Steel mesh welding spot quality online detection method based on machine vision

Abstract

The invention discloses a steel mesh welding spot quality online detection method based on machine vision, relates to the technical field of steel mesh welding spot detection, and solves the problem that most detection methods adopt a global fixed threshold value to carry out welding spot region segmentation. The optimal threshold value is iteratively screened based on gray value probability distribution and inter-class difference indexes, compared with a traditional fixed threshold value method, welding spots and steel mesh surface areas can be distinguished more accurately, area misjudgment caused by uneven illumination and material reflection is avoided, and the welding spot area calibration accuracy is greatly improved; and meanwhile, contour features are constructed through the ratio of child-mother connecting line rotation length measurement, two-dimensional coordinates and midpoint verification are combined, irregularity of welding spot contours is converted into quantitative indexes, refined capture of welding spot form differences is achieved, distinguishing between normal welding spots and abnormal welding spots is more objective, and manual subjective judgment errors are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of steel mesh welding point detection, in particular to a steel mesh welding point quality online detection method based on machine vision. BACKGROUND

[0002] As a key structural component in the fields of highways, high-speed rails, roads and house buildings, the welding point quality of the steel mesh directly determines the connection strength, stability and service life of the overall product. In particular in the fields of highways, high-speed rails, roads and house buildings, welding point defects (such as virtual welding, continuous tin, missed welding and position deviation) may cause safety hazards. Therefore, efficient and accurate detection of the welding point quality of the steel mesh is a core control link in industrial production. At present, steel mesh welding point detection mainly relies on traditional manual detection and early machine vision detection. Among them, manual detection is completed by operators observing with the naked eye combined with magnifying glasses, calipers and other tools, which not only consumes a large amount of labor cost, but also has very low detection efficiency, and is difficult to adapt to the high-speed flow requirement of modern installation. More importantly, manual detection is easily affected by subjective experience, visual fatigue, environmental light and other factors, and has insufficient recognition accuracy for subtle defects (such as small pores and edge virtual sticking), high misjudgment rate and high omission rate, and cannot guarantee the consistency and reliability of the detection results. Although the existing early machine vision detection technology improves the efficiency to some extent, it still has significant technical bottlenecks. On the one hand, most detection methods use global fixed threshold for welding point region segmentation, which easily causes the welding point and the background region to be indistinctly divided when there are uneven illumination, material reflection difference or slight pollution on the surface of the steel mesh, resulting in inaccurate welding point calibration. On the other hand, the determination of welding point abnormalities mainly depends on single geometric features (such as area and perimeter), lacks fine quantitative description of the welding point contour shape, and is difficult to effectively identify implicit defects such as "irregular contour but size up to standard", and no standardized comparison reference is established in the abnormal area screening process, which easily causes "overdetermination" or "delayed determination". SUMMARY

[0003] In view of the deficiencies of the prior art, the present application provides a steel mesh welding point quality online detection method based on machine vision, which solves the problem of using global fixed threshold for welding point region segmentation in most detection methods.

