A method for identifying girth welds in pipeline inspection

Abstract

A method for identifying girth welds in pipeline detection is disclosed. The method comprises the following steps: a plurality of magnetic flux leakage sensors are evenly arranged in the pipeline, a mileage sensor is set in the circumference of the pipeline, and original magnetic flux leakage data and mileage wheel data of multiple channels are synchronously collected in a sampling manner with equal time intervals; a binary map of pipeline magnetic flux leakage distribution is obtained by median filtering, and the morphological features of weld data are enhanced by asymmetric closing operation; straight line segments in the binary map are extracted by Hough transformation to determine whether there are missed welds. The method adopts a two-dimensional data analysis method to perform a similar image processing method on the data to enhance the data, highlight the morphological features of uneven parts in the magnetic flux leakage sensor data, and capture the girth welds therefrom. The method can determine the girth welds in a two-dimensional space to ensure accuracy, and has low requirements on sensor calibration and computer computing power. In addition, a circumferential periodic extension method is adopted in the judgment to improve the reliability of girth weld identification.

Description

Technical Field

[0001] The invention relates to a girth weld recognition method, in particular to a girth weld recognition method in pipeline detection, and belongs to the technical field of pipeline girth weld detection. Background Art

[0002] Pipeline transportation is a very important transportation method. It is mainly used in the transportation of energy, chemical industry and water resources. The maintenance and monitoring of pipelines requires the use of dedicated pipeline detection equipment. After using pipeline detection equipment to detect a section of pipeline, it is necessary to analyze the collected data. The analysis is mainly to determine whether there are defects in the pipeline, locate the defects, and then repair them through construction. Since pipelines are usually very long, the accuracy of positioning is very important. Because pipeline construction requires excavation, inaccurate positioning has a great impact on construction efficiency.

[0003] Pipeline sensors are used to detect pipelines. The sensors are evenly distributed around the detector in the pipeline. All sensors are sampled synchronously. Data is collected by high-frequency equal time interval sampling. The obtained data can be restored to data distributed according to distance with the help of mileage wheel data. The overall data is equivalent to cutting the cylindrical pipeline wall into two-dimensional distribution data with the length of the pipeline and the width of the pipeline circumference, which is used to characterize the defect distribution of the pipeline. This method judges the waveform of a single leakage magnetic sensor separately. When the leakage magnetic sensor fluctuates, it is judged that there may be a weld at this position, and then the waveforms of other sensors are judged. When the total number of sensors that fluctuate near this position (usually a distance of 10mm is set) reaches a set value, it is judged whether the number of sensors that fluctuate at this time reaches the set value. If it reaches the set value, it is judged that there is a weld here.

[0004] Another way is to use a convolutional neural network training model for judgment. The data is cut in the circumferential direction, which is equivalent to obtaining the two-dimensional magnetic signal distribution of the horizontal axis, the vertical axis, and the entire pipeline. Then, by marking the girth weld, a convolutional neural network that can be used to identify the girth weld of the pipeline is trained. The defects of this method are: 1) Due to the use of machine learning, a large number of samples are required to ensure that there is no overfitting, and because the training model requires a lot of computing power, it is necessary to use a special computer for calculation to complete it; 2) The stability of the detector is high. When a new model (caliber) of the detector is replaced, or when the probe sensor of the same caliber detector is replaced, and the sensor calibration is not performed, it will affect the judgment accuracy of the model. Therefore, the steps usually used in this method are to obtain a new detector, the detector must enter the pipeline to collect data for a certain distance, manually mark the data, import the data into the model for training, and then cooperate with this detector to realize the identification of the girth weld. The initial workload is large and time-consuming and laborious. Summary of the invention

[0005] The purpose of the present invention is to provide a method for identifying girth welds in pipeline detection, which can judge girth welds in a two-dimensional space to ensure accuracy, and has low requirements for sensor calibration and computer computing power. In addition, a circumferential periodic extension method is used in the judgment to improve the reliability of girth weld identification.

