CN114549836B - Accurate alignment method for radiographic image and internal detection information of long-distance pipeline - Google Patents
Accurate alignment method for radiographic image and internal detection information of long-distance pipeline Download PDFInfo
- Publication number
- CN114549836B CN114549836B CN202210143692.XA CN202210143692A CN114549836B CN 114549836 B CN114549836 B CN 114549836B CN 202210143692 A CN202210143692 A CN 202210143692A CN 114549836 B CN114549836 B CN 114549836B
- Authority
- CN
- China
- Prior art keywords
- weld
- upstream
- downstream
- image
- loss
- 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.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 59
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000011144 upstream manufacturing Methods 0.000 claims abstract description 73
- 238000010276 construction Methods 0.000 claims abstract description 20
- 230000011218 segmentation Effects 0.000 claims abstract description 5
- 238000003466 welding Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 6
- 238000013136 deep learning model Methods 0.000 claims description 4
- 230000007547 defect Effects 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 239000011324 bead Substances 0.000 claims description 3
- 230000002950 deficient Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000009659 non-destructive testing Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000009412 basement excavation Methods 0.000 abstract description 11
- 238000004364 calculation method Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 abstract description 2
- 238000012423 maintenance Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 5
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 239000003209 petroleum derivative Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N23/00—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00
- G01N23/02—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by transmitting the radiation through the material
- G01N23/04—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by transmitting the radiation through the material and forming images of the material
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Chemical & Material Sciences (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention relates to the field of pipeline safety maintenance, in particular to a method for accurately aligning a long-distance pipeline radiographic image with internal detection information, which comprises the following specific processes: firstly, extracting corresponding information of the number of the inner detection girth weld and the number of the weld crater in the construction period. And then determining the weld type, upstream intersection point and downstream intersection point information in the inner detection girth weld list. And then, constructing a deep learning example segmentation model based on a data driving method and reasoning the industrial negative spiral weld joint and the straight weld joint region to finish the calculation of the position deviation of the upstream and downstream intersection points. And finally, determining the alignment result of the inner detection data and the industrial negative film through the deviation type. The method provided by the invention greatly improves the accuracy of excavation and reduces the excavation cost.
Description
Technical Field
The invention relates to the field of pipeline safety maintenance, in particular to a method for accurately aligning a long-distance pipeline radiographic image with internal detection information.
Background
The long-distance pipeline is a main mode for transporting petroleum and natural gas, and has important significance for ensuring the safe operation of the pipeline. At present, some pipelines are in service for many years and are influenced by various factors such as environment, and the welded joint area of the pipeline is most likely to fail, so that fuel leakage is caused and serious damage is generated. In order to prevent the occurrence of pipeline transportation accidents, excavation analysis is required for pipelines buried underground. However, the excavation cost is high, and through relevant data, some pipelines complete data alignment work, but the annual excavation accuracy is lower than 98%. The excavation accuracy rate not only affects the quality inspection progress of the girth weld, but also increases the excavation cost. It is therefore important to achieve a precise alignment method.
The traditional data alignment method uses internal detection data and construction period construction data, and adopts a mode of one-to-one correspondence of pipe joint characteristics and lengths to determine a construction period welded junction number corresponding to the internal detection number. The accuracy of data alignment is reduced due to images of reasons such as reversed weld port numbering sequence, similar pipe joint length, the same port and the like in the construction period.
According to research, on the basis of data alignment, the method for carrying out data deep alignment by using the internal detection data and the intersection point position of the spiral weld (or the straight weld) and the girth weld in the negative digital image can effectively avoid the influence of factors such as reversed sequence of welded junction numbers, similar pipe joint length, same opening and the like on the data alignment accuracy in the construction period, improve the excavation accuracy and reduce the excavation cost.
Disclosure of Invention
Aiming at the problems that the existing traditional data alignment method is low in recognition efficiency and low in recognition accuracy, the invention provides a method for accurately aligning a long-distance pipeline radiographic image with internal detection information, and the method for accurately aligning the long-distance pipeline radiographic image with the internal detection information can achieve higher data alignment accuracy finally.
