A method, device and medium for identifying longitudinal defects based on a steel pipe ring array phased array image

Abstract

The application discloses a kind of based on steel pipe ring array phased array image longitudinal defect identification method, equipment and medium, it is relevant to image analysis processing technical field, including, acquisition steel pipe longitudinal detection image and ring array phased array detection data, the steel pipe profile in steel pipe longitudinal detection image is identified, the axial position of steel pipe profile is regarded as registration datum, steel pipe longitudinal detection image under different visual angle and ring array phased array detection data are corrected to same image coordinate, form steel pipe longitudinal image registration information;According to steel pipe longitudinal image registration information, pixel position corresponding processing is carried out to the steel pipe longitudinal detection image after correction, ring array phased array detection data is converted into the ring array phased array imaging information corresponding to steel pipe axial position, the image pixel corresponding to same steel pipe axial position under different visual angle is fused with ring array phased array imaging information, generates steel pipe longitudinal reconstruction image information.The application realizes defect authenticity screening, type identification and axial position association.

Description

Technical Field

[0001] This invention relates to the field of image analysis and processing technology, and in particular to a method, device and medium for longitudinal defect identification based on phased array images of steel pipe ring arrays. Background Technology

[0002] With the increasing application of steel pipes in oil and gas transportation, pressure vessels, and structural components, the accuracy of longitudinal defect detection directly affects their service safety and quality assessment. Current steel pipe defect detection methods are gradually evolving from manual visual inspection, magnetic particle testing, eddy current testing, and ultrasonic testing to industrial visual inspection and image recognition inspection. This typically involves acquiring images of the steel pipe surface and combining them with image recognition techniques such as grayscale enhancement, edge extraction, image segmentation, and feature classification to identify abnormal areas such as cracks and scratches. Some technologies employ multi-view image acquisition and image recognition analysis to improve the coverage of the steel pipe surface and the efficiency of online inspection.

[0003] Existing image recognition-based technologies for detecting longitudinal defects in steel pipes still have shortcomings: On the one hand, longitudinal images of steel pipes from different perspectives have inconsistent coordinates and inaccurate axial position correspondences, which makes it easy for the same longitudinal defect to be segmented into multiple discontinuous regions during image recognition; on the other hand, existing image recognition methods focus on the segmentation of brightness and darkness anomalies in a single image, lacking joint analysis of the axial continuity of longitudinal defects, cross-viewpoint consistency, and strip morphology changes, which can easily lead to misjudging illumination fluctuations, surface textures, and isolated noise as longitudinal defects, and affect the stable correlation between defect type and the axial position of the steel pipe. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a longitudinal defect recognition method based on steel pipe ring array phased array images to solve the problems of inaccurate axial correspondence of multi-view steel pipe longitudinal images and insufficient accuracy in longitudinal defect recognition and positioning in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for longitudinal defect recognition based on steel pipe ring array phased array images, comprising: acquiring longitudinal inspection images of steel pipes and ring array phased array inspection data; identifying the steel pipe contour in the longitudinal inspection images of steel pipes and using the axial position of the steel pipe contour as a registration reference; correcting the longitudinal inspection images of steel pipes and the ring array phased array inspection data from different viewpoints to the same image coordinates to form steel pipe longitudinal image registration information; performing pixel position correspondence processing on the corrected longitudinal inspection images of steel pipes based on the steel pipe longitudinal image registration information; converting the ring array phased array inspection data into ring array phased array imaging information corresponding to the axial position of the steel pipe; fusing the image pixels corresponding to the same axial position of the steel pipe from different viewpoints with the ring array phased array imaging information to generate steel pipe longitudinal reconstructed image information; extracting abnormal brightness and darkness variation regions that extend continuously along the axial position of the steel pipe from the steel pipe longitudinal reconstructed image information, and combining them with... The abnormal echo response corresponding to the axial position of the steel pipe in the ring phased array imaging information is used to segment the image based on the grayscale difference between the abnormal brightness change area and the surrounding background image area, obtaining abnormal image regions and marking them as longitudinal defect candidate image region information. From the longitudinal defect candidate image region information, longitudinal defect candidate image region information corresponding to the same axial position of the steel pipe from different viewpoints is read, and combined with the abnormal echo response corresponding to the axial position of the steel pipe in the ring phased array imaging information, the longitudinal defect candidate image region information with corresponding extension directions and consistent abnormal echo responses is image matched to obtain effective longitudinal defect image region information. Strip morphology features are extracted from the effective longitudinal defect image region information, and the defect type is identified based on the strip morphology features and abnormal echo response. The defect type is then associated with the axial position of the steel pipe corresponding to the effective longitudinal defect image region information to generate longitudinal defect image recognition and positioning information for the steel pipe.

[0007] As a preferred embodiment of the image recognition-based method for identifying longitudinal defects in steel pipes according to the present invention, the specific steps for forming longitudinal image registration information of the steel pipe are as follows: Longitudinal detection images of steel pipes from different perspectives and corresponding axial position detection data of steel pipes from ring array phased array are collected. After grayscale standardization of the longitudinal detection images of steel pipes, the boundary between the steel pipe image area and the background image area is identified. The axial position of the steel pipe contour is determined based on the contour center line of the steel pipe image area. Then, the axial sampling position of the steel pipe corresponding to the ring array phased array detection data is matched with the axial position of the steel pipe contour to form axial position reference information. Based on the axial position reference information, the image coordinate transformation relationship between the longitudinal inspection images of the steel pipe under different perspectives is established, and the axial position correspondence between the ring array phased array inspection data and the same image coordinate is determined. Then, the image coordinate transformation relationship is used to correct the longitudinal inspection images of the steel pipe under different perspectives to the same image coordinate, and the ring array phased array inspection data is mapped to the same image coordinate according to the axial position correspondence, thus forming the longitudinal image registration information of the steel pipe.

[0008] As a preferred embodiment of the image recognition-based method for identifying longitudinal defects in steel pipes according to the present invention, the specific steps for generating longitudinal reconstructed image information of the steel pipe are as follows: The corrected longitudinal detection images of the steel pipe under different viewpoints, the corresponding image coordinate transformation relationship, the ring array phased array detection data and the axial position correspondence relationship are read from the longitudinal image registration information of the steel pipe. The pixel grid of the corrected longitudinal detection images of the steel pipe is read to obtain the image pixel information under different viewpoints. The detection response corresponding to the axial position of each steel pipe in the ring array phased array detection data is read according to the axial position correspondence relationship, and then the detection response is converted into the ring array phased array imaging information corresponding to the axial position of the steel pipe. Using the image coordinate transformation relationship as a reference, the axial position of the steel pipe is used to read the image pixels corresponding to the same axial position of the steel pipe and within the overlapping projection range of adjacent viewpoints from the image pixel information of different viewpoints, and the imaging response corresponding to the same axial position of the steel pipe is read from the ring array phased array imaging information; the image pixels corresponding to the same axial position of the steel pipe and the imaging response are fused to generate the longitudinal reconstruction image information of the steel pipe.

