Seamless steel tube quality detection system and method based on image processing

Abstract

The invention belongs to the technical field of seamless steel tube quality detection, and provides a seamless steel tube quality detection system and method based on image processing, and the method comprises the steps: extracting design parameters in a seamless steel tube design file, and recognizing a wall thickness mutation region in a steel tube in combination with the actual feature data of the seamless steel tube; based on the wall thickness abrupt change area, acquiring an image corresponding to a leak detection defect found by historical secondary detection, constructing a historical leak detection defect image library, and analyzing a defect leak detection risk of the wall thickness abrupt change area; according to the method, cost waste caused by blind increase of the overall sampling density is avoided, enhanced detection of the missing detection core area is achieved, the defect missing detection rate of the wall thickness sudden change area, especially the boundary sensitive area, is remarkably reduced, and the precision and efficiency of seamless steel pipe quality detection are improved.

Description

Technical Field

[0001] This invention belongs to the field of seamless steel pipe quality inspection technology, specifically a seamless steel pipe quality inspection system and method based on image processing. Background Technology

[0002] Seamless steel pipes, with their excellent mechanical properties, sealing performance, and corrosion resistance, are widely used in key industrial fields such as petrochemicals, engineering machinery, nuclear power, aerospace, and marine engineering. They are core basic components ensuring the safe and stable operation of various equipment and systems. In actual service, seamless steel pipes often face complex conditions such as high pressure, high temperature, heavy load, and media corrosion. The uniformity of their wall thickness and defects such as microcracks, metal accumulation, and segregation on their internal and surface directly affect the load-bearing capacity and service life of the components. If these defects are not detected in time, they can easily lead to major safety accidents such as rupture, leakage, and explosion, causing not only huge economic losses but also potentially endangering human lives. Therefore, accurate and efficient quality testing of seamless steel pipes is crucial.

[0003] However, existing detection technologies still have many shortcomings: First, they do not effectively integrate the data of missed defects from historical secondary inspections, lack in-depth analysis of the types and characteristics of missed defects, and assess the risk of missed defects only from a single dimension, failing to take into account both the coverage of the types of missed defects and the frequency of high incidence in the region, resulting in a one-sided and unreliable assessment result; Second, they lack precise methods for delineating the boundaries within the region of abrupt changes in wall thickness, making it impossible to identify the boundary-sensitive areas where missed defects are highly concentrated, leading to unreasonable setting of the detection sampling density; Third, they do not consider the gradient change characteristics of the region of abrupt changes in wall thickness in the axial and circumferential directions, making it difficult to accurately mark the initiation and stability boundaries of the abrupt changes through gradient synthesis and surface fitting, further exacerbating the risk of missed defects.

[0004] Therefore, the present invention provides a seamless steel pipe quality inspection system and method based on image processing. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] In a first aspect, the present invention provides a seamless steel pipe quality inspection method based on image processing, comprising the following steps: Extract design parameters from seamless steel pipe design documents and combine them with actual characteristic data of seamless steel pipes to identify areas of abrupt changes in wall thickness within the steel pipe; Based on the region of abrupt change in wall thickness, images of defects that were missed during historical secondary inspections were obtained, a historical image library of missed defects was constructed, and the risk of missed defects in the region of abrupt change in wall thickness was analyzed. If the risk of missing defects in areas with abrupt changes in wall thickness is high, a double boundary marking method is used based on wall thickness data to mark the initiation boundary and the stable boundary of the change, delineate the boundary sensitive area, analyze the concentration of missed defects in the boundary sensitive area, and determine whether the missed defects are in the boundary sensitive area. If the focus is on the boundary-sensitive area, determine the density adjustment coefficient, adjust the quality inspection sampling density of the boundary-sensitive area, and reduce the missed detection of defects in the boundary-sensitive area within the region of sudden change in wall thickness.

[0007] As a further aspect of the present invention: the process for identifying the region of abrupt change in wall thickness is as follows: The core parameters of the steel pipe design documents are obtained from the CAD drawings. The design parameters include the standard wall thickness value and the actual characteristic data of the steel pipe, including the actual wall thickness value. The seamless steel pipe is divided into several grid units according to the grid pattern. Measurement points are set at the intersection of the grid units, and the actual wall thickness value of each measurement point of the seamless steel pipe is obtained by ultrasonic thickness measurement. The wall thickness deviation at each measurement point is obtained by subtracting the actual wall thickness value from the standard wall thickness value. If the wall thickness deviation at the measurement point is greater than or equal to the wall thickness deviation limit, the corresponding measurement point is recorded as the thickening abrupt change point. All thickening abrupt change points in the seamless steel pipe are extracted, and cluster analysis is performed on the thickening abrupt change points using 8-neighbor connected domains to cluster all connected thickening abrupt change points into wall thickness abrupt change regions.

[0008] As a further aspect of the present invention: the process of constructing the historical missed defect image library is as follows: Extract images and actual wall thickness data corresponding to missed defects found in historical secondary inspections, and classify them according to defect type; For each type of missed defect, three core image features are extracted, including: edge features, grayscale features, and morphological features; Based on defect type and image features, a historical database of missed defect images is constructed.