The specific technical scheme is as follows:
a method for precisely aligning a long-distance pipeline radiographic image with internal detection information comprises the following steps:
step 1: extracting corresponding information of the number of the inner detection girth weld and the number of the weld leg in the construction period, and obtaining an original weld bead image;
Step 2: extracting weld type, upstream intersection point and downstream intersection point information in an inner detection girth weld list;
Step 3: constructing a spiral weld and straight weld deep learning model;
Step 3-1: the nondestructive testing engineer marks the areas of the spiral weld and the straight weld to obtain area marking information, and a training set is formed
Wherein, X k is the kth image, Y k is the label area mask corresponding thereto;
step 3-2: generalizing the training set image, adopting image splicing, image scaling, image folding and other modes, and making corresponding treatment on the label to obtain an expanded data set
Step 3-3: performing and feature enhancement based on the image relief operator, and realizing the relief operator according to the following formula;
P'(i,j)＝P(i-1,j-1)-P(i+1,j+1)+TH，
wherein P' (i, j) represents the pixel value at j of the ith row after optimization, and TH selects 128; the feature-enhanced image set can thus be represented as
Step 3-4: an example segmentation model is built based on a clean-up set neural network CSPDARKNET and a backbone network; wherein the loss process comprises three parts of classification, regression and mask; l=l cls+Lres+Lmask, wherein the classification loss L cls is calculated as a cross entropy lossWhere p i denotes that the anchor point is determined to be defective at probability,Representing a real probability value, and taking 0 if the predicted position in the anchor frame is 1 and if the predicted position is 1; regression loss was calculated based on IoU loss, L res =1-IoU, where/>Wherein S p is a predicted position, S t is a true position; the total loss function is the classified superposition loss and regression loss; constructing mask loss L mask and exact multiple classification and regression loss to generate final loss, wherein/>Classifying the loss for each pixel; the final Loss is as follows loss=l rpn+Lcls+Lres+Lmask:
Step 3-5: finally, based on the back propagation network parameters of the gradient descent method, a defect detection model M 1 is obtained when the loss tends to be stable;
Step 4: automatically extracting the positions of the upstream and downstream intersection points and the types of the upstream and downstream welding seams in the digital image of the negative film;
Step 5: converting the ruler position of the upstream and downstream intersection points in the ledger into corresponding clock point positions, and calculating the position deviation of the upstream and downstream intersection points;
Step 6: and comprehensively analyzing the alignment result according to the deviation situation.
The step4 specifically comprises the following steps:
Step 4-1: feeding the industrial negative image enhanced based on the relief operator into a network model and obtaining an inference area omega = M 1 (X (P (i, j))), wherein the spiral weld area is omega 1 (X, y), the straight weld area is omega 2 (X, y), and the girth weld area is omega 3 (X, y);
Step 4-2: determining the type of an upstream and a downstream intersection points and an upstream and a downstream welding seams in the digital image of the negative film; the type of intersection is a spiral weld, wherein the upstream and downstream intersection is (x1,y1)＝min(Ω1∩Ω3),(x2,y2)＝max(Ω1∩Ω3),, the type of intersection is a straight weld, and the position of the upstream and downstream intersection is (x3,y3)＝min(Ω2∩Ω3),(x4,y4)＝max(Ω2∩Ω3);
Step 4-3: calculating the position deviation of the upstream and downstream intersection points, the upstream and downstream deviation delta 2＝(x'2,y'2)-(x'1,y'1 of the spiral weld joint), and the upstream and downstream deviation delta 2＝(x'2,y'2)-(x'1,y'1 of the straight weld joint.
The step 5 specifically comprises the following steps:
Step 5-1: constructing a signature recognition model M 1 based on the signature data according to the step 3;
Step 5-2: acquiring an industrial negative film signature position based on M 1, and calculating a signature interval to which the upstream and downstream intersection point positions in the step 4-2 belong;
Step 5-3: according to the ruler position of the intersection point of the upstream and downstream in the standing book, converting the ruler position into the corresponding clock position to be respectively recorded as α(x1,y1),α(x2,y2),α(x3,y3),α(x4,y4);
Step 5-4: determining the position of the intersection point of the spiral weld joint and the straight hash weld joint in the internal detection information and marking the position as β(x1,y1),β(x2,y2),β(x3,y3),β(x4,y4);
Step 5-5: comparing the digital image of the negative film in the construction period with the internal detection information, and determining the type of the spiral weld; when the weld type is not recognized by the digital image of the negative film in the construction period, determining as unknown;
Step 5-6: and calculating the upstream and downstream intersection errors of the digitized image of the negative film in the construction period and the internal detection information, and expressing the intersection errors by time, wherein the intersection errors are denoted as delta=beta (x, y) -alpha (x, y).
