A method for detecting surface defects of liquid storage tank weld

Abstract

The present invention provides a method for detecting surface defects of a liquid reservoir weld, comprising a weld point cloud extraction step, obtaining a liquid reservoir point cloud set in a three-dimensional coordinate system to locate a weld area to obtain an expected point cloud; a weld point cloud correction step, calculating the current contour line curvature through a slope formula according to two different points on a single contour line, and performing plane correction on the expected point cloud according to the current contour line curvature to obtain a corrected point cloud; a template point cloud synthesis step, synthesizing a defect-free weld point cloud into a template point cloud; a weld point cloud registration step, registering the correction point cloud and the template point cloud to obtain a registered point cloud; a weld defect judgment step, judging the distance between the correction point cloud and the template point cloud in the Z-axis direction in the registered point cloud, and comparing it with a preset defect threshold to obtain the weld defect; the present invention has the advantages of being able to correct annular welds distorted by eccentric rotation, reducing the registration error through multiple weld point clouds and a scaling matrix, and achieving effective detection of weld surface defects.

Description

Technical Field

[0001] The invention relates to the technical field of weld detection, and more particularly to a method for detecting surface defects of a weld of a liquid storage vessel. Background Art

[0002] The liquid reservoir is an important part of the air conditioning system of low-energy vehicles at this stage. The sealing and reliability affect the performance and life of the air conditioning system. Welding is a key process in the production of liquid reservoirs, and the quality of welding directly affects the use effect of the liquid reservoir. However, since the welding process is prone to holes, arc pits, undercuts and other defects due to external reasons, which deteriorates its performance, it is very necessary to detect the surface and retrograde defects of the liquid reservoir weld. However, the current research on the identification of weld defects mainly focuses on non-destructive testing X-ray film detection, while the direct detection of weld surface defects, especially the line detection of annular welds, is still blank.

[0003] During the acquisition of the reservoir weld, due to mechanical vibration and the deflection of the axis when the cylindrical reservoir rotates, the depth image of the collected annular weld has distortion and height fluctuation problems, which greatly affects the consistency of the image. Some existing correction methods mainly use random sampling consistency methods, but such methods cannot effectively correct these existing problems. In the process of detecting weld defects, the existing methods mainly use a single contour line for detection. The arc crater defect height is inconsistent, and it is difficult to detect using a single contour line. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention aims to provide a method for detecting surface defects of a weld seam of a liquid reservoir.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for detecting surface defects of a liquid storage tank weld comprises the following steps:

[0007] The step of extracting the weld point cloud is to construct a three-dimensional coordinate system, obtain a point cloud set of the liquid reservoir in the three-dimensional coordinate system, and locate the weld area through an edge detection strategy to obtain the expected point cloud;

[0008] The step of correcting the weld point cloud is to obtain two different points on a single contour line in the liquid reservoir point cloud set, calculate the curvature of the current contour line through a slope formula, and perform plane correction on the expected point cloud according to the curvature of the current contour line to obtain a corrected point cloud;

[0009] Template point cloud synthesis step, obtaining defect-free weld point cloud and obtaining a point cloud set in the same area as the weld to be inspected as a template point cloud through a feature matching strategy;

[0010] A weld point cloud registration step, registering the correction point cloud and the template point cloud to obtain a registration point cloud;

[0011] The weld defect judgment step is to judge the distance between the correction point cloud and the template point cloud in the Z-axis direction in the registration point cloud, and compare it with a preset defect threshold to obtain the weld defect.

[0012] Furthermore, the plane correction includes calculating the distance value between two different points, and then obtaining the expected change of each point when the current contour line slope is zero according to the current contour line curvature and the distance value, performing a first correction on each contour line in the expected point cloud according to the expected change to obtain a first contour line group, and then performing a second correction on each contour line in the first contour line group according to the average height of the contour line plane as a reference height to obtain a corrected point cloud.

[0013] Furthermore, the slope formula is configured as:

[0014] ,

[0015] in, is the curvature of the current contour, is the number of contour lines, is a point on the contour line, and They are the height values ​​of two points on the current contour line with an interval of n on the z-axis;

[0016] The expected change amount is configured as:

[0017] ,

[0018] in, , , They are respectively the height of the contour point after the transformation, the height of the contour point before the change, and the changed value of the contour point.