[0006] In order to achieve the above object, the present invention provides a method for identifying girth welds in pipeline detection, comprising the following steps:

[0007] Step 1: Evenly arrange several magnetic flux leakage sensors in the pipeline, set up mileage sensors around the pipeline, and synchronously collect multi-channel raw magnetic flux leakage data and mileage wheel data by sampling at equal time intervals;

[0008] Step 2: First, convert the odometer wheel data into actual mileage values, and then obtain the leakage magnetic data of the mileage coordinates through time alignment; compare the odometer wheel data within the set time range, calculate the number of pulses of each odometer wheel data, select the odometer wheel corresponding to the maximum number of pulses, calculate the difference of this odometer wheel, and obtain the mileage pulse distribution of the odometer wheel; calculate the actual mileage value of the odometer wheel according to the mileage pulse distribution; reconstruct the data by making the leakage magnetic data and the actual mileage value at the same time correspond one by one;

[0009] Step 3: Use the median filter method to process the data in step 2 and separate the leakage magnetic data representing different pipe sections, as follows:

[0010] Step 31: When a magnetic flux leakage sensor of an internal detector is tested before delivery, under natural conditions, the value of the magnetic flux leakage sensor of this specification fluctuates around the center value N, and the fluctuation range is [-M / 2, M / 2];

[0011] Step 32: performing median filtering on the magnetic flux leakage data processed in step 2 according to the channel to obtain the median value of the magnetic flux leakage data of each channel;

[0012] Step 33: Draw an intermediate value curve, wherein the intermediate value curve includes weld data and pipeline flat portion data, and the ratio of the line width of the weld data to the line width of the pipeline flat portion is 1:20;

[0013] Step 4: Subtract the leakage magnetic flux data of step 2 from the intermediate value data of step 3, and obtain the pipeline leakage magnetic flux distribution binary map through the threshold, specifically:

[0014] Step 41: Determine a threshold value M according to the fluctuation range of the magnetic flux leakage data obtained in step 31;

[0015] Step 42: Subtract the leakage magnetic flux data obtained in step 2 from the intermediate value data obtained in step 3, and then perform threshold binarization processing to obtain a binarized image, assign a color to the part exceeding the threshold value to represent the weld data, and assign another color to the part not exceeding the threshold value to represent the flat part of the pipeline;

[0016] Step 5: Use asymmetric closing operation to process the binary image and strengthen the morphological characteristics of the weld data in the detection data; use asymmetric closing operation, set the expansion scale larger than the corrosion scale, and connect multiple connected domains to each other; according to the average width W of the pipeline weld under test and the minimum mileage resolution s of the detector, the side length value L of the square expansion kernel can be set to W / s, and an odd number is taken upward, and the side length of the square corrosion kernel is L-2;

[0017] Step 6: Use Hough transform to extract the straight line segment in the binary image and determine whether it is a girth weld. If it is a girth weld, register the weld information in the weld list;

[0018] Step 61: Find straight line segments with a length of 2 / 3 of the pipe circumference, and these straight line segments are uniformly extracted as girth welds;

[0019] Step 62: Since the girth weld is perpendicular to the extension direction of the pipeline, according to the inclination angle existing during the operation of the detector, the straight line segment satisfying the following formula is also extracted:

[0020] |(x2-x1) / (y2-y1)|<tan(5°)

[0021] Where (x1, y1) and (x2, y2) are the coordinates of the two endpoints of the straight line segment;

[0022] Step 63: Remove multiple straight line segments in the same weld obtained in Steps 61 and 62, and only keep one straight line segment to represent the circumferential weld:

[0023] First, represent the straight line segment with the mileage coordinate of the midpoint of the straight line segment, denoted as X, and X = (x1 + x2) / 2. The data corresponding to the serial number is Xn, which is represented by the set {Xn}, and n is the number of straight line segments;

[0024] Then, sort the data in the set {Xn}, select the first data X1, and remove all X data whose distance to X1 is less than d. The calculation formula of d is:

[0025] d = w / 2s

[0026] In the formula: w is the shortest distance between two circumferential welds in the pipeline construction standard;

[0027] For the first data X1, the subset {Xm1||Xm1 - X1| < d} of {Xn}, then all Xm belong to the same weld as X1 and are removed. Only keep the first data X1 to represent this weld; after {Xn} removes {Xm1||Xm1 - X1| < d}, continue to loop this process until all elements in the set {Xn} are traversed. Finally, a set containing a list of mileage values is obtained, and each mileage value corresponds to a circumferential weld;

[0028] Step 7: Determine whether there is a possibility of missed inspection of welds in the weld list information, and reinspect and record the missed inspection positions.