[0009] As a preferred embodiment of the image recognition-based method for identifying longitudinal defects in steel pipes according to the present invention, the specific steps for obtaining abnormal image regions and marking them as candidate image regions for longitudinal defects are as follows: The longitudinal reconstruction image of the steel pipe and the imaging information of the ring phased array are read from the longitudinal reconstruction image information of the steel pipe. The background gray-level changes in the longitudinal reconstruction image of the steel pipe are continuously and smoothly estimated along the axial position of the steel pipe. The difference between the longitudinal reconstruction image of the steel pipe and the continuously smoothed estimated background gray-level changes is compared to obtain the bright and dark change regions with obvious gray-level differences and continuous extension along the axial position of the steel pipe. Then, the imaging response corresponding to the axial position of the steel pipe is read from the imaging information of the ring phased array. The axial position of the steel pipe corresponding to the bright and dark change region is matched with the imaging response, and the abnormal echo response corresponding to the bright and dark change region is extracted to form the abnormal bright and dark change region information. Based on the information of abnormal brightness and darkness change areas, the grayscale difference between the abnormal brightness and darkness change areas and the surrounding background image areas is compared. Combined with the abnormal echo response of the corresponding steel pipe axial position, the image positions with obvious grayscale differences and prominent abnormal echo responses are marked. The marked grayscale change positions are connected to form a segmentation boundary, and the abnormal brightness and darkness change areas are extracted. Then, the discrete parts that do not extend continuously along the steel pipe axial position and whose abnormal echo responses are discontinuous are removed to obtain the abnormal image regions. The abnormal image regions, the corresponding steel pipe axial positions, and the abnormal echo responses are recorded to form longitudinal defect candidate image region information.

[0010] As a preferred embodiment of the image recognition-based method for identifying longitudinal defects in steel pipes according to the present invention, the specific steps of reading longitudinal defect candidate image region information corresponding to the same axial position of the steel pipe from different viewpoints from the longitudinal defect candidate image region information, and combining it with the abnormal echo response corresponding to the axial position of the steel pipe in the ring array phased array imaging information, are as follows: Read the region location attributes and abnormal echo responses corresponding to each longitudinal defect candidate image region from the longitudinal defect candidate image region information, and perform topological coding according to the front-back relationship of each longitudinal defect candidate image region in the axial position of the steel pipe within the same viewpoint. Then, associate and record the abnormal echo responses with the corresponding axial positions of the steel pipe to form axial topological information of the candidate region. The axial start and end positions and front and back order of each longitudinal defect candidate image region are determined by using the axial topology information of the candidate region. The longitudinal defect candidate image regions with overlapping axial intervals and consistent front and back order under different viewpoints are aligned, and the intersection of the axial intervals after alignment is determined as the overlapping axial range. Then, the abnormal echo response of the corresponding axial position of the steel pipe within the overlapping axial range is read from the ring array phased array imaging information to form abnormal echo response information.

[0011] As a preferred embodiment of the image recognition-based method for identifying longitudinal defects in steel pipes according to the present invention, the specific steps for obtaining effective longitudinal defect image region information are as follows: Within the overlapping axial range, the change of the center extension line of the corresponding longitudinal defect candidate image region is tracked to obtain the change of the extension direction. The abnormal echo response of the corresponding steel pipe axial position is read from the ring phased array imaging information. The change of the extension direction, the abnormal echo response and the axial topology information of the candidate region are compared and the cross-view candidate region matching consistency value is calculated. When the cross-view candidate region matching consistency value meets the matching requirements, the longitudinal defect candidate image region information with corresponding extension directions and consistent abnormal echo responses is determined as the candidate region direction matching information. Based on the cross-view candidate region matching consistency value in the candidate region direction matching information, the longitudinal defect candidate image region information that meets the matching requirements and has consistent abnormal echo response is retained and determined as the valid longitudinal defect image region information.

[0012] As a preferred embodiment of the image recognition-based method for identifying longitudinal defects in steel pipes according to the present invention, the specific steps for extracting strip morphological features from the effective longitudinal defect image region information are as follows: The effective longitudinal defect image region, the corresponding axial position of the steel pipe, and the abnormal echo response are read from the effective longitudinal defect image region information. The center extension line of the effective longitudinal defect image region is used as the reference. The transverse grayscale sections of the effective longitudinal defect image region are sequentially cut along the axial position of the steel pipe and arranged in order. Then, the axial position of the steel pipe corresponding to each transverse grayscale section and the abnormal echo response are recorded to form the strip skeleton status information. The transverse grayscale sections arranged along the axial position of the steel pipe in the strip skeleton status information are taken as the processing objects. The boundary position differences and boundary grayscale differences in adjacent transverse grayscale sections are compared in turn to obtain the width change and boundary grayscale change of the effective longitudinal defect image area along the axial position of the steel pipe. Then, the abnormal echo response of the corresponding axial position of the steel pipe is read from the strip skeleton status information, and the abnormal echo response is recorded in correspondence with the width change and boundary grayscale change to form the strip morphological features.

[0013] As a preferred embodiment of the image recognition-based method for identifying longitudinal defects in steel pipes according to the present invention, the specific steps for generating image recognition and positioning information of longitudinal defects in steel pipes are as follows: The defect type is determined by utilizing the width and boundary grayscale variations in the strip morphology features, combined with the abnormal echo response at the corresponding axial position of the steel pipe. When the effective longitudinal defect image area maintains a narrow and continuous width along the axial position of the steel pipe and the grayscale variation at the end is obvious, the corresponding defect type is determined to be a longitudinal crack. When the effective longitudinal defect image area maintains a wide and continuous width along the axial position of the steel pipe and is close to the contour boundary of the steel pipe image area, the corresponding defect type is determined to be a longitudinal scratch. After the defect type is determined, the axial position of the steel pipe and the abnormal echo response corresponding to the strip morphology features are read from the effective longitudinal defect image area information. The defect type, the axial position of the steel pipe, the effective longitudinal defect image area and the abnormal echo response are recorded accordingly to generate the longitudinal defect image recognition and positioning information of the steel pipe.

[0014] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the image recognition-based longitudinal defect identification method for steel pipes as described in the first aspect of the present invention.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the image recognition-based longitudinal defect identification method for steel pipes as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By performing image recognition registration on longitudinal inspection images of steel pipes from different perspectives according to the axial position of the steel pipe contour, and fusing image pixels at the same axial position of the steel pipe, the corresponding expression of multi-view images under a unified image coordinate system is realized, thereby improving the consistency of image recognition position and the accuracy of candidate region extraction; by performing axial topological coding, determining the overlapping axial range, and cross-view image recognition matching judgment on the longitudinal defect candidate image region, and extracting the strip morphological features of the effective longitudinal defect image region, the authenticity screening, type identification, and axial position association of defects are realized, thereby reducing false detections and improving the reliability and positioning stability of longitudinal defect image recognition. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Fig. 1 This is a flowchart of a method for identifying longitudinal defects in steel pipes based on image recognition.

[0019] Fig. 2 A flowchart for generating longitudinal image registration information for steel pipes.

[0020] Fig. 3 A flowchart for generating candidate image region information for longitudinal defects.

[0021] Fig. 4 A flowchart for generating image recognition and location information for longitudinal defects in steel pipes. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figs. 1-4 This is one embodiment of the present invention, which provides a method for identifying longitudinal defects based on phased array images of steel pipe ring arrays, comprising the following steps: S1. Collect longitudinal inspection images of steel pipes and phased array inspection data of ring array, identify the outline of steel pipes in the longitudinal inspection images of steel pipes, and use the axial position of the steel pipe outline as the registration reference. Correct the longitudinal inspection images of steel pipes and phased array inspection data of ring array from different perspectives to the same image coordinates to form longitudinal image registration information of steel pipes.

[0026] Longitudinal inspection images of steel pipes from different perspectives and corresponding phased array inspection data of steel pipe axial positions are collected. After grayscale standardization of the longitudinal inspection images of steel pipes, the boundary between the steel pipe image region and the background image region is identified. The axial position of the steel pipe contour is determined based on the contour center line of the steel pipe image region. Then, the axial sampling position of the steel pipe corresponding to the phased array inspection data is matched with the axial position of the steel pipe contour to form axial position reference information.