The step 6 specifically comprises the following steps:
Case 1: the following cases were found to be automatic alignment
D. The type of the upstream and downstream welding lines in the film image and the inner detection report is the same, and the error of the upstream and downstream intersection points is less than 1h;
e. Only one position is identified by the type of the upstream and downstream welding lines in the film image, the type of the internal detection report is the same, and meanwhile, the error of the upstream and downstream intersection points is less than 1h;
f. Only one position is identified by the type of the upstream and downstream weld joints in the film image, only one intersection point information is in the internal detection report, and the intersection point error is smaller than 1h;
Case 2: the following cases were found to be misaligned:
c. When the type of the identified upstream and downstream weld joints is different at any place;
d. the error of the intersection point between the upstream and the downstream is larger than 1h;
Only one intersection point information exists in the internal detection report, and meanwhile, the intersection point error is more than or equal to 1h; when the intersection error is greater than or equal to 1h, 0 point is required to be calibrated through the negative image, the alignment is carried out again, and the remark is noted as 0 point recalibration;
Case 3: the conclusion is to infer alignment when the following occurs: and when the weld type in the film image is not identified, and the weld junctions of the weld type are automatically aligned at the upstream and downstream sides.
Compared with the prior art, the invention has the following beneficial technical effects:
The invention firstly extracts the corresponding information of the number of the internal detection girth weld and the number of the weld crater in the construction period. And then determining the weld type, upstream intersection point and downstream intersection point information in the inner detection girth weld list. And then, constructing a deep learning example segmentation model based on a data driving method and reasoning the industrial negative spiral weld joint and the straight weld joint region to finish the calculation of the position deviation of the upstream and downstream intersection points. And finally, determining the alignment result of the inner detection data and the industrial negative film through the deviation type. The method provided by the invention greatly improves the accuracy of excavation and reduces the excavation cost. The artificial intelligence method reduces the manual intervention and greatly improves the operation efficiency.
Drawings
FIG. 1 is a diagram of a method for precisely aligning a radiographic image of a long-distance pipeline with internal detection information according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a relief image of a straight weld and a circumferential weld, according to the method for precisely aligning a radiographic image of a long-distance pipeline with internal detection information according to an embodiment of the present invention;
FIG. 3 is a diagram of intelligent recognition results of straight welds and girth welds, according to the method for accurately aligning radiographic images and internal detection information of long-distance pipelines provided by the embodiment of the invention;
fig. 4 is a diagram of intelligent recognition results of spiral weld and girth weld in a method for precisely aligning radiographic images and internal detection information of a long-distance pipeline according to an embodiment of the present invention.
Detailed Description
The present invention will be described in detail with reference to the following examples and drawings, but the scope of the present invention is not limited by the examples and drawings.
A method for accurately aligning a long-distance pipeline radiographic image with internal detection information mainly comprises the following steps: the method comprises the steps of roughly aligning a part of the structure, establishing a deep learning model of the spiral weld and the straight weld, finely aligning the part, and a whole structure diagram is shown in figure 1, wherein the roughly aligned mode is firstly adopted to realize primary alignment based on the weld neck label and internal detection data, then, establishing a deep network to realize extraction of the spiral weld and the straight weld, and finally, completing deep alignment through the deduced position.
Step 1: and extracting corresponding information of the number of the inner detection girth weld and the number of the weld opening in the construction period, and acquiring an original weld bead image for rough alignment.
Step 1-1: the original internal detection data and the industrial negative data information are input.
Step 1-2: the method is characterized in that the inner detection data pipe section characteristics and lengths are adopted, and the construction period welded junction numbers corresponding to the inner detection numbers are determined in a one-to-one correspondence mode to perform coarse alignment.
Step 2: and extracting weld type, upstream intersection point and downstream intersection point information in the inner detection girth weld list.