[0029] The process of reinspecting and recording the missed inspection positions described in Step 7 is as follows:

[0030] The length of a single-section pipeline is limited. After generating the weld list, scan the overall distribution of circumferential welds. When no circumferential weld is found after a length greater than the longest length of a single-section pipeline, re-extract the data in this interval and repeat Steps 4 - 6; at the same time, lower the threshold set in Step 4. If a weld is found, supplement it to the list. If no weld is found, add the midpoint coordinates of this section of the pipeline as weld data to the list and mark that there is a suspected undetected weld here for manual review by data personnel.

[0031] Compared with the prior art, the present invention evenly arranges several magnetic flux leakage sensors in the pipeline, sets mileage sensors around the pipeline, and synchronously collects multi-channel raw magnetic flux leakage data and mileage wheel data by sampling at equal time intervals; then uses median filtering to obtain the binary map of pipeline magnetic flux leakage distribution, and uses asymmetric closing operation to enhance the morphological characteristics of weld data; uses Hough transform to extract straight line segments in the binary map to determine whether there is a missed weld. The present invention uses a two-dimensional data analysis method to perform a similar image processing method on the data to enhance the data, highlight the morphological characteristics of the uneven part in the magnetic flux leakage sensor data, and capture the ring weld from it. It can make judgments in a two-dimensional space, ensuring its accuracy, and because of the use of early enhancement methods such as adaptive thresholds, the sensor calibration requirements are not high. At the same time, the final judgment method in the present invention has a low dimension, and the judgment is made from the overall morphology of the uneven connected domain, which does not require high computing power of the computer. In addition, the circumferential period extension method is used in the judgment, and its reliability is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a flow chart of the present invention;

[0033] Figure 2 Schematic diagram of the fluctuation of magnetic leakage data near a certain weld in an embodiment of the present invention;

[0034] Figure 3 is a grayscale image processed by an adaptive threshold in an embodiment of the present invention;

[0035] Figure 4 A schematic diagram of a girth weld extracted using Hough transform in an embodiment of the present invention;

[0036] Figure 5 This is the fluctuation of magnetic flux leakage data near a certain weld in an embodiment of the present invention. In the figure, there is a short pipe and the positions of two ring welds are very close.

[0037] Figure 6 Schematic diagram of a weld signal converted into a grayscale image through an adaptive threshold in an embodiment of the present invention;

[0038] Figure 7 Schematic diagram of two welds extracted by signal enhancement in an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The present invention will be further described below in conjunction with the accompanying drawings.

[0040] like Figure 1 As shown, a method for identifying girth welds in pipeline inspection includes the following steps:

[0041] Step 1: Evenly arrange several magnetic flux leakage sensors in the pipeline, set up mileage sensors around the pipeline, and synchronously collect multi-channel raw magnetic flux leakage data and mileage wheel data by sampling at equal time intervals;

[0042] Step 2: First, convert the odometer wheel data into actual mileage values, and then obtain the leakage magnetic data of the mileage coordinates through time alignment; compare the odometer wheel data within the set time range, calculate the number of pulses of each odometer wheel data, select the odometer wheel corresponding to the maximum number of pulses, calculate the difference of this odometer wheel, and obtain the mileage pulse distribution of the odometer wheel; calculate the actual mileage value of the odometer wheel according to the mileage pulse distribution; reconstruct the data by making the leakage magnetic data and the actual mileage value at the same time correspond one by one;

[0043] Step 3: Use the median filter method to process the data in step 2 and separate the leakage magnetic data representing different pipe sections, as follows:

[0044] Step 31: When a magnetic flux leakage sensor of an internal detector is tested before delivery, under natural conditions, the value of the magnetic flux leakage sensor of this specification fluctuates around the center value N, and the fluctuation range is [-M / 2, M / 2];