[0027] The specific process includes: image acquisition equipment is positioned at different observation positions around the circumference of the steel pipe and faces the outer surface of the steel pipe to acquire images of the outer surface of the steel pipe along its length within the image imaging range, obtaining longitudinal inspection images of the steel pipe from different perspectives; a ring array phased array detection device is arranged around the circumference of the steel pipe and acquires ring array phased array detection data corresponding to the axial position of the steel pipe; grayscale values ​​of pixels in the longitudinal inspection images of the steel pipe are standardized, the grayscale distribution range of the longitudinal inspection images of the steel pipe is statistically analyzed, and linear mapping is performed according to a unified grayscale expression range to ensure that the longitudinal inspection images of the steel pipe from different perspectives have consistent characteristics. Grayscale representation: In the longitudinal inspection image of the steel pipe after grayscale normalization, the boundary position is identified based on the grayscale change between the steel pipe image region and the background image region, and the continuous boundary positions are connected to form the boundary between the steel pipe image region and the background image region; the middle position of the boundary is continuously determined along the length direction of the steel pipe within the steel pipe image region to obtain the contour center line of the steel pipe image region, and the axial position of the steel pipe corresponding to the contour center line of the steel pipe image region is determined as the axial position of the steel pipe contour. The axial sampling position of the steel pipe corresponding to the ring array phased array detection data is correlated with the axial position of the steel pipe contour to form the axial position reference information.

[0028] It should be noted that the grayscale distribution range refers to the range between the minimum and maximum grayscale values ​​of pixels in the longitudinal inspection image of the steel pipe. The unified grayscale expression range is 0 to 255. During grayscale standardization, the pixel grayscale values ​​in the longitudinal inspection image of the steel pipe are linearly stretched according to the minimum and maximum grayscale values ​​so that the longitudinal inspection images of the steel pipe from different perspectives are mapped to the unified grayscale expression range of 0 to 255.

[0029] Based on the axial position reference information, the image coordinate transformation relationship between the longitudinal inspection images of the steel pipe under different perspectives is established, and the axial position correspondence between the ring array phased array inspection data and the same image coordinate is determined. Then, the image coordinate transformation relationship is used to correct the longitudinal inspection images of the steel pipe under different perspectives to the same image coordinate, and the ring array phased array inspection data is mapped to the same image coordinate according to the axial position correspondence, thus forming the longitudinal image registration information of the steel pipe.

[0030] The specific process includes: reading the contour centerline of the steel pipe image region from different viewpoints and the axial sampling position of the steel pipe corresponding to the ring array phased array detection data from the axial position reference information; extracting the image coordinate position corresponding to each axial position of the steel pipe along the contour centerline of the steel pipe image region; comparing the image coordinate positions representing the same axial position of the steel pipe from different viewpoints to obtain the position offset, direction deviation, and scale difference between the longitudinal detection images of the steel pipe from different viewpoints; determining the image coordinate correction parameters based on the position offset, direction deviation, and scale difference; and forming the image coordinate transformation relationship between the longitudinal detection images of the steel pipe from different viewpoints based on the image coordinate correction parameters; and corresponding the axial sampling position of the steel pipe corresponding to the ring array phased array detection data with the axial position of the steel pipe in the same image coordinate to determine the axial position correspondence between the ring array phased array detection data and the same image coordinate.

[0031] The position offset, orientation deviation, and scale difference of the longitudinal inspection images of the steel pipe under different viewpoints are read from the image coordinate transformation relationship. The image coordinate positions in the longitudinal inspection images of the steel pipe are translated and corrected according to the position offset. The orientation of the image coordinate positions in the longitudinal inspection images of the steel pipe is corrected according to the orientation deviation. The scale of the image coordinate positions in the longitudinal inspection images of the steel pipe is corrected according to the scale difference. The corrected image coordinate positions are mapped to the same image coordinate system. According to the axial position correspondence between the ring array phased array detection data and the same image coordinate system, the ring array phased array detection data is mapped to the same image coordinate system. The corrected longitudinal inspection images of the steel pipe under different viewpoints, the corresponding image coordinate transformation relationship, the ring array phased array detection data and the axial position correspondence relationship are recorded to form the longitudinal image registration information of the steel pipe.

[0032] S2. Based on the longitudinal image registration information of the steel pipe, the pixel position correspondence processing is performed on the corrected longitudinal detection image of the steel pipe, and the ring array phased array detection data is converted into ring array phased array imaging information corresponding to the axial position of the steel pipe. The image pixels corresponding to the same axial position of the steel pipe under different viewpoints are fused with the ring array phased array imaging information to generate longitudinal reconstruction image information of the steel pipe.

[0033] The corrected longitudinal detection images of the steel pipe under different viewpoints, the corresponding image coordinate transformation relationship, the ring array phased array detection data and the axial position correspondence are read from the longitudinal image registration information of the steel pipe. The pixel grid of the corrected longitudinal detection images of the steel pipe is read to obtain the image pixel information under different viewpoints. The detection response corresponding to the axial position of each steel pipe in the ring array phased array detection data is read according to the axial position correspondence, and then the detection response is converted into the ring array phased array imaging information corresponding to the axial position of the steel pipe.

[0034] The specific process includes: reading the corrected longitudinal detection images of the steel pipe from different perspectives, the corresponding image coordinate transformation relationships, the ring array phased array detection data, and the axial position correspondence from the longitudinal image registration information of the steel pipe; generating pixel grid positions in the corrected longitudinal detection images of the steel pipe according to the row and column positions in the same image coordinate system, and corresponding each pixel grid position with the pixel points in the corrected longitudinal detection images of the steel pipe; reading the pixel grayscale values ​​corresponding to the pixel grid positions point by point along the row and column positions, and recording the pixel grid positions, pixel grayscale values, corresponding axial positions of the steel pipe, and image coordinate transformation relationships accordingly to obtain image pixel information from different perspectives; reading the detection responses corresponding to the axial positions of each steel pipe from the ring array phased array detection data according to the axial position correspondence, and then mapping the detection responses to grayscale imaging data according to the corresponding axial positions of the steel pipe to form the ring array phased array imaging information corresponding to the axial positions of the steel pipe.

[0035] Using the image coordinate transformation relationship as a reference, the axial position of the steel pipe is used to read the image pixels corresponding to the same axial position of the steel pipe and within the overlapping projection range of adjacent viewpoints from the image pixel information of different viewpoints, and the imaging response corresponding to the same axial position of the steel pipe is read from the ring array phased array imaging information; the image pixels corresponding to the same axial position of the steel pipe and the imaging response are fused to generate the longitudinal reconstruction image information of the steel pipe.

[0036] The specific process includes: using image coordinate transformation relationships to transform the pixel positions in the image pixel information from different viewpoints to the same image coordinates; using the axial position of the steel pipe in the same image coordinates as the reading reference, determining the pixel positions to be reconstructed one by one along the axial position of the steel pipe; at each pixel position to be reconstructed, searching for the image pixels corresponding to the same axial position of the steel pipe after transformation from the image pixel information from different viewpoints; determining whether the corresponding image pixels fall within the overlapping projection range of adjacent viewpoints by comparing the projection range boundary, pixel coordinate offset, and overlapping area judgment conditions of the images at the corresponding positions from each viewpoint; using the image pixels within the overlapping projection range of adjacent viewpoints as fusion objects, reading the pixel grayscale values ​​corresponding to the fusion objects, and reading the imaging response corresponding to the same axial position of the steel pipe from the ring array phased array imaging information; averaging and fusing the pixel grayscale values ​​corresponding to the same axial position of the steel pipe, then using the imaging response to strengthen and correct the averaged and fused pixel grayscale values, writing the strengthened and corrected pixel grayscale values ​​into the corresponding pixel positions in the same image coordinates, and continuously filling pixels according to the axial position of the steel pipe to generate longitudinal reconstructed image information of the steel pipe.