Step 3: and constructing a spiral weld and straight weld deep learning model.
Step 3-1: the nondestructive testing engineer marks the areas of the spiral weld and the straight weld to obtain area marking information, and a training set is formedWhere X k is the kth image and Y k is the label region mask corresponding thereto.
Step 3-2: generalizing the training set image, adopting image splicing, image scaling, image folding and other modes, and making corresponding treatment on the label to obtain an expanded data set
Step 3-3: and carrying out and characteristic enhancement based on the image relief operator, and realizing the relief operator according to the following formula.
P '(i, j) =p (i-1, j-1) -P (i+1, j+1) +th, where P' (i, j) represents the pixel value at the i-TH row j after optimization, TH is selected to be 128. The feature-enhanced image set can thus be represented asFig. 2 is a graph of the relief image annotation of the straight weld and the girth weld, and the data set annotated in the graph shows that the image enhanced by the image is more favorable for obtaining the target compared with the original gray scale graph.
Step 3-4: an instance segmentation model is constructed based on the clean-up set neural network CSPDARKNET and the backbone network. Wherein the loss process comprises three parts of classification, regression and mask. L=l cls+Lres+Lmask wherein the classification loss L cls is calculated as cross entropy lossWhere p i represents that the anchor point is determined to be defective at probability,/>And (5) representing a real probability value, and taking 0 if the predicted position in the anchor frame is 1 and if the predicted position is the real defect. Regression loss was calculated based on IoU loss, L res =1-IoU, where/>Where S p is the predicted position and S t is the true position. The total loss function is the classified superposition loss and regression loss. Constructing mask loss L mask and exact multiple classification and regression loss to generate final loss, wherein/>The penalty is classified for each pixel. The final Loss is as follows loss=l rpn+Lcls+Lres+Lmask:
And 3-5, finally, based on the back propagation network parameters of the gradient descent method, obtaining a defect detection model M 1 when the loss tends to be stable.
Step 4: and automatically extracting the positions of the upstream and downstream intersection points and the upstream and downstream weld joint types in the digital image of the negative film.
And 4-1, feeding the industrial negative image enhanced based on the relief operator into a network model and obtaining an inference area omega=M 1 (X (P (i, j))), wherein the spiral weld joint area is omega 1 (X, y), the straight weld joint area is omega 2 (X, y), and the girth weld joint area is omega 3 (X, y), wherein the inference result is shown in figure 3, and the weld joint position can be obviously detected.
And 4-2, determining the type of the upstream and downstream intersection points and the upstream and downstream welding seams in the digital image of the negative film. The type of intersection is a spiral weld, wherein the upstream and downstream intersection is (x1,y1)＝min(Ω1∩Ω3),(x2,y2)＝max(Ω1∩Ω3),, the type of intersection is a straight weld, and the position of the upstream and downstream intersection is (x3,y3)＝min(Ω2∩Ω3),(x4,y4)＝max(Ω2∩Ω3).
And 4-3, calculating the position deviation of the upstream and downstream intersection points, namely the upstream and downstream deviation delta 2＝(x'2,y'2)-(x'1,y'1 of the spiral weld joint and the upstream and downstream deviation delta 2＝(x'2,y'2)-(x'1,y'1 of the straight weld joint.
Step 5: and converting the ruler position of the upstream and downstream intersection points in the ledger into corresponding clock point positions, and calculating the position deviation of the upstream and downstream intersection points.
Step 5-1: a signature recognition model M 1 is constructed on the basis of the signature data according to step 3.
Step 5-2: and (3) acquiring an industrial negative signature position based on M 1, and calculating a signature interval to which the upstream and downstream intersection point positions in the step (4-2) belong.
Step 5-3: according to the ruler position of the intersection point of the upstream and downstream in the standing book, converting the ruler position into the corresponding clock position to be respectively recorded as α(x1,y1),α(x2,y2),α(x3,y3),α(x4,y4).
Step 5-4: determining the position of the intersection point of the spiral weld joint and the straight hash weld joint in the internal detection information and marking the position as β(x1,y1),β(x2,y2),β(x3,y3),β(x4,y4).