[0045] Step 32: performing median filtering on the magnetic flux leakage data processed in step 2 according to the channel to obtain the median value of the magnetic flux leakage data of each channel;

[0046] Step 33: Draw an intermediate value curve, wherein the intermediate value curve includes weld data and pipeline flat portion data, and the ratio of the line width of the weld data to the line width of the pipeline flat portion is 1:20;

[0047] Step 4: Subtract the leakage magnetic flux data of step 2 from the intermediate value data of step 3, and obtain the pipeline leakage magnetic flux distribution binary map through the threshold, specifically:

[0048] Step 41: Determine a threshold value M according to the fluctuation range of the magnetic flux leakage data obtained in step 31;

[0049] Step 42: Subtract the leakage magnetic flux data obtained in step 2 from the intermediate value data obtained in step 3, and then perform threshold binarization processing to obtain a binarized image, assign a color to the part exceeding the threshold value to represent the weld data, and assign another color to the part not exceeding the threshold value to represent the flat part of the pipeline;

[0050] Step 5: Process the binary image using an asymmetric closing operation to enhance the morphological features of the weld data in the detection data; use an asymmetric closing operation, set the dilation scale to be larger than the erosion scale, and connect multiple connected domains to each other; among them, according to the average width W of the weld of the pipeline to be measured and the minimum mileage resolution s of the detector, the side length value L of the square dilation kernel can be set to W / s, rounded up to an odd number, and the side length of the square erosion kernel is L - 2;

[0051] Step 6: Use the Hough transform to extract the straight line segments in the binary image, and determine whether they are circumferential welds. If they are circumferential welds, register the weld information in the weld list;

[0052] Step 61: Find the straight line segments with a length of 2 / 3 of the pipeline circumference, and these straight line segments are uniformly extracted as circumferential welds;

[0053] Step 62: Since the circumferential weld is perpendicular to the pipeline extension direction, according to the tilt angle existing in the operation of the detector, the straight line segments that satisfy the following formula are also extracted together:

[0054] |(x2 - x1) / (y2 - y1)| < tan(5°)

[0055] In the formula, (x1, y1), (x2, y2) are the coordinates of the two endpoints of the straight line segment;

[0056] Step 63: Remove multiple straight line segments in the same weld obtained in Step 61 and Step 62, and only leave one straight line segment to represent the circumferential weld:

[0057] First, represent the straight line segment with the mileage coordinate of the midpoint of the straight line segment, denoted as X, and X = (x1 + x2) / 2. The corresponding serial number data is Xn, and it is represented by the set {Xn}, where n is the number of straight line segments;

[0058] Then sort the data in the set {Xn}, select the first data X1, and remove all X data whose distance to X1 is less than d. The calculation formula of d is:

[0059] d = w / 2s

[0060] In the formula: w is the shortest distance between two circumferential welds in the pipeline construction standard;

[0061] For the first data X1, the subset {Xm1||Xm1 - X1| < d} of {Xn}, then all Xm belong to the same weld as X1 and are removed, leaving only the first data X1 to represent this weld; after removing {Xm1||Xm1 - X1| < d} from {Xn}, continue to loop this process until all elements in the set {Xn} are traversed, and finally obtain a set containing a list of mileage values, and each mileage value corresponds to a circumferential weld;

[0062] Step 7, determine whether there is a possibility of missed welds based on the weld list information, and re-check and record the missed welds.

[0063] The process of rechecking and recording the missed inspection positions described in step 7 is as follows:

[0064] The length of a single section of pipeline is limited. After generating the weld list, scan the overall distribution of circumferential welds. When the length is greater than the longest length of the single section of pipeline and no circumferential welds are found, re-extract the data in this interval and repeat steps 4-6. At the same time, lower the threshold set in step 4. If a weld is found, add it to the list. If no weld is found, add the midpoint coordinates of this section of pipeline as weld data to the list, and mark this place as suspected undetected welds for manual review by data personnel.