[0004] To achieve the above purpose, the present application realizes the following technical scheme: a steel mesh welding point quality online detection method based on machine vision, comprising the following steps: Step one, the overall image of the steel mesh is acquired by the machine vision device, and the acquired overall image is grayed, the gray image is confirmed, the gray value associated with different pixel points in the gray image is verified, the partition threshold is determined, the welding point area in the gray image is confirmed according to the partition threshold, and the specific method is: According to the acquired overall image, the RGB value associated with each image point is confirmed, and then the gray value associated with the corresponding image point is confirmed by using: gray value = R x 0.114 + G x 0.587 + B x 0.299, and the image points are grayed according to different gray values associated with different image points, and the gray image associated with the corresponding overall image is confirmed; The different gray values associated with different gray points in the gray image are marked as H, the total number of pixels N (image width x height) of the gray image is confirmed synchronously, a group of gray values is randomly selected, and the total number Q of the existence of this gray value is determined H The occurrence probability P of this gray value H =Q H ÷N; Select any one gray value as threshold T, mark the gray area with H≤T as foreground area, otherwise, mark it as background area; Confirm the area ratio of the foreground area, using: Confirm the area ratio W0 associated with the foreground area, using: W1=1-W0 Confirm the area ratio W1 associated with the background area; Then calculate the average gray value associated with the foreground area and the background area, using: Confirm the average gray value U0 of the foreground area, and then use: Confirm the average gray value U1 associated with the background area; Confirm the average gray value of the entire gray image, using: UZ=W0×U0+W1×U1 Confirm the average gray value UZ of the entire gray image; Then use: Confirm the overall difference index of the entire gray image, select different threshold T, stop when the difference index is at the maximum value, and mark the selected threshold T as the partition threshold, and mark the gray area with gray value≤partition threshold as the welding point area in the gray image; Step two, the area of the several welding point areas marked in the gray image is verified, the edge contour of each welding point area is confirmed, the contour feature of each edge contour is verified, and the contour feature of each edge contour is verified. According to different contour features associated with different welding point areas, select the deviation area from the several welding point areas, and the specific method is: Based on the marked several soldering point areas in the gray image, the edge profile of each soldering point area is determined, and the two-dimensional coordinates associated with each profile point on the edge profile are confirmed by synchronously combining the two-dimensional coordinate system, and several groups of two-dimensional coordinates associated are processed by mean value, the mean value coordinates are confirmed, and are marked in the edge profile, and are recorded as the area midpoint of the corresponding soldering point area; Then a group of profile points is randomly selected from the edge profile, a connecting line between the profile point and the area midpoint is constructed, the connecting line is extended to another profile point on the edge profile, the connecting line is recorded as a sub-connecting line, a mother connecting line perpendicular to the sub-connecting line is constructed, and the two ends of the mother connecting line are located on the edge profile, the sub-connecting line and the mother connecting line are rotated according to the area midpoint, and the two ends of the sub-connecting line and the mother connecting line are located on the edge profile during the rotation process, the length ratio of the sub-connecting line and the mother connecting line is recorded synchronously, the length value of the sub-connecting line is recorded as L1, the length value of the mother connecting line is recorded as L2, the length ratio is L1 ÷ L2, and the maximum value is selected from the several groups of length ratios confirmed in different rotation processes, the maximum value is recorded as the profile feature associated with the corresponding soldering point area, and the profile features associated with different soldering point areas are determined; According to the different profile features associated with different soldering point areas, the several groups of profile features are sorted in the order from small to large, the feature sorting sequence is confirmed, a sequence segment is randomly selected in the feature sorting sequence, the density feature of the corresponding sequence segment is recorded, the minimum value and the maximum value of the sequence segment are determined, and the sequence difference value is determined, wherein the sequence difference value = maximum value - minimum value, and the total number G of profile features in the sequence segment is determined, and the density feature is recorded as Y1, wherein Y1 is a preset value; The different density features associated with different standard segments are confirmed, and a group of standard segments with the maximum total number G of profile features is selected from the confirmed different density features, the standard segment is recorded as the selected segment, and the soldering point area whose profile feature does not belong to the selected segment is recorded as the deviated area; Step three, the selected deviated area is selected from a group of standard areas in other soldering point areas, and the deviated area and the standard area are comprehensively compared and verified to identify whether the deviated area belongs to an abnormal area, and the specific method is: The deviated area is removed from the several soldering point areas to obtain several other soldering point areas, the profile features associated with the several other soldering point areas are processed by mean value, the feature mean value is confirmed, and a group of soldering point areas with the closest profile feature and feature mean value is selected from the other soldering point areas, and the soldering point area is recorded as the standard area; The region midpoint of the deviated region is coincided with the region midpoint of the standard region, and the deviated region is rotated, and in the rotation process, the coincided region of the deviated region and the standard region is recorded, and the region proportion of the coincided region in the deviated region is recorded synchronously, the region proportion associated with each group of rotation processes is determined in turn, and the maximum value is selected from the determined several groups of region proportions, and is recorded as QYmax, and whether QYmax satisfies QYmax >= 90% is identified, if yes, the deviated region is recorded as an abnormal region, otherwise, no marking is performed.