Step 5-5: and comparing the digital image of the negative film in the construction period with the internal detection information, and determining the type of the spiral weld. And determining as unknown when the weld joint type is not recognized by the digitized image of the negative film in the construction period.
Step 5-6: and calculating the upstream and downstream intersection errors of the digitized image of the negative film in the construction period and the internal detection information, and expressing the intersection errors by time, wherein the intersection errors are denoted as delta=beta (x, y) -alpha (x, y).
Step 6: and comprehensively analyzing the alignment result according to the deviation situation.
Case 1: conclusion is automatic alignment:
a. The type of the upstream and downstream welding lines in the film image and the inner detection report is the same, and the error of the upstream and downstream intersection points is less than 1h;
b. Only one position is identified by the type of the upstream and downstream welding lines in the film image, the type of the internal detection report is the same, and meanwhile, the error of the upstream and downstream intersection points is less than 1h;
c. Only one position is identified by the type of the upstream and downstream weld joints in the film image, only one intersection point information is in the internal detection report, and the intersection point error is smaller than 1h;
Case 2: the following cases were found to be misaligned:
a. when the type of the identified upstream and downstream weld joints is different at any place;
b. The error of the upstream and downstream intersection point is greater than 1h;
c. only one intersection point information exists in the internal detection report, and meanwhile, the intersection point error is more than or equal to 1h.
Note that: when the intersection error is greater than or equal to 1h, the 0 point is calibrated through the film image, realigned, and the remark is noted as "recalibrate 0 point".
Case 3: the conclusion is to infer alignment when the following occurs: and when the weld type in the film image is not identified, and the weld junctions of the weld type are automatically aligned at the upstream and downstream sides.
Claims (4)
1. A method for accurately aligning a long-distance pipeline radiographic image with internal detection information is specifically characterized in that: the method comprises the following steps:
step 1: extracting corresponding information of the number of the inner detection girth weld and the number of the weld leg in the construction period, and obtaining an original weld bead image;
Step 2: extracting weld type, upstream intersection point and downstream intersection point information in an inner detection girth weld list;
Step 3: constructing a spiral weld and straight weld deep learning model;
Step 3-1: the nondestructive testing engineer marks the areas of the spiral weld and the straight weld to obtain area marking information, and a training set is formed
Wherein, X k is the kth image, Y k is the label area mask corresponding thereto;
Step 3-2: generalizing the training set image, adopting image stitching, image scaling, image folding and corresponding processing of the label to obtain an expanded data set
Step 3-3: performing and feature enhancement based on the image relief operator, and realizing the relief operator according to the following formula;
P'(i,j)＝P(i-1,j-1)-P(i+1,j+1)+TH，
wherein P' (i, j) represents the pixel value at j of the ith row after optimization, and TH selects 128; the feature-enhanced image set can thus be represented as
Step 3-4: an example segmentation model is built based on a clean-up set neural network CSPDARKNET and a backbone network; wherein the loss process comprises three parts of classification, regression and mask; l=l cls+Lres+Lmask, wherein the classification loss L cls is calculated as a cross entropy lossWhere p i denotes that the anchor point is determined to be defective at probability,/>Representing a real probability value, and taking 0 if the predicted position in the anchor frame is 1 and if the predicted position is 1; regression loss was calculated based on IoU loss, L res =1-IoU, where/>Wherein S p is a predicted position, S t is a true position; the total loss function is the classified superposition loss and regression loss; constructing mask loss L mask and exact multiple classification and regression loss to generate final loss, whereinClassifying the loss for each pixel; the final Loss is as follows loss=l rpn+Lcls+Lres+Lmask:
Step 3-5: finally, based on the back propagation network parameters of the gradient descent method, a defect detection model M 1 is obtained when the loss tends to be stable;
Step 4: automatically extracting the positions of the upstream and downstream intersection points and the types of the upstream and downstream welding seams in the digital image of the negative film;
Step 5: converting the ruler position of the upstream and downstream intersection points in the ledger into corresponding clock point positions, and calculating the position deviation of the upstream and downstream intersection points;
Step 6: and comprehensively analyzing the alignment result according to the deviation situation.