[0037] It should be noted that the overlapping area determination condition refers to transforming the pixel positions in the image pixel information to the same image coordinate system according to the image coordinate transformation relationship under adjacent viewpoints. When the image pixels corresponding to the same axial position of the steel pipe fall within the projection range of the longitudinal detection images of the steel pipe in two adjacent viewpoints, the corresponding image pixels are determined to be within the overlapping projection range of adjacent viewpoints. The projection range boundary is obtained by transforming the image pixel information under the same viewpoint to the same image coordinate system, extracting the outermost boundary position of the continuously distributed pixel positions along the axial position and vertical direction of the steel pipe, and then continuously connecting the outermost boundary positions.

[0038] S3. Extract the abnormal brightness and darkness variation region that extends continuously along the axial position of the steel pipe from the longitudinal reconstruction image information of the steel pipe, and combine it with the abnormal echo response of the corresponding axial position of the steel pipe in the ring array phased array imaging information. Perform image segmentation based on the gray level difference between the abnormal brightness and darkness variation region and the surrounding background image region to obtain the abnormal image region and mark it as the longitudinal defect candidate image region information.

[0039] The longitudinal reconstruction image of the steel pipe and the imaging information of the ring phased array are read from the longitudinal reconstruction image information of the steel pipe. The background gray-level changes in the longitudinal reconstruction image of the steel pipe are continuously and smoothly estimated along the axial position of the steel pipe. The difference between the longitudinal reconstruction image of the steel pipe and the continuously smoothed estimated background gray-level changes is compared to obtain the bright and dark change regions with obvious gray-level differences and continuous extension along the axial position of the steel pipe. Then, the imaging response corresponding to the axial position of the steel pipe is read from the imaging information of the ring phased array. The axial position of the steel pipe corresponding to the bright and dark change region is matched with the imaging response, and the abnormal echo response corresponding to the bright and dark change region is extracted to form the abnormal bright and dark change region information.

[0040] The specific process includes: reading the longitudinal reconstruction image of the steel pipe, the ring array phased array imaging information, and the image coordinates corresponding to the axial position of the steel pipe from the longitudinal reconstruction image information of the steel pipe; reading the pixel grayscale values ​​of the steel pipe image area segment by segment according to the axial position of the steel pipe in the longitudinal reconstruction image of the steel pipe; statistically analyzing the pixel grayscale values ​​of the adjacent image ranges of the same axial position of the steel pipe to obtain the local background grayscale distribution of the corresponding axial position of the steel pipe; continuously averaging the adjacent local background grayscale distributions along the axial position of the steel pipe to weaken the influence of sudden changes in local brightness on the background grayscale, obtaining the continuously changing background grayscale change along the axial position of the steel pipe, and completing the continuous smooth estimation of the background grayscale change in the longitudinal reconstruction image of the steel pipe; reading the imaging response corresponding to each axial position of the steel pipe from the ring array phased array imaging information, and recording the imaging response in correspondence with the corresponding axial position of the steel pipe.

[0041] In the longitudinal reconstruction image of the steel pipe, the pixel grayscale values ​​within the image region of the steel pipe are read column by column according to the axial position of the steel pipe. The pixel grayscale values ​​at the same axial position of the steel pipe are compared with the background grayscale changes after continuous smoothing estimation to obtain the grayscale differences at each axial position of the steel pipe. The grayscale differences are continuously compared along the axial position of the steel pipe, and the image positions where the grayscale difference is consistently higher than the background grayscale changes in the adjacent image range are marked as abnormal bright / dark positions. The abnormal bright / dark positions that appear consecutively at adjacent axial positions of the steel pipe are connected. Connectivity analysis is performed on the connected abnormal bright / dark positions to determine the length of the connected abnormal bright / dark positions in the longitudinal reconstruction image of the steel pipe. The pixel distribution range in the directional direction and the pixel distribution range in the width direction are determined. Then, the axial position of the steel pipe is used as a reference direction to determine whether the abnormal brightness and darkness positions after connection are continuously extended along the axial position of the steel pipe. The abnormal region pixel positions that are continuously distributed along the axial position of the steel pipe are retained, and the scattered image positions that are not continuously distributed along the axial position of the steel pipe are deleted. The retained abnormal region pixel positions are combined into a brightness and darkness variation region that is continuously extended along the axial position of the steel pipe. Then, the imaging response of the brightness and darkness variation region corresponding to the axial position of the steel pipe is read from the phased array imaging information. The brightness and darkness variation region with abnormal imaging response is determined as the region corresponding to the abnormal echo response, thus forming the abnormal brightness and darkness variation region information.

[0042] It should be noted that a significant grayscale difference is determined by the absolute difference between the pixel grayscale value at each image location within a neighboring image range and the background grayscale change after continuous smoothing estimation. The average and standard deviation of the absolute difference are calculated, and the sum of the average and twice the standard deviation is determined as the grayscale difference judgment threshold. When the uniform grayscale expression range is 0 to 255, if the grayscale difference judgment threshold is less than twenty grayscale levels, it is determined according to twenty grayscale levels. When the grayscale difference at the corresponding image location in the longitudinal reconstruction image of the steel pipe is greater than the grayscale difference judgment threshold, and the grayscale difference at no less than three consecutive adjacent axial positions of the steel pipe is greater than the grayscale difference judgment threshold, the corresponding image location is judged to have a significant grayscale difference.

[0043] Based on the information of abnormal brightness and darkness change areas, the grayscale difference between the abnormal brightness and darkness change areas and the surrounding background image areas is compared. Combined with the abnormal echo response of the corresponding steel pipe axial position, the image positions with obvious grayscale differences and prominent abnormal echo responses are marked. The marked grayscale change positions are connected to form a segmentation boundary, and the abnormal brightness and darkness change areas are extracted. Then, the discrete parts that do not extend continuously along the steel pipe axial position and whose abnormal echo responses are discontinuous are removed to obtain the abnormal image regions. The abnormal image regions, the corresponding steel pipe axial positions, and the abnormal echo responses are recorded to form longitudinal defect candidate image region information.

[0044] The specific process includes: based on the information of abnormal brightness and darkness change areas, determining the image location range corresponding to the abnormal brightness and darkness change areas in the longitudinal reconstruction image of the steel pipe; searching outwards along the outer perimeter of the abnormal brightness and darkness change areas for the adjacent background image regions; reading pixel grayscale values ​​point by point within the abnormal brightness and darkness change areas; reading pixel grayscale values ​​point by point within the surrounding background image regions for positions adjacent to the abnormal brightness and darkness change areas; comparing the pixel grayscale values ​​of the abnormal brightness and darkness change areas with the pixel grayscale values ​​of the surrounding background image regions to determine the boundary image positions where the grayscale differences are prominent; and starting from the abnormal brightness and darkness change areas... The abnormal echo response of the corresponding axial position of the steel pipe is read from the information. The boundary image position with prominent gray-level difference is compared with the abnormal echo response of the corresponding axial position of the steel pipe. The image position with obvious gray-level difference and prominent abnormal echo response is marked. The marked image position is determined as the gray-level change position. The gray-level change position is continuously connected according to the adjacent distribution relationship of the gray-level change position on the periphery of the abnormal brightness change area to form the segmentation boundary surrounding the abnormal brightness change area. Based on the segmentation boundary, the image content within the range surrounded by the segmentation boundary is extracted in the longitudinal reconstruction image of the steel pipe to obtain the cropped abnormal brightness change area.