[0019] Furthermore, the second correction configuration is:

[0020] ,

[0021] ,

[0022] ,

[0023] in, is the plane average height of each contour line in the first contour line group, is the compensation height of the contour line, is the second corrected height, The plane area for determining the reference height.

[0024] Furthermore, the feature matching strategy includes a rough matching step and a fine matching step.

[0025] In the coarse registration sub-step, the defect-free weld point cloud is gray-processed, and the weld position in the defect-free weld point cloud after gray-processing is preliminarily located by the SIFT method;

[0026] In the fine registration sub-step, the defect-free weld point cloud after the weld position is located is precisely registered by the ICP method, and the template point cloud is obtained by voxel downsampling and smoothing methods.

[0027] Furthermore, the weld point cloud registration step also includes a weld template adaptation sub-step.

[0028] The weld template adaptation sub-step scales the weld size in the template point cloud by a scaling matrix to obtain a template point cloud with the same weld size as that in the correction point cloud. The scaling matrix is ​​configured as:

[0029] ,

[0030] ,

[0031] ,

[0032] in, and The scaling matrices are The scaling factor of the Y-axis and Z-axis in the three-dimensional coordinate system. is the template point cloud before scaling transformation, is the template point cloud after scaling transformation, and They are the width and height of the weld in the corrected point cloud, and They are the width and height of the weld in the template point cloud respectively.

[0033] Furthermore, the weld defect judgment step includes a distance comparison strategy, which includes using the minimum spanning tree method to find the two points with the smallest distance in the XY plane in the template point cloud for each point in the correction point cloud in the registration point cloud, and then calculating the distance between the two points in the Z-axis direction as the point distance, and comparing it with a preset defect threshold. If the point distance is greater than the defect threshold, the point is classified as a weld defect point.

[0034] Furthermore, the distance comparison strategy is configured as follows:

[0035] ,

[0036] ,

[0037] ,

[0038] ,

[0039] ,

[0040] in, To correct any point in the registration point cloud, is the point set of the template point cloud after scaling transformation, Points in the template point cloud The point with the smallest distance in the XY plane, is the point distance, is the defect threshold, is the weld defect point set.

[0041] Furthermore, the edge detection strategy includes marking null value points in the liquid reservoir point cloud set, and fitting the first non-null value point in each row to obtain the liquid reservoir edge straight line.

[0042] Furthermore, in the weld point cloud extraction step, the expected point cloud is obtained by cutting the liquid reservoir point cloud set using the liquid reservoir edge straight line as a dividing line.

[0043] The beneficial effects of the present invention are as follows: through optical characteristics, the collected point cloud plane is corrected by calculating the slope of the contour line, and the surface defects of the reservoir weld are extracted using the template matching method. Specifically, the surface contour line information collected by the line laser camera during the rolling imaging of the reservoir is used to calculate the slope of the plane area of ​​a single contour line. According to the slope and the position information of each point on the contour line, the slopes of each contour line are unified, and the height of each contour line is adjusted to achieve plane correction. Then, multiple normal weld samples are aligned and fused into one sample, and then down-sampled and smoothed to produce a standard weld template. The standard weld template is scaled according to the width and height of the weld to be detected, and then the two are aligned. Then, the height information of each point between the weld to be detected and the standard weld template is calculated, and the defect threshold is set to achieve the extraction of the weld defect area, which can quickly and accurately complete the detection and feedback of weld surface defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of the method for detecting surface defects of a liquid reservoir weld of the present invention;

[0045] Figure 2 This is a comparison diagram of the edge extraction of the liquid reservoir and the ROI area cropping of the weld in the present invention;

[0046] Figure 3 It is a flow chart of the method for plane correction of collected point cloud in the present invention;

[0047] Figure 4 It is the effect diagram of the plane correction in the present invention;

[0048] Figure 5 It is the effect diagram of the weld registration and defect extraction of the liquid reservoir in the present invention. DETAILED DESCRIPTION