2. The method for precisely aligning the radiographic image and the internal detection information of the long-distance pipeline according to claim 1, wherein the method comprises the following steps: the step 4 specifically comprises the following steps:
Step 4-1: feeding the industrial negative image enhanced based on the relief operator into a network model and obtaining an inference area omega = M 1 (X (P (i, j))), wherein the spiral weld area is omega 1 (X, y), the straight weld area is omega 2 (X, y), and the girth weld area is omega 3 (X, y);
Step 4-2: determining the type of an upstream and a downstream intersection points and an upstream and a downstream welding seams in the digital image of the negative film; the type of intersection is a spiral weld, wherein the upstream and downstream intersection is (x1,y1)＝min(Ω1∩Ω3),(x2,y2)＝max(Ω1∩Ω3),, the type of intersection is a straight weld, and the position of the upstream and downstream intersection is (x3,y3)＝min(Ω2∩Ω3),(x4,y4)＝max(Ω2∩Ω3);
Step 4-3: calculating the position deviation of the upstream and downstream intersection points, the upstream and downstream deviation delta 2＝(x'2,y'2)-(x'1,y'1 of the spiral weld joint), and the upstream and downstream deviation delta 2＝(x'2,y'2)-(x'1,y'1 of the straight weld joint.
3. The method for precisely aligning the radiographic image and the internal detection information of the long-distance pipeline according to claim 1, wherein the method comprises the following steps: the step 5 specifically comprises the following steps:
Step 5-1: constructing a signature recognition model M 1 based on the signature data according to the step 3;
Step 5-2: acquiring an industrial negative film signature position based on M 1, and calculating a signature interval to which the upstream and downstream intersection point positions in the step 4-2 belong;
Step 5-3: according to the ruler position of the intersection point of the upstream and downstream in the standing book, converting the ruler position into the corresponding clock position to be respectively recorded as α(x1,y1),α(x2,y2),α(x3,y3),α(x4,y4);
Step 5-4: determining the position of the intersection point of the spiral weld joint and the straight hash weld joint in the internal detection information and marking the position as β(x1,y1),β(x2,y2),β(x3,y3),β(x4,y4);
Step 5-5: comparing the digital image of the negative film in the construction period with the internal detection information, and determining the type of the spiral weld; when the weld type is not recognized by the digital image of the negative film in the construction period, determining as unknown;
Step 5-6: and calculating the upstream and downstream intersection errors of the digitized image of the negative film in the construction period and the internal detection information, and expressing the intersection errors by time, wherein the intersection errors are denoted as delta=beta (x, y) -alpha (x, y).
4. The method for precisely aligning the radiographic image and the internal detection information of the long-distance pipeline according to claim 1, wherein the method comprises the following steps: the step 6 specifically comprises the following steps:
Case 1: the following cases were found to be automatic alignment
A. The type of the upstream and downstream welding lines in the film image and the inner detection report is the same, and the error of the upstream and downstream intersection points is less than 1h;
b. Only one position is identified by the type of the upstream and downstream welding lines in the film image, the type of the internal detection report is the same, and meanwhile, the error of the upstream and downstream intersection points is less than 1h;
c. Only one position is identified by the type of the upstream and downstream weld joints in the film image, only one intersection point information is in the internal detection report, and the intersection point error is smaller than 1h;
Case 2: the following cases were found to be misaligned:
a. when the type of the identified upstream and downstream weld joints is different at any place;
b. the error of the intersection point between the upstream and the downstream is larger than 1h;
Only one intersection point information exists in the internal detection report, and meanwhile, the intersection point error is more than or equal to 1h; when the intersection error is greater than or equal to 1h, 0 point is required to be calibrated through the negative image, the alignment is carried out again, and the remark is noted as 0 point recalibration;
