CN114964007A - Visual measurement and surface defect detection method for weld size - Google Patents

Abstract

A visual measurement and surface defect detection method for weld size belongs to the field of visual detection algorithms. The visual inspection technology also has the problems of low detection precision and low defect identification rate, and provides a method for visual measurement of weld dimension and detection of surface defects, which comprises the following steps: building a hardware detection system; calibrating a detection system; acquiring a structural light stripe image by using the constructed and calibrated detection system, and preprocessing the structural light stripe image; a structured light central line positioning method and characteristic point extraction; and (4) completing the reconstruction and defect identification of the detected weld joint material by adopting a three-dimensional reconstruction algorithm. The invention ensures to realize higher operation speed on the premise of reducing the operation amount of the algorithm as much as possible under the condition of ensuring that the detection system is stable and has higher precision.

Description

A method for visual measurement of weld size and surface defect detection

technical field

The invention relates to a visual inspection method, in particular to a method for visual measurement of weld seam size and surface defect detection.

Background technique

Welding involves a variety of industrial fields, including aerospace, transportation, construction, metal fabrication and other industries. During the welding process, due to various environmental factors and improper operation, the welding quality cannot reach a satisfactory level. Weld defect detection is an essential link. The traditional weld defect detection is mainly completed by manual detection. This kind of judgment completely depends on the human eye, which often has poor detection accuracy and takes a long time. Moreover, as the detection time increases, the increase of human eye fatigue will lead to a decrease in detection accuracy. , the efficiency is poor.

In recent years, with the continuous development of computer technology and digital image processing technology, it is the current research need to obtain image recognition through machine vision. Machine vision is equivalent to the "eyes" of robots, which involves optical imaging, artificial intelligence, Mechatronics and other technologies are devices that use optical devices and non-contact sensors to obtain characteristic images of objects to be measured, and obtain information from the images through computers, process them in real time, and then realize detection and control devices. Machine vision detection technology has the advantages of safety and reliability, high detection accuracy, and can operate in complex environments. It has a wide range of applications, involving strip steel, mobile phone screens, textiles and many other industries.

At present, visual inspection mainly focuses on object surface detection, recognition and target tracking. Weld seam inspection mainly focuses on the surface of weld seam. The main problems are:

The problem that requires a combination of multiple detection methods; or it is necessary to take pictures of the weld from different angles, and it is necessary to ensure that the experimental environment is adequately illuminated. Although the weld detection can be better achieved, the experimental process relies too much on the repeated operation of the experimenter. , the problem of poor robustness; or the problem that only the detection of weld undercut defects can be realized, and the detection effect needs to be improved; or the defect expression effect is not intuitive enough, and the defect position can only be represented by the outline schematic diagram; or the use of point The laser is used for inspection, and the area needs to be inspected more than 10 times to determine the defect. The inspection process is complicated and needs to be further improved.

Therefore, the existing visual inspection technology still has problems such as low detection accuracy and low defect recognition rate.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the problems of low detection accuracy and low defect recognition rate in visual inspection technology, and propose a method for visual measurement of weld seam size and surface defect detection.

A method for visual measurement of weld seam size and surface defect detection, the method is realized by the following steps:

Step 1. Build a hardware detection system;

The hardware system includes a single-line laser transmitter, a CCD camera, a precision displacement table and a computer; the line laser emitted by the single-line laser transmitter is projected onto the surface of the weld, and the measured weld material on the precision displacement table is driven by a stepping motor to move to realize the single-line laser The transmitter scans the weld;

The second step is to calibrate the detection system;

The system calibration mainly includes two parts: camera calibration and structure cursor calibration; accurate internal parameters and external parameters of the camera are obtained through camera calibration. The internal parameters include the center point, focal length and distortion parameters of the camera. The external parameters are the relative camera in world coordinates Position; obtain the relative positional relationship between the structured light emitted by the single-line laser transmitter and the CCD camera through structured light calibration;

Step 3, using the built and calibrated detection system to obtain a structured light fringe image, and perform a preprocessing operation on the structured light fringe image;

The preprocessing mainly includes the grayscale change, binarization, image filtering, image segmentation, morphological processing and ROI extraction of the structured light fringe image;

Step 4. Structured light centerline positioning method and feature point extraction;

The improved sub-pixel localization algorithm is used to complete the centerline location and extraction; the geometric features of the centerline of the structured light are determined by extracting feature points;

Step 5. Use the three-dimensional reconstruction algorithm to complete the reconstruction and defect identification of the tested weld material;

Perform data analysis on the 3D reconstructed weld image, and judge whether the weld meets the requirements of use according to the obtained 3D reconstructed image, as well as the definition of weld size and defects. Including pores and undercuts on the weld surface.