[0049] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. The same parts are represented by the same reference numerals. It should be noted that the words "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to directions in the accompanying drawings, and the words "bottom surface" and "top surface", "inner" and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0050] Due to mechanical vibration and the deflection of the axis when the cylindrical liquid reservoir rotates, the depth image of the collected annular weld has distortion and height fluctuation problems, which greatly affects the consistency of the image. Some existing correction methods mainly use random sampling consistency methods, but such methods cannot effectively correct these existing problems. In the process of detecting weld defects, the existing methods mainly use a single contour line for detection. The arc pit defect height is inconsistent and it is difficult to use a single contour line for detection. Therefore, the present invention designs such a liquid reservoir weld surface defect detection method, such as Figure 1 As shown, the following steps are included:

[0051] The step of extracting weld point cloud is to construct a three-dimensional coordinate system, obtain the reservoir point cloud set in the three-dimensional coordinate system, and locate the weld area through the edge detection strategy to obtain the expected point cloud. Specifically, Figure 2 As shown, since the collected point cloud is an ordered point cloud and the position of the liquid reservoir is fixed relative to the camera during the scanning process, the approximate position of the weld can be determined by the position of the non-null value point, and the point cloud data is processed line by line along the contour line direction, the position of the first non-null value point in each line is extracted, and a straight line is fitted according to the distribution of the extracted points. The position of the straight line is the position of the edge of the device body. Since the position of the weld is fixed relative to the edge of the device body, the area of ​​the weld is cropped by creating a region of interest (ROI) or the straight line of the edge of the liquid reservoir as a dividing line, and some redundant point clouds are removed to obtain the expected point cloud, and then the expected point cloud is processed by the existing statistical filtering method.

[0052] The step of welding point cloud correction is to obtain two different points on a single contour line in the reservoir point cloud set and calculate the curvature of the current contour line through the slope formula. The slope formula is configured as:

[0053] ,

[0054] in, is the curvature of the current contour, is the number of contour lines, i.e. the number of contour lines, belonging to the contour line set of the collected reservoir point cloud ,Right now , is the point on the contour line, including the point number, point and Point The range needs to meet , , It is The total number of points on the contour line, and They are the height values ​​of two points on the current contour line with an interval of n on the z-axis;

[0055] The expected point cloud is plane-corrected according to the curvature of the current contour line to obtain the corrected point cloud. Specifically, a fixed point on the contour line is taken to obtain the relative height of the current contour line. Since the distance between two points along the contour line of the ordered point cloud is the same, all points on the contour line can be updated according to the curvature and relative height of the current contour line to make all contour line curvatures and point pairs have the same height.

[0056] The plane correction process is as follows: Figure 3 As shown, it includes calculating the distance value between two different points, where the two selected different points belong to the points on the body plane that do not include the weld, and then obtaining the expected change of each point when the current contour line slope is zero according to the curvature of the current contour line and the distance value. The expected change is configured as:

[0057] ,

[0058] in, , , are the height of the contour point after the transformation, the height of the contour point before the transformation, and the value of the change of the contour point;

[0059] According to the expected change, each contour line in the expected point cloud is corrected for the first time to obtain the first contour line group. Since there is still a difference in the height of each contour line in the z-axis direction, each contour line in the first contour line group is corrected for the second time based on the average height of its contour line plane as the reference height to obtain the second contour line group. Each contour line in the second contour line group is composed of the midpoint of the corrected point cloud, and the second correction configuration is:

[0060] ,

[0061] ,

[0062] ,

[0063] in, is the plane average height of each contour line in the first contour line group, is the compensation height of the contour line, is the second corrected height, To determine the plane area of ​​the reference height, the first point on the contour line is generally selected, that is, ,like Figure 4 As shown, the originally tilted and distorted plane is restored to flatness.