[0009] As a further aspect of the present invention: the process for analyzing the risk of missed defect detection in the region of abrupt change in wall thickness is as follows: Extract the types and number of missed defects from historical secondary detections in areas of abrupt changes in wall thickness from the historical missed defect image database, and analyze them to determine the ratio of missed defect types to the ratio of missed defect numbers. The defect omission coefficient is obtained by multiplying the ratio of missed defect types with the ratio of missed defect quantities. If the defect omission coefficient is less than or equal to the defect omission coefficient threshold, the risk of defect omission in the wall thickness abrupt change region is low; otherwise, the risk of defect omission in the wall thickness abrupt change region is high.

[0010] As a further aspect of the present invention: the process for determining the ratio of missed defect types to the ratio of missed defect quantities is as follows: The number of undetected defect types in the area of ​​abrupt wall thickness change is counted and compared with the total number of undetected defect types in seamless steel pipes to obtain the ratio of undetected defect types in the area of ​​abrupt wall thickness change. The number of defects missed in the wall thickness abrupt change area is counted and compared with the total number of defects missed in seamless steel pipes to obtain the defect missed number ratio in the wall thickness abrupt change area.

[0011] As a further aspect of the present invention: the process of delineating the boundary sensitive area is as follows: Using one end face or process baseline of the seamless steel pipe as the starting zero point, a linear encoder is used to collect the wall thickness data of each measurement point with an ultrasonic thickness gauge, while recording the axial displacement value of the measurement point along the length of the steel pipe, which is the axial distance coordinate of the measurement point; using a preset baseline on the surface of the steel pipe as the starting point, the sampling angle is consistent with the grid division angle, and the angle of the measurement point along the circumference of the seamless steel pipe is located by a combination of an indexing plate and a rotary encoder, which is the circumferential angle of the measurement point; The axial distance coordinates, circumferential angles, and actual wall thickness values ​​of each measurement point are bound and stored to form a three-dimensional coordinate point set of the measurement points; The circumferential angles of the three-dimensional coordinate set of the measurement points are converted into Cartesian coordinates to form a unified three-dimensional rectangular coordinate system. The three-dimensional rectangular coordinates of the measurement points are then analyzed and processed to obtain the mutation initiation boundary and mutation stability boundary. The region between the mutation initiation boundary and the mutation stability boundary is designated as the boundary sensitive region.

[0012] As a further aspect of the present invention: the process for determining the mutation initiation boundary and the mutation stability boundary is as follows: Based on any measurement point, the difference between the actual wall thickness value of the current measurement point and the actual wall thickness value of the adjacent axial measurement point is calculated to obtain the axial wall thickness deviation. The distance between the current measurement point and the corresponding adjacent axial measurement point is obtained and recorded as the axial distance. The ratio of the axial wall thickness deviation to the axial distance is calculated to obtain the axial wall thickness gradient of the measurement point. The difference between the actual wall thickness value of the current measurement point and the actual wall thickness value of the adjacent circumferential measurement point is calculated to obtain the circumferential wall thickness deviation. The distance between the current measurement point and the corresponding adjacent circumferential measurement point is obtained and recorded as the circumferential distance. The ratio of the circumferential wall thickness deviation to the circumferential distance is calculated to obtain the circumferential wall thickness gradient of the measurement point. The gradient amplitude calculation formula is used to vector synthesize the axial wall thickness gradient and the circumferential wall thickness gradient of the steel pipe to obtain the comprehensive gradient of the measurement point. If the overall gradient is greater than or equal to the maximum value of the overall gradient, then the corresponding measurement point is recorded as a candidate point for rapid thickening. If the combined gradient is less than or equal to the minimum value of the combined gradient, then the corresponding measurement point is recorded as a candidate point for thickening stability. All rapid thickening candidate points are integrated into a mutation initiation boundary candidate point set, and all thickening stable candidate points are integrated into a mutation stable boundary point set. For the candidate point set of mutation initiation boundary and the stable point set of mutation, B-spline three-dimensional surface fitting is used to map the three-dimensional coordinates of all measurement points in the candidate point set of mutation initiation boundary and the stable point set of mutation into the B-spline parameter domain, and the surface of the rapidly thickened region and the surface of the stable thickened region are obtained respectively. The parameter points are the coordinates of the parameter domain of the B-spline surface. For the fitted rapidly thickening region surface, the three-dimensional gradient vector of the parameter points in the B-spline parameter domain is calculated, the Euclidean norm of the three-dimensional gradient vector is calculated, the gradient magnitude is obtained, and the set of points with gradient magnitude greater than or equal to the maximum value of the comprehensive gradient is selected to form the gradient abrupt change zone of the surface. The contour line tracking algorithm is used to extract continuous closed contour lines in the gradient abrupt change zone, which are the abrupt change initiation boundaries. For the fitted stable thickened region surface, the set of points whose gradient magnitude is less than or equal to the minimum value of the comprehensive gradient is selected to form the gradient steady zone of the surface. The contour line tracking algorithm is used to extract continuous closed contour lines in the gradient steady zone, which are the abrupt stable boundaries.

[0013] As a further aspect of the present invention: the process of determining whether the missed defects are concentrated in the boundary sensitive area is as follows: The number of undetected defects in the boundary sensitive area is counted and compared with the total number of defects in the wall thickness abrupt change area to obtain the proportion of undetected defects in the boundary sensitive area. The number of defects missed in the boundary sensitive area is counted and compared with the volume of the boundary sensitive area to obtain the unit volume missed detection density of the boundary sensitive area. The number of defects missed in the non-boundary sensitive area is counted and compared with the volume of the non-boundary sensitive area to obtain the unit volume missed detection density of the non-boundary sensitive area. The unit volume missed detection density of the boundary sensitive area is then compared with the unit volume missed detection density of the non-boundary sensitive area to obtain the missed detection density ratio. The proportion of missed defects at the boundary is multiplied by the ratio of missed defects density to obtain the concentration coefficient of missed defects in the boundary sensitive area. If the concentration coefficient of missed defects is greater than or equal to the threshold of the concentration coefficient of missed defects, it means that the missed defects are concentrated in the boundary sensitive area; otherwise, it means that the missed defects are not concentrated in the boundary sensitive area.

[0014] As a further aspect of the present invention: the process of determining the density adjustment coefficient is as follows: Set an allowable false negative rate, test the false negative rate at different sampling densities in non-boundary sensitive areas, and screen out the minimum sampling density that meets the allowable false negative rate as the benchmark sampling density; In the boundary sensitive area, the false negative rate under different sampling densities is tested, and the minimum sampling density that meets the allowable false negative rate is selected and denoted as the sensitive sampling density. The density adjustment coefficient of the boundary sensitive area is obtained by comparing the sensitive sampling density with the reference sampling density.

[0015] Secondly, the present invention also provides a seamless steel pipe quality inspection system based on image processing, comprising the following modules: Region identification module: Extracts design parameters from the seamless steel pipe design file, combines them with the actual characteristic data of the seamless steel pipe, and identifies regions in the steel pipe where the wall thickness changes abruptly. Missed Detection Risk Assessment Module: Based on the wall thickness abrupt change region, obtain images corresponding to the missed defects found in the historical secondary inspection, build a historical missed defect image library, and analyze the missed defect risk in the wall thickness abrupt change region; Boundary delineation module: If the risk of missing defects in the area of ​​sudden change in wall thickness is high, based on the wall thickness data, the double boundary marking method is used to mark the initiation boundary and the stable boundary of the change, delineate the boundary sensitive area, analyze the concentration of missed defects in the boundary sensitive area, and determine whether the missed defects are in the boundary sensitive area. Density adjustment module: If the focus is on the boundary sensitive area, determine the density adjustment coefficient, adjust the quality inspection sampling density of the boundary sensitive area, and reduce the missed detection of defects in the boundary sensitive area within the wall thickness change area.

[0016] The beneficial effects of this invention are as follows: 1. This invention combines the design parameters of seamless steel pipes with actual data obtained from ultrasonic thickness measurement. Through grid division, deviation comparison, and connected component clustering analysis, it achieves accurate identification of regions with abrupt changes in wall thickness, clarifying the core focus area for defect omission risk assessment. Simultaneously, based on historical secondary inspection data of missed defects, it classifies defects by type and extracts three core image features—edge, grayscale, and morphology—to construct a database. The defect omission coefficient is quantified by multiplying the ratio of missed defect types by the ratio of their number. This approach balances the coverage of missed defect types and the frequency of high-risk areas, while also scientifically quantifying the risk of missed defects in regions with abrupt changes in wall thickness. It effectively avoids the limitations of a single assessment dimension, providing reliable support for subsequent targeted optimization of inspection strategies, focusing on high-risk areas for enhanced inspection, and reducing the overall defect omission rate of seamless steel pipes.

[0017] 2. This invention addresses the high risk of missed defects in areas with abrupt changes in wall thickness. It binds the three-dimensional coordinates of measurement points and converts them into unified rectangular coordinates. A comprehensive gradient is synthesized by combining axial and circumferential wall thickness gradient vectors. Through B-spline three-dimensional surface fitting and contour line tracing, the initiation and stability boundaries of the abrupt changes are accurately marked, and sensitive boundary areas are delineated. Then, the concentration coefficient of missed defects is obtained by multiplying the proportion of missed defects at the boundary by the ratio of missed defect density, quantifying the degree of clustering of missed defects in sensitive areas. This effectively eliminates interference from regional volume differences and accurately locates the core high-incidence areas of missed defects. Based on the allowable missed defect rate, the baseline sampling density for non-boundary sensitive areas and the sensitive sampling density for boundary sensitive areas are determined separately. The sampling density in sensitive areas is adjusted specifically using a density adjustment coefficient. This avoids the cost waste of blindly increasing the overall sampling density and achieves enhanced detection of core missed defect areas, significantly reducing the defect missed rate in areas with abrupt changes in wall thickness, especially in boundary sensitive areas, and improving the accuracy and efficiency of seamless steel pipe quality inspection. Attached Figure Description

[0018] The invention will now be further described with reference to the accompanying drawings.

[0019] Figure 1 This is a flowchart of the steps of the seamless steel pipe quality inspection method based on image processing according to an embodiment of the present invention; Figure 2 This is a system block diagram of the seamless steel pipe quality inspection system based on image processing according to an embodiment of the present invention. Detailed Implementation

[0020] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0021] Example 1 Please see Figure 1 As shown in the embodiment of the present invention, the seamless steel pipe quality inspection method based on image processing includes the following steps: Step 1: Extract the design parameters from the seamless steel pipe design file, and combine them with the actual characteristic data of the seamless steel pipe to identify the areas of abrupt changes in wall thickness in the steel pipe; In this step, the core parameters of the steel pipe design documents are obtained from the CAD drawings. The design parameters include the standard wall thickness value, and the actual characteristic data of the steel pipe include the actual wall thickness value. The seamless steel pipe is divided into several grid units according to the grid pattern. Measurement points are set at the intersection of the grid units, and the actual wall thickness value of each measurement point of the seamless steel pipe is obtained by ultrasonic thickness measurement. The wall thickness deviation at each measurement point is obtained by subtracting the actual wall thickness value from the standard wall thickness value. If the wall thickness deviation at the measurement point is greater than or equal to the wall thickness deviation limit, the corresponding measurement point is recorded as the thickening abrupt change point. If the wall thickness deviation at the measurement point is less than the wall thickness deviation limit, the corresponding measurement point is recorded as a normal point. It is understandable that the wall thickness deviation limit is set by those skilled in the art based on the characteristics of seamless steel pipes and historical experience; All thickening abrupt change points in the seamless steel pipe are extracted, and cluster analysis is performed on the thickening abrupt change points using 8-neighbor connected domains to cluster all connected thickening abrupt change points into wall thickness abrupt change regions. Step 2: Based on the wall thickness abrupt change region, obtain the images corresponding to the missed defects found in the historical secondary inspection, construct a historical missed defect image library, and analyze the defect missed risk in the wall thickness abrupt change region; In this step, the historical defect miss data includes the number of defects that were missed in areas of abrupt changes in wall thickness within a historical period, including but not limited to 1 year, 3 years, and 5 years. It should be noted that historical defect omission refers to defects that were not detected during the initial inspection of the seamless steel pipe but were discovered during a subsequent second inspection. Extract images and actual wall thickness data corresponding to the missed defects found in the historical secondary inspections, and classify them according to the defect type (microcracks, metal accumulation, segregation); For each type of missed defect, three core image features are extracted, including: edge features: edge length and curvature after Canny edge detection; grayscale features: average grayscale and grayscale standard deviation of the defect area; and morphological features: roundness and aspect ratio of the defect. Based on defect type and image features, a historical missed defect image library is constructed. From the historical missed defect image library, the types and number of missed defects in the historical secondary detection of the wall thickness change area are extracted. The number of missed defect types in the wall thickness change area is counted and compared with the total number of missed defect types in the seamless steel pipe to obtain the ratio of missed defect types in the wall thickness change area. The number of defects missed in the wall thickness abrupt region was counted and compared with the total number of defects missed in seamless steel pipes to obtain the defect missed number ratio in the wall thickness abrupt region. The defect omission coefficient is obtained by multiplying the ratio of missed defect types with the ratio of missed defect quantities. It should be noted that the missed defect type ratio represents the proportion of defect types that were missed during the initial inspection among all possible defect types in a seamless steel pipe, reflecting the degree of missed detection of different types of defects in the wall thickness abrupt change area; the missed defect number ratio represents the proportion of missed detection frequency in the wall thickness abrupt change area among all missed defects in the entire seamless steel pipe, reflecting whether the wall thickness abrupt change area is a high-incidence area for missed detection, and the contribution of this area to the overall missed detection. At the same time, it considers the number of missed defect types and the frequency of missed detection, quantifying the comprehensive missed detection risk of the wall thickness abrupt change area relative to the entire steel pipe. The defect omission coefficient is compared with the defect omission coefficient threshold. If the defect omission coefficient is less than or equal to the defect omission coefficient threshold, the risk of defect omission in the wall thickness abrupt change area is low; if the defect omission coefficient is greater than the defect omission coefficient threshold, the risk of defect omission in the wall thickness abrupt change area is high. The technical solution of this invention is as follows: Design parameters are extracted from the seamless steel pipe design documents, and combined with actual characteristic data of the seamless steel pipe to identify areas of abrupt changes in wall thickness. Based on these areas, images corresponding to missed defects discovered during historical secondary inspections are obtained, a historical image database of missed defects is constructed, and the risk of missed defects in these areas is analyzed. This invention combines seamless steel pipe design parameters with actual data obtained from ultrasonic thickness measurement. Through grid division, deviation comparison, and connected component clustering analysis, it achieves accurate identification of areas of abrupt changes in wall thickness, clarifying the core focus area for defect missed detection risk assessment. Simultaneously, relying on historical secondary inspection data of missed defects, a database is constructed by classifying defects by type and extracting three core image features: edge, grayscale, and morphology. The defect missed detection coefficient is quantified by multiplying the ratio of missed defect types by the ratio of their number. This approach balances the coverage of missed defect types and the frequency of high-incidence areas, while also scientifically quantifying the risk of missed defects in areas of abrupt changes in wall thickness. This effectively avoids the limitations of a single assessment dimension and provides reliable support for subsequent targeted optimization of inspection strategies, focusing on high-risk areas for enhanced inspection, and reducing the overall defect missed rate of seamless steel pipes.

[0022] Example 2 Please see Figure 1 As shown in the embodiment of the present invention, the seamless steel pipe quality inspection method based on image processing further includes the following steps: Step 3: If the risk of missing defects in the area of ​​sudden change in wall thickness is high, based on the wall thickness data, the double boundary marking method is used to mark the initiation boundary and the stable boundary of the change, delineate the boundary sensitive area, analyze the concentration of missed defects in the boundary sensitive area, and determine whether the missed defects are concentrated in the boundary sensitive area. Using one end face or process baseline of the seamless steel pipe as the starting zero point, a linear encoder is used to collect the wall thickness data of each measurement point with an ultrasonic thickness gauge, while recording the axial displacement value of the measurement point along the length of the steel pipe, which is the axial distance coordinate of the measurement point; using a preset baseline on the surface of the steel pipe as the starting point, the sampling angle is consistent with the grid division angle, and the angle of the measurement point along the circumference of the seamless steel pipe is located by a combination of an indexing plate and a rotary encoder, which is the circumferential angle of the measurement point; The axial distance coordinates, circumferential angles, and actual wall thickness values ​​of each measurement point are bound and stored to form a three-dimensional coordinate point set of the measurement points; The circumferential angles in the set of three-dimensional coordinate points of the measurement points are converted into Cartesian coordinates to form a unified three-dimensional rectangular coordinate system. Based on any measurement point, the difference between the actual wall thickness value of the current measurement point and the actual wall thickness value of the adjacent axial measurement point is calculated to obtain the axial wall thickness deviation. The distance between the current measurement point and the corresponding adjacent axial measurement point is obtained and recorded as the axial distance. The ratio of the axial wall thickness deviation to the axial distance is calculated to obtain the axial wall thickness gradient of the measurement point. The difference between the actual wall thickness value of the current measurement point and the actual wall thickness value of the adjacent circumferential measurement point is calculated to obtain the circumferential wall thickness deviation. The distance between the current measurement point and the corresponding adjacent circumferential measurement point is obtained and recorded as the circumferential distance. The ratio of the circumferential wall thickness deviation to the circumferential distance is calculated to obtain the circumferential wall thickness gradient of the measurement point. The gradient amplitude calculation formula is used to vector synthesize the axial wall thickness gradient and the circumferential wall thickness gradient of the steel pipe to obtain the comprehensive gradient of the measurement point. The comprehensive gradient maximum and comprehensive gradient minimum are set by those skilled in the art based on historical experience. If the comprehensive gradient is greater than or equal to the comprehensive gradient maximum, the corresponding measurement point is recorded as a candidate point for rapid thickening. If the combined gradient is less than the maximum value of the combined gradient and greater than the minimum value of the combined gradient, then the corresponding measurement point is recorded as a normal thickening point. If the combined gradient is less than or equal to the minimum value of the combined gradient, then the corresponding measurement point is recorded as a candidate point for thickening stability. All rapid thickening candidate points are integrated into a mutation initiation boundary candidate point set, and all thickening stable candidate points are integrated into a mutation stable boundary point set. For the candidate point set of mutation initiation boundary and the stable point set of mutation, B-spline three-dimensional surface fitting is used to map the three-dimensional coordinates of all measurement points in the candidate point set of mutation initiation boundary and the stable point set of mutation into the B-spline parameter domain, and the surface of the rapidly thickened region and the surface of the stable thickened region are obtained respectively. The parameter points are the coordinates of the parameter domain of the B-spline surface. For the fitted rapidly thickening region surface, the three-dimensional gradient vector of the parameter points in the B-spline parameter domain is calculated, the Euclidean norm of the three-dimensional gradient vector is calculated, the gradient magnitude is obtained, and the set of points with gradient magnitude greater than or equal to the maximum value of the comprehensive gradient is selected to form the gradient abrupt change zone of the surface. The contour line tracking algorithm is used to extract continuous closed contour lines in the gradient abrupt change zone, which are the abrupt change initiation boundaries. For the fitted stable thickened region surface, the set of points whose gradient magnitude is less than or equal to the minimum value of the comprehensive gradient is selected to form the gradient steady zone of the surface. The contour line tracking algorithm is used to extract continuous closed contour lines in the gradient steady zone, which are the abrupt stable boundaries. The region between the mutation initiation boundary and the mutation stability boundary is designated as the boundary sensitive region, and the region outside the mutation stability boundary within the wall thickness mutation region is designated as the non-boundary sensitive region. The number of undetected defects in the boundary sensitive area is counted and compared with the total number of defects in the wall thickness abrupt change area to obtain the proportion of undetected defects in the boundary sensitive area. The number of defects missed in the boundary sensitive area is counted and compared with the volume of the boundary sensitive area to obtain the unit volume missed detection density of the boundary sensitive area. The number of defects missed in the non-boundary sensitive area is counted and compared with the volume of the non-boundary sensitive area to obtain the unit volume missed detection density of the non-boundary sensitive area. The unit volume missed detection density of the boundary sensitive area is then compared with the unit volume missed detection density of the non-boundary sensitive area to obtain the missed detection density ratio. The product of the percentage of missed detections at the boundary and the percentage of missed detections density is used to obtain the concentration factor of missed detections in the boundary sensitive area. It should be noted that the percentage of missed defects at the boundary indicates what proportion of all missed defects in the entire thickness abrupt change region occur within the boundary-sensitive area, reflecting the contribution of the sensitive area to the total number of missed defects in the abrupt change region; the missed defect density ratio indicates the multiple of the missed defect frequency per unit volume in the boundary-sensitive area relative to the non-sensitive area, eliminating the interference of regional volume differences, and reflecting the concentration of missed defects in the sensitive area. While the boundary-sensitive area bears the majority of the missed defects in the thickness abrupt change region, its missed defect density is significantly higher than the overall concentration in the non-sensitive area. The larger the missed defect concentration coefficient, the more likely the missed defects are to preferentially accumulate in the boundary-sensitive area, which is the core high-incidence area of ​​missed defects in the thickness abrupt change region; the smaller the missed defect concentration coefficient, the less likely the missed defects are to be concentrated in the boundary-sensitive area within the abrupt change region. If the concentration factor of missed detection is greater than or equal to the threshold of the concentration factor of missed detection, it means that the missed defects are concentrated in the boundary sensitive area; If the concentration factor of missed detection is less than the threshold of the concentration factor of missed detection, it means that the missed defects are not concentrated in the boundary sensitive area; Step 4: If the defects are concentrated in the boundary sensitive area, determine the density adjustment coefficient, adjust the quality inspection sampling density in the boundary sensitive area, and reduce the missed defects in the boundary sensitive area within the wall thickness change area. Set an allowable false negative rate, test the false negative rate at different sampling densities in non-boundary sensitive areas, and screen out the minimum sampling density that meets the allowable false negative rate as the benchmark sampling density; In the boundary sensitive area, the false negative rate under different sampling densities is tested, and the minimum sampling density that meets the allowable false negative rate is selected and denoted as the sensitive sampling density. The density adjustment coefficient of the boundary sensitive area is obtained by comparing the sensitive sampling density with the reference sampling density. Based on the density adjustment coefficient, the density adjustment coefficient is determined and the quality inspection sampling density of the boundary sensitive area is adjusted to reduce the missed detection of defects in the boundary sensitive area within the wall thickness change area. The technical solution of this invention is as follows: If the risk of missing defects in areas with abrupt changes in wall thickness is high, based on wall thickness data, a double boundary marking method is used to mark the initiation boundary and the stable boundary of the abrupt change, delineate the boundary sensitive area, analyze the concentration of missed defects in the boundary sensitive area, and determine whether the missed defects are in the boundary sensitive area; if they are concentrated in the boundary sensitive area, a density adjustment coefficient is determined, and the quality inspection sampling density in the boundary sensitive area is adjusted to reduce the missed defects in the boundary sensitive area within the wall thickness abrupt change area; For the high risk of missing defects in areas with abrupt changes in wall thickness, this invention binds the three-dimensional coordinates of the measurement points and converts them into unified rectangular coordinates, combines the axial and circumferential wall thickness gradient vectors to synthesize a comprehensive gradient, and accurately marks it through B-spline three-dimensional surface fitting and contour line tracing. The system identifies the initiation and stabilization boundaries of mutations and delineates sensitive boundary areas. A concentration coefficient for missed defects is obtained by multiplying the percentage of missed defects at the boundary by the ratio of missed defect density. This quantifies the degree of clustering of missed defects in sensitive areas, effectively eliminating interference from regional volume differences and accurately locating core high-incidence areas of missed defects. Using the allowable missed defect rate as a benchmark, the baseline sampling density for non-boundary sensitive areas and the sensitive sampling density for boundary sensitive areas are determined. A density adjustment coefficient is used to specifically adjust the detection sampling density in sensitive areas. This avoids the cost waste of blindly increasing the overall sampling density while achieving enhanced detection of core areas of missed defects. This significantly reduces the defect missed rate in areas of wall thickness mutations, especially in boundary sensitive areas, thus improving the accuracy and efficiency of seamless steel pipe quality inspection.

[0023] Example 3 Based on the same inventive concept as the image processing-based seamless steel pipe quality inspection method in the foregoing embodiments, such as Figure 2 As shown, this application provides a seamless steel pipe quality inspection system based on image processing, wherein the system specifically includes the following modules: Region identification module: Extracts design parameters from the seamless steel pipe design file, combines them with the actual characteristic data of the seamless steel pipe, and identifies regions in the steel pipe where the wall thickness changes abruptly. Missed Detection Risk Assessment Module: Based on the wall thickness abrupt change region, obtain images corresponding to the missed defects found in the historical secondary inspection, build a historical missed defect image library, and analyze the missed defect risk in the wall thickness abrupt change region; Boundary delineation module: If the risk of missing defects in the area of ​​sudden change in wall thickness is high, based on the wall thickness data, the double boundary marking method is used to mark the initiation boundary and the stable boundary of the change, delineate the boundary sensitive area, analyze the concentration of missed defects in the boundary sensitive area, and determine whether the missed defects are in the boundary sensitive area. Density adjustment module: If the focus is on the boundary sensitive area, determine the density adjustment coefficient, adjust the quality inspection sampling density of the boundary sensitive area, and reduce the missed detection of defects in the boundary sensitive area within the wall thickness change area.

[0024] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A seamless steel pipe quality inspection method based on image processing, characterized in that: include: Extract design parameters from seamless steel pipe design documents and combine them with actual characteristic data of seamless steel pipes to identify areas of abrupt changes in wall thickness within the steel pipe; Based on the region of abrupt change in wall thickness, images of defects that were missed during historical secondary inspections were obtained, a historical image library of missed defects was constructed, and the risk of missed defects in the region of abrupt change in wall thickness was analyzed. If the risk of missing defects in areas with abrupt changes in wall thickness is high, a double boundary marking method is used based on wall thickness data to mark the initiation boundary and the stable boundary of the change, delineate the boundary sensitive area, analyze the concentration of missed defects in the boundary sensitive area, and determine whether the missed defects are in the boundary sensitive area. If the focus is on the boundary-sensitive area, determine the density adjustment coefficient, adjust the quality inspection sampling density of the boundary-sensitive area, and reduce the missed detection of defects in the boundary-sensitive area within the region of sudden change in wall thickness.

2. The seamless steel pipe quality inspection method based on image processing according to claim 1, characterized in that: The process for identifying the region of abrupt change in wall thickness is as follows: The core parameters of the steel pipe design documents are obtained from the CAD drawings. The design parameters include the standard wall thickness value and the actual characteristic data of the steel pipe, including the actual wall thickness value. The seamless steel pipe is divided into several grid units according to the grid pattern. Measurement points are set at the intersection of the grid units, and the actual wall thickness value of each measurement point of the seamless steel pipe is obtained by ultrasonic thickness measurement. The wall thickness deviation at each measurement point is obtained by subtracting the actual wall thickness value from the standard wall thickness value. If the wall thickness deviation at the measurement point is greater than or equal to the wall thickness deviation limit, the corresponding measurement point is recorded as the thickening abrupt change point. All thickening abrupt change points in the seamless steel pipe are extracted, and cluster analysis is performed on the thickening abrupt change points using 8-neighbor connected domains to cluster all connected thickening abrupt change points into wall thickness abrupt change regions.

3. The seamless steel pipe quality inspection method based on image processing according to claim 1, characterized in that: The process of constructing the historical missed defect image library is as follows: Extract images and actual wall thickness data corresponding to missed defects found in historical secondary inspections, and classify them according to defect type; For each type of missed defect, three core image features are extracted, including: edge features, grayscale features, and morphological features. Based on defect type and image features, a historical database of missed defect images is constructed.

4. The seamless steel pipe quality inspection method based on image processing according to claim 3, characterized in that: The process for analyzing the risk of missed defects in regions with abrupt changes in wall thickness is as follows: Extract the types and number of missed defects from historical secondary detections in areas of abrupt changes in wall thickness from the historical missed defect image database, and analyze them to determine the ratio of missed defect types to the ratio of missed defect numbers. The defect omission coefficient is obtained by multiplying the ratio of missed defect types with the ratio of missed defect quantities. If the defect omission coefficient is less than or equal to the defect omission coefficient threshold, the risk of defect omission in the wall thickness abrupt change region is low; otherwise, the risk of defect omission in the wall thickness abrupt change region is high.

5. The seamless steel pipe quality inspection method based on image processing according to claim 4, characterized in that: The process for determining the ratio of missed defect types to the ratio of missed defect quantities is as follows: The number of undetected defect types in the area of ​​abrupt wall thickness change is counted and compared with the total number of undetected defect types in seamless steel pipes to obtain the ratio of undetected defect types in the area of ​​abrupt wall thickness change. The number of defects missed in the wall thickness abrupt change area is counted and compared with the total number of defects missed in seamless steel pipes to obtain the defect missed number ratio in the wall thickness abrupt change area.

6. The seamless steel pipe quality inspection method based on image processing according to claim 1, characterized in that: The process of delineating the boundary sensitive area is as follows: Using one end face or process baseline of the seamless steel pipe as the starting zero point, a linear encoder is used to collect the wall thickness data of each measurement point with an ultrasonic thickness gauge, while recording the axial displacement value of the measurement point along the length of the steel pipe, which is the axial distance coordinate of the measurement point; using a preset baseline on the surface of the steel pipe as the starting point, the sampling angle is consistent with the grid division angle, and the angle of the measurement point along the circumference of the seamless steel pipe is located by a combination of an indexing plate and a rotary encoder, which is the circumferential angle of the measurement point; The axial distance coordinates, circumferential angles, and actual wall thickness values ​​of each measurement point are bound and stored to form a three-dimensional coordinate point set of the measurement points; The circumferential angles of the three-dimensional coordinate set of the measurement points are converted into Cartesian coordinates to form a unified three-dimensional rectangular coordinate system. The three-dimensional rectangular coordinates of the measurement points are then analyzed and processed to obtain the mutation initiation boundary and mutation stability boundary. The region between the mutation initiation boundary and the mutation stability boundary is designated as the boundary sensitive region.

7. The seamless steel pipe quality inspection method based on image processing according to claim 6, characterized in that: The process for determining the mutation initiation boundary and mutation stability boundary is as follows: Based on any measurement point, the difference between the actual wall thickness value of the current measurement point and the actual wall thickness value of the adjacent axial measurement point is calculated to obtain the axial wall thickness deviation. The distance between the current measurement point and the corresponding adjacent axial measurement point is obtained and recorded as the axial distance. The ratio of the axial wall thickness deviation to the axial distance is calculated to obtain the axial wall thickness gradient of the measurement point. The difference between the actual wall thickness value of the current measurement point and the actual wall thickness value of the adjacent circumferential measurement point is calculated to obtain the circumferential wall thickness deviation. The distance between the current measurement point and the corresponding adjacent circumferential measurement point is obtained and recorded as the circumferential distance. The ratio of the circumferential wall thickness deviation to the circumferential distance is calculated to obtain the circumferential wall thickness gradient of the measurement point. The gradient amplitude calculation formula is used to vector synthesize the axial wall thickness gradient and the circumferential wall thickness gradient of the steel pipe to obtain the comprehensive gradient of the measurement point. If the overall gradient is greater than or equal to the maximum value of the overall gradient, then the corresponding measurement point is recorded as a candidate point for rapid thickening. If the combined gradient is less than or equal to the minimum value of the combined gradient, then the corresponding measurement point is recorded as a candidate point for thickening stability. All rapid thickening candidate points are integrated into a mutation initiation boundary candidate point set, and all thickening stable candidate points are integrated into a mutation stable boundary point set. For the candidate point set of mutation initiation boundary and the stable point set of mutation, B-spline three-dimensional surface fitting is used to map the three-dimensional coordinates of all measurement points in the candidate point set of mutation initiation boundary and the stable point set of mutation into the B-spline parameter domain, and the surface of the rapidly thickened region and the surface of the stable thickened region are obtained respectively. The parameter points are the coordinates of the parameter domain of the B-spline surface. For the fitted rapidly thickening region surface, the three-dimensional gradient vector of the parameter points in the B-spline parameter domain is calculated, the Euclidean norm of the three-dimensional gradient vector is calculated, the gradient magnitude is obtained, and the set of points with gradient magnitude greater than or equal to the maximum value of the comprehensive gradient is selected to form the gradient abrupt change zone of the surface. The contour line tracking algorithm is used to extract continuous closed contour lines in the gradient abrupt change zone, which are the abrupt change initiation boundaries. For the fitted stable thickened region surface, the set of points whose gradient magnitude is less than or equal to the minimum value of the comprehensive gradient is selected to form the gradient steady zone of the surface. The contour line tracking algorithm is used to extract continuous closed contour lines in the gradient steady zone, which are the abrupt stable boundaries.

8. The seamless steel pipe quality inspection method based on image processing according to claim 6, characterized in that: The process of determining whether missed defects are concentrated in the boundary sensitive area is as follows: The number of undetected defects in the boundary sensitive area is counted and compared with the total number of defects in the wall thickness abrupt change area to obtain the proportion of undetected defects in the boundary sensitive area. The number of defects missed in the boundary sensitive area is counted and compared with the volume of the boundary sensitive area to obtain the unit volume missed detection density of the boundary sensitive area. The number of defects missed in the non-boundary sensitive area is counted and compared with the volume of the non-boundary sensitive area to obtain the unit volume missed detection density of the non-boundary sensitive area. The unit volume missed detection density of the boundary sensitive area is then compared with the unit volume missed detection density of the non-boundary sensitive area to obtain the missed detection density ratio. The proportion of missed defects at the boundary is multiplied by the ratio of missed defects density to obtain the concentration coefficient of missed defects in the boundary sensitive area. If the concentration coefficient of missed defects is greater than or equal to the threshold of the concentration coefficient of missed defects, it means that the missed defects are concentrated in the boundary sensitive area; otherwise, it means that the missed defects are not concentrated in the boundary sensitive area.

9. The seamless steel pipe quality inspection method based on image processing according to claim 8, characterized in that: The process of determining the density adjustment coefficient is as follows: Set an allowable false negative rate, test the false negative rate at different sampling densities in non-boundary sensitive areas, and screen out the minimum sampling density that meets the allowable false negative rate as the benchmark sampling density; In the boundary sensitive area, the false negative rate under different sampling densities is tested, and the minimum sampling density that meets the allowable false negative rate is selected and denoted as the sensitive sampling density. The density adjustment coefficient of the boundary sensitive area is obtained by comparing the sensitive sampling density with the reference sampling density.

10. A seamless steel pipe quality inspection system based on image processing, characterized in that, The system is used to perform the method according to any one of claims 1-9, the system comprising: Region identification module: Extracts design parameters from the seamless steel pipe design file, combines them with the actual characteristic data of the seamless steel pipe, and identifies regions in the steel pipe where the wall thickness changes abruptly. Missed Detection Risk Assessment Module: Based on the wall thickness abrupt change region, obtain images corresponding to the missed defects found in the historical secondary inspection, build a historical missed defect image library, and analyze the missed defect risk in the wall thickness abrupt change region; Boundary delineation module: If the risk of missing defects in the area of ​​sudden change in wall thickness is high, based on the wall thickness data, the double boundary marking method is used to mark the initiation boundary and the stable boundary of the change, delineate the boundary sensitive area, analyze the concentration of missed defects in the boundary sensitive area, and determine whether the missed defects are in the boundary sensitive area. Density adjustment module: If the focus is on the boundary sensitive area, determine the density adjustment coefficient, adjust the quality inspection sampling density of the boundary sensitive area, and reduce the missed detection of defects in the boundary sensitive area within the wall thickness change area.

Images