[0065] An embodiment of the present invention is provided below, comprising the following steps:

[0066] Step 1: Collect the original magnetic flux leakage data and mileage wheel data. Use the composite magnetic flux leakage probe (or magnetic flux leakage sensor) to synchronously collect the original magnetic flux leakage data in equal time intervals. At the same time, use the mileage sensor to synchronously sample the mileage data information in the pipeline. Figure 2 and Figure 5 is the collected raw data of magnetic leakage;

[0067] Step 2: Generate magnetic flux leakage data of mileage coordinates. The mileage data sampled by the mileage sensor in step 1 is a pulse signal. According to the number of pulses corresponding to one rotation of the mileage wheel and the circumference of the mileage wheel, the arc on the mileage wheel corresponding to two adjacent pulses is calculated to obtain the minimum unit of the mileage value. The pulse number and the minimum unit are multiplied to convert the pulse signal into the actual corresponding mileage value. In this embodiment, three mileage sensors are arranged around the pipeline. The number of pulses passed by the three mileage wheels is calculated within a certain time range (for example, 5s). The wheel with the largest number of pulses is selected, and the difference of this wheel is calculated to obtain the distribution of mileage pulses. According to the mileage pulse distribution and the minimum unit, the actual mileage value traveled by the selected mileage wheel can be calculated. The magnetic flux leakage data of the same time is reconstructed by corresponding the actual mileage value one by one.

[0068] During data reconstruction, when the differential value of the selected mileage wheel is less than 1, a frame of data representing the current mileage is selected as the actual mileage value; when the differential value is N>1, the data of this frame is copied N times to compensate for the mileage difference, ensuring that the actual mileage value of the overall data is evenly distributed.

[0069] Step 3: Perform median filtering on the magnetic flux leakage data by channel to obtain the magnetic flux leakage median value data of each channel.

[0070] Since different sections of the tested pipeline may have differences in material, wall thickness, inner diameter, etc., these differences will affect the base value of the leakage magnetic detection, that is, when the pipeline is flat, the values ​​measured in different sections will also be different, which will affect the setting of the overall pipeline discrimination parameter threshold. Through experiments, this embodiment knows that the fluctuation of the detector is in a relatively stable range, in [-2,2], but the center value of the fluctuation will change with the above differences. In order to improve the adaptability of the analysis, this embodiment introduces median filtering to deal with these differences. For a detector with a sampling frequency of 2000Hz, assuming that the mileage wheel generates 256 pulses per week and the diameter of the mileage wheel is 80mm, its mileage resolution is 0.98mm, and the weld is usually around 10mm. The one-dimensional median of 101 points in each channel is calculated to ensure that the weld data line width is much smaller than the flat part. After the median is calculated, the overall trend of the curve is obtained to distinguish different pipe sections.

[0071] Step 4: Subtract the leakage magnetic flux data of step 2 from the intermediate value data of step 3, and obtain the pipeline leakage magnetic flux distribution binary map through the threshold, such as Figure 3 and Figure 6 As shown;

[0072] In this embodiment, the threshold is set to 4, and the leakage magnetic data obtained in step 2 is subtracted from the intermediate value data obtained in step 3. After threshold binarization processing, the part exceeding the threshold is assigned white. In the binary diagram, the vertical direction represents the distribution of sensors arranged circumferentially within the detector, and the horizontal direction represents the mileage position at which the data is collected.

[0073] Step 5: Perform an asymmetric closing operation on the image to strengthen the detection and close the uneven parts;

[0074] Since the image directly obtained by binarization will have broken white parts due to the unevenness of the weld itself, in order to connect the broken parts and prevent the scale of the white part from changing too much, a closing operation is chosen here. The general closing operation has a better effect on the noise points inside a connected domain, but it is not good for the connection between two connected domains. An asymmetric closing operation is used here, that is, the scale of expansion is larger than the scale of corrosion, so that the connection between multiple connected domains can be achieved. In order to ensure the longitudinal connection, the expansion kernel is a rectangle with a horizontal size of 7 and a vertical size of 9, and the corrosion kernel is a rectangle with a horizontal size of 5 and a vertical size of 7, resulting in the following Figure 4 and Figure 7 Results shown.