[0045] The captured abnormal brightness variation areas are checked for continuity along the axial position of the steel pipe. The axial position of the steel pipe corresponding to each image location is read point by point within the captured abnormal brightness variation areas. Abnormal echo responses for the corresponding axial positions are read from the abnormal brightness variation area information, and the image locations are arranged sequentially according to the axial position of the steel pipe. Image locations with boundary contact, pixel adjacency, or region overlap at adjacent axial positions of the steel pipe, and whose corresponding abnormal echo responses appear consecutively, are connected to obtain a set of image locations continuously distributed along the axial position of the steel pipe. Image locations that cannot form a continuous connection with the preceding and following axial positions of the steel pipe, or whose corresponding abnormal echo responses... Isolated image locations that do not appear continuously are marked, and these marked isolated image locations are considered as discrete parts that do not extend continuously along the axial direction of the steel pipe and whose abnormal echo responses are discontinuous. The discrete parts are removed from the extracted abnormal brightness and darkness variation areas, and the image locations that are continuously connected along the axial direction of the steel pipe and whose abnormal echo responses appear continuously are retained to obtain the abnormal image region. The axial position of the steel pipe covered by the abnormal image region in the longitudinal reconstruction image of the steel pipe is read from the mapping relationship between the pixel coordinates of the longitudinal reconstruction image of the steel pipe and the axial position of the steel pipe. The abnormal image region, the corresponding axial position of the steel pipe, and the abnormal echo response are then recorded to form longitudinal defect candidate image region information.

[0046] It should be noted that the longitudinal defect candidate image region information includes the image location range corresponding to the abnormal image region, the axial position of the steel pipe, the region boundary position, the viewpoint, and the continuous extension state. The image location range is used to indicate the coverage position of the abnormal image region in the longitudinal reconstruction image of the steel pipe. The axial position of the steel pipe is used to indicate the position of the abnormal image region along the length of the steel pipe. The region boundary position is used to indicate the segmentation boundary of the abnormal image region. The pixel gray value and gray value difference are used to indicate the brightness and darkness changes of the abnormal image region. The viewpoint is used to indicate the source of the longitudinal detection image of the steel pipe corresponding to the abnormal image region. The continuous extension state is used to indicate the continuous distribution of the abnormal image region along the axial position of the steel pipe.

[0047] S4. Read the longitudinal defect candidate image region information corresponding to the same axial position of the steel pipe from different perspectives from the longitudinal defect candidate image region information, and combine it with the abnormal echo response corresponding to the axial position of the steel pipe in the ring array phased array imaging information. Perform image matching on the longitudinal defect candidate image region information with corresponding extension directions and consistent abnormal echo responses to obtain effective longitudinal defect image region information.

[0048] The region location attributes and abnormal echo responses of each longitudinal defect candidate image region are read from the longitudinal defect candidate image region information. Topological coding is performed according to the front-back relationship of each longitudinal defect candidate image region in the axial position of the steel pipe within the same viewpoint. Then, the abnormal echo responses are associated with the corresponding axial positions of the steel pipe to form axial topological information of the candidate region.

[0049] The specific process includes: reading the image position range, corresponding axial position of the steel pipe, viewing angle, and abnormal echo response of each longitudinal defect candidate image region in the longitudinal reconstruction image of the steel pipe from the longitudinal defect candidate image region information; using the image position range, corresponding axial position of the steel pipe, viewing angle, and abnormal echo response as the region positioning attributes corresponding to each longitudinal defect candidate image region; within the same viewing angle, determining the axial start position and axial end position of each longitudinal defect candidate image region based on the axial position of the steel pipe corresponding to each longitudinal defect candidate image region, and arranging each longitudinal defect candidate image region in order from front to back according to the axial start position; assigning front and back order values ​​to each arranged longitudinal defect candidate image region, recording the front and back relationship between adjacent arranged longitudinal defect candidate image regions, and then saving the image position range, axial start position, axial end position, front and back order value, viewing angle, abnormal echo response, and the axial position of the steel pipe corresponding to the abnormal echo response of each longitudinal defect candidate image region to form the axial topology information of the candidate region.

[0050] The axial start and end positions and front and back order of each longitudinal defect candidate image region are determined by using the axial topology information of the candidate region. The longitudinal defect candidate image regions with overlapping axial intervals and consistent front and back order under different viewpoints are aligned, and the intersection of the axial intervals after alignment is determined as the overlapping axial range. Then, the abnormal echo response of the corresponding axial position of the steel pipe within the overlapping axial range is read from the ring array phased array imaging information to form abnormal echo response information.

[0051] The specific process includes: sequentially reading the axial start position, axial end position, sequential order, and abnormal echo response of each longitudinal defect candidate image region from the axial topology information of the candidate region; determining the axial position range of the steel pipe between the axial start position and the axial end position as the corresponding axial interval; comparing the axial intervals of the longitudinal defect candidate image regions from different viewpoints; when the axial start position of one longitudinal defect candidate image region is before the axial end position of another longitudinal defect candidate image region from a different viewpoint, and the axial end position of one longitudinal defect candidate image region is before the axial start position of another longitudinal defect candidate image region from a different viewpoint... After the initial position, it is determined that the axial intervals under different viewpoints overlap. Based on the overlap of the axial intervals, the order of the longitudinal defect candidate image regions in the axial topology information of the candidate regions under different viewpoints is compared. The longitudinal defect candidate image regions with overlapping axial intervals and consistent order are aligned. The axial position range of the steel pipe covered by the aligned intervals is determined as the overlapping axial range. The abnormal echo response corresponding to the axial position of the steel pipe within the overlapping axial range is read from the ring phased array imaging information. The abnormal echo response is then recorded in correspondence with the axial position of the steel pipe within the overlapping axial range to form abnormal echo response information.

[0052] Within the overlapping axial range, the change of the center extension line of the corresponding longitudinal defect candidate image region is tracked to obtain the change of the extension direction. The abnormal echo response of the corresponding steel pipe axial position is read from the ring phased array imaging information. The change of extension direction, abnormal echo response and axial topology information of the candidate region are compared and the cross-view candidate region matching consistency value is calculated. When the cross-view candidate region matching consistency value meets the matching requirements, the longitudinal defect candidate image region information with corresponding extension directions and consistent abnormal echo response is determined as the candidate region direction matching information.

[0053] The specific process includes: within the overlapping axial range, reading the image positions corresponding to the longitudinal defect candidate image regions for interval alignment according to the axial position of the steel pipe, and determining the image position center of the longitudinal defect candidate image region at each axial position of the steel pipe; continuously connecting the image position centers at adjacent axial positions of the steel pipe to obtain the center extension line of the longitudinal defect candidate image region, and obtaining the extension direction change according to the connection direction change between adjacent image position centers; reading the abnormal echo response of the corresponding axial position of the steel pipe from the ring phased array imaging information, comparing the abnormal echo responses corresponding to the longitudinal defect candidate image regions under different viewpoints to obtain the abnormal echo response differences; reading the axial start and end positions and the front and back order of the corresponding longitudinal defect candidate image regions from the axial topology information of the candidate regions, processing the overlapping axial range length, extension direction change difference, abnormal echo response difference and front and back order difference with the same dimension, and then weighting and summarizing according to the degree of axial overlap, the degree of extension direction consistency, the degree of abnormal echo response consistency and the degree of axial topology consistency to obtain the cross-view candidate region matching consistency value.