[0064] For example: contour midpoint represents the area where the weld is above the plane, and the other points represent the plane area;

[0065] 1. Initial height

[0066] First contour line: ;

[0067] Second contour line: ;

[0068] 2. Slope calculation

[0069] Pick the first point and Point , interval point ;

[0070] ;

[0071] ;

[0072] 3. Calculate the change

[0073] According to the formula ;

[0074] First contour line: ;

[0075] Second contour line: ;

[0076] 4. Updated height

[0077] According to the formula ;

[0078] First contour line: ;

[0079] Second contour line: ;

[0080] 5. Highly unified adjustment

[0081] Target height Updated average of the first points of all contour lines:

[0082] ;

[0083] Calculate the height difference of each contour line :

[0084] First contour line: ;

[0085] Second contour line: ;

[0086] Final height after the second correction: ;

[0087] Result: First contour line: ;

[0088] First contour line: .

[0089] In the template point cloud synthesis step, the defect-free weld point cloud is obtained and a point cloud set in the same area as the weld to be inspected is obtained as the template point cloud through a feature matching strategy, wherein the feature matching strategy includes a coarse matching sub-step and a fine matching sub-step. In the coarse matching sub-step, the defect-free weld point cloud is gray-processed and the weld position in the gray-processed defect-free weld point cloud is preliminarily located by the SIFT method; in the fine matching sub-step, the defect-free weld point cloud after the weld position is located is finely aligned by the ICP method, and the template point cloud is obtained by voxel downsampling and smoothing methods.

[0090] In the weld point cloud registration step, the correction point cloud and the template point cloud are registered to obtain the registration point cloud. In order to ensure good registration capability for welds of different specifications and reduce the error after registration, the template weld needs to be adjusted according to the width and height of the correction point cloud before the correction point cloud is registered with the template point cloud. Therefore, the weld template adaptation sub-step is also included. The weld size in the template point cloud is scaled by the scaling matrix to obtain a template point cloud with the same weld size as the correction point cloud. The scaling matrix is ​​configured as:

[0091] ,

[0092] ,

[0093] ,

[0094] in, and The scaling matrices are The scaling factor of the Y-axis and Z-axis in the three-dimensional coordinate system. is the template point cloud before scaling transformation, is the template point cloud after scaling transformation, and They are the width and height of the weld in the corrected point cloud, and are the width and height of the weld in the template point cloud, respectively. Applying the scaling matrix to the template point cloud can significantly reduce the matching error caused by welds of different specifications. The registration process of the template point cloud and the weld point cloud to be detected is as follows: Figure 5 As shown, after registration, there is an obvious difference in the height between the corrected point cloud defect and the template point cloud area, and the defects can be extracted by setting the defect threshold. The above-mentioned corrected point clouds are all weld point clouds to be detected.

[0095] The weld defect judgment step is to judge the distance between the correction point cloud and the template point cloud in the Z-axis direction in the registration point cloud, and compare it with the preset defect threshold to obtain the weld defect; specifically, the weld defect judgment step includes a distance comparison strategy, which includes finding the two points with the smallest distance in the XY plane in the template point cloud for each point in the correction point cloud through the minimum spanning tree method in the registration point cloud, and then calculating the distance between the two points in the Z-axis direction as the point distance, and comparing it with the preset defect threshold. If the point distance is greater than the defect threshold, the point is classified as a weld defect point; wherein the distance comparison strategy is configured as follows:

[0096] ,

[0097] ,

[0098] ,

[0099] ,

[0100] ,

[0101] in, To correct any point in the registration point cloud, is the point set of the template point cloud after scaling transformation, Points in the template point cloud The point with the smallest distance in the XY plane, is the point distance, is the defect threshold, is the weld defect point set, As the point cloud of the weld being tested One point, in Less than Defect Set ,set up ,when When, the decision point It is a weld defect.