[0075] Step 6: Extract the white part features in the binary image, determine whether it is a girth weld, and register the weld information into the weld list;

[0076] In the binary image, the white part is the uneven part of the pipeline, which contains the characteristics of the pipeline, including defects, pipeline components and weld information. The girth weld is obviously different from other features in shape. The girth weld is perpendicular to the pipeline direction and its length is equal to the circumference of the pipeline. In the binary image, the girth weld is a longitudinal straight line segment. Hough transform is used here to extract the straight line features and to achieve weld screening by setting some filtering conditions.

[0077] Hough transform extracts straight line segments from the white part. As long as the length direction of these straight line segments is within the white area, they can be extracted, so some conditions need to be set for screening. The data obtained by Hough transform are the two endpoints of the straight line segment (x1, y1) and (x2, y2).

[0078] In this embodiment, the length is first used to extract the straight line segment (condition 1). The length of the weld is the circumference of the pipeline. Considering the possibility of connection breakage, the length is set to 2 / 3 of the circumference. For example, for a detector with 264 magnetic flux leakage channels, the straight line obtained by Hough transform has a minimum length of 176. This step can be directly set in Hough transform, and the obtained line segment is the one that meets the conditions.

[0079] In this embodiment, the slope angle is used to extract the straight line segment (condition 2). The circumferential weld is perpendicular to the extension direction of the pipeline. Considering that the detector sometimes has a certain inclination during operation, the angle swing range is set to 5°.

[0080] |(x2-x1) / (y2-y1)|<tan(5°)

[0081] This embodiment uses the weld width to process the line segments selected by length and inclination angle:

[0082] Since the weld has a certain width, there must be a situation where there are still multiple straight lines collected by the Hough transform in the same weld after conditions 1 and 2. At this time, the line segment data is processed, and each data is represented by two endpoints (x1, y1) and (x2, y2). Each data is represented by (x1+x2) / 2, recorded as X, and the corresponding serial number data is Xn, that is, the mileage coordinates of the midpoint of the line segment are used to represent this line segment, represented by the set {Xn}, and then the mileage data is sorted, the first data X1 is selected, and all X data with a distance less than d to X1 are removed. Usually the weld width is about 10mm, and the closest adjacent welds are about 200mm. Here, d is set to the number of pixels corresponding to 100mm to distinguish line segments belonging to different welds. Take the mileage resolution of 0.98mm as an example:

[0083] d=100 / 0.98

[0084] For X1, the subset {Xm1||Xm1 - X1| < d} of {Xn} is considered such that all Xm belong to the same weld seam as X1 and should be removed, leaving only X1 to represent this weld seam. After removing {Xm1||Xm1 - X1| < d} from {Xn}, this process is continued in a loop until all elements in the set {Xn} are traversed. The resulting set is a list of mileage values, with each mileage value corresponding to a circumferential weld seam.

[0085] Step 7: Determine whether there is a possibility of undetected weld seams in the weld seam list information, and reinspect and record the undetected positions.

[0086] There is a maximum value for the spacing between weld seams. The longest single-section pipeline for laying pipelines does not exceed 13 meters. After generating the weld seam list, the overall circumferential weld seam distribution will be scanned. When there is no circumferential weld seam for more than 15 meters, the data in this interval will be re-extracted, and steps 4 - 6 will be performed again. The threshold value in step 4 will be reduced from 4 to 3. If a weld seam is found, it will be added to the list. If no weld seam is found, the midpoint coordinates of this section of the pipeline will be used as weld seam data and added to the list, and it will be marked that there is a suspected undetected weld seam here for data personnel to manually review.