[0054] After calculating the cross-view candidate region matching consistency value, the cross-view candidate region matching consistency value corresponding to the longitudinal defect candidate image regions that are interval aligned under different views is compared with the matching requirements. When the cross-view candidate region matching consistency value meets the matching requirements, it is determined that the corresponding longitudinal defect candidate image regions have a matching relationship in terms of coincident axial range, extension direction change, abnormal echo response and candidate region axial topology information. The longitudinal defect candidate image region information with corresponding extension directions and consistent abnormal echo responses, the corresponding coincident axial range, extension direction change, abnormal echo response and cross-view candidate region matching consistency value are recorded and determined as candidate region direction matching information.

[0055] It should be noted that the matching requirement refers to the cross-view candidate region matching consistency value being no less than the preset matching threshold, and the corresponding longitudinal defect candidate image region having the same front-to-back order in the axial topological information of the candidate region. The preset matching threshold is 0.72. When the cross-view candidate region matching consistency value is greater than or equal to 0.72, the corresponding longitudinal defect candidate image region information is determined to meet the matching requirement. The weight coefficients used for weighted summarization are statistically obtained from the cross-view matching information of the labeled longitudinal defect images of steel pipes. The axial overlap, extension direction consistency, and axial topological consistency are compared with the correct matching information, and normalized according to the contribution. The axial overlap and extension direction consistency have a significant impact on cross-view matching, and their weight coefficients are both determined to be 0.4. The axial topological consistency is used to assist in verifying the front-to-back order, and its weight coefficient is determined to be 0.2.

[0056] The expression for calculating the consistency value of cross-view candidate region matching is as follows:

[0057] in, Candidate image regions for longitudinal defects Candidate image regions for longitudinal defects Cross-view candidate region matching consistency value; Number the candidate image regions for longitudinal defects from a single viewpoint; From another perspective and region Numbering of candidate image regions for matching longitudinal defects; For the region With the region The length of the overlapping axial range at the axial position of the steel pipe; For the region axial length; For the region axial length; To prevent extremely small positive numbers with a denominator of zero; For the region The direction angle of extension; For the region The direction angle of extension; For the region The sequential values ​​in the axial topology information of the candidate region; For the region The order of values ​​in the axial topology information of the candidate region.

[0058] It should be noted that, It is a region Axial start and end positions and regions After aligning the start and end positions of the axes, the length of the intersection of the two axial intervals is taken; It is a region The result is obtained by subtracting the axial start position from the axial end position in the axial topology information of the candidate region; It is a region The starting and ending coordinates of the center extension line on the axial position of the steel pipe are determined; It is the region in the axial topology information of the candidate region The axial starting position of the steel pipe is determined.

[0059] Furthermore, before calculating the cross-view candidate region matching consistency value expression, the dimensions of the coincident axial range length, axial length, extension direction angle, and sequential values ​​involved in the calculation have been unified. The coincident axial range length and axial length are converted into dimensionless values ​​through the axial coincidence ratio, the difference between extension direction angles is converted into dimensionless values ​​through cosine consistency, and the difference between sequential values ​​is converted into dimensionless values ​​through sequence difference suppression. This ensures that the degree of axial coincidence, the degree of consistency of extension direction, and the degree of axial topological consistency are within a comparable and unified numerical range. Then, the values ​​are weighted and summarized according to the corresponding weight coefficients to obtain the cross-view candidate region matching consistency value.

[0060] Based on the cross-view candidate region matching consistency value in the candidate region direction matching information, the longitudinal defect candidate image region information that meets the matching requirements and has consistent abnormal echo response is retained and determined as the valid longitudinal defect image region information.

[0061] The specific process includes: sequentially reading longitudinal defect candidate image region information for interval alignment from the candidate region directional matching information, and comparing it with the corresponding cross-view candidate region matching consistency value, overlapping axial range, extension direction change, abnormal echo response, and candidate region axial topology information; grouping longitudinal defect candidate image region information according to the axial position of the steel pipe, and using longitudinal defect candidate image region information corresponding to the same axial position range of the steel pipe as the same screening object; comparing the cross-view candidate region matching consistency value in the same screening object with the matching requirements, and performing a consistency comparison of abnormal echo responses, retaining longitudinal defect candidate regions whose cross-view candidate region matching consistency value meets the matching requirements and whose abnormal echo responses are consistent. For candidate image regions of defects, exclude longitudinal defect candidate image regions whose cross-view candidate region matching consistency value does not meet the matching requirements or whose abnormal echo responses are inconsistent; when there are multiple longitudinal defect candidate image regions that meet the matching requirements and have consistent abnormal echo responses within the same axial position range of the steel pipe, retain the longitudinal defect candidate image regions with higher cross-view candidate region matching consistency values, more complete overlapping axial ranges, and more continuous abnormal echo responses; record the corresponding image position range, steel pipe axial position, overlapping axial range, extension direction changes, and abnormal echo responses of the retained longitudinal defect candidate image regions to determine them as valid longitudinal defect image regions.

[0062] S5. Extract strip morphology features from the effective longitudinal defect image area information, identify the defect type based on the strip morphology features and abnormal echo response, and associate the defect type with the axial position of the steel pipe corresponding to the effective longitudinal defect image area information to generate longitudinal defect image recognition and positioning information of the steel pipe.

[0063] The effective longitudinal defect image region, the corresponding axial position of the steel pipe, and the abnormal echo response are read from the effective longitudinal defect image region information. The center extension line of the effective longitudinal defect image region is used as the reference. The transverse grayscale sections of the effective longitudinal defect image region are sequentially intercepted along the axial position of the steel pipe and arranged in order. Then, the axial position of the steel pipe corresponding to each transverse grayscale section is recorded in correspondence with the abnormal echo response to form the strip skeleton status information.

[0064] The specific process includes: reading the effective longitudinal defect image region, the corresponding axial position of the steel pipe, the abnormal echo response, and the center extension line of the effective longitudinal defect image region from the effective longitudinal defect image region information; reading the image position and pixel grayscale value point by point within the effective longitudinal defect image region according to the axial position of the steel pipe; using the center extension line of the effective longitudinal defect image region as the cutting reference; cutting the pixel grayscale distribution within the effective longitudinal defect image region at each axial position of the steel pipe along the direction perpendicular to the center extension line to obtain the transverse grayscale cross section corresponding to the axial position of the steel pipe; arranging each transverse grayscale cross section in the order of the axial position of the steel pipe from front to back; and recording the axial position of the steel pipe, the cross section boundary position, the cross section pixel grayscale value, and the abnormal echo response corresponding to each transverse grayscale cross section to form the strip skeleton state information.

[0065] The transverse grayscale sections arranged along the axial position of the steel pipe in the strip skeleton status information are taken as the processing objects. The boundary position differences and boundary grayscale differences in adjacent transverse grayscale sections are compared in turn to obtain the width change and boundary grayscale change of the effective longitudinal defect image area along the axial position of the steel pipe. Then, the abnormal echo response of the corresponding axial position of the steel pipe is read from the strip skeleton status information, and the abnormal echo response is recorded in correspondence with the width change and boundary grayscale change to form the strip morphological features.

[0066] The specific process includes: reading the transverse grayscale cross sections, cross-section boundary positions, cross-section pixel grayscale values, and abnormal echo responses arranged according to the axial position of the steel pipe from the strip skeleton status information; taking the transverse grayscale cross sections as the processing object; determining the width of the corresponding transverse grayscale cross section based on the cross-section boundary positions on both sides of each transverse grayscale cross section; comparing the width differences of adjacent transverse grayscale cross sections in the order of the axial position of the steel pipe to obtain the width variation of the effective longitudinal defect image region along the axial position of the steel pipe; reading the corresponding pixel grayscale values ​​from the cross-section boundary positions of adjacent transverse grayscale cross sections; comparing the boundary grayscale differences between adjacent transverse grayscale cross sections to obtain the boundary grayscale variation of the effective longitudinal defect image region along the axial position of the steel pipe; reading the abnormal echo responses corresponding to each axial position of the steel pipe from the strip skeleton status information; and recording the abnormal echo responses in correspondence with the width variation and boundary grayscale variation at the corresponding axial position of the steel pipe to form strip morphological features.

