain a denoised steel pipe welding image, and the quality detection of welded pipe welding seams is realized through a neural network;
the method for obtaining inter-domain non-correlation of the extremely bright area where the extremely bright point is located according to the distribution situation of the communication area where the edge point of each direction of the extremely bright point is located comprises the following steps:
for edge points in all directions of the extremely bright point, calculating Euclidean distances between the edge points and the extremely bright point, calculating Euclidean distance average values of the Euclidean distances of all the edge points of the extremely bright point, and taking the square of the difference between the Euclidean distances of the edge points and the Euclidean distance average value as an index of an exponential function taking a natural constant as a base;
and calculating the product of the square of the difference value of the area of the connected domain where the edge point is positioned and the connected domain where the extremely bright point is positioned and the exponential function, and taking the average value of the product of the edge points in all directions of the extremely bright point as the inter-domain uncorrelation of the extremely bright area where the extremely bright point is positioned.
2. The method for detecting the quality of welded pipe welds based on artificial intelligence according to claim 1, wherein said obtaining the extreme bright point in the welded image of the steel pipe comprises:
and for each pixel point of the steel pipe welding image, if the gray value of the pixel point is larger than the gray values of all the pixel points in the neighborhood, the pixel point is marked as an extremely bright point.
3. The method for detecting the quality of a welded pipe weld according to claim 1, wherein the step of obtaining a sequence of pixels of the extremely bright spots in each direction comprises:
the method comprises the steps of taking an extremely bright point as a center, extending towards eight neighborhood directions, and selecting N pixel points in each direction to form a pixel point sequence of each direction of the extremely bright point, wherein N is a preset parameter.
4. The method for detecting the quality of a welded pipe weld based on artificial intelligence according to claim 1, wherein the obtaining the edge points of the extremely bright areas in each direction according to the pixel point sequences of the extremely bright points in each direction comprises:
and for the pixel point sequences in all directions of the extremely bright points, calculating the gray value difference absolute value of each adjacent pixel point in the pixel point sequences, and taking the next pixel point in the adjacent pixel points with the gray value difference absolute value larger than a first threshold value as the edge point of the extremely bright area in the direction of the pixel point sequences.
5. The method for detecting the quality of a welded pipe weld according to claim 1, wherein the obtaining the noise probability of the extremely bright area where the extremely bright point is located according to the distribution of the edge points of the extremely bright area where the extremely bright point is located in each direction comprises:
the Euclidean distance between the extremely bright point and the extremely bright region edge point in each direction is obtained, the average value of the Euclidean distances of the extremely bright region edge points in all directions of the extremely bright point is calculated, the sum of the squares of the difference values of the Euclidean distances of the extremely bright region edge points in all directions of the extremely bright point and the average value is calculated, and the reciprocal of the product of the sum and the average value is used as the noise possibility of the extremely bright region where the extremely bright point is located.
6. The artificial intelligence based welded pipe weld quality detection method of claim 1, wherein the calculating the principal direction of the binary image comprises:
and performing straight line fitting on each edge straight line of the binary image to obtain each fitted straight line, taking the included angle between each fitted straight line and the rightward direction of the horizontal direction as the approximate direction of each edge straight line, and taking the approximate direction average value of all the edge straight lines of the binary image as the main direction of the binary image.
7. The method for detecting the quality of a welded pipe weld based on artificial intelligence according to claim 1, wherein the obtaining edge points of the extremely bright spots in each direction according to the main direction and the auxiliary direction of the extremely bright spots comprises:
and marking the edge points of other connected areas except the connected area where the extreme light point is located, which are firstly contacted by the straight line passing through the main direction and the auxiliary direction of the extreme light point, as the edge points of all directions of the extreme light point.
8. The method for detecting the quality of a welded pipe weld based on artificial intelligence according to claim 1, wherein the obtaining the noise connected domain trend of the extremely bright area where the extremely bright point is located according to the noise probability and the inter-domain non-correlation of the extremely bright area where the extremely bright point is located comprises:
the noise connected domain trend of the extremely bright region where the extremely bright point is located is the product of the noise probability of the extremely bright region where the extremely bright point is located and the inter-domain uncorrelation.