[0005] The application provides a steel mesh welding point quality online detection method based on machine vision. Compared with the prior art, the method has the following beneficial effects: Compared with the traditional fixed threshold method, the optimal threshold is iteratively screened based on the gray value probability distribution and the inter-class difference index, the welding point and the steel mesh surface region can be more accurately distinguished, the region misjudgment caused by uneven illumination and material reflection is avoided, and the welding point region calibration accuracy is greatly improved; meanwhile, the "sub-mother continuous line rotation length ratio" is used to construct the contour feature, the irregularity of the welding point contour is converted into a quantitative index by combining the two-dimensional coordinates and the midpoint check, the fine capture of the welding point shape difference is realized, the distinction between normal and abnormal welding points is more objective, and the manual subjective judgment error is reduced. The layered check logic of "first screening the deviated region, and then comparing with the standard region" is adopted, the abnormality checking range is initially narrowed through the contour feature, and then the "region midpoint coincidence + rotation maximum coincidence proportion" is used for secondary accurate determination, so that the subtle abnormalities are avoided from being missed, and the normal deviation is prevented from being misjudged as an abnormality; meanwhile, the key parameters (such as the density feature threshold Y1 and the coincidence proportion 90%) can be flexibly adjusted according to the steel mesh specifications and production requirements, the detection standards in different scenes are adapted, the practicality is high, and the diversified welding point quality control requirements can be effectively met. BRIEF DESCRIPTION OF DRAWINGS

[0006] Figure 1 The application method flowchart is shown in the figure. DETAILED DESCRIPTION

[0007] The technical solutions in the embodiments of the application will be clearly and completely described in combination with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0008] First embodiment Please refer to Figure 1The application provides a steel mesh welding point quality online detection method based on machine vision, comprising the following steps: Step one, the overall image of the steel mesh is acquired by the machine vision device, the acquired overall image is subjected to gray scale processing, the gray scale image is confirmed, the gray scale values associated with different pixel points in the gray scale image are subjected to feature verification, the partition threshold is determined, the welding point area in the gray scale image is confirmed according to the partition threshold, specifically, the so-called partition threshold is to confirm the difference degree between corresponding gray scale points according to the comprehensive performance between gray scale values, the reasonable threshold is confirmed according to the difference degree, and the welding point area existing in the steel mesh is one-by-one calibrated, and the welding point area and the surface area of the steel mesh have obvious difference; Step two, the edge contour of each welding point area is confirmed by subjecting the several welding point areas marked in the gray scale image to area verification, the edge contour is subjected to feature verification, the contour feature of each group of edge contours is confirmed, the deviated area is selected from the several welding point areas according to different contour features associated with different welding point areas, specifically, each different welding point area has different contour features, and under normal circumstances, the contour features associated with different welding point areas should be relatively consistent, when the contour features of a certain welding point area have large deviation, the deviated area can be directly determined according to such features, and the deviated area is subjected to subsequent abnormal judgment; Step three, the selected deviated area is selected from a group of standard areas in other welding point areas, and the deviated area and the standard area are subjected to comprehensive comparison and verification, whether the deviated area belongs to an abnormal area is identified, and real-time display is performed.

[0009] As the second embodiment of the application, mainly aiming at the confirmation process of the gray scale image and the confirmation process of the welding point area in step one, specifically comprising: The specific way of confirming the gray scale image is: According to the acquired overall image, the RGB value associated with each image point is confirmed, then the gray scale value associated with the corresponding image point is confirmed by using: gray scale value = R x 0.114 + G x 0.587 + B x 0.299, and the image points are subjected to gray scale processing according to different gray scale values associated with different image points, and the gray scale image associated with the corresponding overall image is confirmed, specifically, different image points have different gray scale values, which can effectively confirm the gray scale image associated with the corresponding overall image; The specific way of confirming the welding point area in the gray scale image is: The different gray scale values associated with different gray scale points in the gray scale image are denoted as H, the total number N (image width x height) of the gray scale image is confirmed synchronously, a group of gray scale values are randomly selected, and the total number Q of the gray scale values is determined H The occurrence probability P of the gray scale value H =QH ÷N; Select any one gray value as the threshold T, the gray area of H≤T is recorded as the foreground area, otherwise, it is recorded as the background area; Confirm the area ratio of the foreground area, use: Confirm the area ratio W0 associated with the foreground area, use: W1=1-W0 Confirm the area ratio W1 associated with the background area; Recalculate the average gray value associated with the foreground area and the background area, use: Confirm the average gray value U0 of the foreground area, and then use: Confirm the average gray value U1 associated with the background area; Confirm the average gray value of the entire gray image, use: UZ=W0×U0+W1×U1 Confirm the average gray value UZ of the entire gray image; Then use: Confirm the overall difference index of the entire gray image, by selecting different threshold T, stop when the difference index is at the maximum value, and record the selected threshold T as the partition threshold, and record the gray value ≤ the partition threshold as the weld point area, and mark it in the gray image in turn; Specifically, in the corresponding gray image, different image points have different gray values, and in the corresponding area, selecting a most reasonable threshold can effectively distinguish the weld point area and the steel mesh area in the corresponding image area, and further multiple verification of different weld point areas can mark the abnormal areas in multiple weld point areas, and complete the analysis and processing process of different weld point areas.

[0010] As a third embodiment of the present application, it mainly aims at the specific way of selecting the deviated area from the weld point area in step two, and the specific selection method includes the following steps: Based on the several weld point areas marked in the gray image, determine the edge contour of each weld point area, and simultaneously combine the two-dimensional coordinate system to confirm the two-dimensional coordinates associated with each contour point on the edge contour, and perform mean value processing on the associated several groups of two-dimensional coordinates to confirm the mean value coordinates and simultaneously mark them in the edge contour as the area center point of the corresponding weld point area; Then a group of contour points is randomly selected from the edge contour, a connecting line between the contour points and the region center point is constructed, the connecting line is extended to another contour point on the edge contour, the connecting line is recorded as a child connecting line, a parent connecting line perpendicular to the child connecting line is constructed, and the two end points of the parent connecting line are located on the edge contour, the child connecting line and the parent connecting line are rotated synchronously according to the region center point, and in the rotation process, the two end points of the child connecting line and the parent connecting line are located on the edge contour, the length ratio of the child connecting line and the parent connecting line is recorded synchronously, the length value of the child connecting line is recorded as L1, the length value of the parent connecting line is recorded as L2, and the length ratio is L1 ÷ L2, then the length ratios associated with different rotation processes are confirmed in turn, the maximum value is selected from the confirmed several groups of length ratios, and the selected maximum value is recorded as the contour feature associated with the corresponding weld point region, and the contour features associated with different weld point regions are determined, specifically, different weld point regions have different contour features, and in the corresponding weld point region, the rotation process of the child and parent connecting lines can be quickly determined according to the generated connecting line and perpendicular rotation, so as to realize the specific length ratio of the child and parent connecting lines, and complete the determination process of the corresponding contour feature, facilitating subsequent comparison and verification of several groups of weld point regions. The specific way of selecting the deviated region from the several weld point regions is: According to different contour features associated with different weld point regions, the multiple groups of contour features are sorted in the order from small to large, the feature sequence is confirmed, a sequence segment is randomly selected in the feature sequence, the density feature of the corresponding sequence segment is recorded, the minimum value and the maximum value of the sequence segment are determined, and the sequence difference value is determined, wherein the sequence difference value = maximum value - minimum value, and the total number G of contour features in the sequence segment is determined, that is, sequence difference value ÷ G = density feature, and the sequence segment that satisfies density feature ≤ Y1 is recorded as a standard segment, wherein Y1 is a preset value, and its specific value is determined by the operator according to experience; The different density features associated with different standard segments are confirmed, and the standard segment with the maximum total number G of contour features is selected from the confirmed different density features, the standard segment is recorded as a selected segment, and the weld point region whose contour feature does not belong to the selected segment is recorded as a deviated region; Specifically, the deviated region is a region that deviates greatly from other weld point regions, and the standard segment can be quickly determined according to the contour feature sorting and confirmation, and the value segment with the maximum total number of contour features is selected from the standard segment, and the corresponding value segment belongs to a most standard value segment; As the third embodiment of the present application, the confirmation process of the abnormal region in step three includes the following specific steps: The deviated area is removed from the several welding point areas to obtain several other welding point areas, the contour features associated with the several other welding point areas are subjected to mean value processing, the feature mean value is confirmed, a group of welding point areas with contour features closest to the feature mean value is selected from the other welding point areas, and the welding point area is recorded as a standard area; The area midpoint of the deviated area is overlapped with the area midpoint of the standard area, the deviated area is rotated, and the overlapping area of the deviated area and the standard area is recorded in the rotation process, the area proportion of the overlapping area in the deviated area is recorded synchronously, the area proportion associated with each group of rotation processes is determined in turn, the maximum value is selected from the determined several groups of area proportions, and is recorded as QYmax, whether QYmax satisfies QYmax≥90% is identified, if yes, the deviated area is recorded as an abnormal area, otherwise, no marking is performed. Specifically, in the corresponding evaluation process, if the deviated area and the standard area still have large differences when subjected to the overlapping check, it means that the corresponding welding point area has a large deviation, manual inspection is required, it is an abnormal area, and the welding point area needs to be re-welded.