Claims

1. A method for identifying girth welds in pipeline inspection, characterized in that: The following steps are involved: Step 1: Evenly arrange several magnetic flux leakage sensors in the pipeline, set up mileage sensors around the pipeline, and synchronously collect multi-channel raw magnetic flux leakage data and mileage wheel data by sampling at equal time intervals; Step 2: First, convert the odometer wheel data into actual mileage values, and then obtain the leakage magnetic data of the mileage coordinates through time alignment; compare the odometer wheel data within the set time range, calculate the number of pulses of each odometer wheel data, select the odometer wheel corresponding to the maximum number of pulses, calculate the difference of this odometer wheel, and obtain the mileage pulse distribution of this odometer wheel; Calculate the actual mileage value of the mileage wheel according to the mileage pulse distribution; reconstruct the data by making a one-to-one correspondence between the leakage magnetic data and the actual mileage value at the same time; Step 3: Use the median filter method to process the data in step 2 and separate the leakage magnetic data representing different pipe sections, as follows: Step 31: When a magnetic flux leakage sensor of an internal detector is tested before delivery, under natural conditions, the value of the magnetic flux leakage sensor of this specification fluctuates around the center value N, and the fluctuation range is [-M / 2, M / 2]; Step 32: performing median filtering on the magnetic flux leakage data processed in step 2 according to the channel to obtain the median value of the magnetic flux leakage data of each channel; Step 33: Draw an intermediate value curve, wherein the intermediate value curve includes weld data and pipeline flat portion data, and the ratio of the line width of the weld data to the line width of the pipeline flat portion is 1:20; Step 4: Subtract the leakage magnetic flux data of step 2 from the intermediate value data of step 3, and obtain the pipeline leakage magnetic flux distribution binary map through the threshold, specifically: Step 41: Determine a threshold value M according to the fluctuation range of the magnetic flux leakage data obtained in step 31; Step 42: Subtract the leakage magnetic flux data obtained in step 2 from the intermediate value data obtained in step 3, and then perform threshold binarization processing to obtain a binarized image, assign a color to the part exceeding the threshold value to represent the weld data, and assign another color to the part not exceeding the threshold value to represent the flat part of the pipeline; Step 5: Use asymmetric closing operation to process the binary image and strengthen the morphological characteristics of the weld data in the detection data; use asymmetric closing operation, set the expansion scale larger than the corrosion scale, and connect multiple connected domains to each other; wherein, according to the average width W of the pipeline weld under test and the minimum mileage resolution s of the detector, set the side length value L of the square expansion kernel to W / s, take an odd number upward, and the side length of the square corrosion kernel to L-2; Step 6: Use Hough transform to extract the straight line segment in the binary image and determine whether it is a girth weld. If it is a girth weld, register the weld information in the weld list; Step 61: Find straight line segments with a length of 2 / 3 of the pipe circumference, and these straight line segments are uniformly extracted as girth welds; Step 62: Since the girth weld is perpendicular to the extension direction of the pipeline, according to the inclination angle existing during the operation of the detector, the straight line segment satisfying the following formula is also extracted: |(x2-x1) / (y2-y1)|<tan(5°) Where (x1, y1) and (x2, y2) are the coordinates of the two endpoints of the straight line segment; Step 63: Remove multiple straight line segments in the same weld obtained in Steps 61 and 62, and only keep one straight line segment to represent the circumferential weld: First, represent the straight line segment with the mileage coordinate of the midpoint of the straight line segment, denoted as X, and X = (x1 + x2) / 2. The corresponding serial number data is Xn, which is represented by the set {Xn}, where n is the number of straight line segments; Then, sort the data in the set {Xn}, select the first data X1, and remove all X data whose distance to X1 is less than d. The calculation formula for d is: d = w / 2s In the formula: w is the shortest distance between two circumferential welds in the pipeline construction standard; For the first data X1, the subset of {Xn} is {Xm1||Xm1 - X1| < d}. Then all Xm belong to the same weld as X1 and are removed, only leaving the first data X1 to represent this weld. After {Xn} removes {Xm1||Xm1 - X1| < d}, continue to loop this process until all elements in the set {Xn} are traversed, and finally obtain a set containing a list of mileage values, with each mileage value corresponding to a circumferential weld; Step 7: Determine whether there is a possibility of missed welds in the weld list information, and reinspect and record the missed positions.

2. A method for identifying girth welds in pipeline inspection according to claim 1, characterized in that: The process of reinspecting and recording the missed positions described in Step 7 is as follows: The length of a single section of the pipeline is limited. After generating the weld list, scan the overall circumferential weld distribution. When no circumferential weld is found after exceeding the longest length of a single section of the pipeline, re-extract the data in this interval and repeat Steps 4 - 6; at the same time, lower the threshold set in Step 4. If a weld is found, supplement it to the list. If no weld is found, add the midpoint coordinates of this section of the pipeline as weld data to the list and mark that there is a suspected undetected weld here for manual review by data personnel.

Images