[0067] The defect type is determined by utilizing the width and boundary grayscale variations in the strip morphology features, combined with the abnormal echo response at the corresponding axial position of the steel pipe. When the effective longitudinal defect image area maintains a narrow and continuous width along the axial position of the steel pipe and the grayscale variation at the end is obvious, the corresponding defect type is determined to be a longitudinal crack. When the effective longitudinal defect image area maintains a wide and continuous width along the axial position of the steel pipe and is close to the contour boundary of the steel pipe image area, the corresponding defect type is determined to be a longitudinal scratch.

[0068] The specific process includes: extracting the width variation, boundary grayscale variation, and corresponding abnormal echo response of the effective longitudinal defect image region along the axial position of the steel pipe from the strip morphological features; comparing the width difference of adjacent transverse grayscale sections segment by segment according to the axial position of the steel pipe to determine whether the width of the effective longitudinal defect image region remains stable along the axial position of the continuous steel pipe; when the effective longitudinal defect image region is elongated along the axial position of the continuous steel pipe, the width of the transverse grayscale section remains relatively small, and the width variation of adjacent transverse grayscale sections is small, while the boundary grayscale variation at both ends of the effective longitudinal defect image region is higher than that at the middle extension position, and corresponding... When abnormal echo responses occur in a concentrated manner along the axial position of the steel pipe, the effective longitudinal defect image region is determined to be narrow and continuous along the axial position of the steel pipe with obvious grayscale changes at the ends, and the corresponding defect type is determined to be a longitudinal crack. When the effective longitudinal defect image region is distributed as a wide strip along the continuous axial position of the steel pipe, the transverse grayscale cross-section width is continuously large, and the image position range of the effective longitudinal defect image region is close to the contour boundary of the steel pipe image region, and the abnormal echo response at the corresponding axial position of the steel pipe is continuously distributed along the surface, the effective longitudinal defect image region is determined to be wide and continuous along the axial position of the steel pipe and close to the contour boundary of the steel pipe image region, and the corresponding defect type is determined to be a longitudinal scratch.

[0069] It should be noted that the rules for determining narrow-width continuity and wide-width continuity are as follows: the width of the transverse grayscale section and the connection status of adjacent transverse grayscale sections are read from the strip morphology features; when the width of the transverse grayscale section at the axial position of the continuous steel pipe is relatively small, and adjacent transverse grayscale sections maintain boundary connection or pixel adjacency, the effective longitudinal defect image area is determined to maintain narrow-width continuity; when the width of the transverse grayscale section at the axial position of the continuous steel pipe is relatively large, and adjacent transverse grayscale sections maintain boundary connection or pixel adjacency, the effective longitudinal defect image area is determined to maintain wide-width continuity.

[0070] After the defect type is determined, the axial position of the steel pipe and the abnormal echo response corresponding to the strip morphology features are read from the effective longitudinal defect image area information. The defect type, the axial position of the steel pipe, the effective longitudinal defect image area and the abnormal echo response are recorded accordingly to generate the longitudinal defect image recognition and positioning information of the steel pipe.

[0071] The specific process includes, after the defect type is determined, reading the image position range, axial position of the steel pipe, and abnormal echo response corresponding to the effective longitudinal defect image region from the effective longitudinal defect image region information, and determining the axial position of the steel pipe corresponding to the strip morphology feature as the positioning position corresponding to the defect type based on the correspondence between the strip morphology feature and the effective longitudinal defect image region; recording the defect type, axial position of the steel pipe, effective longitudinal defect image region, and abnormal echo response according to the same correspondence, and mapping each defect type to a specific effective longitudinal defect image region, axial position of the steel pipe, and abnormal echo response to generate longitudinal defect image recognition and positioning information for the steel pipe.

[0072] This embodiment also provides a computer device applicable to the image recognition-based method for identifying longitudinal defects in steel pipes, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the image recognition-based method for identifying longitudinal defects in steel pipes as proposed in the above embodiment.

[0073] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0074] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the image recognition-based longitudinal defect identification method for steel pipes as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0075] In summary, this invention achieves improved image recognition consistency and candidate region extraction accuracy by: registering longitudinal inspection images of steel pipes from different perspectives according to the axial position of the steel pipe contour, and fusing image pixels at the same axial position of the steel pipe; and by performing axial topological coding, determining the overlapping axial range, and cross-view image recognition matching on the longitudinal defect candidate image region, and extracting the strip morphological features of the effective longitudinal defect image region. This enables defect authenticity screening, type identification, and axial position association, thereby reducing false detections and improving the reliability and positioning stability of longitudinal defect image recognition.

[0076] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for longitudinal defect recognition based on phased array images of steel pipe ring arrays, characterized in that, include: Longitudinal inspection images of steel pipes and phased array inspection data of ring array are collected. The outline of the steel pipe in the longitudinal inspection image is identified, and the axial position of the steel pipe outline is used as the registration reference. The longitudinal inspection images of steel pipes and phased array inspection data of ring array from different perspectives are corrected to the same image coordinates to form longitudinal image registration information of steel pipes. Based on the longitudinal image registration information of the steel pipe, the pixel position correspondence processing is performed on the corrected longitudinal detection image of the steel pipe, and the ring array phased array detection data is converted into ring array phased array imaging information corresponding to the axial position of the steel pipe. The image pixels corresponding to the same axial position of the steel pipe under different viewpoints are fused with the ring array phased array imaging information to generate longitudinal reconstruction image information of the steel pipe. The abnormal brightness and darkness variation regions that extend continuously along the axial position of the steel pipe are extracted from the longitudinal reconstruction image information of the steel pipe. Combined with the abnormal echo response of the corresponding axial position of the steel pipe in the ring array phased array imaging information, the image is segmented according to the gray level difference between the abnormal brightness and darkness variation regions and the surrounding background image regions to obtain the abnormal image regions and mark them as longitudinal defect candidate image region information. The longitudinal defect candidate image region information corresponding to the same axial position of the steel pipe from different perspectives is read from the longitudinal defect candidate image region information. Combined with the abnormal echo response corresponding to the axial position of the steel pipe in the ring array phased array imaging information, the longitudinal defect candidate image region information with corresponding extension directions and consistent abnormal echo response is image matched to obtain effective longitudinal defect image region information. Strip morphology features are extracted from the effective longitudinal defect image region information. The defect type is identified based on the strip morphology features and abnormal echo response. The defect type is then associated with the axial position of the steel pipe corresponding to the effective longitudinal defect image region information to generate longitudinal defect image recognition and positioning information for the steel pipe.

2. The method for identifying longitudinal defects in steel pipes based on image recognition as described in claim 1, characterized in that, The specific steps for forming the longitudinal image registration information of the steel pipe are as follows: Longitudinal detection images of steel pipes from different perspectives and corresponding axial position detection data of steel pipes from ring array phased array are collected. After grayscale standardization of the longitudinal detection images of steel pipes, the boundary between the steel pipe image area and the background image area is identified. The axial position of the steel pipe contour is determined based on the contour center line of the steel pipe image area. Then, the axial sampling position of the steel pipe corresponding to the ring array phased array detection data is matched with the axial position of the steel pipe contour to form axial position reference information. Based on the axial position reference information, the image coordinate transformation relationship between the longitudinal inspection images of the steel pipe under different perspectives is established, and the axial position correspondence between the ring array phased array inspection data and the same image coordinate is determined. Then, the image coordinate transformation relationship is used to correct the longitudinal inspection images of the steel pipe under different perspectives to the same image coordinate, and the ring array phased array inspection data is mapped to the same image coordinate according to the axial position correspondence, thus forming the longitudinal image registration information of the steel pipe.