Case 3: the conclusion is to infer alignment when the following occurs: and when the weld type in the film image is not identified, and the weld junctions of the weld type are automatically aligned at the upstream and downstream sides.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210143692.XA CN114549836B (en) | 2022-02-17 | 2022-02-17 | Accurate alignment method for radiographic image and internal detection information of long-distance pipeline |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210143692.XA CN114549836B (en) | 2022-02-17 | 2022-02-17 | Accurate alignment method for radiographic image and internal detection information of long-distance pipeline |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114549836A CN114549836A (en) | 2022-05-27 |
| CN114549836B true CN114549836B (en) | 2024-05-24 |
Family
ID=81675203
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210143692.XA Active CN114549836B (en) | 2022-02-17 | 2022-02-17 | Accurate alignment method for radiographic image and internal detection information of long-distance pipeline |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114549836B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108665452A (en) * | 2018-05-09 | 2018-10-16 | 广东大鹏液化天然气有限公司 | A kind of pipeline-weld film scanning storage and identification of Weld Defects and its system based on big data |
| CN111461569A (en) * | 2020-04-17 | 2020-07-28 | 广东大鹏液化天然气有限公司 | Safety management method for pipeline body |
| CN112734685A (en) * | 2019-10-14 | 2021-04-30 | 中国石油天然气股份有限公司 | Pipeline weld joint information identification method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10970856B2 (en) * | 2018-12-27 | 2021-04-06 | Baidu Usa Llc | Joint learning of geometry and motion with three-dimensional holistic understanding |
-
2022
- 2022-02-17 CN CN202210143692.XA patent/CN114549836B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108665452A (en) * | 2018-05-09 | 2018-10-16 | 广东大鹏液化天然气有限公司 | A kind of pipeline-weld film scanning storage and identification of Weld Defects and its system based on big data |
| CN112734685A (en) * | 2019-10-14 | 2021-04-30 | 中国石油天然气股份有限公司 | Pipeline weld joint information identification method |
| CN111461569A (en) * | 2020-04-17 | 2020-07-28 | 广东大鹏液化天然气有限公司 | Safety management method for pipeline body |
Non-Patent Citations (1)
| Title |
|---|
| 油气长输管道大数据整合技术方法研究;金剑;朱学山;刘艳阳;李宁;邹斌;;油气田地面工程;20200217(02);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114549836A (en) | 2022-05-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110992349A (en) | Underground pipeline abnormity automatic positioning and identification method based on deep learning | |
| WO2020133639A1 (en) | Intelligent analysis system for magnetic flux leakage detection data in pipeline | |
| CN115525636B (en) | Multi-wheel alignment method and device for detection data in magnetic leakage of gas long-distance pipeline | |
| CN117218083A (en) | Improved YOLOv8 steel pipe weld defect detection method | |
| CN113763363A (en) | Drainage pipeline fracture defect detection and grade evaluation method | |
| CN111665342A (en) | Detection data analysis system in pipeline | |
| CN108804537B (en) | Method for internally detecting data alignment | |
| Liu et al. | An intelligent defect detection approach based on cascade attention network under complex magnetic flux leakage signals | |
| CN115995056A (en) | Automatic bridge disease identification method based on deep learning | |
| CN114549836B (en) | Accurate alignment method for radiographic image and internal detection information of long-distance pipeline | |
| CN114119505A (en) | Method and device for detecting chip adhesion area defects | |
| Li et al. | RGB-thermal fusion network for leakage detection of crude oil transmission pipes | |
| CN116029979A (en) | Cloth flaw visual detection method based on improved Yolov4 | |
| CN113780111B (en) | Pipeline connector defect accurate identification method based on optimized YOLOv3 algorithm | |
| CN111289526A (en) | Method and system for detecting defects on inner surface of gas conveying pipeline | |
| CN109767426B (en) | Shield tunnel water leakage detection method based on image feature recognition | |
| KR20230127511A (en) | Real-time evaluation system and method of rock state based on tunnel face image by using deep learning | |
| CN112329590B (en) | Pipeline assembly detection system and detection method | |
| KR102047251B1 (en) | Bid support system and method regarding offshore oil and gas facilities | |
| CN112215301A (en) | Image straight line detection method based on convolutional neural network | |
| CN115266932A (en) | Gearbox gear shaft flaw detection method | |
| CN115082650A (en) | Implementation method of automatic pipeline defect labeling tool based on convolutional neural network | |
| CN115049600A (en) | Intelligent identification system and method for small sample pipeline defects | |
| Amri et al. | Implementation of Underwater Image Enhancement for Corrosion Pipeline Inspection (UIECPI) | |
| CN110516269A (en) | A kind of optimization for submarine pipeline engineering critical evaluation refers to strain process |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |