Weld seam false defect recognition and filtering system based on image processing

Abstract

The present application relates to the technical field of image data processing, and more particularly to a welding seam pseudo-defect identification and filtering system based on image processing, which acquires a surface image of a local area of a welding seam through an image acquisition module; labels a sub-area of the welding seam as a photosensitive dominant sub-area or a photosensitive non-dominant sub-area through a region labeling module; screens a defect confidence suspected sub-area according to a comparison result of texture characteristic quantization values of the photosensitive dominant sub-area and the photosensitive non-dominant sub-area through a preliminary identification module; analyzes a spatial distribution of the defect confidence suspected sub-area through a feature matching module to determine a defect filtering type of the local area of the welding seam; and determines a defect filtering strategy for the local area of the welding seam through an identification and filtering module. Thus, the difference between pseudo-defects and real defects in optical response rules is utilized, a targeted area differentiation analysis mechanism is adopted, and the targeting of the filtering strategy is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing, and in particular to a welding seam pseudo-defect identification and filtering system based on image processing. BACKGROUND

[0002] In the field of pipeline transportation, the quality of the welding seam, as the core part of the structural connection, directly determines the safety and reliability of the equipment operation. Accurate identification of welding seam defects is a key link in industrial quality detection. Current welding seam detection systems based on image processing usually acquire welding seam surface images through a camera, and then use image segmentation, feature extraction, pattern recognition, and other algorithms to identify defect areas. However, in actual detection scenarios, the corner parts of the welding fusion line are prone to form point-like or strip-like high-light areas. This abnormal image area caused by optical reflection characteristics is not an actual defect of the welding seam. It has a high similarity in image gray-scale features with real defects, and is easily misjudged as a real defect by the detection system.

[0003] The formation of pseudo-defects is related to the pose of the image capture lens and the welding metal surface. When the relative angle between the plane where the local area of the welding seam is located and the camera meets certain conditions, mirror reflection will occur, and then the pseudo-defect phenomenon with abnormal image features will be formed. This regularity provides a technical breakthrough for the identification and filtering of pseudo-defects.

[0004] For example, Chinese Patent Publication No. CN119963489A discloses a pseudo-defect identification model construction method, a pseudo-defect identification method, and corresponding devices. The method includes: S11, processing a welding seam radiographic film containing pseudo-defects to obtain a digital image containing pseudo-defects; S12, cropping the digital image to obtain a plurality of sub-images of a predetermined size, and obtaining sub-images containing pseudo-defects as samples; S13, based on all samples, respectively obtaining first samples and second samples according to a predetermined proportion; S14, constructing a detection network, training the detection network based on the first samples to obtain a detection model, verifying the detection model based on the second samples, and when the verification result meets a predetermined index, using the detection model as a pseudo-defect identification model.

[0005] The existing technology also has the following problems:

[0006] The existing technology does not consider the phenomenon of mirror reflection of the corner of the welding fusion line due to the specific pose relationship between the camera and the metal surface, which seriously interferes with the recognition accuracy of the detection system. The existing technology cannot identify the light sensitivity difference of the material surface, and does not fully utilize the difference in optical response law between pseudo-defects and real defects, resulting in insufficient targeting of the filtering strategy. SUMMARY

[0007] To this end, the present application provides a weld seam pseudo-defect identification and filtering system based on image processing, to overcome the problem that the prior art cannot identify the photosensitive difference of the material surface, and does not fully utilize the difference in optical response law between pseudo-defects and real defects, resulting in insufficient targeting of the filtering strategy.

[0008] To achieve the above-mentioned purpose, the present application provides a weld seam pseudo-defect identification and filtering system based on image processing, comprising:

[0009] An image acquisition module, comprising a camera for acquiring a surface image of a local area of a weld seam and a light source for illuminating the weld seam;

[0010] A region labeling module connected with the image acquisition module, for labeling a weld seam sub-region as a photosensitive dominant sub-region or a photosensitive non-dominant sub-region according to an image photosensitive coefficient of the weld seam sub-region in the local area of the weld seam;

[0011] Wherein, the image photosensitive coefficient is determined according to the pose state of the plane where the weld seam sub-region is located and the camera;

[0012] A primary identification module connected with the region labeling module, for determining the comparison result of the texture feature quantization values of the photosensitive dominant sub-region and the photosensitive non-dominant sub-region, to screen a defect confidence suspected sub-region;

[0013] A feature matching module connected with the primary identification module, for analyzing the spatial distribution of the defect confidence suspected sub-region, to determine a defect filtering type of the local area of the weld seam;

[0014] An identification and filtering module connected with the image acquisition module and the feature matching module respectively, for determining a defect filtering strategy for the local area of the weld seam according to the defect filtering type.

[0015] Further, the region labeling module is used to determine the image photosensitive coefficient of the weld seam sub-region, wherein,

[0016] The region labeling module is used to divide the local area of the weld seam into a plurality of triangular weld seam sub-regions, and extract the coordinates of the three vertices of the weld seam sub-region in the geodetic coordinate system according to the point cloud data to determine the region plane corresponding to the weld seam sub-region;

[0017] The region labeling module determines the sine value of the included angle between the normal vector of the region plane and the central axis of the view angle of the camera as the image photosensitive coefficient.

[0018] Further, the region labeling module is used to label the weld seam sub-region as a photosensitive dominant sub-region or a photosensitive non-dominant sub-region, wherein,

[0019] The region labeling module is configured to compare the image photosensitivity coefficient with a preset image photosensitivity reference coefficient.

[0020] If the image photosensitivity coefficient is greater than the image photosensitivity reference coefficient, the region labeling module is configured to label a weld sub-region as a photosensitivity dominant sub-region.

[0021] If the image photosensitivity coefficient is less than or equal to the image photosensitivity reference coefficient, the region labeling module is configured to label a weld sub-region as a photosensitivity non-dominant sub-region.

[0022] Further, the initial identification module is configured to determine a comparison result of texture feature quantization values of the photosensitivity dominant sub-region and the photosensitivity non-dominant sub-region, wherein,

[0023] The initial identification module is configured to determine a gray value standard deviation of a plurality of pixel points in each photosensitivity dominant sub-region as a first texture feature quantization value, and determine a gray value standard deviation of pixel points in all photosensitivity non-dominant sub-regions as a second texture feature quantization value.

[0024] The initial identification module is configured to calculate a ratio of the first texture feature quantization value to the second texture feature quantization value.

[0025] Further, the initial identification module is configured to screen out a defect confidence suspected sub-region according to a comparison result of the ratio and a confidence suspected screening condition, wherein,

[0026] If the ratio satisfies the confidence suspected screening condition, the initial identification module marks the photosensitivity dominant sub-region as a defect confidence suspected sub-region.

[0027] The confidence suspected screening condition is that the ratio is greater than a preset ratio threshold value.

[0028] Further, the feature matching module is configured to analyze a spatial distribution of the defect confidence suspected sub-region, wherein,

[0029] The feature matching module is configured to obtain center coordinates of all defect confidence suspected sub-regions, and determine a distribution position relative to a non-photosensitivity region reference axis based on the center coordinates.

[0030] The non-photosensitivity region reference axis passes through a region center of a region surrounded by the photosensitivity non-dominant sub-region, and is parallel to an axis of the welded pipe.

[0031] Further, the feature matching module is configured to determine a defect filtering type of a local region of the weld, wherein,

[0032] If the center coordinates of the defect confidence suspected sub-region are distributed on both sides of the non-photosensitivity region reference axis, respectively, the feature matching module determines that the local region of the weld is of a first defect filtering type.

[0033] If the center coordinates of the defect confidence suspected sub-regions are only distributed on one side of the non-photosensitive region reference axis, the feature matching module determines that the local region of the weld is of a second defect filtering type.

[0034] Further, the identification filtering module is configured to determine a defect filtering strategy for the local region of the weld, wherein,

[0035] If the local region of the weld is of the first defect filtering type, the identification filtering module determines the defect filtering strategy for the local region of the weld as obtaining a plurality of frames of first images of the local region of the weld, and determining whether to determine the local region of the weld as a false defect region of the weld according to a gray scale change of a pixel point in the first images.

[0036] If the local region of the weld is of the second defect filtering type, the identification filtering module obtains a plurality of frames of second images of the local region of the weld, and determines whether to determine the local region of the weld as a false defect region of the weld according to a gray scale change of a pixel point in the second images.

[0037] The light source brightness corresponding to each frame of the first images is different from each other, and the camera shooting position corresponding to each frame of the second images is different from each other.

[0038] Further, the identification filtering module is configured to determine whether to determine the local region of the weld as a false defect region of the weld according to the gray scale change of the pixel point in the first images, wherein,

[0039] The identification filtering module is configured to obtain a difference value of the average gray scale of the pixel points on both sides of the non-photosensitive region reference axis, and if the difference value corresponding to each frame of the first images is within a preset difference value range, the identification filtering module determines to determine the local region of the weld as a false defect region of the weld.

[0040] Further, the identification filtering module is configured to determine whether to determine the local region of the weld as a false defect region of the weld according to the gray scale change of the pixel point in the second images, wherein,

[0041] The identification filtering module is configured to obtain a change amount of the average gray scale of the pixel points in the adjacent frames of the second images in the region where the center coordinates of the defect confidence suspected sub-regions are distributed, and if the change amount is greater than a preset change amount threshold, the identification filtering module determines to determine the local region of the weld as a false defect region of the weld.

[0042] Compared with the prior art, the present application has the beneficial effects that the present application acquires the surface image of the local area of the weld through the image acquisition module; the sub-area of the weld is marked as a photosensitive dominant sub-area or a photosensitive non-dominant sub-area through the area marking module; the defect confidence suspected sub-area is screened according to the comparison result of the texture characteristic quantitative value of the photosensitive dominant sub-area and the photosensitive non-dominant sub-area through the primary identification module; the spatial distribution of the defect confidence suspected sub-area is analyzed to determine the defect filtering type of the local area of the weld through the feature matching module; and the defect filtering strategy for the local area of the weld is determined through the identification filtering module. Further, the difference in optical response law between the pseudo defect and the real defect is utilized, a targeted area differentiation analysis mechanism is adopted, and the filtering strategy is improved in pertinence.

[0043] Further, in the present application, the mirror reflection phenomenon of the weld fusion line corner is caused by the reflection of the light source light on the metal surface and the convergence of the camera lens when the local area of the weld and the camera form a specific pose. The normal vector of the area plane is the core information representing the spatial orientation of the plane, and the central axis of the camera's viewing angle clearly indicates the direction of the light received by the camera. The angle between the two reflects the photosensitive intensity of the sub-area surface to the camera: the larger the angle, the higher the matching degree of the area plane orientation and the camera receiving direction, and the stronger the intensity of the mirror reflection. The sine value of the angle is used as the image photosensitive coefficient to establish a linear correlation between the angle change and the photosensitive intensity change, so that the photosensitive coefficient can intuitively and accurately quantify the photosensitive characteristic difference of the sub-area, and the pseudo defect phenomenon is identified through the pose correlation characteristics.

[0044] Further, in the present application, the parts such as the weld fusion line corner that are prone to pseudo defects have a large offset angle between the sub-area normal vector and the lens central axis due to the special relationship between the surface plane and the camera pose, and the corresponding image photosensitive coefficient is higher, which is easy to cause mirror reflection and form pseudo defect visual features. The areas with smaller photosensitive coefficients have stable weld surface posture and uniform photosensitivity, which are closer to the inherent texture environment of real defects. By comparing the quantitative threshold values, the photosensitive difference areas caused by the illumination angle and the inherent surface feature areas are separated, solving the problem that the prior art cannot distinguish the photosensitive difference.

[0045] Further, in the present application, the mirror reflection of the weld fusion line corner causes the reflection path of the light on the weld surface to have symmetry characteristics; or the spatial distribution is concentrated on one side of the reference axis, without symmetry characteristics. By distributing the suspected sub-area on both sides of the axis to determine the first defect filtering type and distributing it on one side to determine the second defect filtering type, the difference in spatial distribution between the defect and the real defect can be accurately captured, and through the division of the distribution type, a targeted area differentiation analysis mechanism is realized, improving the reliability of the weld defect detection.

[0046] Further, the present application identifies the filtering module by controlling the light source brightness gradient to collect multiple frames of first images, calculates the difference value of the average gray value of the pixel points on both sides of the reference axis in each frame of image, and judges whether the difference value falls within the preset range, when all the difference values meet the interval requirement, it shows that the gray contrast of the region is caused by consistent visual illusion of light interference, not the inherent characteristics of real defects, and then it can be determined that the region is a weld pseudo-defect region; the difference of optical response law is used to realize the accurate distinction between pseudo-defects and real defects.

[0047] Further, the present application adjusts the camera shooting position to obtain multiple frames of second images, and constructs a verification dimension of view angle change: for pseudo-defects, the change of camera shooting position will directly lead to the offset of light reflection path, and the gray mutation region originally formed by directional reflection will appear obvious brightness attenuation or morphological offset, which is reflected in the image data as the dramatic change of the average gray of the suspected region pixel points in adjacent frames; for real defects, the stability of its physical structure makes the gray contrast between the surface defects and the surrounding area consistent under different shooting angles, and the average gray change of the suspected region in adjacent frames will maintain at a low level. When the average gray change of the suspected region in adjacent frames of second images is greater than a preset threshold, it shows that the gray feature of the region has strong sensitivity to the change of view angle, which meets the cause of the visual illusion of pseudo-defects depending on the pose, and the difference of optical response law is used to realize the accurate distinction between pseudo-defects and real defects. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 The system block diagram of the weld pseudo-defect identification and filtering system based on image processing of the embodiment of the present application;

[0049] Figure 2 The angle diagram between the normal vector of the region plane and the central axis of the view angle of the camera of the embodiment of the present application;

[0050] Figure 3 The logic flow chart of marking the photosensitive dominant sub-region or the photosensitive non-dominant sub-region of the embodiment of the present application;

[0051] Figure 4 The logic flow chart of judging the defect filtering type of the embodiment of the present application;

[0052] In the figure: 1-camera, 2-central axis of view angle, 3-weld sub-region, 4-normal vector of region plane. DETAILED DESCRIPTION

[0053] In order to make the purpose and advantages of the present application more clear and explicit, the present application is further described below combined with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present application, and do not limit the present application.

[0054] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. Those skilled in the art will appreciate that the embodiments are only used to explain the technical principles of the present application, and are not intended to limit the protection scope of the present application.

[0055] It should be noted that, in the description of the present application, the terms indicating the direction or positional relationship of "upper", "lower", "inner", "outer" and the like are based on the direction or positional relationship shown in the drawings, which is only for the convenience of description, and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application.

[0056] In addition, it should also be noted that, in the description of the present application, unless otherwise explicitly specified and limited, the term "connection" should be understood broadly, for example, it can be a fixed connection, or a detachable connection, or an integral connection; it can be a mechanical connection, or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or a communication within two elements. Those skilled in the art can understand the specific meaning of the above-mentioned terms in the present application according to the specific circumstances.

[0057] Please refer to Figure 1 The system block diagram of the weld false defect identification and filtering system based on image processing of the embodiment of the present application is shown, and the weld false defect identification and filtering system based on image processing of the present application comprises:

[0058] An image acquisition module, which comprises a camera for acquiring a surface image of a local area of a weld and a light source for irradiating the weld;

[0059] The present application does not limit the camera, which can adopt a CMOS industrial camera with a resolution of ≥12 million pixels.

[0060] The present application does not limit the specific structure of the light source, and in the present application, a ring-shaped LED light source can be adopted, and the brightness can be adjusted within the range of 2000-10000 Lux.

[0061] A region labeling module connected with the image acquisition module, for labeling a weld sub-region in the local area of the weld as a photosensitive dominant sub-region or a photosensitive non-dominant sub-region according to an image photosensitive coefficient of the weld sub-region;

[0062] Wherein, the image photosensitive coefficient is determined according to the pose state of the plane where the weld sub-region is located and the camera;

[0063] Specifically, the present application does not limit the structure of the region labeling module, which can be an image processor, and the weld sub-region in the local area of the weld is labeled and processed by calculating the image photosensitive coefficient.

[0064] A primary identification module, connected with the region labeling module, is configured to determine a comparison result of texture feature quantization values of the photosensitive dominant sub-region and the photosensitive non-dominant sub-region, so as to screen the defect confidence suspected sub-region;

[0065] A feature matching module, connected with the primary identification module, is configured to analyze the spatial distribution of the defect confidence suspected sub-region, so as to determine the defect filtering type of the local region of the weld;

[0066] An identification filtering module, connected with the image acquisition module and the feature matching module respectively, is configured to determine the defect filtering strategy for the local region of the weld according to the defect filtering type.

[0067] Specifically, the specific structure of the primary identification module, the feature matching module and the identification filtering module is not limited, and each unit thereof can be constituted by a logic component. Each module is connected through an Ethernet / IP industrial Ethernet to ensure real-time data transmission. The logic component can be a field programmable logic component, a microprocessor, a processor used in a computer, etc., which will not be described here.

[0068] Specifically, the region labeling module is configured to determine the image photosensitive coefficient of the weld sub-region, wherein,

[0069] The region labeling module is configured to divide the local region of the weld into a plurality of triangular weld sub-regions, and to determine the region plane corresponding to the weld sub-region according to the coordinates of the three vertices of the weld sub-region in the geodetic coordinate system extracted from the point cloud data.

[0070] The region labeling module determines the sine value of the included angle between the normal vector of the region plane and the central axis of the camera view angle as the image photosensitive coefficient.

[0071] Please refer to Figure 2 The included angle between the central axis 2 of the camera 1 view angle and the normal vector 4 of the region plane corresponding to the weld sub-region 3 is θ, and the image photosensitive coefficient is sinθ. The three vertices of the weld sub-region 3 are a, b and c.

[0072] In the implementation of the present application, the region labeling module collects the point cloud data of the local region of the weld through a laser point cloud scanner, extracts the coordinates of the three vertices a, b and c of the weld sub-region from the point cloud data, and constructs two coplanar direction vectors pointing to b and c respectively with vertex a as the base point. The normal vector of the region plane is a vector perpendicular to any two non-collinear vectors in the plane, which can be obtained by vector cross multiplication operation. This is a mathematical calculation means, and the specific calculation method of the normal vector of the region plane will not be described here.

[0073] In the implementation of the present application, the local area of the weld in the single frame image is divided into several triangular weld sub-areas, and the length of the side of the triangular weld sub-area is determined according to the diameter of the welded pipe, and preferably, the length of the side is 1 / n of the diameter of the welded pipe, and the value of n is [8, 15].

[0074] In the implementation of the present application, the coordinates of the three vertices of the weld sub-area are obtained by a laser point cloud scanner, the X-axis of the geodetic coordinate system is parallel to the axis of the welded pipe, and the Y-axis is parallel to the horizontal ground.

[0075] In the present application, the local area of the weld is divided into several triangular sub-areas based on the geometric characteristic that a triangle has a uniquely determined plane, and compared with other polygons, the triangular sub-area can accurately lock the spatial pose of the plane through the coordinates of the three vertices.

[0076] As can be understood by those skilled in the art from the generation mechanism of the pseudo defect, the mirror reflection phenomenon of the weld fusion line corner is caused by the fact that when the local area of the weld and the camera form a specific pose, the light reflected by the metal surface converges into the camera lens, the normal vector of the area plane is the core information representing the spatial orientation of the plane, the central axis of the view angle of the camera clearly indicates the direction of the light received by the camera, and the included angle between the two reflects the matching degree of the surface of the weld sub-area to the camera: the larger the included angle, the higher the matching degree of the orientation of the area plane to the receiving direction of the camera, and the greater the intensity of the mirror reflection. The sine value of the included angle is used as the image light sensitivity coefficient because the sine function is monotonically increasing in the range of 0 to 90 degrees, which can establish a linear correlation between the change of the included angle and the change of the light sensitivity, so that the light sensitivity coefficient can intuitively and accurately quantify the difference in light sensitivity of the sub-area, and the pseudo defect phenomenon can be identified through the pose correlation characteristic.

[0077] Please refer to Figure 3 As shown in the figure, it is a logic flow chart for labeling the light-sensitive dominant sub-area or the light-insensitive sub-area according to the embodiment of the present application. The area labeling module is used to label the weld sub-area as a light-sensitive dominant sub-area or a light-insensitive sub-area, wherein,

[0078] The area labeling module is used to compare the image light sensitivity coefficient with the preset image light sensitivity reference coefficient.

[0079] If the image light sensitivity coefficient is greater than the image light sensitivity reference coefficient, the area labeling module is used to label the weld sub-area as a light-sensitive dominant sub-area.

[0080] If the image light sensitivity coefficient is less than or equal to the image light sensitivity reference coefficient, the area labeling module is used to label the weld sub-area as a light-insensitive sub-area.

[0081] In the implementation of the present application, the preset image light-sensing reference coefficient can be set by a person skilled in the art according to the light reflection characteristics of the actual material surface. The smaller the image light-sensing reference coefficient set for the material with a smoother surface is. Here, some common image light-sensing reference coefficients of metal materials are provided. The image light-sensing reference coefficient of aluminum alloy material can be set to 0.5, the image light-sensing reference coefficient of stainless steel material can be set to 0.55, and the image light-sensing reference coefficient of carbon steel material can be set to 0.7.

[0082] It can be understood that, in the present application, the parts such as the weld fusion line corners which are prone to false defects have a special relationship between the surface plane and the camera pose, the offset angle between the normal vector of the sub-region and the lens axis is large, the corresponding image light-sensing coefficient is higher, and the mirror reflection is easy to form a visual feature similar to a defect. The areas with smaller light-sensing coefficients have stable weld surface poses, uniform light-sensing, and are closer to the inherent texture environment of the real defects. By comparing the quantitative threshold values, the light-sensing difference areas caused by the light angle are separated from the inherent surface feature areas, and the problem that the prior art cannot distinguish the light-sensing difference is solved.

[0083] Specifically, the preliminary identification module is configured to determine a comparison result of the texture feature quantization values of the light-sensing dominant sub-region and the light-sensing non-dominant sub-region, wherein,

[0084] The preliminary identification module is configured to determine the gray value standard deviation of a plurality of pixel points in each light-sensing dominant sub-region as a first texture feature quantization value, and determine the gray value standard deviation of the pixel points in all light-sensing non-dominant sub-regions as a second texture feature quantization value.

[0085] The preliminary identification module is configured to calculate the ratio of the first texture feature quantization value to the second texture feature quantization value.

[0086] In the implementation of the present application, 10 pixel points can be selected in each light-sensing dominant sub-region and light-sensing non-dominant sub-region, and the gray value standard deviation thereof is calculated as the first texture feature quantization value of each light-sensing dominant sub-region. The gray value standard deviation of the pixel points in all light-sensing non-dominant sub-regions is collected as the second texture feature quantization value.

[0087] In the implementation of the present application, a common grayscale algorithm in the image processing field can be used to convert a color image into a grayscale image. A commonly used method is to calculate the gray value of each pixel point by using a weighted average method. The red, green, and blue three-channel pixel values are used to calculate the gray value of the pixel point by using the weighted average method, which is the prior art, and will not be described here.

[0088] Specifically, the preliminary identification module is configured to screen out a defect confidence suspected sub-region according to a comparison result of the ratio and a confidence suspected screening condition, wherein,

[0089] If the ratio does not satisfy the confidence suspicion screening condition, the initial identification module does not mark the photosensitive dominant sub-region;

[0090] If the ratio satisfies the confidence suspicion screening condition, the initial identification module marks the photosensitive dominant sub-region as a defect confidence suspicion sub-region.

[0091] The confidence suspicion screening condition is that the ratio is greater than a preset ratio threshold.

[0092] In the implementation of the present application, the preset ratio threshold R0 is determined according to the data tested and recorded in advance. 30 groups of defect-free weld samples are selected in advance, the average value R of the ratio of the first texture feature quantization value to the second texture feature quantization value in each group of samples is calculated, and the ratio threshold R0 is determined according to the calculated average value R of the ratio. The ratio threshold R0 = δ × R, δ is the value factor of the ratio threshold, and the value range of δ is [1.1, 1.3]. Preferably, the value of δ is 1.2.

[0093] It can be understood that, in the present application, the photosensitive non-dominant sub-region is uniform in photosensitivity, and the pixel gray scale distribution is concentrated, so the gray scale value standard deviation can be used as the reference difference of the normal texture of the weld. If the photosensitive dominant sub-region is a false defect, local gray scale mutation will occur due to light difference. If it is a real defect, inherent rough texture will be formed due to the material surface such as cracks, pores, etc., and local gray scale mutation also exists. By calculating the ratio of the first and second quantization values, when the ratio is greater than the preset threshold, it indicates that the surface texture gray scale dispersion of the photosensitive dominant sub-region deviates from the normal reference, and the possibility of existing defects is great, realizing the screening of suspected defects through photosensitive difference.

[0094] In actual weld image recognition, the gray scale mutation of the false defect is caused by the mirror reflection caused by the light angle. This reflection only changes the local pixel brightness, which will make the gray scale value standard deviation at a high level. The real defects such as cracks, pores and slag inclusion will form inherent and stable rough texture structure due to the destruction of the material continuity, which will make the gray scale value standard deviation at a high level.

[0095] Specifically, the feature matching module is used to analyze the spatial distribution of the defect confidence suspicion sub-region, wherein,

[0096] The feature matching module is used to obtain the center coordinates of all defect confidence suspicion sub-regions, and determine the distribution position relative to the non-photosensitive area reference axis based on the center coordinates.

[0097] The non-photosensitive area reference axis passes through the region center of the region surrounded by the photosensitive non-dominant sub-region, and is parallel to the axis of the welded pipe.

[0098] In the embodiment of the present application, the center coordinates of the defect confidence suspected sub-region are determined according to the geometric center of the contour pixel point coordinates, and the area center of the area surrounded by the photosensitive non-expressive sub-region is determined through the contour pixel coordinates, which is a common technical means of the image edge algorithm, and will not be described here.

[0099] It can be understood that the center of the area surrounded by the photosensitive non-expressive sub-region and the reference axis parallel to the pipe axis can be used as the spatial reference of the weld surface because the photosensitive non-expressive sub-region has no optical interference. The spatial distribution of the pseudo defect is often affected by the illumination angle and presents certain regular characteristics. By obtaining the center coordinates of the defect confidence suspected sub-region and judging the distribution position of the center coordinates relative to the reference axis, the spatial characteristics of the suspected region can be quantitatively classified.

[0100] Please refer to Figure 4 It is a logic flow chart for determining the defect filtering type of the embodiment of the present application. The feature matching module is used to determine the defect filtering type of the local area of the weld, wherein,

[0101] If the center coordinates of the defect confidence suspected sub-region are distributed on both sides of the non-photosensitive area reference axis, the feature matching module determines that the local area of the weld is of the first defect filtering type.

[0102] If the center coordinates of the defect confidence suspected sub-region are only distributed on one side of the non-photosensitive area reference axis, the feature matching module determines that the local area of the weld is of the second defect filtering type.

[0103] It can be understood that the reflection path of the light on the weld surface caused by the mirror reflection of the weld fusion line corner may have symmetry characteristics, or the spatial distribution may be concentrated on one side of the reference axis without symmetry characteristics. By determining the first defect filtering type when the suspected sub-region is distributed on both sides of the axis and determining the second defect filtering type when the suspected sub-region is distributed on one side, the difference between the defect and the real defect in the spatial distribution rule can be accurately captured, the targeted area differentiation analysis mechanism is realized through the distribution type division, and the reliability of the weld defect detection is improved.

[0104] Specifically, the identification filtering module is used to determine the defect filtering strategy for the local area of the weld, wherein,

[0105] If the local area of the weld is of the first defect filtering type, the defect filtering strategy of the identification filtering module for the local area of the weld is to obtain a plurality of frames of first images of the local area of the weld, and to determine whether the local area of the weld is determined as a pseudo defect area of the weld according to the gray scale change of the pixel points in the first images.

[0106] If the local area of the weld is of the second defect filtering type, the identification filtering module acquires a plurality of frames of the second images of the local area of the weld, and determines whether the local area of the weld is a false defect area of the weld according to the gray scale variation of the pixel points of the second images.

[0107] The light source brightness corresponding to each frame of the first images is different from each other, and the camera shooting positions corresponding to each frame of the second images are different from each other.

[0108] In the implementation of the present application, the light source brightness range corresponding to the first images is 2000-10000 Lux, the adjacent frame brightness difference is 1500 Lux, and the frame number of a single group of the first images is 5.

[0109] In the implementation of the present application, the camera is offset in the horizontal direction perpendicular to the pipe axis, the offset direction is the direction close to the defect confidence suspected sub-area, the single offset distance is 1 cm, and the frame number of a single group of the second images is 5.

[0110] Specifically, the identification filtering module is used to determine whether the local area of the weld is a false defect area of the weld according to the gray scale variation of the pixel points in the first images.

[0111] The identification filtering module is used to acquire the difference between the average values of the pixel points on both sides of the non-photosensitive area reference axis, and if the difference corresponding to each frame of the first images is within a preset difference range, the identification filtering module determines that the local area of the weld is a false defect area of the weld.

[0112] In the implementation of the present application, different difference ranges can be set for different materials, and for stainless steel material, the difference range can be [15, 20].

[0113] For example, for a local area of a stainless steel weld, 5 frames of first images with different brightness are sequentially acquired, and the brightness is 2000 Lux, 3500 Lux, 5000 Lux, 6500 Lux and 8000 Lux, respectively.

[0114] The gray scale average values of the pixel points on both sides under different brightness are calculated respectively, and under the condition of 2000 Lux, the gray scale average value of the pixel points on the A side is G1, and the gray scale average value of the pixel points on the B side is G2, the difference G between the gray scale average value of the pixel points on the A side and the gray scale average value of the pixel points on the B side is calculated, and the difference between the gray scale average value of the pixel points on the A side and the gray scale average value of the pixel points on the B side under each brightness condition is calculated in the same way. 12 =G1-G2, and the difference between the gray scale average value of the pixel points on the A side and the gray scale average value of the pixel points on the B side under each brightness condition is calculated in the same way.

[0115] It can be understood that if the area is a symmetrical mirror reflection of the weld fusion line corner which causes a false defect, its gray scale feature completely depends on the external light intensity, and under different brightness illumination, the gray scale contrast on both sides of the axis will show a stable regularity, and the difference between the average values of the gray scales of the pixel points on both sides will not fluctuate greatly due to brightness adjustment, but will always be in a fixed interval; if it is a real defect, the gray scale difference is caused by the inherent discontinuity of the physical structure of the weld surface, which is little affected by the change of light intensity, and due to the irregularity of the defect itself, the gray scale difference on both sides of the axis under different brightness will deviate from the stable interval of the false defect.

[0116] Based on the above reasons, the identification filtering module acquires multiple frames of first images by controlling the brightness gradient of the light source, calculates the difference between the average values of the gray scales of the pixel points on both sides of the reference axis in each frame of image, and judges whether the difference falls within a preset range, when the difference of all frames meets the interval requirement, it indicates that the gray scale contrast of the area is caused by consistent visual illusion caused by light interference, rather than the inherent characteristics of real defects, and then it can be determined that the area is a weld false defect area; the difference in optical response law is used to realize the accurate distinction between false defects and real defects.

[0117] Specifically, the identification filtering module is used to determine whether the local area of the weld is a weld false defect area according to the gray scale change of the pixel points of the second image.

[0118] The identification filtering module is used to obtain the change amount of the average gray scale of the pixel points of the area in the adjacent frame of the second image, which has the center coordinates of the defect confidence suspected sub-area; if the change amount is greater than a preset change amount threshold, the identification filtering module determines that the local area of the weld is a weld false defect area.

[0119] In the implementation of the present application, different change amount thresholds of the average gray scale of the pixel points in the adjacent frame of the second image can be set for different materials, and for stainless steel materials, the change amount threshold can be 10.

[0120] It can be understood that the suspected sub-area corresponding to the second defect filtering type of the present application is concentrated on one side of the reference axis of the non-photosensitive area, and such suspected defects are not easy to distinguish whether they are caused by real defects or directional visual illusion caused by matching of light at a specific angle and single-sided protruding structure, but real defects are not affected by the change of the camera shooting position.

[0121] Based on the above reasons, the present application obtains multiple second images by adjusting the camera shooting position, and constructs a verification dimension of the change of the visual angle: for the pseudo defects, the change of the camera shooting position will directly cause the offset of the light reflection path, and the originally formed gray mutation area due to the directional reflection will appear obvious brightness attenuation or morphological offset, which is reflected on the image data as the dramatic change of the average gray of the pixel points in the suspected area in the adjacent frames; while for the real defects, the stability of the physical structure makes the gray contrast between the surface defects and the surrounding area consistent under different shooting angles, and the average gray change of the suspected area in the adjacent frames will be maintained at a low level. When the average gray change of the suspected area in the adjacent second images is greater than a preset threshold, it indicates that the gray feature of the area has strong sensitivity to the change of the visual angle, which conforms to the regularity of the visual illusion of the pseudo defects depending on the pose, and accordingly the area can be determined as the pseudo defect area of the weld; otherwise, it is more likely to be a real defect.

[0122] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical solutions after the changes or replacements will fall within the protection scope of the present application.

[0123] The above is only the preferred embodiment of the present application, and is not used to limit the present application; for those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A weld seam false defect identification and filtering system based on image processing, characterized in that, include: The image acquisition module includes a camera for acquiring surface images of a local area of ​​the weld and a light source for illuminating the weld. The region labeling module, which is connected to the image acquisition module, is used to label the weld sub-region as a photosensitive dominant sub-region or a photosensitive non-photosensitive sub-region based on the image photosensitive coefficient of the weld sub-region within the local area of ​​the weld. The image sensitivity coefficient is determined based on the plane containing the weld sub-region and the pose of the camera. The region labeling module is used to divide the local area of ​​the weld into several triangular weld sub-regions, and extract the coordinates of the three vertices of the weld sub-region in the geodetic coordinate system based on the point cloud data to determine the region plane corresponding to the weld sub-region. The region labeling module determines the image photosensitive coefficient as the sine value of the angle between the normal vector of the region plane and the central axis of the camera's field of view. The initial identification module, which is connected to the region labeling module, is used to determine the comparison results of the texture feature quantization values ​​of the photosensitive dominant sub-region and the photosensitive non-dominant sub-region, so as to screen the suspected defect confidence sub-regions. The feature matching module, which is connected to the initial identification module, is used to analyze the spatial distribution of suspected defect confidence sub-regions in order to determine the defect filtering type of the local area of ​​the weld. The identification and filtering module is connected to the image acquisition module and the feature matching module respectively, and is used to determine the defect filtering strategy for the local area of ​​the weld according to the defect filtering type.

2. The weld seam false defect identification and filtering system based on image processing according to claim 1, characterized in that, The region labeling module is used to label weld sub-regions as either photosensitive dominant sub-regions or photosensitive non-dominant sub-regions, wherein... The region labeling module is used to compare the image photosensitivity coefficient with a preset image photosensitivity reference coefficient; If the image photosensitive coefficient is greater than the image photosensitive reference coefficient, the region labeling module is used to label the weld sub-region as a photosensitive sub-region. If the image sensitivity coefficient is less than or equal to the image sensitivity reference coefficient, the region labeling module is used to label the weld sub-region as a non-photosensitive sub-region.

3. The image processing-based weld seam false defect identification and filtering system according to claim 2, characterized in that, The initial identification module is used to determine the comparison result of the texture feature quantization values ​​between the photosensitive dominant sub-region and the photosensitive non-dominant sub-region, wherein... The initial identification module is used to determine the standard deviation of gray values ​​of several pixels in each photosensitive dominant sub-region as the first texture feature quantization value; and to determine the standard deviation of gray values ​​of pixels in all photosensitive non-dominant sub-regions as the second texture feature quantization value; The initial identification module is used to calculate the ratio of the first texture feature quantization value to the second texture feature quantization value.

4. The image processing-based weld seam false defect identification and filtering system according to claim 3, characterized in that, The initial identification module is used to filter out sub-regions with suspected defects based on the comparison results between the ratio and the confidence-based suspected screening criteria. If the ratio satisfies the confidence-suspect screening condition, the initial identification module marks the photosensitive sub-region as a confidence-suspect sub-region of defects; The confidence threshold for screening is that the ratio is greater than a preset ratio threshold.

5. The image processing-based weld seam false defect identification and filtering system according to claim 4, characterized in that, The feature matching module is used to analyze the spatial distribution of suspected defect confidence sub-regions, wherein... The feature matching module is used to obtain the center coordinates of all suspected defect sub-regions and determine their distribution positions relative to the reference axis of the non-photosensitive area based on the center coordinates. The reference axis of the non-photosensitive region passes through the center of the region enclosed by the non-photosensitive sub-regions and is parallel to the axis of the welded pipe.

6. The image processing-based weld seam false defect identification and filtering system according to claim 5, characterized in that, The feature matching module is used to determine the defect filtering type in a local area of ​​the weld, wherein... If the center coordinates of the suspected defect confidence sub-regions are distributed on both sides of the non-photosensitive area reference axis, the feature matching module determines that the local area of ​​the weld is the first defect filtering type. If the center coordinates of the suspected defect confidence sub-region are only distributed on one side of the non-photosensitive area reference axis, the feature matching module determines that the local area of ​​the weld is the second defect filtering type.

7. The image processing-based weld seam false defect identification and filtering system according to claim 6, characterized in that, The identification and filtering module is used to determine a defect filtering strategy for local areas of the weld, wherein... If the local area of ​​the weld is the first type of defect filtering, the defect filtering strategy of the identification and filtering module for the local area of ​​the weld is to acquire several frames of the first image of the local area of ​​the weld, and determine whether to identify the local area of ​​the weld as a false defect area of ​​the weld based on the grayscale change of the pixels in the first image. If the local area of ​​the weld is of the second defect filtering type, the identification and filtering module acquires several frames of second images of the local area of ​​the weld, and determines whether to identify the local area of ​​the weld as a false defect area of ​​the weld based on the pixel grayscale change of the second image. The light source brightness corresponding to the first image in each frame is different, and the camera shooting position corresponding to the second image in each frame is different.

8. The image processing-based weld seam false defect identification and filtering system according to claim 7, characterized in that, The identification and filtering module is used to determine whether to classify the local area of ​​the weld as a false defect area based on the grayscale changes of pixels in the first image. The identification and filtering module is used to obtain the difference in the average grayscale value of pixels on both sides of the reference axis of the non-photosensitive area; if the difference corresponding to the first image of each frame is within the preset difference range, the identification and filtering module determines that the local area of ​​the weld is a false defect area of ​​the weld.

9. The image processing-based weld seam false defect identification and filtering system according to claim 7, characterized in that, The identification and filtering module is used to determine whether to classify the local area of ​​the weld as a false defect area based on the pixel grayscale changes in the second image. The identification and filtering module is used to obtain the change in the average gray level of pixels in the region with the center coordinates of the suspected defective sub-region in the adjacent frame of the second image; if the change is greater than the preset change threshold, the identification and filtering module determines that the local area of ​​the weld is a false defect area of ​​the weld.

Images