Preferably, the process of calibrating the detection system described in step 2 is:

The camera projection measurement model is designed as shown in formula (2-1);

where n is a constant scale coefficient, [u, v] ^T is the image coordinate of the target point, R is a 3×3 rotation matrix, t is a 3×1 translation matrix, M ₁ , M ₂ are the internal and external parameters of the camera , f _x and f _y are the focal length parameters of the camera in the x and y directions respectively, c _x and c _y are the main points of imaging; [x _c , y _c , z _c ] ^T is the three-dimensional coordinate in the camera coordinate system;

则二者的数学关系式如式(2-2) 所示；The coordinates are (x, y), and the coordinates considering the distortion in the actual situation are

Then the mathematical relationship between the two is shown in formula (2-2);

Among them, _δx and _δy represent the nonlinear distortion value, and the relationship between the nonlinear distortion value and the non-distortion point is shown in formula (2-3);

where k ₁ and k ₂ are radial distortion coefficients;

After the punctuation of the camera is completed, the internal and external parameters of the camera are obtained, the light plane of the laser is calibrated, and the connection between the light plane and the image plane is established through the feature points on the light bar, so that the feature points on the light bar are obtained. Coordinate values on the image plane coordinate system; usually, the ball valve takes an image of a light bar at two different heights, and extracts the center point of the light bar as a feature point. In the image plane coordinate system, the feature point is (u _i , v _i ), the transformation equation of formula (2-4) can be obtained from the camera calibration parameter equation of formula (2-1);

为光平面中特征点的坐标，令M＝M₁M₂，且由相机标定可求得M，其表示如(2-5)所示；In the formula,

is the coordinate of the feature point in the light plane, let M=M ₁ M ₂ , and M can be obtained by the camera calibration, and its expression is shown in (2-5);

Expanding formula (2-4), formula (3-9) can be obtained;

Equation (3-10) can be obtained by removing n from (2-6);

Collecting two light strip feature points at different heights, and fitting all points into a light plane, the light plane equation of the light plane in the image plane coordinate system can be obtained, thereby obtaining the relative positional relationship between the laser and the camera.

Preferably, the image preprocessing step includes:

Using the triangulation method, the laser of the single-line laser is obliquely incident on the welding seam, the industrial CCD camera is located directly above the welding seam, the welding seam part is in the center of the field of view of the CCD camera, and the line laser emitted by the single-line laser is projected onto the surface of the object, and after reflection, The lens of the CCD camera receives the reflected light from the surface of the object to be measured, and forms a light bar on the lens of the CCD camera; based on the height difference between the weld and the master, the reflected light on the lens also shifts. According to the triangle similarity principle And some distance parameters inside the camera and between the camera and the laser can obtain the specific size of the weld;

Let the angle between the laser beam and the measured reference plane be γ, the angle between the reflected beam AA' and the normal line is θ, the angle between the reflected beam AA' and the normal line is θ, and the angle between the laser beam and the photosensitive unit inside the CCD is ω, and the distance from the incident light point A to the center point O of the lens is L1, the imaging distance between the imaging point A' and the center point of the lens is L2; make two vertical lines, and the vertical feet are C and D respectively; when the change in height difference between the surfaces of the object to be measured is y, the light spot on the corresponding CCD The moving distance is x; use the triangle similarity theorem to find the size of y according to formula (3-1);

将以上参数代入式(3-2)整理后可得到测量距离公式(1-2)；where |B′D|=x sinω, |BC|=|AB|sin(θ+γ), |OA′|=L2, |OA|=L1, |DA′|=xcosω, |AC|=|AB |cos(θ+γ),

After substituting the above parameters into formula (3-2), the measurement distance formula (1-2) can be obtained;

The image processing part specifically includes two parts: image preprocessing for structured light images and extraction of structured light feature points; among them, image preprocessing includes grayscale changes, filtering and denoising, threshold segmentation, morphological processing, etc.; structured light Feature extraction includes centerline extraction, feature point extraction, and finally three-dimensional reconstruction; the present application improves the traditional denoising algorithm to achieve effective noise removal;

First, the image to be tested is collected by the camera; the image is converted into a grayscale image that is easy to process, and the noise in the image is eliminated by using improved image filtering technology; then image segmentation is performed, the purpose of which is to separate the areas with the same characteristics in the image. Finally, the target part is extracted by classification; the purpose of morphological processing is to repair the broken structured light and ensure the continuity of structured light. The content of morphological processing is mainly corrosion and expansion; the most important purpose of ROI extraction is to reduce The burden of image processing increases the speed of image processing, and can avoid the interference of non-ROI areas.