9. The method for detecting the quality of welded pipe welds based on artificial intelligence according to claim 1, wherein the acquiring the trends of the pixels of the welded pipe images comprises:
for each pixel point of the steel pipe welding image, when the pixel point is in an extremely bright area in the binary image, taking the trend of a noise point connected area in the extremely bright area as the trend of the pixel point; when the pixel is not in the extremely bright area in the binary image, the tendency of the pixel is marked as 0.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311300696.5A CN117036356A (en) | 2023-10-10 | 2023-10-10 | Welded pipe weld quality detection method based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311300696.5A CN117036356A (en) | 2023-10-10 | 2023-10-10 | Welded pipe weld quality detection method based on artificial intelligence |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117036356A true CN117036356A (en) | 2023-11-10 |
Family
ID=88624879
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311300696.5A Withdrawn CN117036356A (en) | 2023-10-10 | 2023-10-10 | Welded pipe weld quality detection method based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117036356A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117314916A (en) * | 2023-11-29 | 2023-12-29 | 宝鸡市钛程金属复合材料有限公司 | Explosion welding detection method for metal composite plate based on artificial intelligence |
| CN117830300A (en) * | 2024-03-04 | 2024-04-05 | 新奥新能源工程技术有限公司 | Visual-based gas pipeline appearance quality detection method |
-
2023
- 2023-10-10 CN CN202311300696.5A patent/CN117036356A/en not_active Withdrawn
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117314916A (en) * | 2023-11-29 | 2023-12-29 | 宝鸡市钛程金属复合材料有限公司 | Explosion welding detection method for metal composite plate based on artificial intelligence |
| CN117314916B (en) * | 2023-11-29 | 2024-01-30 | 宝鸡市钛程金属复合材料有限公司 | Explosion welding detection method for metal composite plate based on artificial intelligence |
| CN117830300A (en) * | 2024-03-04 | 2024-04-05 | 新奥新能源工程技术有限公司 | Visual-based gas pipeline appearance quality detection method |
| CN117830300B (en) * | 2024-03-04 | 2024-05-14 | 新奥新能源工程技术有限公司 | Visual-based gas pipeline appearance quality detection method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117036356A (en) | Welded pipe weld quality detection method based on artificial intelligence | |
| CN107610111B (en) | deep learning-based welding spot image detection method | |
| Shah et al. | Butt welding joints recognition and location identification by using local thresholding | |
| CN113362326A (en) | Method and device for detecting welding spot defects of battery | |
| CN109191459B (en) | Automatic identification and rating method for continuous casting billet macrostructure center segregation defect | |
| CN111696107B (en) | Molten pool contour image extraction method for realizing closed connected domain | |
| CN113421261B (en) | Structural member production process defect detection method based on image processing | |
| CN111667462B (en) | Automatic detection method and system for small and weak targets in welding defects | |
| CN114972343B (en) | Method for detecting production abnormity of laser welding product | |
| CN116703907B (en) | Machine vision-based method for detecting surface defects of automobile castings | |
| CN111127402A (en) | Visual detection method for welding quality of robot | |
| CN104865277B (en) | Method for automatically identifying pipe weld defect based on double-wall double-projection transillumination image | |
| CN115205290B (en) | Online detection method and system for PCB production process | |
| CN113496483B (en) | Weld seam air hole defect detection method based on image processing | |
| WO2020110667A1 (en) | Surface defect detecting method, surface defect detecting device, method for manufacturing steel material, steel material quality control method, steel material manufacturing equipment, method for creating surface defect determination model, and surface defect determination model | |
| CN110751628B (en) | Ultrasonic image-based automatic weld defect identification method | |
| CN113409313A (en) | Wheel weld surface defect detection method based on computer vision | |
| CN117095004B (en) | Excavator walking frame main body welding deformation detection method based on computer vision | |
| CN109859177A (en) | Industrial x-ray image assessment method and device based on deep learning | |
| CN110717872A (en) | Method and system for extracting characteristic points of V-shaped welding seam image under laser-assisted positioning | |
| CN108921828B (en) | Method for identifying insignificant weld joint in complex scene | |
| Li et al. | Detection algorithm of defects on polyethylene gas pipe using image recognition | |
| CN115063422A (en) | Intelligent detection method for container welding quality | |
| CN115330774A (en) | Welding image molten pool edge detection method | |
| Lee et al. | Review on the recent welding research with application of CNN-based deep learning part II: Model evaluation and visualizations |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20231110 |
|
| WW01 | Invention patent application withdrawn after publication |