3. The image recognition-based method for identifying longitudinal defects in steel pipes as described in claim 2, characterized in that, The specific steps for generating the longitudinal reconstructed image information of the steel pipe are as follows: The longitudinal detection images of the steel pipe under different perspectives, the corresponding image coordinate transformation relationship, the ring array phased array detection data and the axial position correspondence are read from the longitudinal image registration information of the steel pipe. The pixel grid is read from the longitudinal detection images of the steel pipe to obtain the image pixel information under different perspectives. According to the axial position correspondence, the detection response of each steel pipe corresponding to the axial position in the ring array phased array detection data is read, and then the detection response is converted into the ring array phased array imaging information corresponding to the axial position of the steel pipe. Using the image coordinate transformation relationship as a reference, the axial position of the steel pipe is used to read the image pixels corresponding to the same axial position of the steel pipe and within the overlapping projection range of adjacent viewpoints from the image pixel information of different viewpoints, and the imaging response corresponding to the same axial position of the steel pipe is read from the ring array phased array imaging information; the image pixels corresponding to the same axial position of the steel pipe and the imaging response are fused to generate the longitudinal reconstruction image information of the steel pipe.

4. The image recognition-based method for identifying longitudinal defects in steel pipes as described in claim 1 or 3, characterized in that, The specific steps for obtaining abnormal image regions and marking them as candidate image regions for longitudinal defects are as follows: The longitudinal reconstruction image of the steel pipe and the ring phased array imaging information are read from the longitudinal reconstruction image information of the steel pipe. The background gray level change in the longitudinal reconstruction image of the steel pipe is continuously and smoothly estimated along the axial position of the steel pipe. The difference between the longitudinal reconstruction image of the steel pipe and the background gray level change after continuous smooth estimation is compared to obtain the bright and dark change area with obvious gray level difference and continuous extension along the axial position of the steel pipe. Then, the imaging response corresponding to the axial position of the steel pipe is read from the phased array imaging information of the ring array. The axial position of the steel pipe corresponding to the brightness change area is matched with the imaging response, and the abnormal echo response corresponding to the brightness change area is extracted to form the abnormal brightness change area information. Based on the information of abnormal brightness and darkness change areas, the grayscale difference between the abnormal brightness and darkness change areas and the surrounding background image areas is compared. Combined with the abnormal echo response of the corresponding steel pipe axial position, the image positions with obvious grayscale differences and prominent abnormal echo responses are marked. The marked grayscale change positions are connected to form a segmentation boundary, and the abnormal brightness and darkness change areas are extracted. Then, the discrete parts that do not extend continuously along the steel pipe axial position and whose abnormal echo responses are discontinuous are removed to obtain the abnormal image regions. The abnormal image regions, the corresponding steel pipe axial positions, and the abnormal echo responses are recorded to form longitudinal defect candidate image region information.

5. The image recognition-based method for identifying longitudinal defects in steel pipes as described in claim 4, characterized in that, The specific steps for reading longitudinal defect candidate image region information corresponding to the same axial position of the steel pipe from different viewpoints from the longitudinal defect candidate image region information, and combining it with the abnormal echo response corresponding to the axial position of the steel pipe in the ring array phased array imaging information, are as follows: Read the region location attributes and abnormal echo responses corresponding to each longitudinal defect candidate image region from the longitudinal defect candidate image region information, and perform topological coding according to the front-back relationship of each longitudinal defect candidate image region in the axial position of the steel pipe within the same viewpoint. Then, associate and record the abnormal echo responses with the corresponding axial positions of the steel pipe to form axial topological information of the candidate region. The axial start and end positions and front and back order of each longitudinal defect candidate image region are determined by using the axial topology information of the candidate region. The longitudinal defect candidate image regions with overlapping axial intervals and consistent front and back order under different viewpoints are aligned, and the intersection of the axial intervals after alignment is determined as the overlapping axial range. Then, the abnormal echo response of the corresponding axial position of the steel pipe within the overlapping axial range is read from the ring array phased array imaging information to form abnormal echo response information.

6. The image recognition-based method for identifying longitudinal defects in steel pipes as described in claim 5, characterized in that, The specific steps for obtaining effective longitudinal defect image region information are as follows: Within the overlapping axial range, the change of the center extension line of the corresponding longitudinal defect candidate image region is tracked to obtain the change of the extension direction. The abnormal echo response of the corresponding steel pipe axial position is read from the ring array phased array imaging information. The change of the extension direction, the abnormal echo response and the axial topology information of the candidate region are compared and the cross-view candidate region matching consistency value is calculated. When the consistency value of cross-view candidate region matching meets the matching requirements, the longitudinal defect candidate image region information with corresponding extension directions and consistent abnormal echo response is determined as the candidate region direction matching information. Based on the cross-view candidate region matching consistency value in the candidate region direction matching information, the longitudinal defect candidate image region information that meets the matching requirements and has consistent abnormal echo response is retained and determined as the valid longitudinal defect image region information.

7. The image recognition-based method for identifying longitudinal defects in steel pipes as described in claim 6, characterized in that, The specific steps for extracting stripe morphological features from effective longitudinal defect image region information are as follows: The effective longitudinal defect image region, the corresponding axial position of the steel pipe, and the abnormal echo response are read from the effective longitudinal defect image region information. The center extension line of the effective longitudinal defect image region is used as the reference. The transverse grayscale sections of the effective longitudinal defect image region are sequentially cut along the axial position of the steel pipe and arranged in order. Then, the axial position of the steel pipe corresponding to each transverse grayscale section and the abnormal echo response are recorded to form the strip skeleton status information. The transverse grayscale sections arranged along the axial position of the steel pipe in the strip skeleton status information are taken as the processing objects. The boundary position differences and boundary grayscale differences in adjacent transverse grayscale sections are compared in turn to obtain the width change and boundary grayscale change of the effective longitudinal defect image area along the axial position of the steel pipe. Then, the abnormal echo response of the corresponding axial position of the steel pipe is read from the strip skeleton status information, and the abnormal echo response is recorded in correspondence with the width change and boundary grayscale change to form the strip morphological features.

8. The image recognition-based method for identifying longitudinal defects in steel pipes as described in claim 1 or 7, characterized in that, The specific steps for generating image recognition and location information of longitudinal defects in steel pipes are as follows: The defect type is determined by utilizing the width and boundary grayscale variations in the strip morphology features, combined with the abnormal echo response at the corresponding axial position of the steel pipe. When the effective longitudinal defect image area maintains a narrow and continuous width along the axial position of the steel pipe and the grayscale variation at the end is obvious, the corresponding defect type is determined to be a longitudinal crack. When the effective longitudinal defect image area maintains a wide and continuous width along the axial position of the steel pipe and is close to the contour boundary of the steel pipe image area, the corresponding defect type is determined to be a longitudinal scratch. After the defect type is determined, the axial position of the steel pipe and the abnormal echo response corresponding to the strip morphology features are read from the effective longitudinal defect image area information. The defect type, the axial position of the steel pipe, the effective longitudinal defect image area and the abnormal echo response are recorded accordingly to generate the longitudinal defect image recognition and positioning information of the steel pipe.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the image recognition-based longitudinal defect identification method for steel pipes as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the image recognition-based longitudinal defect identification method for steel pipes as described in any one of claims 1 to 8.

Images