Preferably, the three-dimensional reconstruction step is:

It is also important to select a suitable feature point extraction method; after obtaining the processed structured light image, the object to be measured moves in a specific direction and speed through a stepping motor on the worktable, and the CCD camera works at the same time to continuously butt the surface of the weld. Scanning is performed to obtain disparity maps, and finally 3D reconstruction is performed.

Preferably, the detection of the weld is mainly divided into two parts, one is the measurement of the surface size of the weld, including the measurement of the weld width and the excess height of the weld. Therefore, it is necessary to judge whether these physical quantities exceed the limit and determine whether it affects the quality of the weld; the second is to detect defects such as cracks and pores on the weld surface to determine whether the defects exist.

The beneficial effects of the present invention are:

The hardware system mainly includes single-line laser transmitter, CCD camera, precision displacement stage, computer image processing and other parts. The detection principle of the system is: under the condition that the environment is not affected by other conditions and human beings, the displacement stage moves in a specific direction and speed. The camera is used to convert the workpiece to be inspected into an image signal and send it to the computer for processing. Then, after image preprocessing, a clear and low-noise structured light image is obtained. After processing according to the three-dimensional measurement algorithm, the three-dimensional information of the weld is obtained and the weld is reconstructed. Defect judgment is carried out to determine whether it meets the requirements of the weld standard.

The invention designs and realizes a welding seam defect detection system on the metal surface by extracting the characteristic points of the welding seam and combining the characteristics of the welding seam defects. The specific goals it is expected to achieve are as follows:

1) Build a single-line structured light scanning 3D measurement system. Through system calibration, the laser and CCD camera can achieve the accuracy of structured light measurement, and the collected weld image has the characteristics of high resolution and low noise.

2) In image processing, improve the appropriate image denoising algorithm according to the actual situation to ensure that the image is clear and the noise is low.

3) Through the structured light 3D measurement system, the measurement of the weld width and residual height is completed; through the structured light 3D data, the 3D topography of the weld is constructed, and then the cracks, pores, undercuts, etc. are constructed through the selection of stripe feature points. Weld defects.

4) Improve the center fringe positioning algorithm, effectively extract the center line and feature points of the structured light image, lay the foundation for the three-dimensional reconstruction of the weld and defect identification, and realize the identification of weld defects. On a structured light image, the width of a single line of structured light will cover multiple pixel widths, and there is at least one pixel error. The improved sub-pixel positioning technology and corresponding algorithm can effectively improve the positioning accuracy of the center stripe.

5) Under the condition of ensuring the stability of the detection system and the high precision, and on the premise of reducing the amount of arithmetic operations as much as possible, it is guaranteed to achieve a higher operation speed.

Description of drawings

Fig. 1 is the principle schematic diagram of the method of the present invention;

Fig. 2 is the calibration flow chart involved in the present invention;

Fig. 3 is the oblique shot structure diagram that the present invention relates to;

Fig. 4 is the schematic diagram of the oblique type measuring beam involved in the present invention;

5 is a flowchart of the structured light map preprocessing involved in the present invention;

FIG. 6 is a flowchart of the three-dimensional reconstruction involved in the present invention.

Detailed ways

Specific implementation one:

A method for visual measurement of weld size and surface defect detection in this embodiment, as shown in FIG. 1 , is implemented through the following steps:

Step 1. Build a hardware detection system;

The hardware system includes a single-line laser transmitter, a CCD camera, a precision displacement table and a computer; the line laser emitted by the single-line laser transmitter is projected onto the surface of the weld, and the measured weld material on the precision displacement table is driven by a stepping motor to move to realize the single-line laser The transmitter scans the weld, and the CCD camera captures a structured light fringe image for each step of the stepper motor, which is used to: transmit the acquired structured light fringe image to the computer for subsequent image processing, and use the triangulation method The measurement principle obtains the three-dimensional data of the structured light fringe image;

The second step is to calibrate the detection system;

The system calibration mainly includes two parts: camera calibration and structure cursor calibration; accurate internal parameters and external parameters of the camera are obtained through camera calibration. The internal parameters include the center point, focal length and distortion parameters of the camera. The external parameters are the relative camera in world coordinates Position; obtain the relative positional relationship between the structured light emitted by the single-line laser transmitter and the CCD camera through structured light calibration, which provides an accurate parameter basis for subsequent 3D measurements;

Step 3, using the built and calibrated detection system to obtain a structured light fringe image, and perform a preprocessing operation on the structured light fringe image;

The preprocessing mainly includes the gray level change, binarization, image filtering, image segmentation, morphological processing and ROI (region of interest) extraction of the structured light fringe image;

Step 4. Structured light centerline positioning method and feature point extraction;

The extraction of centerline location is the key to improve the accuracy of 3D measurement and reconstruction. Since the preprocessed structured light stripes have a certain width, generally dozens of pixels wide, an improved sub-pixel localization algorithm is used to complete the extraction of the centerline to improve the 3D measurement accuracy and reduce the data At the same time, the connectivity and characteristics of the light bar cannot be destroyed; the geometric characteristics of the center line of the structured light are determined by extracting feature points, providing effective 3D data for 3D reconstruction;

Step 5. Use the three-dimensional reconstruction algorithm to complete the reconstruction and defect identification of the tested weld material;

Perform data analysis on the 3D reconstructed weld image, and judge whether the weld meets the requirements of use according to the obtained 3D reconstructed image and the definition of weld size and defects. The weld size includes parameters such as the width and height of the weld. , Defects include defects such as pores and undercuts on the weld surface.

Specific implementation two:

Different from the specific embodiment 1, in a method for visual measurement of weld seam size and surface defect detection in this embodiment, the process of calibrating the detection system described in step 2 is as follows:

Camera calibration is mainly divided into traditional calibration method, adaptive calibration method and active vision calibration method. The traditional camera calibration method is to obtain the world three-dimensional coordinates of the feature points required to calibrate the target through a one-dimensional, two-dimensional, or three-dimensional calibration target under the geometric imaging model of a certain camera, and then use the camera to shoot several objects. The photos of the calibration targets at different positions are obtained through image processing and a series of mathematical transformations to obtain the mapping relationship between the two-dimensional coordinates of the image and the three-dimensional coordinates of the world, and the internal and external parameters of the camera are obtained. The advantage is that the calibration accuracy is high, but a high-precision calibration target is required. The self-calibration method is to obtain the camera parameters by using the camera to shoot multiple different images of a static scene at different positions when it is not suitable to use the calibration target. The advantage of this method is that it has a wide range of applications and strong flexibility, but the calibration accuracy is low and it is easily affected by noise. The active vision camera calibration also does not need to calibrate the target, but uses the precision console to control the camera to perform specific movements such as translation, rotation, etc. In this process, multiple associated images are captured, and the motion parameters of these images and the camera are used to calculate Get the internal and external parameters of the camera. Although the method has good robustness and high calibration accuracy, the experimental conditions are harsh and sufficient movement space is required.

Based on the high requirements for welding seam measurement accuracy and the existing experimental conditions, the Zhang Zhengyou calibration method in the traditional camera calibration was selected to obtain the internal parameters of the camera. Then the nonlinear distortion factor is introduced into the calibration model to further improve the calibration accuracy. The camera projection measurement model is designed as shown in formula (2-1);

where n is a constant scale coefficient, [u, v] ^T is the image coordinate of the target point, R is a 3×3 rotation matrix, t is a 3×1 translation matrix, M ₁ , M ₂ are the internal and external parameters of the camera , f _x and f _y are the focal length parameters of the camera in the x and y directions, respectively, c _x and c _y are the imaging principal points (the intersection of the optical axis and the imaging plane); [x _c , y _c , z _c ] ^T is the camera coordinate 3D coordinates under the system;

则二者的数学关系式如式(2-2)所示；In the ideal case, the coordinate without considering the distortion is (x, y), and in the actual case, the coordinate considering the distortion is

Then the mathematical relationship between the two is shown in formula (2-2);

Among them, _δx and _δy represent the nonlinear distortion value, and the relationship between the nonlinear distortion value and the non-distortion point is shown in formula (2-3);

Among them, k ₁ and k ₂ are radial distortion coefficients. In the process of camera calibration, the nonlinear distortion factors also include centrifugal distortion and thin prism distortion. However, introducing too many nonlinear distortion parameters cannot improve the accuracy of camera calibration, but will lead to The calibration result is unstable, so this application only considers radial distortion;

After the punctuation of the camera is completed, the internal and external parameters of the camera are obtained, the light plane of the laser is calibrated, and the connection between the light plane and the image plane is established through the feature points on the light bar, so that the feature points on the light bar are obtained. Coordinate values on the image plane coordinate system; usually, the ball valve takes an image of a light bar at two different heights, and extracts the center point of the light bar as a feature point. In the image plane coordinate system, the feature point is (u _i , v _i ), the transformation equation of formula (2-4) can be obtained from the camera calibration parameter equation of formula (2-1);

为光平面中特征点的坐标，令M＝M₁M₂，且由相机标定可求得M，其表示如(2-5)所示；In the formula,

is the coordinate of the feature point in the light plane, let M=M ₁ M ₂ , and M can be obtained by the camera calibration, and its expression is shown in (2-5);

Expanding formula (2-4), formula (3-9) can be obtained;

Equation (3-10) can be obtained by removing n from (2-6);

Collect the feature points of the light strip at two different heights respectively, and fit all the points into a light plane, and then the light plane equation of the light plane in the image plane coordinate system can be obtained, so as to obtain the relative position relationship between the laser and the camera; The system calibration flow chart is shown in Figure 2.

Specific implementation three:

Different from the specific embodiment 2, in a method for visual measurement of weld seam size and surface defect detection in this embodiment, the image preprocessing steps include:

Using the triangulation method, the schematic diagram of the structure of its vision sensor is shown in Figure 3. The laser oblique incidence of the single-line laser irradiates the welding seam, the industrial CCD camera is located directly above the welding seam, and the welding seam part is in the center of the CCD camera's field of view, so as to reduce the distortion of the collected image and improve the measurement accuracy.

The line laser emitted by the single-line laser is projected onto the surface of the object, and after reflection, the lens of the CCD camera receives the reflected light from the surface of the object to be measured, and forms a light bar on the lens of the CCD camera; based on the height difference between the weld and the master, in the The reflected light on the lens is also displaced. According to the similarity principle of the triangle and some distance parameters between the camera and the laser, the specific size of the weld can be obtained; the principle of triangulation is shown in Figure 4;

Let the angle between the laser beam and the measured reference plane be γ, the angle between the reflected beam AA' and the normal line is θ, the angle between the reflected beam AA' and the normal line is θ, and the angle between the laser beam and the photosensitive unit inside the CCD is ω, and the distance from the incident light point A to the center point O of the lens is L1, the imaging distance between the imaging point A' and the center point of the lens is L2; make two vertical lines, and the vertical feet are C and D respectively; when the change in height difference between the surfaces of the object to be measured is y, the light spot on the corresponding CCD The moving distance is x; use the triangle similarity theorem to find the size of y according to formula (3-1);

将以上参数代入式(3-2)整理后可得到测量距离公式(1-2)；where |B′D|=x sinω, |BC|=|AB|sin(θ+γ), |OA′|=L2, |OA|=L1, |DA′|=xcosω, |AC|=|AB |cos(θ+γ),

After substituting the above parameters into formula (3-2), the measurement distance formula (1-2) can be obtained;

The image processing part specifically includes two parts: image preprocessing for structured light images and extraction of structured light feature points; among them, image preprocessing includes grayscale changes, filtering and denoising, threshold segmentation, morphological processing, etc.; structured light Feature extraction includes centerline extraction, feature point extraction, and finally three-dimensional reconstruction; the present application improves the traditional denoising algorithm to effectively remove noise; the algorithm flowchart for structured light image processing is shown in Figure 3.

First, the image to be tested is collected by the camera; the image is converted into a grayscale image that is easy to process, and the noise in the image is eliminated by using improved image filtering technology; then image segmentation is performed, the purpose of which is to separate the areas with the same characteristics in the image. Finally, the target part is extracted after classification; the purpose of morphological processing is to repair the broken structured light and ensure the continuity of structured light. The content of morphological processing is mainly corrosion and expansion; ROI (region of interest) extraction is the most important The purpose is to reduce the burden of image processing, improve the speed of image processing, and avoid the interference of non-ROI areas.

Specific implementation four:

The difference from the third embodiment is that in a method for visual measurement of weld seam size and surface defect detection in this embodiment, the three-dimensional reconstruction steps are:

The 3D reconstruction process is shown in Figure 4. After the preprocessed image is obtained through the above steps, a suitable algorithm is used to extract the centerline of the structured light image according to the characteristics of the image. Each method such as the curve fitting method has its advantages and disadvantages and usage scenarios. In order to improve the extraction accuracy of the centerline of structured light, this application proposes an improved sub-pixel localization algorithm; the extraction of feature points has the same purpose as the extraction of the centerline. It is to prepare for improving the accuracy of the 3D reconstructed image, so it is also important to select a suitable feature point extraction method; after obtaining the processed structured light image, the object to be measured passes through the stepper motor on the worktable at a specific direction and speed. Moving, while the CCD camera is working, the surface of the weld is continuously scanned to obtain a parallax map, and finally a three-dimensional reconstruction is performed.

Figure 4 3D reconstruction flow chart.

Specific implementation five:

The difference from the fourth embodiment is that, in a method for visual measurement of weld seam size and surface defect detection in this embodiment, the detection of weld seam is mainly divided into two parts, one is the measurement of weld seam surface size. , including measuring the weld width (melt width) and weld height (remaining height) of the weld. Uneven values of these two values will lead to uneven bearing capacity and stress at the weld, so it is necessary to judge whether these physical quantities exceed The second is to detect defects such as cracks and pores on the surface of the weld to determine whether the defects exist or not.

The embodiment of the present invention announces the preferred embodiment, but is not limited to this, those of ordinary skill in the art can easily understand the spirit of the present invention according to the above-mentioned embodiment, and make different extensions and changes, However, as long as they do not depart from the spirit of the present invention, they are all within the protection scope of the present invention.

Claims (5)

1. A method for visual measurement of weld seam size and surface defect detection, characterized in that: the method is realized by the following steps: Step 1. Build a hardware detection system; The hardware system includes a single-line laser transmitter, a CCD camera, a precision displacement table and a computer; the line laser emitted by the single-line laser transmitter is projected onto the surface of the weld, and the measured weld material on the precision displacement table is driven by a stepping motor to move to realize the single-line laser The transmitter scans the weld; The second step is to calibrate the detection system; The system calibration mainly includes two parts: camera calibration and structure cursor calibration; accurate internal parameters and external parameters of the camera are obtained through camera calibration. The internal parameters include the center point, focal length and distortion parameters of the camera. The external parameters are the relative camera in world coordinates Position; obtain the relative positional relationship between the structured light emitted by the single-line laser transmitter and the CCD camera through structured light calibration; Step 3, using the built and calibrated detection system to obtain a structured light fringe image, and perform a preprocessing operation on the structured light fringe image; The preprocessing mainly includes the grayscale change, binarization, image filtering, image segmentation, morphological processing and ROI extraction of the structured light fringe image; Step 4. Structured light centerline positioning method and feature point extraction; The improved sub-pixel localization algorithm is used to complete the centerline location and extraction; the geometric features of the centerline of the structured light are determined by extracting feature points; Step 5. Use the three-dimensional reconstruction algorithm to complete the reconstruction and defect identification of the tested weld material; Perform data analysis on the 3D reconstructed weld image, and judge whether the weld meets the requirements of use according to the obtained 3D reconstructed image, as well as the definition of weld size and defects. Including pores and undercuts on the weld surface. 2. a kind of weld size visual measurement and surface defect detection method according to claim 1, is characterized in that: the process of carrying out the calibration of the detection system described in step 2 is: The camera projection measurement model is designed as shown in formula (2-1);

where n is a constant scale coefficient, [u, v] ^T is the image coordinate of the target point, R is a 3×3 rotation matrix, t is a 3×1 translation matrix, M ₁ , M ₂ are the internal and external parameters of the camera , f _x and f _y are the focal length parameters of the camera in the x and y directions respectively, c _x and c _y are the main points of imaging; [x _c , y _c , z _c ] ^T is the three-dimensional coordinate in the camera coordinate system; The coordinates are (x, y), and the coordinates considering the distortion in the actual situation are

Then the mathematical relationship between the two is shown in formula (2-2);

Among them, _δx and _δy represent the nonlinear distortion value, and the relationship between the nonlinear distortion value and the non-distortion point is shown in formula (2-3);

where k ₁ and k ₂ are radial distortion coefficients; After the punctuation of the camera is completed, the internal and external parameters of the camera are obtained, the light plane of the laser is calibrated, and the connection between the light plane and the image plane is established through the feature points on the light bar, so that the feature points on the light bar are obtained. Coordinate values on the image plane coordinate system; usually, the ball valve takes an image of a light bar at two different heights, and extracts the center point of the light bar as a feature point. In the image plane coordinate system, the feature point is (u _i , v _i ), the transformation equation of formula (2-4) can be obtained from the camera calibration parameter equation of formula (2-1);

In the formula,

is the coordinate of the feature point in the light plane, let M=M ₁ M ₂ , and M can be obtained by the camera calibration, and its expression is shown in (2-5);

Expanding formula (2-4), formula (3-9) can be obtained;

Equation (3-10) can be obtained by removing n from (2-6);

Collecting two light strip feature points at different heights, and fitting all points into a light plane, the light plane equation of the light plane in the image plane coordinate system can be obtained, thereby obtaining the relative positional relationship between the laser and the camera. 3. A method for visual measurement of weld size and surface defect detection according to claim 2, wherein the image preprocessing step comprises: Using the triangulation method, the laser of the single-line laser is obliquely incident on the welding seam, the industrial CCD camera is located directly above the welding seam, the welding seam part is in the center of the field of view of the CCD camera, and the line laser emitted by the single-line laser is projected onto the surface of the object, and after reflection, The lens of the CCD camera receives the reflected light from the surface of the object to be measured, and forms a light bar on the lens of the CCD camera; based on the height difference between the weld and the master, the reflected light on the lens also shifts. According to the triangle similarity principle And some distance parameters inside the camera and between the camera and the laser can obtain the specific size of the weld; Let the angle between the laser beam and the measured reference plane be γ, the angle between the reflected beam AA' and the normal line is θ, the angle between the reflected beam AA' and the normal line is θ, and the angle between the laser beam and the photosensitive unit inside the CCD is ω, and the distance from the incident light point A to the center point O of the lens is L1, the imaging distance between the imaging point A' and the center point of the lens is L2; make two vertical lines, and the vertical feet are C and D respectively; when the change in height difference between the surfaces of the object to be measured is y, the light spot on the corresponding CCD The moving distance is x; use the triangle similarity theorem to find the size of y according to formula (3-1);

where |B′D|=x sin ω, |BC|=|AB|sin(θ+γ), |OA′|=L2, |OA|=L1, |DA′|=x cosω, |AC|= |AB|cos(θ+γ),

After substituting the above parameters into formula (3-2), the measurement distance formula (1-2) can be obtained;

The image processing part specifically includes two parts: image preprocessing for structured light images and extraction of structured light feature points; among them, image preprocessing includes grayscale changes, filtering and denoising, threshold segmentation, morphological processing, etc.; structured light Feature extraction includes centerline extraction, feature point extraction, and finally three-dimensional reconstruction; the present application improves the traditional denoising algorithm to achieve effective noise removal; First, the image to be tested is collected by the camera; the image is converted into a grayscale image that is easy to process, and the noise in the image is eliminated by using improved image filtering technology; then image segmentation is performed, the purpose of which is to separate the areas with the same characteristics in the image. Finally, the target part is extracted by classification; the purpose of morphological processing is to repair the broken structured light and ensure the continuity of structured light. The content of morphological processing is mainly corrosion and expansion; the most important purpose of ROI extraction is to reduce The burden of image processing increases the speed of image processing, and can avoid the interference of non-ROI areas. 4. a kind of weld size visual measurement and surface defect detection method according to claim 3, is characterized in that: described three-dimensional reconstruction step is: It is also important to select a suitable feature point extraction method; after obtaining the processed structured light image, the object to be measured moves in a specific direction and speed through a stepping motor on the worktable, and the CCD camera works at the same time to continuously butt the surface of the weld. Scanning is performed to obtain disparity maps, and finally 3D reconstruction is performed. 5. A method for visual measurement of weld seam size and surface defect detection according to claim 4, characterized in that: the described detection of the weld seam is mainly divided into two parts, one is the measurement of the surface size of the weld seam , including measuring the melted width and residual height of the weld. The unevenness of these two values will lead to uneven bearing capacity and stress at the weld. Therefore, it is necessary to judge whether these physical quantities exceed the limit and determine whether it affects the quality of the weld; two It is to detect defects such as cracks and pores on the surface of the weld to determine whether the defects exist.

Images