[0011] Part of the data in the above formula is subjected to numerical calculation after dimension removal, and the contents not described in detail in the specification all belong to the prior art known to those skilled in the art.

[0012] The above embodiments are only used to illustrate the technical method of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A machine vision-based online inspection method for weld point quality of steel mesh, characterized in that, Includes the following steps: Step 1: Acquire the overall image of the steel mesh using machine vision equipment, and perform grayscale processing on the acquired overall image to confirm the grayscale image. Then, perform feature verification on the grayscale values ​​associated with different pixels in the grayscale image to determine the partition threshold. Based on the partition threshold, confirm the welding point area in the grayscale image. Step 2: Perform region verification on several solder joint areas marked in the grayscale image to confirm the edge contour of each solder joint area, and perform feature verification on the edge contour to confirm the contour features of each set of edge contours. Based on the different contour features associated with different solder joint areas, select the deviation area from several solder joint areas. Step 3: Select a set of standard areas from other solder joint areas for the selected deviation area, and compare and verify the deviation area with the standard areas to identify whether the deviation area belongs to the abnormal area.

2. The online inspection method for weld point quality of steel mesh based on machine vision according to claim 1, characterized in that, In step one, the specific method for confirming the grayscale image is as follows: Based on the acquired overall image, the RGB value associated with each image point is confirmed. Then, the gray value associated with the corresponding image point is confirmed using the formula: gray value = R×0.114 + G×0.587 + B×0.

299. Based on the different gray values ​​associated with different image points, the image points are converted to grayscale to confirm the grayscale image associated with the corresponding overall image.

3. The online inspection method for steel mesh weld quality based on machine vision according to claim 2, characterized in that, In step one, the specific method for identifying the solder joint area within the grayscale image is as follows: The different gray values ​​associated with different gray points within a grayscale image are denoted as H. The total number of pixels N (image width × height) of the grayscale image is simultaneously confirmed. A set of gray values ​​is randomly selected, and the total number Q of these gray values ​​is determined. H The probability P of this grayscale value appearing H =Q H ÷N; Select any gray value and denote it as the threshold T. The gray area with H≤T is denoteed as the foreground area, and the gray area with H≤T is denoteed as the background area. To determine the area proportion of the foreground region, the following method is used: Confirm the area proportion W0 associated with the foreground area, and use W1 = 1 - W0 to confirm the area proportion W1 associated with the background area; Next, calculate the average gray level associated with the foreground and background regions, using: After confirming the average gray level U0 of the foreground region, then use: Confirm the average gray level U1 associated with the background area; Next, the average gray level of the entire grayscale image is confirmed using the formula: UZ = W0 × U0 + W1 × U1. Re-adopted: Confirm the overall difference index of the entire grayscale image. Select different thresholds T until the difference index reaches its maximum value. Record the selected threshold T as the partition threshold. Record the grayscale areas with grayscale values ​​≤ the partition threshold as solder joint areas and mark them sequentially within the grayscale image.