[0102] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for detecting surface defects of a liquid reservoir weld, characterized in that: The steps include: The step of extracting the weld point cloud is to construct a three-dimensional coordinate system, obtain a point cloud set of the liquid reservoir in the three-dimensional coordinate system, and locate the weld area through an edge detection strategy to obtain the expected point cloud; The step of correcting the weld point cloud is to obtain two different points on a single contour line in the liquid reservoir point cloud set, calculate the curvature of the current contour line through a slope formula, and perform plane correction on the expected point cloud according to the curvature of the current contour line to obtain a corrected point cloud; The plane correction includes calculating the distance value between two different points, and then obtaining the expected change of each point when the slope of the current contour line is zero according to the curvature of the current contour line and the distance value, performing a first correction on each contour line in the expected point cloud according to the expected change to obtain a first contour line group, and then performing a second correction on each contour line in the first contour line group according to the average height of the contour line plane as a reference height to obtain a corrected point cloud; Template point cloud synthesis step, obtaining defect-free weld point cloud and obtaining a point cloud set in the same area as the weld to be inspected as a template point cloud through a feature matching strategy; A weld point cloud registration step, registering the correction point cloud and the template point cloud to obtain a registration point cloud; The weld defect judgment step is to judge the distance between the correction point cloud and the template point cloud in the Z-axis direction in the registration point cloud, and compare it with a preset defect threshold to obtain the weld defect.

2. The method for detecting surface defects of a liquid reservoir weld according to claim 1, characterized in that: The slope formula is configured as: ， in, is the curvature of the current contour, is the number of contour lines, is a point on the contour line, and They are the height values ​​of two points on the current contour line with an interval of n on the z-axis; The expected change amount is configured as: ， in, , , They are respectively the height of the contour point after the transformation, the height of the contour point before the change, and the changed value of the contour point.

3. A method for detecting surface defects of a liquid reservoir weld according to claim 2, characterized in that: The second correction configuration is: ， ， ， in, is the plane average height of each contour line in the first contour line group, is the compensation height of the contour line, is the second corrected height, The plane area for determining the reference height.

4. The method for detecting surface defects of a liquid reservoir weld according to claim 1, characterized in that: The feature matching strategy includes a rough matching step and a fine matching step. In the coarse registration sub-step, the defect-free weld point cloud is gray-processed, and the weld position in the defect-free weld point cloud after gray-processing is preliminarily located by the SIFT method; In the fine registration sub-step, the defect-free weld point cloud after the weld position is located is precisely registered by the ICP method, and the template point cloud is obtained by voxel downsampling and smoothing methods.

5. A method for detecting surface defects of a liquid reservoir weld according to claim 1 or 3, characterized in that: The weld point cloud registration step also includes a weld template adaptation sub-step. The weld template adaptation sub-step scales the weld size in the template point cloud by a scaling matrix to obtain a template point cloud with the same weld size as that in the correction point cloud. The scaling matrix is ​​configured as: ， ， ， in, and The scaling matrices are The scaling factor of the Y-axis and Z-axis in the three-dimensional coordinate system. is the template point cloud before scaling transformation, is the template point cloud after scaling transformation, and They are the width and height of the weld in the corrected point cloud, and They are the width and height of the weld in the template point cloud respectively.

6. A method for detecting surface defects of a liquid reservoir weld according to claim 5, characterized in that: The weld defect judgment step includes a distance comparison strategy, which includes finding the two points with the smallest distance in the XY plane from the template point cloud for each point in the correction point cloud through the minimum spanning tree method in the registration point cloud, and then calculating the distance between the two points in the Z-axis direction as the point distance, and comparing it with a preset defect threshold. If the point distance is greater than the defect threshold, the point is classified as a weld defect point.

7. A method for detecting surface defects of a liquid reservoir weld according to claim 6, characterized in that: The distance comparison strategy is configured as follows: ， ， ， ， ， in, To correct any point in the registration point cloud, is the point set of the template point cloud after scaling transformation, Points in the template point cloud At the point with the smallest distance in the XY plane, is the point distance, is the defect threshold, is the weld defect point set.

8. The method for detecting surface defects of a liquid reservoir weld according to claim 1, characterized in that: The edge detection strategy includes marking null value points in the liquid reservoir point cloud set, and fitting the first non-null value point in each row to obtain the liquid reservoir edge straight line.

9. A method for detecting surface defects of a liquid reservoir weld according to claim 8, characterized in that: The weld point cloud extraction step is to cut the liquid reservoir point cloud set using the liquid reservoir edge straight line as a dividing line to obtain the expected point cloud.

Images