4. The online inspection method for weld point quality of steel mesh based on machine vision according to claim 1, characterized in that, In step two, the specific method for confirming the contour features of the weld area is as follows: Based on several solder joint regions marked in the grayscale image, the edge contour of each solder joint region is determined, and simultaneously combined with the two-dimensional coordinate system, the two-dimensional coordinates associated with each contour point on the edge contour are confirmed. The average value of the associated two-dimensional coordinates is then processed to confirm the average coordinates, which are then simultaneously marked in the edge contour and recorded as the region center of the corresponding solder joint region. Next, randomly select a set of contour points from the edge contour, construct a line connecting this contour point to the center of the region, and extend this line to another contour point on the edge contour. This line is called the sub-line. Then, construct a set of parent lines perpendicular to the sub-lines, with both ends of the parent lines located on the edge contour. Rotate the sub-lines and parent lines synchronously according to the center of the region. During the rotation, both ends of the sub-lines and parent lines are located on the edge contour. Record the length ratio of the sub-lines and parent lines synchronously. Record the length value of the sub-lines as L1 and the length value of the parent lines as L2. Use the formula: length ratio = L1 ÷ L2. Then, confirm the length ratio associated with different rotation processes in turn. Select the maximum value from the confirmed sets of length ratios. Record the selected maximum value as the contour feature associated with the corresponding solder joint area. Determine the contour features associated with different solder joint areas.

5. The online inspection method for steel mesh weld quality based on machine vision according to claim 4, characterized in that, In step two, the specific method for selecting the deviation area from several solder joint areas is as follows: Based on the different contour features associated with different weld point areas, multiple sets of contour features are sorted in ascending order of value to confirm the feature sorting sequence. Within the feature sorting sequence, a sequence segment is randomly selected, and the density feature of the corresponding sequence segment is recorded. The minimum and maximum values ​​of the sequence segment are determined to determine the sequence difference, where the sequence difference = maximum value - minimum value. The total number of contour features G within the sequence segment is determined, using the formula: sequence difference ÷ G = density feature. Sequence segments that satisfy density feature ≤ Y1 are recorded as standard segments, where Y1 is a preset value. The different density features associated with different standard segments are identified, and from the identified different density features, the standard segments with the largest total number G of contour features are selected. This standard segment is recorded as the selected segment, and the solder joint area whose contour features do not belong to this selected segment is recorded as the deviation area.

6. The online inspection method for weld point quality of steel mesh based on machine vision according to claim 1, characterized in that, In step three, the specific method for confirming the abnormal area is as follows: The deviation areas are removed from several solder joint areas to obtain several other solder joint areas. The contour features associated with these other solder joint areas are averaged to confirm the feature mean. Then, the set of solder joint areas whose contour features are closest to the feature mean is selected from the other solder joint areas and this set of solder joint areas is recorded as the standard area. Align the midpoint of the deviation area with the midpoint of the standard area, and rotate the deviation area. During the rotation, record the overlap area between the deviation area and the standard area, and simultaneously record the percentage of the overlap area located within the deviation area. Determine the percentage of the area associated with each rotation process in turn, and select the maximum value from the determined percentages, denoted as QYmax. Identify whether QYmax satisfies: QYmax≥90%. If it does, then record this deviation area as an abnormal area.

7. The online inspection method for weld point quality of steel mesh based on machine vision according to claim 6, characterized in that, If QYmax does not meet the condition: QYmax≥90%, then no marking will be performed.

Images